ICTD2009 Proceedings

3rd International Conference on
Information and Communication Technologies and Development
PROCEEDINGS
April 17-19, 2009
Carnegie Mellon University in Qatar
Education City, Doha, Qatar
Host Organization:
Organizing Partner:
Platinum Sponsors:
Gold Sponsors:
Bronze Sponsors:
Media Partner:
Technical Sponsors:
Table of Contents
Introduction 1
A Global Empirical Evaluation of New Communication Technology 3

Use and Democratic Tendency
Victoria Stodden and Patrick Meier
A Review of the Research on Mobile Use by Micro and Small 17

Enterprises (MSEs)
Jonathan Donner and Marcela Escobari
An Evaluation of the Use of ICT within Primary Education in Malawi 27

David Hollow and Paola Masperi
Claim Mobile: Engaging Conflicting Stakeholder Requirements in 35

Healthcare in Uganda
Melissa R. Ho, Emmanuel K. Owusu and Paul Aoki
Computer Games in the Developing World: The Value of Non- 46

Instrumental Engagement with ICTs, or Taking Play Seriously
Beth E. Kolko and Cynthis Putnam
Content Creation and Dissemination by-and-for Users in Rural Areas 56

Sheetal K. Agarwal, Arun Kumar, Amit Anil Nanavati, Nitendra Rajput
E for Express: “Seeing” the Indian State through ICTD 66

Renee Kuriyan and Isha Ray
Evaluating the Accuracy of Data collection on Mobile Phones: A Study 74

of Forms, SMS, and Voice
Somani Patnaik, Emma Brunskill and William Thies
FOLKSOMAPS – Towards Community Driven Intelligent Maps for 85

Developing Regions
Arun Kumar, Dipanjan Chakraborty, Himanshu Chauhan
HIV Health Information Access using Spoken Dialogue Systems: 95

Touchtone vs. Speech
Aditi Sharma Grover, Madelaine Plauché, Etienne Barnard, Christiaan Kuun
ICT4What? – Using the Choice Framework to Operationalize the 108

Capability Approach to Development
Dorothea Kleine
ICTD for Healthcare in Ghana: Two Parallel Case Studies 118
Rowena Luk, Matei Zaharia, Melissa Ho, Brian Levine and Paul M. Aoki
Improving Child Literacy in Africa: Experiments with an Automated 129

Reading Tutor
G. Ayorkor Mills-Tettey, Jack Mostow, M. Bernadine Dias, Tracy Morrison
Sweet, Sarah M. Belousov, M. Frederick Dias, Haijun Gong
Improving Literacy in Rural India: Cellphone Games in an After-School 139

Program
Matthew Kim, Anuj Kumar, Shirley Jain, Akhil Mathur, and John Canny
Kelsa+: Digital Literacy for Low-Income Office Workers 150

Aishwarya Lakshmi Ratan, Sambit Satpathy, Lilian Zia, Kentaro Toyama,
Sean Blagsvedt, Udai Singh Pawar, Thanuja Subramaniam
Mapping the Dynamics of Social Enterprises and ICTD in Cambodia 163

Kelly Hutchinson and Alemayehu Molla
Political Incentives and Policy Outcomes: Who Benefits from Technology- 173
Enabled Service Centers?
Jennifer Bussell
Results from a Study of Impact of E-government Projects in India 183

Subhash C. Bhatnagar and Nupur Singh
The Contribution of User-Based Subsidies to the Impact and Sustainability 192

of Telecenters – the eCenter Project in Kyrgyzstan
Michael L. Best, Dhanaraj Thakur and Beth E. Kolko
Poster Papers 201
A Speech Enabled Indian Language Text-to-Braille Transliteration 201

System
Tirthankar Dasgupta and Anupam Basu
Analyzing Statistical Relationships between Global Indicators through 212

Visualization
Prabath Gunawardane, Erin Middleton, Suresh Lodha, Ben Crow and
James Davis
ATMosphere: A System for ATM Microdeposit Services in Rural Contexts 222

Michael Paik and Lakshminarayanan Subramanian
Building a Transportation Information System Using Only GPS and Basic 233
SMS Infrastructure
Ruth E. Anderson, Anthony Poon, Caitlin Lustig, Waylon Brunette, Gaetano
Borriello, Beth E. Kolko
Challenges in Health Information Systems Integration: Zanzibar 243

Experience
Edwin Nyella
Cross Technology Comparison for Information Services in Rural 252

Bangladesh
Faheem Hussain and Rahul Tongia
Decentralization, Clientelism and Popular Participation - Is There a 267

Role for ICTs to Improve Local Governance?
Björn-Sören Gigler
Design and Deployment of a Blood Safety Monitoring Tool 280

S. Thomas, A. Asuntogun, J. Pitman, B. Mulenga, and S. Vempala
Dimensions of IT Literacy in an Arab Region Study in Barkha(Oman) 288

Sherif M. Aziz
Emergency Communication and System Design: The Case of Indian 300

Ocean Tsunami
R. Chen, J. Coles, J. Lee, and F. R. Rao
Empowering Muslim Youth through Computer Education, Access, Use: 310

A Gender Analysis
Farida Khan and Rehanan Ghadially
eServices Provisioning in a Community Development Context through a 320

JADE MAS Platform
Mamello Thinyane, Alfredo Terzoli, Peter Clayton
Extending the Technology-Community-Management Model to Disaster 328

Recovery in Asia
Arul Chib and A.L.E. Komathi
Featherweight Multimedia for Information Dissemination 337

Gerry Chu, Sambit Satpathy, Kentaro Toyama, Rikin Gandhi, Ravin
Balakrishnan, S. Raghu Menon
ICT Governance in Higher Education: Case Study of the Rise and Fall 348
of Open Source in a gulf University
Sofiane M. Sahyraoui
ICTD State of the Union: Where Have We Reached and Where Are We 357
Headed?
Rabin Patra, Joyojeet Pal, Sergiu Nedevschi
Information Communication Techology and Sustainable Communities in 367

Africa: The Case of the Niger Delta Region of Nigeria (Feb. 2009)
Uduak A. Okon
Integrating Health Information Systems in Sierra Leone 379

Johan Sæbø, Edem Kwame Kossi, Romain Tohouri Golly-Kobrissa, Ola
Titlestad, Jørn Braa
Mobile Telephony Access and Usage in Africa 392

A. Chabossou, C. Stork, M. stork, Z. Zahonogo
Numeric Paper Forms for NGOs 406

Gusharan Singh, Leah Findlater, Kentaro Toyama, Scott Helmer, Rikin Gandhi,
Ravin Balakrishnan
Rajnikant’s Laptop: Computers and Development in Popular Indian 417

Cinema
Joyojeet Pal
Regulatory Independence and Wireless Market Development: A 427

Comparative Analysis of Two African Nations
Annemijn F. van Gorp and Carleen F. Maitland
Social Enterprises: A Vocational Entrepreneurship Framework for 437

Street Youth
Paul Javid, Kentaro Toyama, Manna Biswas
Speech vs. Touch-Tone: Telephony Interfaces for Information Access 447

by Low Literate Users
Jahanzeb Sherwani, Sooraj Palijo, Sarwat Mirza, Tanveer Ahmed, Nosheen
Ali, Roni Rosenfeld
The Case for SmartTrack 458

Michael Paik, Ashlesh Sharma, Arthur Meacham, Giulio Quarta, Philip
Trahanas, Brian Levine, Mary Ann Hopkins, Barbara Rapchak,
Lakshminarayanan Subramanian
Uses of Mobile Phones in Post-Conflict Liberia 468

Michael L. Best, Edem Wornyo, Thomas N. Smyth and John Etherton
Demos 478
An Automated Braille Writing Tutor with Multilingual Exercises 478

and Educational Games
M. Bernadine Dias, M. Freddie Dias, Sarah Belousov, Mohammed Kaleemur
Rahman, Saurabh Sanghvi, Imran Fanaswala, Wael Ghazzawi, Ameer
Abdulsalam, Noura El-Moughny, and S. Raghu Menon
Boosting European Market Access to Malian Mango Growers 479

Saskia Harmen
Creating a Mobile-Phone Based Geographic Surveillance System 480

for Asian Flu
Yibo Lin and Claire Heffernan
Design of a Blood Flow System 481

A. Osuntogun, S. Thomas, J. Pitman, S. Basavaraju, B. Mulenga and
S. Vempala
DISHA: DISease and Health Awareness for Children on Multiple 482

Input Devices
Mohit Jain, Aakar Gupta, Navkar Samdaria, Praveen Shekhar and Joyojeet
Pal
Freedom Fone: Dial-up Information Service 483

Bev Clark and Brenda Burrell
FrontlineSMS and Ushahidi – A Demo 484

Ken Banks and Erik Hersman
Global Youth Connectivity (GYC) – Using ICT for Peaceful Recovery 485
and Long-term Change
Anne Bertrand
Implementing E-Government Accessible to Illiterate Citizens 486

D. Kettani and A. El Mahdi
Improving Data Quality with Dynamic Forms 487

Kuang Chen, Harr Chen, Neil Conway, Heather Dolan, Joseph M.
Hellerstein, and Tapan S. Parikh
IWB4D – Interactive Whiteboards for Development 488

John Traxler and Lee Griffiths
Livestock, Learning and Diagnostics: New Directions in Veterinary 489
Tele-medicine
Jun Yu and Claire Heffernan
Metamouse: Multiple Mice for Legacy Applications 490

Kurtis Heimerl, Divya Ramachandran, Joyojeet Pal, Eric Brewer,
and Tapan Parikh
Mobile Phone Job Services: Linking Developing-country Youth with 491

Employers, via SMS
Amber Houssain, Mohammad Kilany, and Jacob Korenblum
MultiMath: Numeric Keypads for Math Learning on Shared Personal 492

Computers
Sunil Garg, Charlotte Robinson, Clint Tseng, Heather Underwood, Richard
Anderson, Joyojeet Pal
A New Generation of Open Source Data Collection Tools 493

Yaw Anokwa, Carl Hartung, Adam Lerer, Brian DeRenzi, Gaetano Borriello
RuralScope: An Information System for Tracking Rural Disbursements 494

Sai Gopal Thota, Rabin Ratra, Murali Medisetty, Sivananda Reddy, Vivek
Mungala, Joyojeet Pal
T-Cube Web Interface in support of Real-Time Bio-surveillance Program 495

Artur Bubrawski, Maheshkumar Sabhnani, Michael Knight, Michael Baysek,
Daniel Neill, Saswati Ray, Anna Michalska and Nuwan Waidyanatha
Web Search over Low Bandwidth 496

Jay Chen, Lakshminarayanan Subramanian, and Jinyang Li
Author Index 497

ICTD 2009 – Organizing Committee
Honorary Chairs
Dr. Hessa Sultan Al-Jaber Dr. Kentaro Toyama

Secretary General Assistant Managing Director
ictQATAR Microsoft Research India (MSRI)
Conference Chair
Dr. M. Bernardine Dias

Carnegie Mellon University
Program Committee Chairs
Dr. Richard Heeks Dr. Rahul Tongia

University of Manchester CSTEP, Bangalore
Advisory Board
Dr. Vallampadugai S. Arunachalam Dr. Francois Bar

Center for Study of Science, Technology, University of Southern California
and Policy (CSTEP), Bangalore
Dr. Michael Best Dr. Kenneth Keniston

Georgia Institute of Technology Massachusetts Institute of Technology
Dr. Balaji Parthasarathy Dr. Krithi Ramamritham

International Institute of Information Indian Institute of Technology (IIT),
Technology (IIIT), Bangalore Bom bay
Dr. Raj Reddy Dr. AnnaLee Saxenian

Carnegie Mellon University University of California, Berkeley
Dr. Kentaro Toyama Dr. Ernest Wilson

Microsoft Research India University of Southern California
Senior Program Committee Members
Dr. Francois Bar Dr. Michael Best

University of Southern California Georgia Institute of Technology
Dr. Eric Brewer Dr. Chris Coward

University of California, Berkeley University of Washington
Dr. Robert Davidson Dr. Hernan Galperin
City University of Hong Kong Universidad de San Andrés
Dr. Shirin Madon Dr. Alemayehu Molla

London School of Economics Royal Melbourne Institute of Technology
Dr. Balaji Parthasarathy Dr. Krithi Ramamritham

International Institute of Information Indian Institute of Technology (IIT)-Bombay
Technology (IIIT)-Bangalore
Dr. Kentaro Toyama Dr. Tim Unwin

Microsoft Research India Royal Holloway University of London
Panels and Workshops Chairs
Dr. Joseph Mertz Dr. Joyojeet Pal

Carnegie Mellon University University of Washington
Publications Chairs
Dr. Yonina Cooper Dr. Thrishantha Nanayakkara

Carnegie Mellon University in Qatar Harvard University and University of
Moratuwa, Sri Lanka
Poster Chairs
Dr. Faheem Hussain Dr. Tapan Parikh

Carnegie Mellon University in Qatar University of California, Berkeley
Demo Chairs
Dr. Khaled Harras Bill Thies

Carnegie Mellon University in Qatar Massachusetts Institute of Technology
Regional Chairs
Africa
Dr. Gary Marsden Dr. Tim Waema
University of Capetown, South Africa University of Nairobi, Kenya
Middle East
Dr. Adnan Abu Dayya Dr. Fouad Mrad
Qatar University American University of Beirut
East Asia
Dr. Jack Linchuan Qiu Rinalia Abdul Rahim
The Chinese University of Hong Kong Global Knowledge Partnership, Malaysia
South Asia
Anita Gurumurthy Dr. Umar Saif
IT for Change, India Lahore University of Management
Sciences, Pakistan
South and Central America

Dr. Nicolau Reinhard Dr. Osvaldo Rodriguez
Universidade de São Paulo, Brazil La Planta University, Argentina
North America
Dr. John Bennett Claudia Morrell
University of Colorado at Boulder Multinational Development of Women
in Technology
Europe
John Traxler
University of Wolverhampton, UK
Oceania
Christina Higa Dr. Esther Batiri Williams
University of Hawaii The University of the South Pacific
Technical Program Committee Members
Dr. Jessica Aalami Dr. Salam Abdallah

University of California, Berkeley Abu Dhabi University
Dr. Reuben Abraham Dr. Erwin Alampay

Indian School of Business, Hyderabad University of Philippines
Dr. Richard Anderson Dr. Peng HwaAng

University of Washington Nanyang Technological University
Akhtar Badshah Dr. V. Balaji

Microsoft ICRISAT
Dr. Anupam Basu Dr. John Bennett

Indian Institute of Technology University of Colorado, Boulder
(IIT)-Kharagpur, India
Dr. Subhash Bhatnagar Robert Bichler
Indian Institute of Management, University of Salzburg
Ahmedabad
Dr. John Canny Dr. Royal Colle

University of California, Berkeley Cornell University
Dr. Rahul De Dr. Andy Dearden

Indian Institute of Management, Bangalore Sheffield Hallam University
Dr. Uday Desai Dr. Antonio Diaz

Indian Institute of Technology University of Auckland
(IIT)-Bombay
Dr. Jonathan Donner Dr. Ayman Elnaggar

Microsoft Research Sultan Qaboos University
Dr. Kevin Fall Dr. Ping Gao

Intel/Berkeley University of Manchester
Dr. Alison Gillwald Dr. Gillian Green

University of Witwatersrand University of Bolton
Pat Hall Dr. Saskia Harmsen

Kathmandu University International Institute for Communication
and Development (IICD)
Dr. Claire Heffernan Dr. Bill Hefley

University of Reading Carnegie Mellon University
Dr. Heather Hudson Dr. Faheem Hussain

University of San Francisco Carnegie Mellon University in Qatar
Mahad Ibrahim Dr. Ashok Jhunjhunwala

University of California, Berkeley Indian Institute of Technology
(IIT)-Madras
Dr. Muhammadou M. O. Kah Matt Kam

American University of Nigeria University of California, Berkeley
Dr. Sherif Kamel Dr. Atreyi Kankanhalli

American University of Cairo National University of Singapore
Dr. Srinivasan Keshav Dr. G. R. Kiran

University of Waterloo London School of Economics
Dr. Dorothea Kleine Dr. Jim Koch
University of London Santa Clara University
Dr. Beth Kolko Richa Kumar

University of Washington Massachusetts Institute of Technology
Dr. Gillian Marcelle Dr. Victor Mbarika

WITS University Southern University and A&M College
Dr. Wagner Meira Dr. Michel Menou

UFMG – Federal University City University of London
at Minas Gerais
Dr. Harekrishna Misra Dr. Amit Mitra

Institute of Rural Management, Anand Cranfield University
Dr. Beda Mutagahywa Dr. Shrikant Naidu

University of Dar es Salaam Motorola Labs, India
Dr. Amit Nanavati Dr. Solomon Negash

IBM Kennesaw State University
Dr. Joyojeet Pal Dr. Tapan Parikh

University of Washington University of California, Berkeley
Dr. Fay Payton Francisco Proenza

North Carolina State University Food and Agriculture Organization
of the United Nations (FAO)
Dr. Ranjini Raghavendra Dr. Nicolau Reinhard

Indian Institute of Management, Bangalore University of Sao Paolo
Marijn Rijken Dr. Osvaldo Rodriguez

TNO (Netherlands Organisation for La Planta University
Applied Scientific Research)
Dr. Roni Rosenfeld Dr. Sofiane Sahraoui

Carnegie Mellon University American University of Sharjah
Dr. Tony Salvador Dr. Maung Sein

Intel University of Agder
Dr. Afzal Sher Jahanzeb Sherwani

Swedish Program on ICT in Developing Carnegie Mellon University
Regions (SPIDER)
Dr. Nirvikar Singh Dr. Hettie Soriyan
University of California, Santa Cruz Obafemi Awolowo University
Dr. Christoph Stork Dr. Eswaran Subrahmanian

Wits University CMU/NIST/CSTEP
Dr. Lakshmi(narayanan) Subramanian Bill Thies

Courant Institute of Mathematical Sciences Massachusetts Institute of Technology
Dr. Cathy Urquhart Dr. Carinade Villiers

University of Auckland University of Pretoria
Dr. Timothy Waema Dr. Ernest Wilson

University of Nairobi University of Southern California
Dr. Adel El Zaim

International Development Research
Centre (IDRC) Canada
Local Organizing Committee Chairs

Murry Evans Elaine Farah
Carnegie Mellon University in Qatar ictQatar
Erin Stewart
Publicity Coordinators
Noha Al Afifi Andrea Zrimsek
Carnegie Mellon University in Qatar Carnegie Mellon University in Qatar
Logistics Coordinators
Renee Barcelona Sarah Belousov
Carnegie Mellon University in Qatar Carnegie Mellon University
Shams Hasan Kara Nesimiuk
Carnegie Mellon University in Qatar Carnegie Mellon University in Qatar
Ermine Teves
Website Coordinators
M. Freddie Dias Daniel Freeman
Carnegie Mellon University Carnegie Mellon University
Local Organizing Committee Members

Ray Corcoran, Bob Gaus, Jim Gartner, Shamila Khader, Aaron Lyvers
1
Introduction to the Proceedings of ICTD2009
Welcome to the 3rd International Conference on Information and Communication

Technologies and Development (ICTD2009). It is with great pleasure that we present
the ICTD2009 proceedings, which include all of the full papers presented at the
conference in Doha, Qatar, held on 17-19 April 2009.
ICTD is the premier series of scholarly conferences on the use of ICTs for
development, spanning technical and social science domains. The call for papers
attracted a record 250 submissions. All papers were put through a double-blind peer-
review process. The Program Co-Chairs assigned papers to our Senior Program
Committee members who oversaw a review process involving three Program
Committee reviewers per paper: one with deep expertise about the subject matter;
another with broad background in the area; and one drawn from an altogether
different discipline. Our continuing hope is that this encourages a convergence of
vocabulary and ideas within the ICTD field, while maintaining the integrity of
different disciplines. The Senior Program Committee members then meta-reviewed
the papers. Authors were allowed a brief rebuttal to reviewer comments before final
acceptance decisions were made and revisions were finalized.
Ultimately, 19 papers were selected for oral presentation, and another 27 papers were
chosen as full papers for poster presentation; an acceptance rate of just over 18%.
These papers represent some of the best work being done in ICTD today. They focus
on a wide variety of development goals, and involve a broad and innovative range of
digital technologies. They draw from all continents of the global South, and focus on
all stages of the ICTD lifecycle: from readiness through design and adoption to use
and impact. They also tell us about all levels, from the individual through
communities and projects to ICTD programmes and policies. We hope that you will
find them an insightful, provocative, and informative contribution to our fast-growing
field of research and practice.
We wish to thank our keynote speakers William H. Gates, Chairman of Microsoft

Corporation and Co-Chairman of the Bill and Melinda Gates Foundation, and Carlos
A. Primo Braga, Director of the Economic Policy and Debt in the Poverty Reduction
and Economic Management Network (PREM) at The World Bank for their insightful
presentations. All the other conference presenters also deserve our gratitude for the
variety of content and insights they added to the conference program.
We also need to thank a number of people without whom the program could not have
been put together. First, we wish to thank our Honorary Chairs, Dr. Hessa Sultan Al-
Jaber, Secretary General of The Supreme Council of Information and Communication
Technology (ictQatar), and Dr. Kentaro Toyama, Assistant Managing Director of
Microsoft Research India (MSRI) for their tremendous support in making this
conference a success. We are of course deeply indebted to our Program Committee:
those who did the hard work of reviewing and our senior PC members who managed
the review process so effectively; all together it is they who ultimately steer the course
of ICTD. We thank the Regional Chairs for their assistance in promoting the
conference, and our Advisory Board which provided guidance and moral support;
particularly Kentaro Toyama as the guiding light of ICTD. We also thank the
Publication Chairs, Yonina Cooper and Thrishantha Nanayakkara, who made this
2
proceedings possible; and Faheem Hussain and Tapan Parikh, who carried the load in
organisation of the poster presentations.
The conference program for ICTD2009 went well beyond papers, and was
significantly enhanced by demonstrations, organised by Bill Thies and Khaled Harras,
and by a series of panels and workshops, organised by Joe Mertz and Joyojeet Pal.
We are very grateful for their input and hard work.
This conference would not have been possible without the tireless efforts of a number
of organizers and volunteers, notably our local organizing committee chairs Elaine
Farah, Murry Evans, and Erin Stewart, and the many others on the local organizing
committee, especially Dean Charles Thorpe, Sarah Belousov, Ermine Teves, Renee
Barcelona, Kara Nesimiuk, Andy Zrimsek, Noha Al Afifi, Shams Hassan, Aaron
Lyvers, Shamila Khader, Ray Corcoran, Bob Gaus, Freddie Dias, Daniel Freeman,
and Jim Gartner. If there are others we did not name explicitly, it is our lapse.
The success of the conference is in part due to our many sponsors and partners. We
are extremely grateful to our organizing partner, The Supreme Council of Information
and Communication Technology (ictQatar), to our media partner Al Jazeera
Children’s Channel (JCC), to our technical sponsors IEEE and ACM, and to our
financial sponsors, the Qatar National Research Fund (Platinum Sponsor); Canada’s
International Development Research Centre (Platinum Sponsor); Qatar Telecom
(Platinum Sponsor); ExxonMobil (Gold Sponsor); Microsoft Corporation (Gold
Sponsor); IBM (Bronze Sponsor); the Computer Science program at Carnegie Mellon
University in Qatar (Bronze Sponsor); and other sponsors who asked not to be
publicly acknowledged.
Finally, we are indebted to Carnegie Mellon University in Qatar for hosting the
conference at their campus in Education City, and to the numerous employees of
Carnegie Mellon University (in both the Pittsburgh and Doha campuses), especially
the TechBridgeWorld team, who truly made this conference a success.
Thanks to everyone who contributed to ICTD2009 in so many ways, and thank you
for participating! We are confident you will find the following papers, abstracts, and
information not only interesting and useful, but the seeds for further research,
innovation, and developmental impact.
Richard Heeks, University of Manchester M. Bernardine Dias, Carnegie Mellon

Rahul Tongia, CSTEP, Bangalore University
Program Chairs Conference Chair
3
A Global Empirical Evaluation of New

Communication Technology Use and Democratic
Tendency
Victoria Stodden Patrick Meier
Berkman Center for Internet and Society Berkman Center for Internet and Society
Harvard Law School Harvard Law School
Cambridge, MA 02138 Cambridge, MA 02138
vcs@stanford.edu patrick.meier@tufts.edu
Abstract—Is the dramatic increase in Internet use associated democracy and regime transitions literature. To be sure, “the
with a commensurate rise in democracy? Few previous studies trouble with the zealots of technology as an instrument of
have drawn on multiple perception-based measures of governance democratic liberation is not that they misconceive technology
to assess the Internet’s effects on the process of democratization.
This paper uses perception-based time series data on “Voice & but that they fail to understand democracy” [2]. ‘In other
Accountability,” “Political Stability,” and “Rule of Law” to pro- words, “it turns out there is no simple general answer to
vide insights into democratic tendency. The results of regression the question: Is the technology democratizing?’ until we have
analysis suggest that the level of “Voice & Accountability” in a made clear what sort of democracy we intend.” We address this
country increases with Internet use, while the level of “Political question first before proceeding with a more detailed literature
Stability” decreases with increasing Internet use. Additionally,
Internet use was found to increase significantly for countries review.
with increasing levels of “Voice & Accountability.” In contrast, Barber’s notion of “strong democracy” comprises the careful
“Rule of Law” was not significantly affected by a country’s level and prudent judgment of citizens who participate in deliber-
of Internet use. Increasing cell phone use did not seem to affect ative, self-governing communities. Schmitter and Karl write
either “Voice & Accountability,” “Political Stability” or “Rule that, “modern political democracy is a system of governance
of Law.” In turn, cell phone use was not affected by any of
these three measures of democratic tendency. When limiting our in which rulers are held accountable for their actions in the
analysis to autocratic regimes, we noted a significant negative public realm by citizens, acting indirectly through the compe-
effect of Internet and cell phone use on “Political Stability” and tition and cooperation of their elected representatives”[3]. The
found that the “Rule of Law” and “Political Stability” metrics two authors emphasize that citizens are the most distinctive
drove ICT adoption. element in democracies. “All regimes have rulers and a public
Index terms—cell phone, democracy, fixed effects model, ICT,
realm, but only to the extent that they are democratic do
internet
they have citizens” [3]. In contemporary studies of democracy
and particularly in pluralist theory, “a vibrant civil society
I. I NTRODUCTION is usually regarded as an essential for good governance
Does the globalization of the Internet have a democratizing and effective democratic consolidation” [4]. In other words,
effect? The question has already been posed by numerous regular elections are not sufficient. As Zakaria noted, illiberal
studies but these have largely taken the form of qualitative case democracies have free elections but citizens remain cut off
studies and/or large theoretical analyses. In terms of a rigorous, from real power due to the lack of civil liberties [5].
quantitative establishment of the democratization effects of the Huber et al. write that the most basic feature of democracy
Internet, however, the jury is still out [1]. At the heart of is power sharing [6]. They identify three clusters of power
this debate, moreover, lies a more fundamental question about as primarily relevant for the chances of democracy: (1) the
the essence of democracy. In fact, “unless we are clear about balance of power in civil society; (2) the balance of power
what democracy means to us, and what kind of democracy we between state and society; and (3) the transnational balance of
envision, technology is as likely to stunt as to enhance the civic power that shape the first two and constrain political decision-
polity” [2]. The purpose of this paper is to contribute more making. By remaining diverse and independent of the state,
rigorous data-driven analysis to the literature on Internet and political participation by civil society acts as a channel of
democracy since “there is no doubt that rigorous and data- public voice and accountability, and a way of challenging
driven analysis of this relationship will benefit scholars and and checking the unbridled power of authoritarian regimes
policymakers alike” [1]. [4]. The structure of state-society relations is equally relevant
Previous research on the topic of Internet and democracy for democracy. As Huber et al. note, “the power of the state
can be characterized as lacking a serious perusal of the needs to be counterbalanced by the organizational strength of
4
the civil society to make democracy possible; the state must when Internet access and cell phone use is significantly more
not be so strong and autonomous from all social forces as prevalent and globalized than in the 1990s. If a statistically
to overpower civil society and rule without accountability.” significant relationship between Internet and democracy does
Clearly then a governing body that fails to follow the “rule exist, then it is more likely to manifest itself now and not in the
of law,” should not be considered democratic [3]. These 1990s. Third, we draw on both Internet and cell phone data per
elements of democracy are not sufficient conditions for a stable 100 inhabitants per 181 countries to assess the impact of the
democracy, but they are necessary and indispensable to the information revolution on democratization. We use regression
persistence of democratic governance. analysis to determine whether Internet or cell phone use has
The italicized terms above represent the fundamentals be- had a correlative effect on measures of democratic tendency.
hind the sort of democracy we intend: active citizen par- We also model whether the collection of democratic measures
ticipation, good governance, accountability, power sharing, has had a correlative effect on Internet or cell phone use.
balance of power and rule of law. The few quantitative studies The paper is structured as follows: the first section reviews
that do exist on Internet and democracy tend to aggregate the current debate and literature on Internet and democracy.
these fundamentals of democracy into a single index. Doing The second section explicates the datasets used in this study
so means these lose important information on how these while the third section formalizes the statistical models em-
individual components of democracy may be affected by the ployed in the regression analysis. Section four reviews the
growing prevalence of global Internet access. Furthermore, results and provides an interpretation of the findings. The fifth
past quantitative and qualitative studies tend to focus primarily and final section concludes the study.
on the impact of the Internet on established democracies. They
also focus on the 1990s almost exclusively, a serious limitation II. L ITERATURE R EVIEW
that remains surprisingly understated in the literature. Equally The Internet and democracy literature comprises two distinct
problematic in the current literature is the interchangeable schools of thought each comprising a host of qualitative
use of the terms “Internet” and “information revolution.” The research and some quantitative inquiry. In this section we
terms are purposefully not differentiated on the basis that the review in some detail the qualitative and quantitative literatures
predominant feature of the information society is the spread that have contributed to both schools of thought over the past
of the Internet. While this is true of Western democracies, it ten to fifteen years. In so doing, we compare and contrast the
is certainly not true for the majority of developing, nondemo- main arguments along with the respective findings. As noted
cratic countries, where cell phones are the most widely spread in the introduction, one common shortcoming of the Internet
communication technology after radios [7]. Indeed, the irony is and democracy literature is the tendency to oversimplify our
that “those who might most benefit from the net’s democratic understanding of democracy. The purpose of this literature
and informational potential are least likely to have access to review is thus to redress this gap evident in previous studies.
it, the tools to gain access, or the educational background to The first school of thought is often referred to as the
take advantage of the tools” [2]. more populist school of thought. This strand of the literature
This paper seeks to redress each of these shortcomings. subscribes to the viewpoint that the Internet has democratic
First, since the boundaries of the term “democracy,” and how relevance and impact [8], [9], [10], [11]. According to these
it is measured, is subject to lively debate, we use multiple authors, the Internet will decentralize access to communication
perception-based measures of governance for our dependent and information while increasing citizen access [12]. Best
variables. Governance indicators provide a better set of proxies and Wade write that “the Internet’s collective characteristics
for the sort of democracy we intend as identified above. We (e.g., low cost, multidirectional capability, etc.), helps make
therefore draw on the following three World Bank indicators: this possible.” We first review the qualitative literature that
(1) Voice and Accountability (VA) measuring perceptions comprises this school of thought followed by quantitative
of the extent to which a country’s citizens are able to par- studies.
ticipate in selecting their government, as well as freedom Dahl previously observed that telecommunications tech-
of expression, freedom of association, and a free media; (2) nologies have a key role in making possible the advanced
Rule of Law (RL) measuring perceptions of the extent to democratic country, where policy is firmly anchored in the
which agents have confidence in and abide by the rules of judgment of the “demos” [13]. In his list of the procedural
society, and in particular the quality of contract enforcement, minimal conditions that must be present for modern political
property rights, the police, and the courts, as well as the democracy to exist, Dahl thus argues that citizens should
likelihood of crime and violence; and (3). Political Stability have the right to seek out alternative sources of information.
and Absence of Violence (PS) measuring perceptions of Rheingold has dubbed the Internet as “the great equalizer”
the likelihood that the government will be destabilized or because it can “equalize the balance of power between citizens
overthrown by unconstitutional or violent means, including and power barons” [14]. The idea here “is that the Internet
politically motivated violence and terrorism. These metrics are will serve as a mass audience, and will politicize them in the
drawn from the World Bank Governance Indicators Research process” [12]. Anderson et al. draw on sociological research
Database (see http://info.worldbank.org/governance/wgi/). to show that electronic networks lead to the “break-down of
Second, we draw on data from 2000 through 2006, a time status-based social structures” and “increased participation in
5
discussion, decision-making, and task processes by those who for example). At the international level, the authors make the
typically are political or economically disadvantaged” [15]. bold claim that the Internet has “contributed to the rise of a
Etzioni sees in the Internet the possibility of an advancement more multicentric world structure in which nation-states have
of the state of public affairs through “teledemocracy” [16]. seen their preeminence lessen and non-governmental actors
Other scholars claim that the Internet will “enable a Jeffer- take the stage” [1]. The salient point here is that groups and
sonian revolution” [17]. Grossman argues that a “third epoch individuals can far more efficiently form coalitions of conse-
of democracy is arriving by the hand of technology,” and quence with a range of powerful collectives. As Best and Wade
claims that a “new, hybrid electronic republic’ is now forming rightly note, there are obvious democratic elements to this,
to displace the creation of Montesquieu, Locke, Madison and including the need for “nation-states to provide democratic
their contemporaries” [18], [12]. The practice of “electronic rights to their citizens so as to build legitimacy on the global
democracy,” according to Browning, will differ substantially stage” [1].
from previous renderings of democracy [19]. Perhaps the most We now turn to the quantitative studies that comprise the
provocative claim associated with this more populist school of first school of thought. One of the earlier statistical studies
thought is the one made by Negroponte, who argues that the on this side of the literature was carried out by Kedzie,
nation-state will evaporate as a consequence of the information who provides an account of how information communication
revolution [20]. Snider suggests that citizens need only be technologies contributed to the “third wave” of democracy
potentially informed in order to hold government accountable [26]. Prior to the fall of the Soviet Union and the prolifer-
[21]. ation of new democracies in Eurasia, the mainstream theory
Barber opines that by favoring decentralization, the mul- of democratization held that democracy followed economic
tiplication of choice, and consumer sovereignty, new tech- growth and development [27]. To be sure, one of the few robust
nologies such as the internet or cell phone have already, findings in the literature is that democracy is more likely in
albeit inadvertently, benefited democratic political culture [2]. more developed countries [28], [29], [30]. Longdregan and
According to Barber, “democracy is a form of government Poole have also shown that the most significant predictor of
that depends on information and communication. It is obvious transitions to authoritarianism is poverty [31], [32]. “In short,
then that new technologies of information and communications after 20 years of observation and analysis during the third wave
can be nurturing to democracy. They can challenge passivity, of academic interest in democratization, we can be reasonably
they can enhance information equality, they can overcome certain that a positive relationship between development and
sectarianism and prejudice, and they can facilitate participation democracy exists, though we do not know why” [33].
in deliberative political processes.” Hill and Hughes argue that Kedzie, however, was more interested in testing another
those who subscribe to the populist school have reasons to be potential causal mechanism, the “dictator’s dilemma” hypoth-
optimistic: “If the mere fact that political discourse against esis, which suggests that the globalization of markets places
repressive governments is taking place is a good in itself, then pressure on authoritarian regimes to keep their countries’
the utopians have reason to celebrate [22]. Perhaps the Internet communication borders open. He reasoned that the ensuing
will bring about a wider democratic revolution in the world” massive flow of information would not only allow for “the
[22]. efficient passage of commercial information, but also for more
Bimber is more cautious, arguing that while the Internet is ’democratic’ information” [1]. As Bimber observes, the most
accelerating the process of issue group formation and action important predictions about the Internet’s impact on politics
(in America), the structure of political power has not been amount to “causal claims regarding the effect of information
revolutionized or qualitatively transformed into a new epoch of flow on political participation and the organization of interests”
democracy [12]. According to Bimber, other scholars believe [12]. Other scholars have made related arguments. Webster,
that the Internet has a “transformative potential” because it for example, writes that the Internet has helped to facilitate
facilitates a kind of “one-to-one interaction among citizens a new form of capitalism called “information capitalism” in
and between citizens and government.” Along these lines, which global labor markets require highly flexible workers
some scholars such as Corrado and Firestone write that the who continuously adapt and learn [34]. Regimes that impose
Internet has the potential to promote “unmediated” commu- restrictions on information capitalism forgo the financial re-
nication and thereby decrease citizens’ reliance on officials turns possible by tapping into the information economy [35],
and organizations [23]. In sum, what distinguishes the populist [36], [37].
enthusiasm for the Internet is the “idea that elites and political In his study, Kedzie employs regression analysis to compare
intermediaries will grow less important” [12]. how much of the variation in democracy is explained by both
In the more contemporary, qualitative literature, Steele and traditional predictors of democracy and the strength of Internet
Stein, argue that the Internet amplifies trends in international diffusion by drawing on data from 144 countries [26]. For
relations [24]. Rosenau and Johnson address the impact of the his set of control variables, Kedzie included more traditional
Internet at both the individual and international level [25]. At predictors of democracy including economic development,
the individual level, the authors argue that the Internet can be education, human development and health. He also included
used as a tool to organize collectively to effect social and polit- indicators of pre-Internet information communication tech-
ical change around the world (see http://www.DigiActive.org, nologies (ICTs). His results suggest that the Internet is indeed
6
a strong predictor of democracy, more so than traditional variation in the Freedom House data, which makes meaningful
determinants of democracy. In a follow up study, Richards statistical analysis more difficult.
assessed the relationship between the Internet and physical In contrast to the populist literature, the second school
integrity [38]. His findings support Kedzie’s. However, the of thought disputes the majority of claims that exist vis-à-vis
latter study faces an important limitation since Kedzie’s (rather the relationship between Internet and democracy. The counter-
simple) longitudinal analysis draws on data from 1993. At arguments are based on both qualitative and quantitative re-
this point during the early 1990s, the Internet was hardly search. In terms of qualitative research, several scholars argue
globalized. that the Internet is merely an extension of the ruling class and
Best and Wade recognize this important short coming in centralized control [39], [40], [41], [34], [42]. According to
Kedzie’s study and therefore explore the global effect of the Neuman, even if the increase of ICTs had led to an increase
Internet on democracy over a ten year period, 1992−2002 [1]. in the motivation to communicate - which he argues has not
They aggregate political and civil rights data from Freedom happened - then ICTs would have become centralized by
House to formulate a democracy index, which serves as their government turning them into social control mechanisms [43].
dependent variable. The number of Internet users per 1, 000 Scholars who subscribe to this school of thought maintain that
represents their independent variable while the following mea- mass media information technologies discourage collective
sures are used as control variables: GDP per capita, education behavior, ”unless the rise in couch potatoes can be considered
and literacy rates, life expectancy, urbanization, prevalence of a social movement” [44], [45], [46].
non-Internet ICTs. Their analysis shows that a statistically In contrast to Snider’s argument about the mere potential of
powerful correlation exists between Internet diffusion and citizens being informed acting as a source of accountability, if
level of democratization. “The more salient observation to power is measured by the potential for “monopoly and control
make, however, is that while economic prevalence and liter- over information and communication, it is evident that the new
acy maintain relatively constant correlations with democracy, technology can become a dangerous facilitator of tyranny”
the correlation for Internet prevalence gradually strengthens, [21], [2]. Indeed, while the Internet may enable citizens to
almost to the same level as economic prevalence” [1]. subvert political hierarchy, Barber notes that with increased
The authors suggest that this dynamic reflects the grow- participation comes the peril of political and economic surveil-
ing significant relationship between Internet prevalence and lance. The populist school of thought is often blind to “how
democracy: “perhaps this is an indication that the Internet easily liberating technologies become tools of repression” [2].
has come of age as a correlate of democracy” [1]. Indeed, Bimber rejects the supposition that the Internet will
they posit that this growth in correlation strength might “be have significant effects on public life, point out that “both
expected given the positive network externalities, the network theory and empirical evidence cast grave doubt on the
effect’ that is a salient property of the Internet” [1]. However, communication-action connection at the core of the populist
the coefficients from the regression analysis reveal that Internet theory” [12]. Lippmann argued that the capacity of ICTs to
usage is only able to predict a minimal amount of the variation recreate politics is constrained by human nature, ie. cognitive
in democracy: “to generate one point of democracy, an extra processing, and not by the technical properties of the media
500 Internet users per 1,000 citizens is needed, or an extra themselves [47]. The Internet, then, is no different than other
$5, 882 of GDP per capita is needed” [1]. The scale of democ- ICTs even if the new medium differs from previous technolo-
racy runs from 2 to 14. In terms of democracy’s traditional gies in a fundamental way, namely allowing social bonding
determinants, GDP was a weak predictor while literacy turned to occur asynchronously. In sum, the Internet is “hardly
out to have no significance whatsoever. The other control producing the first dramatic expansion in communication:
variables used were either insignificant or internally correlated. telephone, radio, and television also expanded communication
While Best and Wade’s important contribution to the litera- profoundly.” There seems no compelling reason to believe
ture on Internet and democracy is one of the few contemporary that the communication capacity of the Net will have such a
quantitative studies carried out thus far, their approach does dramatically different effect than have other advances in point-
face a number of important limitations [1]. For one, their to-point and broadcast communication” [12]. Moreover, Page
democracy index needs to be unpacked and “its constituent argues, new ICTs may very well overcome spatial distance
components, such as freedom of the press, openness of the but his far from sufficient for establishing vibrant forms of
electoral process,” for example, tested against traditional de- political communication and deliberation [48].
terminants of democracy to determine whether one compo- Furthermore, “if democracy is to be understood as delib-
nent provides more explanatory power than others. Another erative and participatory activity on the part of responsible
limitation of their data is the fact that it extends only to citizens, it will have to resist the innovative forms of dem-
2002. This should be updated today due to the rapid pace agoguery that accompany innovative technology and that are
of ICT diffusion over the past several years. In addition, too often overlooked by enthusiasts [2]. Aristotle wrote that
several scholars have criticized the Freedom House data with the basis of a democratic state is liberty. Barber adds that a
regards to conceptualization, measurement and aggregation “free society is free only to the degree that its citizens are
issues (Munck and Verkuilen, 2002; Rydland et al., 2008). informed and that communication among them is open and
Furthermore, as discussed subsequently, there is little to no informed [2]. However, recent research and empirical work
7
confirms that governments increasingly have the upper hand is global in scope and uses time series data from 1992 to
in controlling and regulating the impact of the information 2002. We seek to build on their work by using data from
revolution [49], [50], [51], [52], [53], [54], [55], [56], [57], 2000 to 2006 and using the World Bank Governance metrics
[58]. As Goldsmith observes, “if governments can raise the as our measures of democratic tendency. Best and Wade
cost of Net transactions, they can regulate the transactions” combined the Freedom House metrics of political rights and
[59]. Beilock and Dimitrova found that countries with lower civil liberties as their measure of democratic tendency. As
Freedom House scores for civil liberties had significantly described subsequently we feel the Freedom House data are
lower Internet usage (even when controlling for economic not well suited to a regression study such as this one.
development) [60].
De Mesquita and George Downs also argue that government A. Our Approach to Measuring Democratic Tendency Using
elites (e.g., in Singapore) have learned to “stifle the bottom-up the World Bank Governance Indicators
democratic potential of the Internet and still promote economic Dahl characterizes a government with power vested in a
growth, contrary to Kedzie’s dictator’s dilemma argument” plurality as follows [13]:
[61]. As Bimber notes, the “central theoretical problem for
1) Control over governmental decisions about policy is
the populist claim is the absence of a clear link between
constitutionally vested in elected officials.
increases in information and increases in popular political
2) Elected officials are chosen and peacefully removed
action” [12] To this end, McLuhan’s old dictum may be wrong:
in relatively frequent, fair and free elections in which
“the medium is not the whole message. Content matters, and
coercion is quite limited.
there is simply no overwhelming reason to believe that a
3) Practically all adults have the right to vote in these
new medium will necessarily enhance the political quality
elections.
of communicative content” [12]. In short, “technology need
4) Most adults also have the right to run for the public
not inevitably corrupt democracy, but its potential for benign
offices for which candidates run in these elections.
dominion cannot be ignored” [2].
5) Citizens have an effectively enforced right to freedom
In terms of quantitative analysis, Scheufele and Nisbet’s
of expression, particularly political expression, including
2002 statistical study suggests that the Internet does not
criticism of the officials, the conduct of the government,
increase democracy. “Through linear regression, they find
the prevailing political, economic, and social system, and
that mass media broadcasting (e.g., television, newspapers)
the dominant ideology.
plays a far more effective role than the Internet in promoting
6) They also have access to alternative sources of informa-
democratic citizenship” [1], [62]. Given that an established
tion that are not monopolized by the government or any
body of quantitative research on this topic has yet to material-
other single group.
ize, Scheufele and Nisbet do caution against generalizing the
7) Finally, they have an effectively enforced right to form
results of their study, which focused exclusively on the United
and join autonomous associations, including political
States. Other scholars interested in this line of research have
associations, such as political parties and interest groups,
questioned the supposed direction of causation drummed up
that attempt to influence the government by competing
by the populist school of thought. Using multiple measures
in elections and by other peaceful means.
of regime type, Milner’s statistical analysis demonstrates that,
ceteris paribus, democracies permit much greater online ac- The first four points largely describe procedural aspects of
cess, both in terms of Internet users per capita and Internet a democracy, whereas the last three points delineate the com-
hosts per capita [63]. To this end, the information revolution munication aspect necessary for a well-functioning democratic
may merely be reinforcing pre-existing dynamics. regime. In fact, Diamond goes further and notes that “[s]ome
Milner’s study uses data from 1991 − 2001 to measure the insist on a fairly robust (though still procedural) definition
influence of regime type of adoption of the internet. This study of democracy, like Dahl’s ‘polyarchy.’ By this conception,
attempts to address a slightly different question—whether democracy requires not only free, fair, and competitive elec-
there is a relationship between measures of democracy and tions, but also the freedoms that make them truly meaningful
ICT penetration—but we build on her work by extending the (such as freedom of organization and freedom of expression),
range of years to 2006 (although we begin measurement in alternative sources of information, and institutions to ensure
2000). We adopt a fixed effect model and control for the time that government policies depend on the votes and preferences
component directly in the model. of citizens” [64]. Expanding on Dahl, Diamond has developed
Beilock and Dimitrova develop a model to explain global his own list of characteristics of a democracy [65]:
country differences in Internet use using income, measures of 1) Control of the state and its key decisions and allocations
freedom, region dummies, and development indices [60]. Their lies, in fact as well as in constitutional theory, with
data is a cross section from 2001 and does not take democratic elected officials (and not democratically unaccountable
variables directly into account. actors or foreign powers); in particular, the military is
Best and Wade ask the question closest to that addressed subordinate to the authority of elected civilian officials.
in this paper [1]. They ask whether Internet penetration has 2) Executive power is constrained, constitutionally and in
an effect on the level of democracy in a country. Their study fact, by the autonomous power of other government in-
8
stitutions (such as an independent judiciary, parliament, respondents, as well as thousands of experts working for the
and other mechanisms of horizontal accountability). private sector, NGOs, and public sector agencies” [66].
3) Not only are electoral outcomes uncertain, with a sig- In points eight through ten, Diamond gives a description
nificant opposition vote and the presumption of party of the role of law in a democracy. The World Bank has a
alternation in government, but no group that adheres to governance metric that expresses some of this: “measuring
constitutional principles is denied the right to form a perceptions of the extent to which agents have confidence in
party and contest elections (even if electoral thresholds and abide by the rules of society, and in particular the quality
and other rules exclude small parties from winning of contract enforcement, property rights, the police, and the
representation in parliament). courts, as well as the likelihood of crime and violence” [66].
4) Cultural, ethnic, religious, and other minority groups This approximates the World Banks “Rule of Law” metric and
(as well as historically disadvantaged majorities) are not we propose it as an empirical measure of Diamond’s points
prohibited (legally or in practice) from expressing their eight through ten.
interests in the political process or from speaking their The World Bank also has a metric measuring political sta-
language or practicing their culture. bility: “perceptions of the likelihood that the government will
5) Beyond parties and elections, citizens have multiple, be destabilized or overthrown by unconstitutional or violent
ongoing channels for expression and representation of means, including politically-motivated violence and terrorism”
their interests and values, including diverse, independent [66]. While not enumerated in either Diamond’s or Dahl’s
associations and movements, which they have the free- lists, Amartya Sen postulated an empirical correlation between
dom to form and join. democratic regimes and political stability [67]. He notes both
6) There are alternative sources of information (including the “political incentives provided by democratic governance”
independent media) to which citizens have (politically) to prevent crises and specifically that the “positive role of
unfettered access. political and civil right applies to the prevention of economic
7) Individuals also have substantial freedom of belief, opin- and social disasters in general” [67]. Thus we investigate the
ion, discussion, speech, publication, assembly, demon- World Banks “Political Stability” metric as another measure
stration, and petition. of democratic tendency.
8) Citizens are politically equal under the law (even though In measuring the relationship between ICT penetration and
they are invariably unequal in their political resources). these democratic variables, it is clear that country wealth is a
9) Individual and group liberties are effectively protected confounding factor that sound be taken into account: wealthier
by an independent, nondiscriminatory judiciary, whose countries are both more likely to be democratic and to be
decisions are enforced and respected by other centers of the heavier users of both the Internet and the cell phone. We
power. gathered gross domestic product (GDP) data from 2000 to
10) The rule of law protects citizens from unjustified deten- 2006 from the International Monetary Fund. The GDP data is
tion, exile, terror, torture, and undue interference in their purchasing parity adjusted to be directly comparable between
personal lives not only by the state but also by organized countries.
nonstate or anti-state forces. Diamond notes that country size is highly correlated with
regime type: “countries with populations under one million are
Like Dahl, Diamond includes procedural aspects of a demo- much more likely to be both democracies and liberal democ-
cratic regime (points one through three) and he enshrines what racies. Two-thirds of these countries are liberal democracies,
he considers essential communication requirements in points while only 30 percent of countries with populations over one
four through seven. These latter points can be characterized as million are. Among the larger 150 countries, only half are
approximating a “freedom of expression” or “political voice” democracies, while 70 percent of the small countries are. The
aspect to democracy. We found the World Bank governance countries with populations over one million are about twice as
metric of “Political Voice and Accountability” to represent likely as small states to have an electoral authoritarian regime
Diamond’s notion well in that it measures “perceptions of the and half again as likely to have a closed authoritarian regime.”
extent to which a country’s citizen’s are able to participate in [64], [65]. Because of this, we included population in our
selecting their government, as well as freedom of expression, models to control for country size. The population data for
freedom of association, and a free media” [66]. Like all the 2000 to 2006 was also gathered from the IMF.1
World Bank Governance metrics, it was built from surveys
and other sources of data within each country. These sources B. Limitations of The Data
give an idea of “freedom of belief, opinion, discussion, speech, The ICT data is gathered from the International Telecom-
publication, assembly, demonstration, and petition” present in munications Union (ITU). The ITU requested the number of
the country, although they do not measure the proliferation Internet and cell phone users from each country. This raises
of channels of communication that Diamond enunciates. The a host of questions about the reliability of the data since it
World Bank governance indicators metrics are based on 35
data sources some of which yield “subjective or perceptions- 1 Both the IMF GDP and population data are available at http://www.imf.
based data” including that from “household and firm survey org/external/pubs/ft/weo/2008/01/weodata/
9
is self-reported by the country. Perhaps the country has an activity for 40 of the countries most actively engaged in
incentive to under or over-report? It is plausible that some repressing internet activity [68], [69]. Although not as extreme
countries have more reliable data collection mechanisms in as the case of the Internet, cell phone use can be restricted
place than others. With one report per country per year, the by the government as well, and will differ from country to
data are highly granular, but they do seem to follow steady country. Zuckerman gives several examples of government
trends, and steadily upwards in ICT adoption. crackdowns on mobile phone use: Belarus’s reported shutdown
The World Bank did not calculate governance metrics for of their SMS network in March 2006, reports of Ethiopian
2001. We carried out simple linear interpolation to provide a cell phone blocking during the 2005 election protests, and
data point for each country in this year. We made the decision Cambodian blocking of SMS for two days before their 2007
to do this since there are only six time points from which elections [7].
to interpolate, and only one, the year 2000, falls before our North Korea and Cuba were dropped from the study since
missing data. This creates data that are “too smooth” for year official data is not reported for these countries. It is generally
2001 since they have been created from our pre-existing data known that there is very little internet access in Cuba, and little
and this will make our results appear more precise than they to none in North Korea and these are both regimes with little
in fact are. We argue that this effects is minor since the World democracy. If we had been able to include these countries in
Bank data are themselves aggregated from a large number of our estimation of the models, this would likely have bolstered
sources, and thus less subject to noise than using a single our results.2 Details of the data cleaning and amalgamation
source would be. process are on the study website at http://www.stodden.net/
Other metrics of the level of democratic rights exist, such as ICTD.
the Freedom House “Freedom in the World” metric. Freedom Our population data was obtained from the International
House carries out an annual global survey of political rights Monetary Fund and contains a sparse number os missing
and civil liberties. We choose not to use this as a measurement values. The IMF has made estimates of their missing data
of democratic tendency for two reasons. We felt that the World to complete the dataset.3
Bank Governance Indicators could be well grounded in the
theory of democracy as measures of democracy. Secondly, the III. E MPIRICAL M ODELING
Freedom House measures have some quantitative limitations. Our data comprises a panel containing N different times
A certain amount of inertia is built into the measurements so series each consisting of T observations. The number of
that it is difficult for a country to move much from one year to countries, N , is 181, and T , the number of years in our study,
the next. For both political rights and civil liberties a country is 7. A fixed effects model of our democratic measures’ effect
is scaled from 0 to 7, giving only 8 possible outcomes for on ICT penetration follows:
a country. Combining these two factors leads to a database
that does not shift very much from year to year. In the years ICTit = β0 + β1 RLit + β2 V Ait + β3 P Sit +
of our study, 2000 to 2006, of the 193 countries surveyed by
Freedom House (after subtracting the 9 with missing values β4 P OPit + β5 GDPit + β6 M Fit +
for both political rights and civil liberties for the entire time γt T D + ξi CD + it
series), 105 had no change in their scores for political rights
and 86 had no changes in their civil rights scores. The average i = 1, . . . , N, t = 1, . . . , T.
variance of those that did exhibit some change from 2000 to
2006 was 0.48 for political rights and 0.32 for civil liberties.
In this paper we model the penetration of Internet or cell
This means the majorities of countries, if they changed at all,
phone usage per 100 inhabitants, ICT , as a function of the
changed by perhaps one point on the 8 point scale.
World Bank democratic measures (Rule of Law, Voice and
Having more years of data, including 2007, would improve
Accountability, and Political Stability)4 , country size, country
our modeling. At the time of this writing, the World Bank
wealth, and the male/female ratio in the country.
Governance metrics were not available for 2007.
ICT is one of “Internet Use” or “Cell Phone Use.”5 GDPit
The IMF estimated some of the population number for some
is the per capita Gross Domestic Product for country i at
of the countries. It is likely this has the effect of providing
year t, adjusted for purchasing power parity.6 P OPit is the
population data that is smoother than it would otherwise be.
Note also that both ICT measures, the Internet and Cell Phone 2 Countries with missing values also tended to support our hypothesis:
use, are measures per 100 inhabitants. We emphasize that this Afghanistan, Bhutan, Comoros, Kiribati, Serbia, St. Kitts and Nevis, Timor-
must be carefully noted in interpretation of the regression Leste, and Tonga. Cuba, Iraq, Montenegro, and North Korea simply did not
furnish enough data for inclusion in the study.
results, since we use population as an independent variable. 3 A precise explanation of their data interpolation procedure was not readily
We also note that even though the ITU collects Internet use available. See http://www.econstats.com/weo/V023.htm
4 Available at http://info.worldbank.org/governance/wgi/
statistics for each country, what it really means to use the
5 The data used in this study is available at http://www.itu.int/ITU-
Internet can vary by country due to filtering, censoring, and
D/ict/informationsharing/
other restrictions on access. The OpenNet Intiative at the 6 The data are available at
Berkman Center for Internet and Society monitors the filtering http://www.imf.org/external/pubs/ft/weo/2008/01/weodata/index.aspx
10
population of country i at year t and M Fit is the male/female the same test was performed on the World Bank variables, 54,
gender ratio.7 62, and 60 of 181 rejected nonstationarity for Rule of Law,
β0 is the intercept term for country i, and γt T D and ξi CD Voice and Accountability, and Political Stability, respectively
are a time effect and a country effect, respectively. The effect (about one third of the data). Although there some evidence
of time is controlled for by dummy variables: T Dt is 1 for year of stationarity, it appears that the majority of the time series
t and 0 otherwise. Similarly CDi is a dummy variable that is included in this study are nonstationary, and it’s potential
1 for country i and 0 otherwise, controlling for the differences biasing of coefficient estimates is a concern.10 The typical
between countries. Finally, it is a disturbance term with remedy is the difference the data to remove the nonstationarity.
distribution N (0, σi2 ), which we assume to be uncorrelated Running the regressions in the above equation allowed us to
across country cross sections. In this study there are N = 181 carry out tests on the residuals directly to evaluate the level of
countries and T = 7 years. autocorrelation. There are two regressions to be run in this
study, modeling Internet penetration and cell phone usage.
A. Autocorrelation in Panel Data
Typically the Durbin-Watson test with one lag is used to test
A panel regression model of this type is subject to possible for autocorrelation in the structure of the regression residuals
autocorrelation between subsequent observations because of and the regression with Internet use as a dependent variable
the time series components. In a regression model as described was found to have autocorrelation present, and the cell phone
above, it is possible to ’discover’ what Granger and Newbold penetration regression was not, with Durbin-Watson values of
[70] termed “spurious” relationships between the variables. 1.86 and 2.01 respectively.11
That is, tests of significance on estimated coefficients may The Durbin-Watson statistic is created by calculating d =
PT
indicate a significant result, when in fact none is present. (t −t−1 )2
t=2P
t 2 , where t is the tth residual from the regres-
Granger and Newbold suggest economic time series data may 1 t
sion. It follows that 0 < d < 4. A value of 2 indicates
be especially prone to autocorrelation since they tend to be
no autocorrelation. To test whether the Durbin-Watson test
non-stationary, in that it is not uncommon for the process
statistics could be considered equivalent to 2, the test in the
generating the data to depend on the time it was sampled.
R statistical software package was used [71].
For example, economic time series are commonly subject
The adjusted R2 values were 0.9246 and 0.9199, seeming to
to seasonal or cyclical effects. Time series data that are
fit the Internet regression squarely into Granger and Newbold’s
not stationary will violate the assumptions of least squares
area of caution: a low Durbin-Watson statistic and a high
regression since the variance of the error term will depend
R2 value, and also casts some suspicion on the cell phone
on time and thus introduce a bias into coefficient estimation.8
regression. Granger and Newbold offer that until “a really
Granger and Newbold describe a high R2 value and a low
satisfactory procedure is available, we recommend taking first
Durbin-Watson statistic as warning signs that the estimated
differences of all variables that appear to be highly autocorre-
model may be yielding spurious results.
lated.” (p. 8.) We carried out this operation on both the Internet
Thus it is important to determine whether the time series
and cell phone use regressions because of the high R2 values
data in this study are nonstationary. Our data are sampled
and the nonstationary data in both regressions, even though the
yearly suggesting they may avoid annual cyclical effects.
Durbin-Watson statistic did not suggest autocorrelation among
We analyze 181 countries from 2000 to 2006 (with 2001
the cell phone regression errors. The plots of the residuals
interpolated for the World Bank variables) and thus have 7
for both regressions indicate possible heteroskedasticity. This
values in each time series. As Granger and Newbold mention,
suggests running the following differenced model:
finely sampled time series tend to exacerbate the cyclical
effects and thus nonstationarity in the data.
As is typical, the Augmented Dickey-Fuller test was used to ∆ICTit = β00 + β10 ∆RLit + β20 ∆V Ait + β30 ∆P Sit +
assess nonstationarity in each of our time series. Since Internet
and Cell phone use are increasing rapidly for the vast majority β40 ∆P OPit + β50 ∆GDPit + β60 ∆M Fit +
of the countries in our study, we measured the autoregressive
structure of each time series as stationary around a trend line, γt0 T D + ξi0 CD + it
and modeled with one lag. The null hypothesis is that the
data are nonstationary. The test was applied to the ICT and i = 1, . . . , N, t = 2, . . . , T.
World Bank data at the country level. For the Internet and
mobile phone data the test rejected the null hypothesis of
nonstationarity for 40 and 35 of 181 countries respectively, Running this model for differenced Internet penetration and
meaning that for around 20% of the Internet and cell phone differenced mobile phone use did not improve the Durbin-
penetration time series, nonstationarity is not evident.9 . When
10 Note that a combination of nonstationary time series may in fact be
7 The gender ratio data was obtained from the Census Department’s Inter- stationary. This is termed cointegration.
national Database at http://www.census.gov/ipc/www/idb/tables.html 11 The p-values for the Durbin-Watson test are generated via a bootstrapping
8 For a mathematically precise explanation see [70], p. 2.
method and can fluctuate. In this case the p-values were 0.022 for the Internet
9 The tests were performed at the 10% level regression and 0.962 for the cell phone regression.
11
TABLE I TABLE II
S UMMARY S TATISTICS FOR 2000 DATA S UMMARY S TATISTICS FOR 2006 DATA
Variable Mean Min Max Variable Mean Min Max

Internet Use per 100 7.25 0 45.58 Internet Use per 100 21.66 0.03 92.52
Cell Use per 100 16.02 0 81.73 Cell Use per 100 53.19 0.42 138.06
Rule of Law -0.07 -2.02 1.95 Rule of Law -0.07 -2.00 2.03
Voice & Accountability -0.06 -2.05 1.67 Voice & Accountability -0.07 -2.28 1.72
Political Stability -0.08 -2.73 1.54 Political Stability -0.08 -2.31 1.60
GDP (PPP per capita) 8,998.10 229.36 55,248.25 GDP (PPP per capita) 12,147.70 195.43 76,537.15
POP (millions) 33.20 0.04 1267.43 POP (millions) 35.67 0.051 1314.48
Gender Ratio (M/F) 100.4 85.3 212.3 Gender Ratio (M/F) 100.4 84.3 218.5
TABLE III
R EGRESSION C OEFFICIENT E STIMATES FOR G LOBAL I NTERNET
12
Watson statistics. They became 2.16 and 1.92 respectively. P ENETRATION
The adjusted R2 values were reduced to 0.3686 and 0.4059.
Variable Estimate Standard Error p-value
Both Durbin-Watson statistics reject the null hypothesis of
Rule of Law 0.9018 1.1193 0.4207
no autocorrelation at the 5% level. A common method of Voice & Accountability 0.9122 0.8514 0.2844
controlling for autoregression is using a 2-stage least squares Political Stability -0.3783 0.5345 0.4793
approach [74], [75]. In the first stage, the autoregressive GDP 0.0004 0.0002 0.0972*
POP -0.0488 0.8123 0.9521
structure in the residual is estimated using a model postulating Gender Ratio 0.8212 0.9189 0.3718
that the autocorrelation has a single lag structure, specifically:
i = ρ ∗ i−1 A. Modeling Global ICT Penetration as a Function of Demo-

cratic Tendency
where i is the ith residual from the initial regression. We estimated two panel regressions of ICT penetration with
The first stage allows us to find an estimate of ρ, ρ̂ using controls for autocorrelation as discussed in the preceding sec-
a least-squares model. In stage 2, ρ̂ is used to remove the tion (differencing and the Durbin-Watson correction). Internet
autocorrelation in the variables. Each variable, represented as and cell phone use were modeled as functions of demographic
V AR in the next equation, is then adjusted to create a new variables along with control variables:
variable, adjV AR, using the formula:
∆Internetit = β00 + β10 ∆RLit + β20 ∆V Ait + β30 ∆P Sit +

adjV ARi−1 = V ARi − ρ̂V ARi−1
β40 ∆P OPit + β50 ∆GDPit + β60 ∆M Fit +
Since we are analyzing panel data with both time and
γt0 T D + ξi0 CD + it
country dimensions, implementation of the autocorrelation was
applied at the country level using different estimates of ρ for i = 1, . . . , N, t = 2, . . . , T.
the Internet and cell phone regressions. Carrying this out on the
differenced data increased the Durbin-Watson statistics to 2.24
and 2.18, and autocorrelation is still detected by this test at the The regression coefficients for Internet penetration are given
5% level. The adjusted R2 statistics were 0.4292 and 0.3599. in Table III. The coefficients on the individual country and
This was our final model analyzed in the following section. time dummy variables are not included for space reasons.13
Since autocorrelation has not been eradicated from the data The most significant coefficient was GDP, and none of the
we interpret our results cautiously and look for corroboration. democratic measures were significant. The positive coefficient
on GDP confirms our intuition that wealthier countries have
IV. R ESULTS AND F INDINGS higher levels of Internet use.
The same regression was run for cell phone penetration, and
We used the R statistical package to estimate these models the coefficient estimates are presented in Table IV. Cell phone
(version 2.7.2) [71]. The complete set of code and data used use appears not to associated with wealth as Internet use is,
in this study can be found at http://www.stodden.net/ICTD. and the male to female gender ratio in the country is strongly
Tables I and II present summary statistics of the dependent associated with increases in cell phone use, as is the level of
variables (Internet and cell phone use per 100 country inhab- political stability.
itants) and the World Bank metrics for the countries in 2000 The influence of the gender ratio may be driven by outlier
and 2006 respectively. countries: Most countries had a male/female gender ratio of
a little less than 100, implying slightly more females than
12 Testing the residuals for autocorrelation using the Durbin-Watson test is males in the population. The coefficient of 4.88 implies that
the typical procedure, when the sample is large. It is also possible to test for
cointegration: whether the combination of time series is stationary. See [72] 13 The complete regression results can be found online at http://www.
and [73]. stodden.net/ICTD.
12
TABLE IV TABLE VI
R EGRESSION C OEFFICIENT E STIMATES FOR G LOBAL C ELL P HONE C ELL P HONE U SE PER 100 I NHABITANTS FOR H IGHER G ENDER R ATIO
P ENETRATION C OUNTRIES , 2000 AND 2006
Variable Estimate Standard Error p-value Rank Country 2000 Country 2006
Rule of Law -1.4120 2.0333 0.4876 1 United Arab Emirates 43.98 United Arab Emirates 118.51
Voice & Accountability -2.458 1.5946 0.1236 2 Qatar 19.90 Qatar 109.6
Political Stability 2.2823 1.0044 0.0233** 3 Kuwait 21.74 Kuwait 91.49
GDP -0.0002 0.0005 0.7061 4 Oman 6.63 Maldives 87.88
POP 0.1225 1.4327 0.9319 5 Bahrain 30.61 Bahrain 122.88
Gender Ratio 4.8839 1.7342 0.0050*** 6 Saudi Arabia 6.40 Oman 69.59
7 Maldives 2.83 Saudi Arabia 78.05
TABLE V 8 Bhutan 0 Bhutan 9.77
T OP 10 M ALE /F EMALE R ATIOS , 2000 AND 2006 9 Jordan 7.72 Jordan 74.4
10 Djibouti 0.04 Grenada 44.59
Rank Country 2000 Country 2006
1 United Arab Emirates 212.3 United Arab Emirates 218.5
2 Qatar 198.1 Qatar 202.5 confounding factors.
3 Kuwait 150.3 Kuwait 152.3
4 Oman 131.3 Maldives 127.3 The World Bank measure for political stability is also
5 Bahrain 129.9 Bahrain 126.7 significantly positively correlated with increased cell phone
6 Saudi Arabia 125 Oman 124.7 use. This finding suggests that political instability is related
7 Maldives 117.3 Saudi Arabia 120.5
8 Bhutan 112.3 Bhutan 111.0 to the mass diffusion of cell phone usage. In other words,
9 Jordan 109.9 Jordan 110.2 an increase in cell phone availability could increase the
10 Djibouti 107.1 Grenada 108.1 perceived likelihood that the government will be destabilized
or overthrown by unconstitutional or violent means. In their
statistical analysis, Mansfield and Snyder find that the process
as the gender ratio increases by about 5, cell phone usage per of democratization itself is indeed a destabilizing one [76].
100 inhabitants will increase by one phone. Throughout the “Certainly, the virtues of working democratic structures do not
years studied, roughly 7 of the 10 countries with the highest translate into a carefree path to the stabilization of democracy”
male/female ratio each year were located in the middle east, [77]. This finding also supports the arguments presented by
and the ratios at that end of the distribution dwarfed the other Rosenau and Johnson, as well as Shirky, who opine that the
countries’. As displayed in Table VI, it is plausible some of Internet can be used as a tool by civil society to organize
these values are extreme enough to have a large impact on the collectively to effect political change [25], [78].
regression fit, although why this did not occur in the internet
regression is not clear. It is also possible the very high gender B. Modeling ICT Penetration as a Function of Democratic
ratio values represent a ’middle east effect’ since a number Tendency Among the Most and Least Affluent Countries
of those countries are highly represented in the top 10 gender Examining ICT penetration for different strata of wealth
ratio values. This implies that the coefficient on the gender may help isolate effects that are characteristic of those groups.
ratio variable could represent a high growth in cell phone use Since wealth is a driver of investment one would expect
in the middle east. Table VII gives the 2000 and 2006 cell GDP to play a role in the country’s readiness and ability to
phone data for these countries. The coefficient indicates that adopt new communication technologies [79], [80], [81]. As
as the rate of change in the proportion of men increases, so established in the literature we also found increases in GDP
does the rate of change in cell phone use. to be associated with increases in Internet use. We choose
Interestingly, the greater political stability and the lower the to examine ICT penetration in both the top and bottom 20%
perceived threat of violence, the greater cell phone penetration. of countries by 2006 GDP more closely. Our focus on these
This may represent infrastructural stability if associated with groups, in particular the bottom quintile, is driven by Mansfield
political stability and thus a measure of investor’s confidence. and Snyder’s work theorizing the instability of emergent and
It is not clear why this factor would not therefore also be transitional regimes and the existence of the global digital
associated with an increase in Internet use. Perhaps cell phones divide [82].
are easier to proliferate than access to the Internet and so a Table VII lists the countries that fall into each of these
smaller increase in political stability encourages cell phone groups.
increase before Internet increase. As in the previous section we fit a model with an ICT
As shown in Tables I and II, the average cell phone penetration measure as the explanatory variable, and measures
penetration in 2000 was about 16 phones per 100 inhabitants of democratic tendency and controls as independent variables
and in 2006 it was about 53, nearly a three-fold increase. This for a panel regression over years 2000 to 2006. For the
is a high rate of increase but, notably, the countries listed top wealthiest quintile of countries our model did not yield
in Table VI (those with the highest male to female gender statistically significant results for Internet penetration. Table
ratios) had much higher than average growth in cell phone VIII gives the coefficient estimates. Since these countries are
penetration. This is quantified in the significant coefficient in exceptionally wealthy and relatively stable politically it may
the regression in Table IV, while allowing for the included not be a surprise that GDP is not a driver of Internet use, and
13
TABLE VII TABLE IX

T OP AND B OTTOM 20% OF C OUNTRIES BY GDP IN 2006 I NTERNET P ENETRATION , B OTTOM 20% OF GDP (2006)
Top 20% Countries Bottom 20% Countries Variable Coefficient Standard Error p-value
Qatar Zimbabwe Rule of Law 0.3702 0.4081 0.3660
Luxembourg Congo (Dem. Rep.) Voice & Accountability -0.8115 0.3656 0.0281**
Brunei Darussalam Liberia Political Stability -0.2583 0.1822 0.1585
Norway Burundi GDP -0.0002 0.0014 0.9053
Singapore Guinea-Bissau POP -0.0309 0.2784 0.9117
United States Afghanistan Gender Ratio 0.4026 0.3235 0.2155
Ireland Sierra Leone
Switzerland Niger TABLE X
Hong Kong, China Central African Rep. C ELL P HONE P ENETRATION , T OP 20% OF GDP (2006)
Kuwait Ethiopia
Iceland Malawi
Canada Eritrea Variable Coefficient Standard Error p-value
Netherlands Mozambique Rule of Law 2.354 6.161 0.7029
Austria Togo Voice & Accountability 4.012 4.736 0.3984
Denmark Rwanda Political Stability 5.071 1.887 0.0612*
United Arab Emirates Uganda GDP -0.0008 0.0005 0.1071
Sweden Myanmar POP -2.896 6.483 0.6558
Australia Mali Gender Ratio 2.8300 4.3690 0.5182
Belgium Madagascar
United Kingdom Guinea
Finland Comoros
Germany Tanzania political turmoil and it may be the case that countries with
Japan Nepal higher Voice and Accountability rankings have been reluctant
France Burkina Faso to permit the growth of the Internet in their milieu.
Bahrain Lesotho
Italy Bangladesh Table X gives the coefficient estimates from the panel
Spain Gambia regression for cell phone penetration for the top 20% of
Taiwan, China Haiti wealthiest countries. There is a statistically significant effect in
Greece Zambia
Cyprus Ghana the World Bank metric of Political Stability: greater Political
New Zealand Sao Tomé & Principe Stability is associated with an increase in cell phone use per
Slovenia Benin inhabitant. This result seems intuitive as political stability is
Israel Kenya
Bahamas Senegal historically associated with greater investment in communica-
Korea (Rep.) Cambodia tions infrastructure and is consistent with our earlier regression
Saudi Arabia Chad on cell phone use. This suggests the wealthiest countries may
Czech Republic Côte d’Ivoire
be driving the correlation between political stability and cell
phone penetration. None of the other variables were found
TABLE VIII
I NTERNET P ENETRATION , T OP 20% OF GDP (2006) to have a statistically significant relationship with cell phone
penetration in the wealthiest countries.
Variable Coefficient Standard Error p-value Among the poorest countries, growth in Voice and Ac-
Rule of Law -0.3709 5.4055 0.945 countability had a statistically significant negative effect on
Voice & Accountability 4.771 4.0981 0.246
Political Stability -1.0344 2.2760 0.650 growth in cell phone penetration. Although consistent with
GDP 0.0003 0.0004 0.450 the Internet penetration regression results for this group of
POP 1.6791 5.5724 0.764 countries, the paradox remains as to the increase in per capita
Gender Ratio -1.7979 3.7568 0.633
cell phone use as Voice and Accountability decreases. It is
plausible citizens desire newer forms of ICT when Voice
and Accountability is restricted. Another explanation may be
nor are the measures of democratic tendency, even though our that although not all countries in the bottom quintile are
subsets contains countries with varying levels of autocratic autocratic, a significant proportion are and cell phone use may
control. facilitate the mobilization, organization and coordination of
Among the least wealthy quintile we fit the same panel resistance against autocratic rule. Interestingly, the voice and
regression model as above. Table IX gives the regression coef- accountability metric is not a significant predictor globally, yet
ficient estimates. Our predictors did not yield highly significant is significant for types of ICT among the poorest countries.
coefficient estimates with the exception of the World Bank
voice and Accountability metric. Voice and Accountability V. C ONCLUSIONS AND F UTURE R ESEARCH
is negatively correlated with Internet penetration: implying This paper is the first to our knowledge that uses recent
that when countries notch up in the Voice and Accountability Internet and cell phone use data in an empirical study of
ranking, the use of the Internet increases. This seemingly their relationship to democratic tendency. Previous studies uses
paradoxical finding may be explained when note that our measures of Internet use that ended in 1993 [26], [38]. Best
analysis is restricted the the lowest quintile of country in and Wade’s data reached only to 2002. This paper is also the
wealth. These countries experience disproportionately greater first to the best of our knowledge that measures ICT diffusion
14
TABLE XI
C ELL P HONE P ENETRATION , B OTTOM 20% OF GDP (2006) used globally than the Internet, so this may be a function of
cell technology’s earlier foothold than the Internet’s. As Inter-
Variable Coefficient Standard Error p-value net diffusion catches up to that of cell phone, the democratic
Rule of Law 0.3702 0.4081 0.3660
Voice & Accountability -0.8115 0.3656 0.0281**
metrics may be found to be predictors of this diffusion. This
Political Stability -0.2583 0.1822 0.1585 is not necessarily inconsistent with the populist thread in the
GDP -0.0002 0.0014 0.9053 literature as, according to Bimber, other scholars believe that
POP -0.0309 0.2784 0.9117
Gender Ratio 0.4026 0.3235 0.2155
the Internet may have a “transformative potential” because
of the “one-to-one interaction among citizens and between
citizens and government” it creates [12].
A paradox is created for the populist school in the finding
as a function of democratic indicators. Previous research has that “Voice and Accountability” is negatively correlated with
focused on whether ICT use predicts democratic measures. cell phone diffusion. This can be interpreted in favor of the
These results support Bimbers assertions that the structure argument that ICT diffusion can be centralized by government
of political power has not been revolutionized or transformed turning them into social control mechanisms [43]. As the rate
into a new epoch of democracy [12]. Evidence can be found of Voice and Accountability increases, the rate of diffusion
for both the populist thread in the literature and the notion that of ICTs decreases among the poorest and least developed
ICTs may act as an extension of the ruling class. countries. Rates of cell phone use increase globally with
We found a statistically significant positive relationship decreases in political stability, lending further support to the
between the rate of diffusion of the cell phone and the World thesis of ICTs as an extension of ruling class control.
Bank’s “Political Stability” measure capturing perceptions This paradox may be resolved is we consider Page’s view
regarding the likelihood that a government will be destabilized in 1995, that ICT diffusion may be still too nascent and
or overthrown by unconstitutional or violent means. Political insufficient to generate a well functioning system of political
Stability continues to have a significant positive relationship communication and deliberation [48]. It appears we are still
with the rate of cell phone use among the most affluent 20% too early to expect a close relationship between a vibrant
of countries, but that relationship does not hold among the public sphere and ICTs globally. As Bimber states, the “central
wealthiest countries. The “Voice and Accountability” indicator theoretical problem for the populist claim is the absence of a
which measures perceptions of the extent to which citizens clear link between increases in information and increases in
are able to participate in selecting their government, as well popular political action” [12]. This paper provides evidence
as freedom of expression, association and of the media, of the existence of this problem and the need to develop our
was a significant negative predictor of the rate of cell and understanding of this dynamic further.
Internet diffusion among the poorest quintile of countries. The It would be interesting to tie this research more closely to
diffusion of ICT access did not have any significant influence development, in line with Sen’s reasoning that “Developing
on “Rule of Law” while the increasing availability of cell and strengthening a democratic system is an essential compo-
phones were shown to have no influence on any of the three nent of the process of development,” could extend the empiri-
World Bank indicators. cal analysis in a fruitful direction. This might mean specifically
Recall that the Rule of Law metric measures perceptions of testing whether the order in which political and civil rights are
the extent to which individuals have confidence in and abide extended as a country emerges from an autocracy affects the
by the rules of society—in particular the quality of contract rate of development. This could provide a setting in which to
enforcement, property rights, the police, and the courts as well test the “Lee Thesis,” that political rights should be withheld
as the likelihood of crime and violence. This measure was orig- until economic development is achieved.
inally included in the analysis based on Diamond’s research on It may be valuable to explore empirical issues further.
the characteristics of democracy. However, the findings here Modeling the autocorrelation structure with more than one lag
suggest that the increase in “Rule of Law” perceptions does may help reduce autocorrelation. It would also be interesting
not influence ICT penetration, either positively or negatively. to test for cointegration among these variables. Certainly
In other words, perceptions regarding the “Rule of Law” may documented feedback loops exist between our independent
be framed and influenced by factors other than widespread variables, such as GDP and measures of democracy, and taking
ICT use. this explicitly into account may improve the modeling [83],
The populist school of thought believes ICT diffusion will [26]. It is also plausible that feedback loops exist between
decentralize access to communication and information while ICTs and democratic measures and future modeling could
increasing citizen access [12], while Hill and Hughes claim accommodate this. Further research into the modeling aspects
that perhaps the Internet will bring about “a wider demo- could estimate models including variables that control for
cratic revolution in the world” [22]. In our modeling Internet the different manifestations of cell phone and Internet use in
diffusion was not predicted by our measures of democracy, different countries. A more comprehensive model might ex-
implying this revolution has not yet arrived. The rate of cell plore possible non-linearities: whether countries with low ICT
phone diffusion was predicted by higher rates of the “Political adoption rates have different patterns of democratic tendency
Stability” metric. Recall that cell phone are much more widely than those with high adoption rates.
15
ACKNOWLEDGMENT [30] R. Burkhard and M. Lewis-Beck, “Comparative democracy: the eco-

nomic development thesis,” American Political Science Review, vol. 88,
The authors would like to thank Michael Best and Colin pp. 903–10, 1994.
Maclay for helpful discussion and the anonymous reviewers [31] J. Longdregan and K. Poole, “Poverty, the coup trap, and the seizure of
for constructive criticism. Sadia Ahsanuddin and Kristen Lovin executive power,” World Politics, vol. 42, pp. 151–83, 1990.
[32] ——, “Does high income promote democracy?” World Politics, vol. 49,
provided outstanding research assistance. pp. 1–30, 1996.
[33] B. Geddes, “What do we know about democratization after twenty
R EFERENCES years?” Annual Review of Political Science, vol. 2, pp. 115–44, 1999.
[1] M. Best and K. Wade, “The internet and democracy: Global catalyst or [34] F. Webster, Globalization, information and change. State University of
democratic dud?” June 2006, sam Nunn School of International Affairs. New York Press, 2002.
[2] B. Barber, “Three scenarios for the future of technology and strong [35] R. Reich, The Work of Nations: Preparing ourselves for 21st century
democracy,” Political Science Quarterly, vol. 113, no. 41573-589, 1999. capitalism. Vintage, 1992.
[3] P. Schmitter and T. Karl, “What democracy is . . . and is not,” Journal [36] T. Friedman, The Lexus and the Olive Tree: Understanding Globaliza-
of Democracy, vol. 2, no. 3, 1991. tion. Anchor, 1999.
[4] P. Norris, “Political protest in fragile states,” July 2006, international [37] L. Waverman, M. Meschi, and M. Fuss, “The impact of telecoms on
Political Science Association World Congress. economic growth in developing countries,” Centre for Economic Policy
[5] F. Zakaria, “The rise of illiberal democracies,” Foreign Affairs, Novem- Research, Tech. Rep., March 2005.
ber 1997. [38] D. Richards, Making the national international: Information technology
[6] Huber, D. Rueschemeyer, and J. Stephens, “The impact of economic and government respect for human rights. Albany, NY: State University
development on democracy,” Journal of Economic Perspectives, vol. 7, of New York Press, 2002.
no. 3, pp. 71–85, 1993. [39] D. Barney, Prometheus Wired: The Hope for Democracy in the Age of
[7] E. Zuckerman, “Draft paper on mobile phones and activism,” 2007. Network Technology. University of Chicago Press, 2000.
[Online]. Available: http://www.ethanzuckerman.com/blog/2007/04/09/ [40] A. Mazrui and R. Ostergard, Technology transfer in the computer age:
draft-paper-on-mobile-phones-and-activism/ The African experience. State University of New York Press, 2002.
[8] D. Kidd, Indymedia.org: a new communications commons. Routledge, [41] D. Strienstra, Gender, women’s organizing, and the Internet. Albany,
2003, pp. 47–69. NY: State University of New York Press, 2002.
[9] A. Scott and J. Street, “From media politics to e-protest,” Information, [42] S. Kalathil and T. Boas, Open Networks, Closed Regimes. Carnegie
Communication and Society, vol. 3, no. 2, pp. 215–40, 2000. Endowment for International Peace., 2003.
[10] S. Andrew, “The control revolution: How the internet is putting individ- [43] W. Neuman, The future of the mass audience. Cambridge University
uals in charge and changing the world,” The Century Foundation, Tech. Press, 1991.
Rep., 2000. [44] T. Eyck, “Does information matter? a research note on information
[11] R. Clarke, “Information technology: Weapon of authoritarianism or tool technologies and political protest,” Social Science Journal, vol. 38, pp.
of democracy?” August 1994, iFIP World Congress. 147–160, 2001.
[12] B. Bimber, “The internet and political mobilization - research note on [45] N. Postman, Amusing ourselves to death. New York, NY: Vintage
the 1996 election season,” Social Science Computer Review, vol. 16, Books, 1985.
no. 4, pp. 391–401, 1998. [46] H. Schiller, Culture Inc. Oxford University Press, 1989.
[13] R. Dahl, Democracy and its Critics. New Haven, CT: Yale University, [47] W. Lippmann, Public Opinion. New York, NY: Macmillan, 1934.
1989. [48] Page, Who Deliberates? Mass Media in Modern Democracy. Chicago,
[14] H. Rheingold, Virtual Reality. New York, NY: Summit Books, 1991. IL: University of Chicago Press, 1996.
[15] R. Anderson, T. Bikson, S. Law, and B. Mitchell, “Universal access to [49] G. Cherian, “The limits of singapore’s ‘light touch’ web regulation,”
e-mail: Feasibility and societal implications,” RAND, Tech. Rep., 1995. 2008, politics 2.0 International Conference.
[16] A. Etzioni, The Spirit of Community: Rights, Responsibilities, and the [50] R. Deibert, J. Palfrey, R. Rohozinski, and J. Zittrain, Access Denied:
Communitarian Agenda. New York, NY: Crown Publishers, 1993. The Practice and Policy of Global Internet Filtering. The MIT Press,
[17] J. Pavlik, New Media Technology: Cultural and Commercial Perspective. 2008.
Boston, MA: Allyn and Bacon, 1996. [51] S. Oats, “Internet and democracy in russia,” April 2008, politics 2.0
[18] L. Grossman, The Electronic Republic. New York: Viking, 1995. International Conference.
[19] G. Browning, Electronic Democracy: Using the Internet to Influence [52] R. Singel, “Seeking tighter censorship, repressive states target
Politics. Wilton, CT: Online Inc., 1996. web 2.0 apps,” Wired Magazine, March 2008. [Online]. Available:
[20] N. Negroponte, Being Digital. New York, NY: Vintage, 1995. http://blog.wired.com/business/2008/03/etech-what-happ.html
[21] J. Snider, “New media: Potential information and democratic account- [53] J. Zittrain, The Future of the Internet—And How to Stop It. Yale
ability: A case study of governmental access community media,” August University Press, 2008.
1996, american Political Science Association Convention. [54] S. Mydans, “Myanmar junta unplugs internet,” The International
[22] K. Hill and J. Hughes, “Computer mediated political communica- Herald Tribune, October 2003. [Online]. Available: http://www.iht.com/
tion: The usenet and political communities,” Political Communication, articles/2007/10/04/asia/04info.php
vol. 14, pp. 3–27, 1999. [55] J. Goldsmith and T. Wu, Who Controls the Internet? Illusions of a
[23] A. Corrado and C. Firestone, Elections in Cyberspace: Toward a Borderless World. Oxford University Press, 2006.
New Era in American Politics, A. Corrado and C. Firestone, Eds. [56] L. Lessig, Code: Version 2.0. Basic Books, 2006.
Washington, DC: Aspen Institute, 1996. [57] D. Drezner, “The internet’s effect on state-society relations,” March
[24] C. Steele and A. Stein, Communications revolutions and international 2006, international Studies Association Convention.
politics. Albany, NY: State University of New York Press, 2002. [58] T. Price, Cyber Activism: Advocacy Groups and the Internet. Wash-
[25] J. Rosenau and D. Johnson, Information technology and turbulence in ington, DC: Foundation for Public Affairs, 2000.
world politics. State University of New York Press, 2002. [59] J. Goldsmith, “Regulation of the internet: Three persistent fallacies,”
[26] C. Kedzie, “Communication and democracy: Coincident revolutions Chicago-Kent Law Review, vol. 73, December 1998.
and the emergent dictator’s dilemma,” RAND, Tech. Rep., 1997. [60] R. Beilock and D. Dimitrova, “An exploratory model of inter-country
[Online]. Available: http://www.rand.org/publications/RGSD/RGSD127 internet diffusion,” Telecommunications Policy, vol. 27, pp. 237–252,
[27] S. Lipset, “Some social requisites of democracy: Economic development 2003.
and political legitimacy,” American Political Science Review, vol. 53, pp. [61] B. Mesquita and G. Downs, “Development and democracy,” Foreign
69–105, 1959. Affairs, September 2005.
[28] R. Jackman, “On the relation of economic development to democratic [62] D. Scheufele and M. Nisbet, “Being a citizen online: New opportunities
performance,” American Journal of Political Science, vol. 17, pp. 611– and dead ends,” The Harvard International Journal of Press/Politics,
21, 1973. vol. 7, no. 3, pp. 55–75, 2002.
[29] K. Bollen, “Political democracy and the timing of development,” Amer- [63] H. Milner, “The digital divide: The role of political institutions in tech-
ican Social Review, vol. 44, pp. 572–87, 1979. nology diffusion,” August 2003, american Political Science Association.
16
[64] L. Diamond, “elections without democracy: Thinking about hybrid [76] E. Mansfield and J. Snyder, “Conceptualizing and measuring democ-
regimes,” Journal of Democracy, vol. 13, no. 2, April 2002. racy:evaluating alternative indices,” International Organization, vol. 56,
[65] ——, Developing Democracy: Toward Consolidation. John Hopkins no. 2, pp. 297–337, 2001.
University Press, 1999. [77] H. Barnes, “The contribution of democracy to rebuilding postconflict
[66] D. Kaufmann, A. Kraay, and M. Mastruzzi, “Governance matters vii: societies,” The American Journal of International Law, vol. 95, no. 1,
Aggregate and individual governance indicators 1996-2007,” World pp. 86–101, 2001.
Bank, Tech. Rep., 2007. [78] C. Shirky, Here Comes Everybody: The Power of Organizing without
[67] A. Sen, Development as Freedom. Oxford University Press, 2001. Organizations. Penguin Press, 2008.
[68] “Opennet initiative.” [Online]. Available: http://opennet.net/ [79] C. Argevou, “Ols bias in a nonstationary autoregression,” Information
[69] R. Faris and N. Villeneuve, Access Denied: The Practice and Policy of Technology for Development, vol. 8, no. 1, pp. 15–28, 1998.
Global Internet Filtering. MIT Press, 2008, ch. 1. [80] B. Oyelaran-Oyeyinka and K. Lai, “Internet diffusion in sub-saharn
[70] C. Granger and P. Newbold, “Spurious regressions in econometrics,” africa: A cross-country analysis,” Telecommunication Policy, vol. 8,
Journal of Econometrics, vol. 2, pp. 111–120, 1974. no. 1, pp. 507–527, 1998.
[71] “R.” [Online]. Available: http://www.r-project.org/ [81] J. Gibbs, K. Kraemer, and J. Dedrick, “Environment and policy factors
[72] C. Granger, “Some properties of time series data and their use in shaping global e-commerce diffusion: A cross country comparison,” The
econometric model specification,” Journal of Econometrics, vol. 16, pp. Information Society, vol. 19, no. 1, pp. 5–18, 1993.
121–130, 1981. [82] N. Kshetri and N. Dholakia, ENCYCLOPEDIA OF INFORMATION
[73] ——, “Co-integration and error correction: Representation, estimation SCIENCE AND TECHNOLOGY, 2nd edition. Idea Group Publishing,
and testing,” Econometrica, vol. 2, pp. 251–176, 1987. 2008.
[74] T. Amemiya, “The nonlinear two-stage least squares estimator,” Journal [83] P. Norris, Digital Divide: Civic Engagement, Information Poverty, and
of Econometrics, vol. 2, pp. 105–110, 1974. the Internet Worldwide. Cambridge University Press, 2001.
[75] R. Cumby, J. Huizinga, and M. Obstfeld, “Two-step two-stage least
squares estimation in models with rational expectations,” Journal of
Econometrics, vol. 21, pp. 333–355, 1983.
17
A review of the research on mobile use by micro

and small enterprises (MSEs)
Jonathan Donner and Marcela Escobari
Like landlines, mobile phones allow people to communicate

Abstract— The paper offers a systematic review of 14 studies at a distance and exchange information instantaneously. Thus,
of the use of mobile telephony by micro and small enterprises there is significant potential for mobile use to increase MSE
(MSEs) in the developing world, detailing findings about changes productivity. However, since the dynamics underpinning this
to enterprises’ internal processes and external relationships, and
findings about mobile use vs. traditional landline use. Results
potential are nuanced, and since current supporting evidence
suggest that there is currently more evidence for the benefits of is scarce and methodologically heterogeneous [9], it is
mobile use accruing mostly (but not exclusively) to existing MSEs important to rigorously examine mobile use by MSEs. For
rather than new MSEs, in ways that amplify existing material example, there is a difference between using a mobile to serve
and informational flows rather than transform them. The review existing customers more effectively, and using it to start a new
presents a more complete picture of mobile use by MSEs than business. There is a difference between using a mobile to
was previously available to ICTD researchers, and indentifies
priorities for future research, including comparisons of the
check market prices and using it bypass a middleman who
impact of mobile use across subsectors of MSEs and assessments carries goods to market. Popular press and practitioner reports
of use of advanced services such as mobile banking and mobile generally fail to make these distinctions.
commerce. Fortunately, a small but methodologically diverse set of
research studies have examined mobile use by MSEs in detail.
This paper offers a systematic review of this existing
Index Terms—Business Economics, Communication, ICTD, literature, identifying known patterns of mobile use, as well as
Mobile Communication, MSE, Social Factors
some important gaps in the research.
The review employs distinct foci. First, it offers an
assessment of how mobile use influences the internal process
I. INTRODUCTION
of an enterprise, using Porter’s value chain model [23].
T his paper presents a systematic review of fourteen studies

of the use of mobile telephony by Micro and Small
Enterprises (MSEs) in the developing world [1-14].
Second, it offers a corresponding assessment of how mobile
use influences the network of relationships external to the
enterprise—the value system [23] of producers, traders,
The majority of non-agricultural enterprises in the wholesalers, retailers and end-customers. Finally, it explores
developing world have ten or fewer employees [15, 16]. These two elements unique to mobile communication—the increased
MSEs employ up to 25% of working-age adults in some spatial and temporal mobility afforded by wireless devices,
countries [16], and while the contribution of MSEs to and the resulting blurring of the personal and the professional
aggregate economic growth remains a matter of debate [17], spheres—to assess how MSE mobile use differs from landline
their importance to household livelihoods and poverty use.
alleviation is undeniable. Thus, MSEs are the focus of
programs at many of the world’s largest development II. MICRO AND SMALL ENTERPRISES (MSES)
institutions [18].
Working definitions of MSEs vary from country to country
Since the year 2000, the spread of mobile telephony across
and from researcher to researcher [24]. This analysis defines
the developing world has raised hopes among policymakers
an MSE as any non-farm i enterprise, formal or informal, with
that MSEs will benefit from easier access to
less than 50 employees, including sole proprietorships, part-
telecommunications. The successful entrepreneur, suddenly
time businesses, and home-based businesses. The size
enabled by his mobile phone, has been given a prominent role
thresholds draw on Mead and Leidholm, [16], who note that
in the global development narrative and become a semi-
the absolute majority of such enterprises in the developing
regular fixture in both the popular press [19, 20] and
world are sole proprietorships, and that firms with less than 10
practitioner media [21, 22].
employees substantially outnumber larger enterprises.
A number of factors distinguish the term MSE (micro and
Manuscript version received February 20, 2009. small enterprise) from SME (small and medium enterprise).
J. Donner is with Microsoft Research India, jdonner@microsoft.com
M Escobari is with the Center for International Development, The terms MSE and SME are acronyms, each combining two
marcela_escobari@harvard.edu distinct sizes of enterprises into a single reference. However,
18
with no commonly accepted definitions of the thresholds Particularly focused and powerful evidence appears in
between micro, small, and medium, there are often implicit Jensen’s [10] research on the fishermen of Kerala. Working
conceptual overlaps between the acronyms. with five-year time series data at three fish markets in coastal
Unlike SMEs, the majority of MSEs are informal India, Jensen and his team found that “the adoption of mobile
enterprises. Once again, however, there is no universal phones by fishermen and wholesalers was associated with a
standard to determine what makes an enterprise informal vs. dramatic reduction in price dispersion, the complete
formal [6]. In many cases, even the term entrepreneur may be elimination of waste, and near-perfect adherence to the Law of
a bit of a romantic misnomer. Evidence suggests that among One Price. Both consumer and producer welfare increased.”
MSEs, only a small minority of enterprises are poised for [10: 879] Soon after the introduction of mobile coverage,
growth; most remain small or struggle to survive, and yield a fishermen bought mobiles and accumulated lists of up to 100
low return on labor and capital [17, 25]. buyers in their handsets’ address books; subsequently, while
Though significantly less growth-oriented and productive still at sea, fishermen could call a range of possible landing
(on average) than SMEs and other larger firms [17], MSEs points and buyers in order to determine the best price and best
share a basic similarity with all enterprises; each combines place to sell their catch.
investments in capital with some labor (their own, their By contrast, Jagun, Heeks, and Whalley’s [9] examination
families’ or their employees) in the hopes of yielding a of the mobile’s role in mediating supply chains in the Nigerian
product or service whose market value exceeds the cost of market for traditional hand-woven ceremonial cloth is broad
those inputs. Thus, there has always been a thread in the ICTD in scope. It offers a multidisciplinary literature review, a
literature that seeks to understand how various technologies conceptual framework articulating effects at multiple levels,
could be used advantageously by MSEs [26-28]. Prior to the and a detailed case study. They describe “process” benefits to
widespread introduction of the mobile into the developing mobile use, as calls at a distance reduce the time of trades and
world, the landline’s importance in this regard was already replace costly journeys. They also describe “structural”
clear: impacts; finding no disintermediation of traders, but rather an
intensification of their role. Traders are more likely to have
Phones are the information-related technology that
mobiles than the less prosperous weavers in the supply chain,
has done the most to reduce costs, increase income
and reduce uncertainty and risk. Phones support the and thus are better positioned to coordinate with a wider range
current reality of informal information systems, they of downstream customers and to maintain a more dynamic and
can help extend social and business networks, and responsive set of relationships with weavers. For example,
they clearly substitute for journeys and, in some weavers previously had to pay cash to get their supplies.
cases, for brokers, traders and other business Mobiles give weavers access to credit by enabling calls on
intermediaries. They therefore work “with the grain” their behalf to fabric vendors by traders, who vouch for the
of informality yet at the same time help to eat into the veracity of weavers’ orders, and promise to cover the costs of
problems of insularity that can run alongside. Phones the fabric in advance of the completion of the weavers’ work.
also meet the priority information needs of this group
of communication rather than processing of IV. METHODS AND CODING PROTOCOL
information [27: 18] Many of the studies of mobile use by MSEs are qualitative,
The quotation focuses directly on the basic tasks of running and do not report statistical findings. Even among quantitative
a business—reducing costs, increasing income, managing studies, there is little agreement in terms of dependent and
risk—and links them to core functions of mediated independent variables under scrutiny. Thus, a statistical meta-
communication technologies, particularly the substitution for analysis would not be applicable [30]. Similarly, a method
journeys. As demonstrated elsewhere, [26] the key is designed specifically for comparing ethnographies, such as
increased productivity. reciprocal translation [31] would be unlikely to bridge
qualitative and quantitative studies.
III. STUDIES ON MOBILES AND MSES The analysis draws instead on a systematic review
Recently, studies have emerged that directly address how methodology [32] to aggregate findings across the available
MSEs in the developing world are using mobiles rather than studies. By using a standardized protocol, coding each
landlines or other ICTs. The studies are not as numerous as individual study for the appearance or absence of certain
the enthusiasm in the popular press might suggest. They are a assertions, the review assesses and parsimoniously represents
tiny fraction of the total literature on mobile use in the what the research literature, in aggregate, suggests about
developing world [29]. They have emerged from different mobile use by MSEs. The exercise relies on clearly articulated
disciplines, and, as relative contemporaries, often do not cite eligibility criteria to select studies and on standardized
each other. This section presents two studies representing questions to evaluate them. These two levels of
distinct methodologies and conclusions, to provide an standardization, agreed upon before the formal review
example of the range of available perspectives and to set the commenced, separates the exercise from a conventional list-
stage for the systematic review. based or thematic/narrative literature review.
19
A. Selecting studies and producers, for example), but it became clear that the
Papers were initially identified by online literature and population of existing studies is too small to support that
database searches (using keyword combinations of mobile, inquiry.
cellular, microenterprise, MSE, etc.), and by snowball The final protocol employed three distinct foci. First, it
references from the bibliographies of studies already in hand. assessed the impact of mobile use on the internal process of an
Next, these studies were assessed against a series of eligibility enterprise, using Porter’s value chain model [23], Fig. 1. The
criteria: to be included, studies had to be specific to both value chain comprises the activity inputs into a product or
mobiles and MSEs, report generalizable findings, and contain service: inbound logistics, operations (production), outbound
detailed primary data about mobile use in everyday logistics, marketing and sales, and after-sales service. The
conditions. What started out as a reasonably large body of value chain also includes supporting functions: firm
studies was trimmed back significantly. In order to provide infrastructure, human resources, technology development
additional resources to other researchers, however, this section (knowledge developed or owned by the enterprise), and
lists those excluded papers, along with the rationale for the procurement. Together these activities can create customer
decisions. value in excess of the costs to provide it, yielding profit.
To be included in the review, papers had to be specifically Porter [23: 168], argues that information and communication
focused on mobile phones, which excluded some excellent technologies can be used to improve almost any of these
research on landlines or payphones and MSEs [28, 33, 34]. primary and supporting activities. Although the value chain
The papers also had to be about MSEs, not SMEs. Papers that framework was developed with larger enterprises in mind, it
did not explicitly include sole proprietors and informal can be applied to MSEs, since in small firms the same
enterprises were excluded [35-37]. individual can carry out different business-related activities
We made a more difficult decision to exclude papers that during the day. (Indeed, even in larger firms there is often not
were not generalizable to a wide range of MSEs. An a perfect mapping between the activities and functions in the
important line of research explores how many individuals earn value chain and distinct people or departments). Both
livelihoods in the mobile business itself, by selling airtime, researchers coded individual papers for mentions of the
fixing handsets or operating village phones [38-40]. However, mobile’s role in any of the primary or supporting functions.
these studies treat mobiles as products and services, rather
than enablers of general business processes.
To fit into the evaluation protocol, papers had to offer FIGURE 1: PORTER’S VALUE CHAIN
sufficient details around the use of mobiles to illuminate their
role in these business processes. A few surveys that were
otherwise topically correct did not yield information of this
kind [41], or blurred the lines between mobiles and other ICTs
to the point where assertions about mobiles in particular were
difficult to extract [42, 43]. Reviews without new primary
data were excluded [44, 45]. When multiple papers drew on
the same set of data [46, 47], only one paper was retained.
Finally, the review focused on analyses of mobile use in
everyday settings, rather than proposals for or evaluations of
new pilot technologies [48-51] or programmatic interventions
by NGOs [52]. The development of such technologies and
programs is central to the ICTD field, but such initiatives yield
different forms of evidence about mobile use than those that
examine MSEs operating on their own. note: image released to public domain as per
This limiting exercise forced a trade-off: the remaining http://en.wikipedia.org/wiki/Image:ValueChain.PNG
papers clearly describe some element of the use of mobiles by
MSEs in developing countries, but the population of such
studies is relatively small. Thirteen papers and one book were The second analysis used another Porter framework, the
retained. value system [23] ii , to offer a corresponding assessment of
how mobile use influences the network of interdependencies
B. Evaluation protocol and relationships external to the enterprise, including
The process of developing the evaluation questions was producers, traders, wholesalers, retailers and end-customers.
iterative. It was based mostly on an initial reading of the An initial reading of the papers identified four categories of
documents by the researchers, while also integrating current potential impacts. Some research stresses (a) the increased
narratives in the popular and practitioner literatures. An availability of information in the network; other studies stress
original goal was to code studies according to subcategories of (b) the entry of new actors, particularly buyers and sellers,
MSEs (to discern differences in mobile use between traders into markets. Both factors tend to increase competition, but do
20
so in different ways. One focuses on the actors in the network, V. RESULTS

the other on the information those actors exchange.
Nevertheless, the two changes do not necessarily move in A. Enterprise value chain
tandem; it is one thing to assert that the same set of actors Most studies mention the core processes of marketing and
exchange information at lower cost and higher frequency, sales [1, 2, 4, 6, 7, 9-13]. Analyses ranging from Jensen’s [10]
another to say that markets have expanded. This information- model of searching for the best price for fish to Kamga’s [11]
vs.-actors cut highlights this distinction. Two other categories description of improvements to the local laundry services in
of network impacts can be expressed as assertions that (c) Cote d’Ivoire asserts that mobiles help connect vendors and
mobiles help enterprises cut out middlemen and that (d) buyers, often at a distance and usually at lower cost than an
mobiles help individuals start new businesses. in-person journey. Esselaar et al. [6] report results of a survey
Unlike the value chain analysis (coded for affirmative of SMEs, including 1/3 microenterprises, conducted across 13
mentions only), the value system analysis coded for both countries. “Mobile phones are used more often for keeping in
affirmations and negations of the four potential impacts. After contact with customers and clients” (p 92). This is the highly
the initial reading, we elected to track negations since some of visible, intuitive role of mobiles for small enterprises.
the sources made a point of arguing against one or more of the The picture is sparser for other core processes within the
assertions from the popular and practitioner literature. value chain. Three studies mention inbound and outbound
The third analysis explored how the impact of mobile use logistics [1, 9, 13], particularly Abraham [1], who details how
on small enterprises is or is not different from the impact of fishermen can now use small supply boats (dispatched via
landline telephony on those same kinds of businesses [28, 33, mobile) to stay out fishing longer. Overå [13] describes how
34]. Technological properties of mobile communication make traders in Ghana can time harvests (inbound) and change the
it inherently more prone to adoption by MSEs than landlines: terms of delivery financing (outbound) because of the mobile.
it is cheaper to build towers than lay cable, prepay accounts Operations receives two mentions, again by [1], who notes
have no startup costs, and inexpensive/used handsets are that fishermen use mobiles to coordinate the timing and
readily available. However, in this case the third analysis location of when to drop nets and search for fish. Similarly,
focused narrowly on two differences in use rather than cost or [9] describes how weavers call customers mid-process to
access. revise plans for the garments they are creating.
First, studies were coded for mentions of mobility. Mobility Only studies by Frempong et al. [7] and Molony [11]
is a crucial difference between mobiles and landlines—while mention after-sales service. Molony describes how Tanzanian
landlines connect places to places, mobiles generally connect exporters of carved wood use the mobile to elicit feedback and
people to people, wherever they are and regardless of the time built trust with buyers after (and ideally between) sales.
and situation. This mobility leads to increased individual In terms of crosscutting functions, five studies reference
addressability, and can change how people structure social procurement [1, 6, 7, 9, 13] and address price search by
and economic activity [53-55]. Most relevant to this analysis, buyers of inputs (or by traders). There is little evidence to date
mobility may enable the rise of roaming businesses, just-infor the mobile’s role in transforming the proprietary
time service and what Townsend [56] has called the “real time technology, infrastructure or HR functions of MSEs, perhaps
city”. And yet some evidence suggests that in the developing because these enterprises are too small to invest in these
world many mobiles are purchased as substitutes for landlines, assets. Although studies outside the review [36, 58] provide
rather than complements to them [57]. anecdotes of small employers giving mobiles to employees,
Second, the nature of the mobile as a portable, personal this infrastructure function is of limited utility for tiny firms
device means it is particularly easy to use for both personal and sole-proprietorships.
and business functions during the same day. Thus, studies of
B. Industry value system
the role of mobiles in the lives of MSE operators are often
different from studies of the role of the device in the The second analysis turns the lens outside the enterprise,
businesses themselves. The analysis coded for studies that towards its location in a network of relationships. The most
explore these social functions. common finding links mobile use to an increase in the flow of
Once the protocol was established, each researcher re-read information between actors in the value system [1, 2, 6, 7, 9-
the papers, coding them in isolation. We then compared our 14]. The two primary sub-themes are more frequent or wide-
codes and resolved any discrepancies through discussion. The ranging exchanges of price information [1, 2, 13], and a more
resulting codes are less prone to reflect the bias of a single generalized discussion of increased communication with
reader. Of the 112 cells on the matrix requiring codes, 16 customers [6, 12, 14]. These findings are reflections of the
required discussion to resolve coding discrepancies between frequent references to marketing and sales and procurement
the two researchers. activities in the previous value chain analysis. Reference [9]
mentions an increase in the completeness of the information,
but notes that they saw no increase in quality
21
TABLE I: SUMMARY OF THE MSE AND MOBILE STUDIES IN THE REVIEW
Enterprise Value Chain Industry Value System Uses

Add Add Bypass Start
Studies Cross-cutting Inform- Buyers/ Middle- busi-
Core Processes Functions ation Sellers men nesses Mobility Social
Inbound &
[1] R. Abraham, “Mobile phones and
Outbound Logistics,
economic development: evidence from the Procurement Yes Yes No -- Yes Yes
Operations,
fishing industry in India,”
Marketing & Sales.
[2] J. C. Aker, “Does digital divide or
provide? The impact of cell phones on grain Marketing & Sales -- Yes Yes -- -- Yes --
markets in Niger”
[3] J. Donner, “Microentrepreneurs and
mobiles: An exploration of the uses of
-- -- -- -- -- -- Yes Yes
mobile phones by small business owners in
Rwanda”
[4] J. Donner, “The use of mobile phones by
microentrepreneurs in Kigali, Rwanda: Marketing & Sales -- -- Yes -- -- -- Yes
Changes to social and business networks”
[5] J. Donner, “Customer acquisition among
small and informal businesses in urban
-- -- -- No -- -- -- Yes
India: Comparing face to face, interpersonal,
and mediated channels”
[6] S. Esselaar, C. Stork, A. Ndiwalana, and
M. Deen-Swarra, “ICT usage and its impact
Marketing & Sales Procurement Yes -- -- -- -- Yes
on profitability of SMEs in 13 African
Countries”
[7] G. Frempong, G. Essegbey, and E.
Tetteh, “Survey on the use of mobile Marketing & Sales;
Procurement Yes Yes -- -- -- --
telephones for micro and small business Service
development: The case of Ghana,”
[8] H. Horst and D. Miller, “The Cell
Phone: An Anthropology of -- -- -- -- -- No -- Yes
Communication.”
[9] A. Jagun, R. Heeks, and J. Whalley, Inbound &
“The Impact of Mobile Telephony on Outbound Logistics,
Procurement Yes Yes No -- -- --
Developing Country Micro-Enterprise: A Operations,
Nigerian Case Study” Marketing & Sales
[10] R. Jensen, “The Digital Provide:
Information (Technology), Market
Marketing & Sales -- Yes Yes -- -- Yes --
Performance, and Welfare in the South
Indian Fisheries Sector”
[11] O. Kamga, “Mobile phone in Cote
Marketing & Sales -- Yes -- -- -- Yes Yes
d'Ivoire: uses and self-fulfillment”
[12] T. Molony, “‘I don't trust the phone; it
always lies’: Trust and information and Marketing & Sales;
-- Yes No No -- -- Yes
communication technologies in Tanzanian Service
micro- and small enterprises”
[13] R. Overå, “Networks, distance, and Inbound Logistics,
trust: Telecommunications Development Outbound Logistics, Procurement Yes Yes -- -- Yes Yes
and changing trading practices in Ghana” Marketing & Sales
[14] J. Samuel, N. Shah, and W.
Hadingham, “Mobile Communications in
-- -- Yes Yes -- Yes Yes Yes
South Africa, Tanzania, and Egypt: Results
from Community and Business Surveys”
22
While it is one thing to say that mobile use accelerates the introduced, there is more evidence for changes in degree
flow of information in existing value systems, it is another to (more information, more customers) than for changes in
say that mobile use brings new customers or suppliers into the structure (new channels, new businesses).
market. This is the first element in the systematic review in
C. On attributes of the mobile vs. the landline
which there is some disagreement between the primary
studies. Numerous studies present evidence that mobile use Roughly half of the studies described use cases that take
expands markets by allowing MSEs to reach new customers advantage of mobility. Clearly, fishermen take advantage of
[1, 2, 4, 7, 9, 10, 13, 14]. Of the Keralan fishermen, [10: 891] wireless telecommunications [1, 10] to place and receive calls
explains “while almost all sales before mobile phones were while on the water. This is not only an advantage for
conducted via beach actions, fishermen with phones, often determining which markets to target, but [1] points out that it
carrying lists with the numbers of dozens or even hundreds of also enables fishermen to feel safer while at sea. Traders [2,
potential buyers, would typically call several buyers in 13] use the mobile to be individually addressable wherever
different markets before deciding where to sell their catch”. they are. Reference [11] illustrates the responsiveness of
Similarly, Aker finds that “Grain traders in markets with cell businesses that can serve the customer, 24 hours a day, while
phone coverage search over a greater number of markets, have [13] portrays “availability as comparative advantage”, and
more contacts and sell in more markets. This underscores the argues that this more frequent interaction builds trust between
fact that the primary mechanism by which cell phones affect suppliers and customers.
market efficiency is a reduction in search costs and hence Given that MSE operators often carry their mobiles
transaction costs.” [2: 4-5] In the study, mobile use lowered throughout the day and into the evening, a blending of
price dispersion by 21%, increased profits by 29%. mediated communication for social and instrumental purposes
And yet two other studies specifically argue that the phones often occurs. While some of the papers in the review focus
have done little to introduce new buyers. Donner [5] finds that exclusively on the business functions, others [1, 3-6, 8, 11-14]
MSEs in urban India are much more likely to recruit illustrate this blurring. Blurring occurs at the aggregate level,
customers via face-to-face channels, rather than via a phone —a survey by Donner [4] found that roughly 1/3 of calls made
call. [12] argues that mobiles help accelerate and strengthen by MSE owners in Rwanda were business-related. It also
trusting relationships but only among parties that have already occurs within individual calls—non-business (“chit-chat”)
established a face-to-face bond. exchanges increase trust between clients and customers [12,
Two other general assertions about the impact of mobile 13]. Finally [8] describes the “link up” process in Jamaica, in
use on MSE value systems receive less support from the which individuals retain a roster of numbers of friends, family
studies. None of the studies asserts that mobiles help MSEs and acquaintances that can be tapped periodically for loans or
bypass middlemen. Indeed three of the papers focus small cash gift transfers. This process intermingles social and
specifically on middlemen, wholesalers or traders as economic functions of mobile use.
enterprises, [2, 9, 13] describing how mobiles allow them to
perform their roles more effectively. Another specifically VI. DISCUSSION
emphasizes how producers work with existing middlemen in This paper offers a systematic review of the current
their industries, rather than routing around them. Rather than research on the impact of mobile use on MSEs, applying both
radically restructuring these marketplaces, Molony argues, an internal (value chain) and external (value system)
“mobile phones can be seen as a facilitating technology for perspective. The review finds a pattern of evidence suggesting
existing, trust-based relationships” [12: 78] that mobiles increase the information available to MSEs.
Similarly, there is relatively little evidence for the assertion Some [2, 10] provide quantitative evidence for how this
that mobiles help people start new businesses. Only Samuel et information translates into reduced price variability and higher
al. [14] make this case, reporting that among a sample of profits per actor. The current studies suggest mobiles are most
MSEs in Egypt and South Africa, 26%-29% of businesses useful for streamlining marketing and sales (downstream) and
attributed their start to the availability of the mobile. Taking procurement (upstream) with existing business contacts. In
the opposite position, Horst and Miller [8: 164] argue that some cases, studies suggest that mobile use expands the size
despite some isolated examples to the contrary (taxi drivers of markets by bringing a larger number of buyers and sellers
and musicians), “there is no new spirit of enterprise based on into the marketplace. However not all studies found evidence
either the cell phone or the internet” among the Jamaican that new customers were acquired. Far fewer studies present
households in their study. Nevertheless, they argue that evidence that mobiles enable the creation of new businesses,
despite a dearth of new enterprises, the mobile is essential to or that mobile use re-organizes value systems to allow
the economic survival of those households. By allowing producers to bypass middlemen. Indeed, middlemen are
individuals to leverage broad networks of informal social and positioned to take advantage of mobiles themselves.
financial support through a process Jamaican mobile users call To summarize, the review of the evidence offered across
“link up,” “the phone is not central to making money, but is the thirteen studies suggests that within the MSE sector,
vital to getting money.” [8: 165] benefits of mobile use accrue mostly (but not exclusively) to
In sum, in value systems where mobile telephony is existing enterprises, in ways that amplify and accelerate
23
material and informational flows, rather than fundamentally range of variables for both the independent and dependant
transform them. This summary does not diminish the positive sides of such analyses. Indeed, a quantification of mobile use
utility of mobiles to MSEs, but it places that utility in context by subpopulations of MSEs may begin to close the gap
and in doing so echoes assertions by Castells [54] and Harper between micro-level case studies and research on the impact
[59]. Mobile use by MSEs can be understood as an extension of mobiles on macroeconomic growth [63].
of the information society, not its restructuring [60].
B. Enterprises, livelihoods or lives?
A second theme in the analysis involves mobiles as
complements or substitutes for landlines. Evidence here The conclusions of this review naturally depend on the
remains mixed; while some case studies highlight enterprises methods and theoretical frameworks employed at the primary
that take advantage of “availability as comparative advantage” and secondary stages of analysis. Ethnographies such as [11]
[13], there is insufficient data to determine whether these are and [9] tended to discuss a broader range of uses and impacts
isolated cases or representations of a more universal than focused quantitative tests [2, 5, 10]. Similarly, the lenses
condition. The majority of MSEs may take greater advantage chosen for this systematic review highlight some dimensions
of place-to-place connectivity, than mobility, but this point of MSE use over others. For example, a livelihoods
merits further study. framework [64], instead of enterprise-specific frameworks
from Porter [23] would have emphasized different patterns.
A. Generalization and segmentation The dichotomous treatment of new vs. existing businesses
Clearly, mobiles offer distinct benefits to MSEs— has its limitations. Reference [9] describes the evolution of
everything from more accurate price information and access to some weavers to “coordinator weavers,” suggesting a
new customers to better after-sales service and procurement, transformation in the structural location and internal processes
from increased responsiveness to the opportunity to build trust of some enterprises, but this was an exception among the
at a distance. However, it is currently difficult to determine studies, most of which looked only to existing enterprises for
whether the various impacts and benefits enumerated in this their sample. The methods used by [8] can identify
review accrue equally to all MSEs. Looking across a diversity occupational multiplicity—holding down more than one job at
of MSEs, across nations, industries, and different locations in once—in a way that studies focused specifically on existing
value systems, the current research points toward a MSEs cannot do. Similarly, [7] is able to assert that few
multiplicity of intertwined and sometimes contradictory households start new enterprises thanks to mobile use only
impacts of mobile use, rather than the universal and rather because households rather than MSEs are its primary units of
uncomplicated benefits which have characterized the popular analysis.
rhetoric. These examples reveal that insights about use of mobiles in
To guide future policy or institutional interventions, it would MSEs can come from studies that focus not exclusively on
be helpful to deploy future research against a set of open enterprises, but rather on the individuals who manage them.
questions: Which kinds of MSEs are gaining the most return Focusing on individuals also allows for increased linkage to
on mobile use? Which kinds (and what proportion) of MSEs research on social uses of the mobile.
are poised to find new customers and expand their markets,
C. New applications on the mobile platform
and not simply be more responsive to the ones they already
have? Which kinds (and what proportion) are unlikely to reap Studies have documented how mobiles can enable
any benefit from mobiles, or actually may be threatened by information search and improve communication between
changes in mobile use elsewhere in their value systems? MSEs, customers and suppliers. However, there has been little
These questions remain open because most studies to date evidence to date that suggests mobiles are being used for
have been either sector-specific explorations or broad information storage or processing. As was the case with
aggregate surveys; few studies specifically identify landlines [27], MSEs value voice calls more than any other
differences in mobile use or impact among subpopulations of function on the mobile, and use the calls to augment, rather
MSEs. By contrast, recent studies of mobile use by farmers than replace, face-to-face ties. As [6: 99] explains, mobiles
are identifying factors which differentiate between subgroups, “cannot be used to track inventory, provide cash flow and
for example, between growers of perishable and non- income statements, or even more basically, produce formal
perishable crops [61], by distance from local markets [61], or letters, marketing campaigns, or brochures”.
according to different levels of infrastructural constraint [62]. Recently, however, various systems have appeared that go
An important path for further study would apply similar beyond the voice and peer-to-peer texting functions on the
comparative analyses to assess and predict the impact of handset. These make the handset approximate a PC (with
mobile use by different classes of MSEs. processing happening on the handheld), or as the client in a
These are hefty quantitative tasks. Future designs will client-server model, with primary processing happening
require increased attention to the factors that distinguish elsewhere on the network. These latter models take advantage
subgroups of MSEs as well as careful measurement of of everything from basic SMS [65] to voice to full-blown
desirable outcomes such as productivity, market participation, mobile internet browsing experiences. Two of the more
or revenue growth. However, this review helps to identify a promising applications are distributed marketplaces, such as
24
Cell Bazaar, Manobi, and Tradenet, and mobile- International Conference on Information and Communication
Technologies and Development, (ICTD 2006). Berkeley, CA:
banking/mobile-payments initiatives such as Kenya’s popular IEEE, pp. 48-56, 2006.
M-PESA [66]. In addition, a variety of promising pilots are [2] J. C. Aker, “Does digital divide or provide? The impact of cell
underway, such as [48-51], which promise to yield further phones on grain markets in Niger,” BREAD Working Papers. vol.
functionality, for example, in supply chain management. 177, Durham, NC: Bureau for Research and Economic Analysis of
Development, 2008.
Full evaluations of the use and impact of these services on [3] J. Donner, “Microentrepreneurs and mobiles: An exploration of the
MSEs are not yet available, and are urgently needed. This uses of mobile phones by small business owners in Rwanda,”
baseline analysis can help clarify which impacts these systems Information Technologies and International Development, vol. 2,
pp. 1-21, 2004.
may be having. Findings that suggest, for example, that [4] ___, “The use of mobile phones by microentrepreneurs in Kigali,
mobile trading platforms help MSEs bypass middlemen, Rwanda: Changes to social and business networks” Information
would be even more impressive given this synthesis’ Technologies and International Development, vol. 3, pp. 3-19,
2006.
conclusion that there is limited evidence that basic voice calls [5] ___, “Customer acquisition among small and informal businesses
can have this effect. Similarly, findings that suggest MSEs are in urban India: Comparing face to face, interpersonal, and
using m-banking or m-payments applications to transform mediated channels,” The Electronic Journal of Information
Systems in Developing Countries, vol. 32, pp. 1-16, 2007.
credit relationships or otherwise change the procurement and [6] S. Esselaar, C. Stork, A. Ndiwalana, and M. Deen-Swarra, “ICT
sales functions could be assessed more accurately against the usage and its impact on profitability of SMEs in 13 African
baseline of the voice-based behaviors found so far. countries,” Information Technologies and International
Development, vol. 4, pp. 87-100, 2007.
[7] G. Frempong, G. Essegbey, and E. Tetteh, “Survey on the use of
VII. CONCLUSION mobile telephones for micro and small business development: The
This paper has summarized fourteen primary research case of Ghana,” Institutes of the Council for Scientific and
Industrial Research, Science and Technology Policy Research
studies examining mobile use by MSEs; the research generally (STEPRI), Accra, 2007.
concurs with conventional wisdom—mobile use helps many [8] H. Horst and D. Miller, The cell phone: An anthropology of
MSEs become more productive, particularly but not communication. Oxford: Berg, 2006.
[9] A. Jagun, R. Heeks, and J. Whalley, “The impact of mobile
exclusively via improvements to sales and marketing and telephony on developing country micro-enterprise: A Nigerian case
procurement processes. That said, the review suggests that not study,” Information Technologies and International Development,
all enterprises will prosper from increased access to vol. 4, pp. 47-65, 2008.
[10] R. Jensen, “The digital provide: Information (technology), market
telecommunications, and among those that do, their uses of performance, and welfare in the South Indian fisheries sector,”
mobiles will vary across industries and positions in value Quarterly Journal of Economics, vol. 122, pp. 879-924, 2007.
systems. As mentioned above, current evidence suggests that [11] O. Kamga, “Mobile phone in Cote d'Ivoire: Uses and self-
the benefits of mobile use accrue mostly (but not exclusively) fulfillment,” Proceedings of the International Conference on
Information and Communication Technologies and Development,
to existing enterprises, in ways that amplify and accelerate (ICTD 2006). Berkeley, CA, pp. 184-192, 2006.
material and informational flows, rather than fundamentally [12] T. Molony, “‘I don't trust the phone; it always lies’: Trust and
transform them. information and communication technologies in Tanzanian micro-
and small enterprises,” Information Technologies and
On balance, MSEs are likely to remain unproductive International Development, vol. 3, pp. 67-83, 2006.
relative to larger enterprises [17]. However, the improvements [13] R. Overå, “Networks, distance, and trust: Telecommunications
to productivity associated with mobile use do seem to be development and changing trading practices in Ghana,” World
Development, vol. 34, pp. 1301-1315, 2006.
improving the livelihoods of many individuals in the MSE [14] J. Samuel, N. Shah, and W. Hadingham, “Mobile communications
sector. in South Africa, Tanzania, and Egypt: Results from community and
The results of this review are helpful to the ICTD research business surveys,” in Africa: the impact of mobile phones. vol.
2007, Vodafone, Ed. London: Moving the Debate Forward: The
community in at least three ways. First, by disaggregating and
Vodafone Policy Paper Series #3, 2005, pp. 44-52.
identifying distinct impacts of mobile use, the review provides [15] International Labor Organization, “Key indicators of the labour
a more nuanced and more accurate representation of the value market,” International Labor Organization, Geneva 2003.
of mobile use to MSEs than was previously available. Second, [16] D. C. Mead and C. Leidholm, “The dynamics of micro and small
enterprises in developing countries,” World Development, vol. 26,
the review identifies a skew (in both sample and implied pp. 61-74, 1998.
impacts) towards existing enterprises that should be noted by [17] R. la Porta and A. Shleifer, “The unofficial economy and economic
policymakers who may expect mobiles to create new development,” Brookings Institute, Washington, DC 2008.
[18] D. Snodgrass, “Inventory and analysis of donor-sponsored MSE
businesses and new employment. Finally, the review identifies development programs,” Development Alternatives Inc.,
two priorities for future research: a segmentation and further Washington, DC 2005.
quantification of impacts by subsectors of MSEs, and an [19] Economist, “Calling an end to poverty,” The Economist. vol. 376,
2005, pp. 51-52.
assessment of the use of new non-voice advanced mobile [20] S. LaFraniere, “Cellphones catapult rural Africa to 21st century,”
services (such as mobile banking and mobile marketplaces) by in The New York Times New York, p. 1, 2005.
MSEs [21] USAID, “Using cellular phones in Uganda for rural income
generation and more,” Issue 7 ed. vol. 2006: DOT-COM Alliance
E-Newsletter, 2004.
REFERENCES [22] A. Bhavnani, R. Won-Wai Chiu, S. Janakiram, and P. Silarszky,
[1] R. Abraham, “Mobile phones and economic development: “The role of mobile phones in sustainable rural poverty reduction,”
Evidence from the fishing industry in India,” Proceedings of the
25
World Bank Global Information and Communications Department, Washington, DC: Financial Times Limited / International Finance
Washington, DC 15 June 2008. Corporation (IFC), pp. 43-54, 2007.
[23] M. E. Porter, Competitive advantage: Creating and sustaining [45] A. T. Rashid and L. Elder, “Mobile phones and development: An
superior performance. New York: Free Press, 1985. analysis of IDRC-supported projects,” Electronic Journal of
[24] D. C. Mead and C. Morrison, “The informal sector elephant,” Information Systems in Developing Countries, vol. 36, pp. 1-16,
World Development, vol. 24, pp. 1611-1619, 1998. 2009.
[25] R. Duncombe and R. Heeks, “Enterprise development and [46] T. Molony, “Running out of credit: The limitations of mobile
information and communications in developing countries: telephony in a Tanzanian agricultural marketing system,” Journal
Supporting ‘ICT flyers’,” Manchester: IDPM, University of of Modern African Studies, vol. 46, pp. 1-22, 2008.
Manchester, 2001. [47] ___, “The role of mobile phones in Tanzania’s informal
[26] R. J. Saunders, J. J. Warford, and B. Wellenieus, construction sector: The case of Dar es Salaam” Urban Forum, vol.
Telecommunications and economic development, 2 ed. Baltimore, 19, pp. 175-186, June 2008.
MD: Johns Hopkins University Press, 1994. [48] P. S. Javid and T. S. Parikh, “Augmenting rural supply chains with
[27] R. Duncombe and R. Heeks, “Information and communication a location-enhanced mobile information system,” Proceedings of
technologies and small enterprise in Africa: Findings from the International Conference on Information and Communication
Botswana." vol. 2006 Manchester, UK: Institute for Development Technologies and Development, (ICTD 2006). Berkeley, CA pp.
Policy and Management, University of Manchester, 1999. 110-119, 2006.
[28] ___, “Enterprise across the digital divide: Information systems and [49] A. Kumar, N. Rajput, S. Agarwal, D. Chakraborty, and A. A.
rural microenterprise in Botswana,” Journal of International Nanavati, “Organizing the unorganized - employing IT to empower
Development, vol. 14, pp. 61-74, 2002. the under-privileged,” Proceedings of the 17th international
[29] J. Donner, “Research approaches to mobile use in the developing conference on World Wide Web, Beijing, pp. 935-944, 2008.
world: A review of the literature,” The Information Society, vol. [50] D. Chakraborty, K. Dasgupta, S. Mittal, A. Misra, A. Gupta, E.
24, pp. 140-159, 2008. Newmark, and C. L. Oberle, “Businessfinder: Harnessing presence
[30] R. J. Light and D. B. Pillemer, Summing up: The science of to enable live yellow pages for small, medium and micro mobile
reviewing research. Cambridge, MA: Harvard University Press, businesses,” IEEE Communications Magazine, vol. 45, pp. 144-
1984. 151, 2007.
[31] G. W. Noblit and R. D. Hare, Meta-ethnography: Synthesizing [51] S. Biswas and S. Roy, “A shared ICT infrastructure for Indian
qualitative studies. Newbury Park, CA: Sage, 1988. SME clusters,” in 2nd IEEE/ACM International Conference on
[32] J. H. Littell, J. Corcoran, and V. Pillai, Systematic reviews and Information and Communication Technologies and Development
meta-analysis. New York: Oxford University Press, 2008. (ICTD2007), Bangalore, pp. 221-230, 2007.
[33] D. Souter, N. Scott, C. Garforth, R. Jain, O. Mascararenhas, and K. [52] Food India, “Inter-city marketing network for women micro-
McKemey, “The economic impact of telecommunications on rural entrepreneurs using cell phone,” Food India 2002.
livelihoods and poverty reduction: A study or rural communities in [53] J. E. Katz and M. Aakhus, “Perpetual contact: Mobile
India (Gujarat), Mozambique, and Tanzania,” Commonwealth communication, private talk, public performance,” Cambridge,
Telecommunications Organisation for UK Department for UK: Cambridge University Press, 2002.
International Development, 2005. [54] M. Castells, J. L. Qiu, M. Fernández-Ardèvol, and A. Sey, Mobile
[34] R. Duncombe and R. Heeks, “Information, ICTs and small communication and society: A global perspective (information
enterprise: Findings from Botswana,” Small Scale Enterprises in revolution and global politics). Cambridge, MA: MIT Press, 2007.
Developing and Transitional Economies, H. Katrak and R. [55] C. Licoppe, “'connected' presence: The emergence of a new
Strange, Eds. New York: Palgrave, 2002, pp. 285-304. repertoire for managing social relationships in a changing
[35] J. O. Adeoti and A. I. Adeoti, “Easing the burden of fixed communication technoscape,” Environment and Planning D:
telephone lines on small-scale entrepreneurs in Nigeria: GSM lines Society and Space, vol. 22, pp. 135-156, 2004.
to the rescue,” Telematics and Informatics, vol. 25, pp. 1-18, 2008. [56] A. M. Townsend, “Life in the real-time city: Mobile telephones
[36] R. A. Boadi, R. Boateng, R. Hinson, and R. A. Opoku, and urban metabolism,” Journal of Urban Technology, vol. 7, pp.
“Preliminary insights into m-commerce adoption in Ghana,” 85-104, 2000.
Information Development, vol. 23, pp. 253-265, 2007. [57] J. Hamilton, “Are main lines and mobile phones substitutes or
[37] W. Mei and L. H. Yun, “Perpetual contact for business: Mobile complements? Evidence from Africa,” Telecommunications Policy,
phones and Fujian entrepreneurs,” Proceedings of ANZCA08: vol. 27, pp. 109-133, 2003.
Power and Place, Wellington, NZ, 2008. [58] J. Donner, “The social and economic implications of mobile
[38] S. Aminuzzaman, H. Baldersheim, and I. Jamil, “Talking back: telephony in Rwanda: An ownership/access typology,” in Thumb
Empowerment and mobile phones in rural Bangladesh: A study of culture: The meaning of mobile phones for society, P. Glotz, S.
the village pay phone of Grameen bank,” Contemporary South Bertschi, and C. Locke, Eds. Bielefeld, Germany: Transcript
Asia, vol. 12, pp. 327-348, 2003. Verlag, 2005, pp. 37-52.
[39] H. Ilahiane and J. Sherry, “Joutia: Street vendor entrepreneurship [59] R. Harper, “Are mobiles good or bad for society?,” in Mobile
and the informal economy of information and communication democracy: Essays on society, self and politics, K. Nyíri, Ed.
technologies in morocco,” The Journal of North African Studies, Budapest, Hungary: Passagen Verlag, pp. 185-214, 2003.
vol. 13, pp. 243-255, 2008. [60] J. Donner, “Shrinking fourth world? Mobiles, development, and
[40] J. Lugo and T. Sampson, “E-informality in Venezuela: The 'other inclusion,” in Handbook of mobile communication studies, J. Katz,
path' of technology,” Bulletin of Latin American Research, vol. 27, Ed. Cambridge, MA: MIT Press,2008, pp. 29-42.
pp. 102-118, 2008. [61] M. Muto and T. Yamano, “The impact of mobile phone coverage
[41] B. F. Chogi, “The impact of mobile phone technologies on medium expansion on market participation: Panel data evidence from
and small enterprises/jua kali (MSEs),” Proceedings of CPRsouth2 Uganda,” Japan Bank for International Cooperation, Tokyo 2008.
Chennai, 2007. [62] S. Gandhi, S. Mittal, and G. Tripathi, “The impact of mobiles on
[42] R. O. Opiyo and O. A. K'Akumu, “ICT application in the informal agricultural productivity,” in India: The Impact of Mobile Phones,
sector: The case of the Kariokor market MSE cluster in Nairobi,” The Vodafone Policy Paper Series #9 London: Vodafone, 2009,
Urban Forum, vol. 17, pp. 241-261, 2006. pp. 21-33.
[43] E. D. Moyi, “Networks, information and small enterprises: New [63] L. Waverman, M. Meschi, and M. Fuss, “The impact of telecoms
technologies and the ambiguity of empowerment,” Information on economic growth in developing nations." vol. 2007 London:
Technology for Development, vol. 10, pp. 221-232, 2003. Moving the Debate Forward: The Vodafone Policy Paper Series
[44] K. O. Davis, Cosmas. , “ICTs as appropriate technologies for #3, 2005, pp. 10-23.
African development” IFC/FT First Annual Essay Competition: [64] R. Duncombe, “Using the livelihoods framework to analyze ICT
Business & Development: The Private Path to Prosperity, applications for poverty reduction through microenterprise,”
26
Information Technologies and International Development, vol. 3, Conference on Information and Communication Technologies and
pp. 81-100, 2007. Development (ICTD2007), Bangalore, pp. 89-98, 2007.
[65] R. Veeraraghavan, N. Yasodhar, and K. Toyama, “Warana [66] O. Morawczynski and G. Miscione, “Examining trust in mobile
unwired: Mobile phones replacing PCs in a rural sugarcane banking transactions in Kenya: The case of m-PESA in Kenya,” in
cooperative,” Proceedings of the 2nd IEEE/ACM International IFIP WG 9.4-University of Pretoria Joint Workshop, Pretoria,
South Africa, 2008
i
The analysis includes two studies of fishermen but excludes
farmers
ii
The term ‘value system’ is sometimes used interchangeably
with ‘industry value chain’. This review uses the Porter
nomenclature to distinguish between the intra- and extra-
enterprise systems
27
An Evaluation of the use of ICT within Primary

Education in Malawi
David Hollow and Paola Masperi
the authority of the Ministry of Education, Science and

Abstract— The paper demonstrates how appropriate technology, Technology (MoEST). Funding for the initiative was provided
when combined with quality curriculum-based content, has the by a British company who also developed the educational
potential to have a positive impact on primary education within
software for the portable devices.
developing countries. It documents an evaluation of portable
learning technology from the Ministry of Education, Science and
Technology in Malawi, providing theoretical background and
educational context before detailing the methodology and II. CONTEXT
findings of the study. Significant impacts from the introduction
of the program were increased school attendance, reduced drop- A. Education in Malawi
out rates and improved student and teacher enthusiasm. Free primary education was introduced in Malawi in 1994 as
Alongside this, the audio and video material and interactive
learning techniques offered potential pedagogical benefits in
a result of the commitment made by the country at the World
combining learner-centered and outcome-based activities with Conference on Education for All (EFA) in 1990. Widespread
continuous assessment, helping children to retain information agreement exists regarding the vital place of education within
more effectively and record higher test scores. Major challenges poverty reduction efforts, capacity building and growth
were also identified regarding implementation and sustainability. strategies of developing countries. In Malawi, numerous
These centered around the need for rigorous teacher training, policy, budgetary and multilevel commitments have
classroom integration, appropriate deployment, maintenance,
sustained impact and overall cost-effectiveness. In closing, the
contributed to significant progress in the delivery of primary
paper emphasizes the need for such programs to be driven by education since 1994. This includes enrolment figures rising
educational concerns and recognizes the similar challenges faced from 1.9 million to 3.2 million, the construction of 1,000 new
in many related initiatives. classrooms and roll out of the new curriculum to 5,500
schools [1]. However, Malawi is one of the poorest countries
Index Terms— Education, Evaluation, Malawi, Technology in the world, with a GNI per capita of $230. Life expectancy
at birth is 48 years and 63% of the population live on less than
$2 per day [2]. Within such a context of extreme poverty, the
I. INTRODUCTION rapid advancements in access to free schooling have put the
I N setting the context for the subsequent evaluation the

paper begins with an overview of two distinct spheres,
firstly the current education context in Malawi and secondly
national education system under considerable strain.
The increase in enrolment has led to significant pressure
upon primary schools in Malawi, most notably in regard to
the potential role of technology within education. These two increasing class sizes, a lack of fully qualified teachers,
seemingly disparate themes are brought together through the limited teaching materials and inadequate infrastructure. It is
case study of the Interactive Learning Program. An overview estimated that an extra 8,000 teachers are currently required to
of the program is provided, followed by an explanation of the meet the MoEST target pupil to teacher ratio of 1:60 [3]. In
methodological approach employed in the monitoring and addition, and despite national net enrolment levels of 91%,
evaluation exercise. The findings are then categorized into national drop out rates remain high with only 44% of those
four sections of analysis which inform the concluding that enroll in Standard 1 completing Standard 5 and less than
comments and recommendations. 30% reaching Standard 8, the final year of primary education
The Interactive Learning Program in Malawi comes under [4]. Linked to this and demonstrating the ongoing challenge in
regard to teaching capabilities and educational outcomes is a
Manuscript received September 21, 2008. declining performance in national examinations with a failure
David Hollow is a PhD candidate with the ICT4D Collective in the rate of over 40% [1].
Geography Department at Royal Holloway, University of London. e-mail: The importance of ensuring good quality education
d.m.hollow@rhul.ac.uk
Paola Masperi is an independent development consultant, based in London. alongside improved access [5] has been long recognized in
e-mail: paola@mayamiko.org Malawi and in 1999 it was decided that a comprehensive
28
reform of the curriculum was required due to the based content and sustained classroom integration.
developments that had occurred since 1994 [3]. The relevance
of topics such as democracy, human rights, gender and
HIV/AIDS were increasingly acknowledged and these have III. INTERACTIVE LEARNING PROGRAMME
been systematically incorporated into the new primary school Having outlined the overall context of education within
curriculum in the form of life skills education. Malawi and the theoretical role of technology within
education, attention now turns to the Interactive Learning
B. Technology in Education Program. It considers whether the initiative constitutes a
suitable application of portable technology to support the
There is significant attention focused upon the potential for
provision of basic education within the country. To begin, the
ICT to assist in leapfrogging educational challenges within
specifics of the program are explained, including the
developing nations [6] [7] [8], with much enthusiasm for a
background, technology and pedagogical rationale.
possible technology-enabled ‘breakthrough in learning’ [9].
However, much debate surrounds the question of whether the
infusion of technology into education has actually instigated A. Background to Program
more than incremental changes to the field. Proponents assert The Interactive Learning Program was introduced in 2006
that the last decade have resulted in the emergence of a new and following initial positive feedback the MoEST requested,
landscape for education across the globe [10] with technology in early 2007, that the initiative be scaled up to incorporate 50
positively effecting student motivation [11] and, when primary schools. The schools that the MoEST selected for
initiatives are implemented with fidelity, leading to a participation ranged from Karonga District in the north
significant increase in learning [12]. This is challenged by through to the southernmost districts of Mulanje and
those who argue that achieving structural technological Phalombe. A total of 520 custom-made handheld interactive
change in schools takes much longer than anticipated [13] learning aids were distributed to the participating schools.
with no substantial evidence that the introduction of This totaled ten devices for each school, except for two test
technology has yet caused any fundamental change in either a schools which received 20 devices for the purposes of targeted
developed or developing context [14]. Wagner [15] notes a monitoring and evaluation.
wide variety of outcomes from ICT for education projects in
developing countries, with significant negative impacts
including the reinforcing of dependencies, imposition without B. Technology Utilized
community involvement and collapse due to lack of funding The device in question is a handheld interactive learning aid
or political commitment. (from herein referred to as the learning machine or gadget as
Within this contested environment, the range of technologies named by the participating children), slightly larger than a
theoretically available for deployment within developing mobile phone and able to play video and audio through either
country educational contexts is rapidly expanding, with a a loudspeaker or headphones. Positioned next to the screen are
transition towards increasingly portable, powerful and a selection of buttons which can be pressed in response to
adaptable tools [16]. The new technologies available have questions asked in the lesson being watched. The device has
potential to mark a significant transition beyond an in-built rechargeable lithium battery with power for
conceptualizing e-learning through conventional static between four and six hours of continuous play. A total of 25
computer laboratories [17], emphasising instead the place of lessons in Chichewa and 40 in English are preinstalled and
anytime, anywhere computing [16] and mobile learning [18]. stored on two gigabytes of internal Flash memory, leaving
Such a shift creates opportunity for major changes in the additional room for newly developed lessons to be
application of educational technology, with key potential incorporated. The lessons are designed for use by children in
benefits noted as increased enthusiasm, cooperation, Standard 3, 4 and 5, each lasting approximately 20 minutes
communication and student ownership [19]. and covering a range of curriculum based subjects such as
Despite the potential benefits of learning technologies [20] General Studies, Social Studies, Science, Mathematics,
it is important to recognize that not all solutions which have Geography, Life Skills and English.
contributed to educational advancements in the developed With less than 5% of primary schools across Malawi
world can simply be transferred to a developing world estimated as having access to a reliable supply of mains
context. In order for the potential of ICTs to be realized in electricity [3] the initiative was dependent upon solar charging
improving the quality of education it is important they are systems. Each of the project schools were equipped with a 20-
applied with cultural understanding, local knowledge and Watt solar panel connected to a deep-cycle sealed battery,
sensitivity [11] [21]. Indeed, ‘effective use of ICTs must be generating enough energy to power 20 of the learning
tied to the needs of developing countries and challenge the machines. Teachers were given training regarding appropriate
one size fits all approach of many programmes’ [19;7]. A charging and usage, and instructed to leave the devices to
recognition of this fact emphasises the need for primary focus recharge overnight once every two days.
on appropriate software development linked to curriculum-
29
C. Pedagogy monitoring and evaluation. Where undertaken, there is often

The digital content on the learning machines is designed in strong emphasis placed on hardware-based input indicators
order to enable and promote interactive and outcome-based [25] rather than more complex learning outcomes. In order to
learning, linked to the Malawi national curriculum and avoid this simplification and demonstrate an alterative
actively supporting the government Primary Curriculum approach, a rigorous monitoring, evaluation and impact
Assessment Reform [22]. The lessons are designed to be user assessment structure was built into the program throughout
friendly, with the audio-visual content accessible to illiterate implementation. This involved three field visits from the
or semi-literate users. Instructions from lip synchronized monitoring and evaluation team in September 2007,
cartoon characters explain the functions of the various buttons November 2007 and March 2008.
on the learning machine. Once a lesson title has been selected In addition to being marginalized, monitoring and evaluation
the methodology becomes interactive and learner-centered. of ICT for education programs has rarely embraced creative
Having listened to brief teaching points the learners are tested processes of assessment [26]. Recognizing this, a systems-
on their understanding of the information supplied through the based approach was utilized in designing the methodology
posing of multiple-choice questions answered by pressing [27] [28] [29]. This involved adopting a mixed- and multi-
buttons on the device. At the end of the lesson learners are method approach which focused on process and aimed to
given opportunity to test what they have learnt through engage with multiple stakeholders so as to gain a plurality of
undertaking an overall quiz of ten questions where they are perspectives regarding program impact. It also ensured that
congratulated according to the score attained. This outcome- data gathered was not solely anecdotal but credible,
based approach provides incentive for the children to work dependable and confirmable [30]. Within this overall aim for
conscientiously and allows the teacher to monitor progress. increased creativity and rigor, the research remained subject to
The decision to include lessons in both Chichewa and the standard complexities and constraints of limited time,
English was taken on the basis that early learning content is budget, data and personnel [31] [32].
most effective when communicated in the vernacular language It light of the constraints it was decided by the MoEST that
[23] [24]. The dual language approach equips children for the five representative test schools would serve as a sample of the
challenging transition from Chichewa to English as the 50 participating in the program. The schools selected by the
official medium of school instruction which takes place in MoEST were Chin’gombe, Mwatibu, Mthentera, Mbinzi and
Standard 5. The most conducive learning environment for Dzenza. None of these schools had any form of ICT
using the devices is with groups of between four and six incorporated into the curriculum prior to the introduction of
students. In this context team work, group participation and the program. A total of 15 days were spent in these schools
the development of leadership skills can each be encouraged. due to a recognition of the benefits of prioritizing classroom-
based research [33]. A combination of qualitative and
quantitative methods were utilized and these are now outlined
IV. METHODOLOGICAL APPROACH alongside the rationale for each.
A total of 15 lessons in which the learning machines were
In regard to the overall context of technology within
being used were observed throughout the process [34].
education, Kozma [25;21] concludes after assessing a wide
Teachers were asked to conduct the lesson as they would
variety of studies that there is ‘no consistent relationship
normally, without altering classroom arrangements. Observing
between the mere availability or use of ICT and student
usage of the technology in a normal environment provided an
learning’. Beyond such generic assertions, there are
ideal foundation for the subsequent methods, giving
significant knowledge gaps remaining regarding what kind of
opportunity to discuss successes and suggest ways to work
initiative works and what does not. As InfoDev [11;5]
around challenges with the children, teachers and head-
summarize, ‘despite thousands of impact studies, the impact of
teachers.
ICT use on student achievement remains difficult to measure
Some 15 group interviews were held with children in the
and open to much reasonable debate’. In light of such
five schools, in order to hear what difference the program had
observations all partners in the Interactive Learning Program
made to their lives. The children were asked what they liked
initiative recognized the necessity for a comprehensive
and disliked and improvements they would make for the
monitoring and evaluation exercise, the key objectives of
future. The interviews followed a semi-structured, guided
which were to;
approach [35] [36] facilitated by a teacher and conducted in
• assess the impact of the program on primary education
Chichewa. Participants were selected at random from
• assess the feasibility of the program in each test school
Standard 3 and 4 classes, were aged between 7 and 15, and
• identify program weaknesses for future refinement
had equal gender representation.
• develop both teacher and organizational capacity
Following these group interviews, more detailed
• provide feedback to partners regarding suitability of scale-
conversations were held with individual children in order to
up and sustainability
learn about particular experiences or perspectives that they
A common theme from ICT-enhanced education programs
had expressed. Such unstructured and story-based
within developing countries is the marginalized place of
conversation was useful in identifying unanticipated program
30
impacts [37]. In addition to this, a three stage individual

interview was conducted with the head-teacher of each school
and also with four related officials from the MoEST. These V. ANALYSIS
interviews provided opportunity to obtain input regarding the The monitoring and evaluation exercise was focused on
program feasibility, impact and future direction. assessing three interconnected spheres which cut across the
Conducting focus groups was considered to be the most five previously identified research objectives. These were
appropriate method to gain detailed feedback from teachers impact upon students, impact upon teachers and effectiveness
regarding their view of the program and the difference it had of the technology. Analysis of each sphere is based on the data
made to their lives, including positive and negative impacts gathered using the range of methodological approaches
and potential improvements. A total of 15 focus groups were previously outlined. The analysis is illustrated throughout with
conducted in this safe environment [38] which facilitated quotations from children, teachers, head teachers and MoEST
robust dialogue [39] but ensured no one felt pressurized to officials.
share personal experiences [40].
Building on these, the learning octagon was a research tool
developed specifically for the program in order to stimulate A. Impact upon Students
more detailed discussion between teachers in the focus In assessing the overall impact upon students the study
groups. The octagon providing a pictorial representation to considered the specific impact on four main areas of student
assess the strengths and weaknesses of eight different attendance, enthusiasm, attainment in curriculum subjects and
dimensions to the program by drawing on a combination of attainment in life skills subjects.
the Octagon tool [41] and Most Significant Change approach A significant and universally agreed impact of the program
[42] [43]. This enabled the teachers to collaborate on creating was the increase in school attendance. Teachers reported a
a visual representation of the impacts of the initiative. large increase in class sizes with fewer children absent than
In addition to these methods, and in order to enable on- prior to the intervention. One student from Standard 4
going monitoring on a weekly basis, one teacher from each of explained why this was the case:
the five schools was selected to complete a diary documenting Before the gadgets more pupils were absenting themselves
their experiences of using the learning machines in their from classes but now we encourage our fellow pupils to come
lessons. The diaries were semi-structured and designed to to school and tell them, today if you absent yourself, you will
provide a continuous record of particular strengths, miss using the gadgets. I used to absent myself 50% of the
weaknesses and challenges encountered. time before the gadgets came, now I come to school everyday.
Baseline tests were conducted at stage one and three of the This is a significant achievement within a national context of
research. The objective of the test was to provide a high drop-out rates and absenteeism. For any educational
quantitative assessment of the impact of the program on the intervention to succeed it is necessary for learners to be
attainment of the children regarding both curriculum and life- attending school on a regular basis and the motivating
skills. In each of the five test schools there were 12 children influence of appropriate technology is clearly a contributing
randomly selected from Standard 3 and 4, providing a total of factor in achieving this desired end. Despite strong anecdotal
60 children, 30 boys and 30 girls. Five additional schools evidence, assessing the statistical significance of the change in
were selected to act as a control group with an additional 60 attendance was hampered by a lack of daily attendance
children tested. records maintained in each school, meaning it was impossible
Alongside focusing on the test schools, evaluation to track exact attendance patterns prior to the current term.
questionnaires and equipment feedback forms were distributed Children in the group interviews reported that they were
to all 50 participating schools in order to gain a broader enthusiastic about coming to school now that they were using
understanding of program impact. The questionnaires gave the learning machines. Several students reported having
opportunity for feedback regarding patterns of usage and shared the experience with family and community members
challenges encountered in implementation. The equipment who actively encouraged them not to miss out on the
feedback forms enabled the documentation of technical opportunity to learn with the new technology. As one student
problems encountered throughout the program. from Standard 3 explained:
Each of the methods outlined above enabled a detailed They [parents] said to me that you have to work hard and
perspective to be developed regarding the strengths and make sure you do not run away from lessons when you are
weaknesses of the program. The monitoring and evaluation using the computer. I used to run away from class but now I
methodology employed was participatory throughout [44] and have changed my behavior … because I am attracted to the
engaged with children, teachers, headmasters, community gadgets and if I miss the chance to use it, it will never come
leaders, government officials and civil society representatives. again.
Each school was visited at least three times throughout the six Other children emphasized what they had learnt through the
month research period and this allowed for a progressive lessons and how this related to what they hoped to do in the
approach and the development of good relationships with a future. One student from Standard 5 reported the impact of a
number of stakeholders. particular lesson regarding counting:
31
Now I know how to add and subtract – I could see the water. I did not know this before the gadgets.
pictures and I was able to subtract. This makes a foundation Children were also able to talk about culturally sensitive
for me to be able to work in a bank – this is what I want to do subjects such as HIV/AIDS and explain what they had learnt
in the future. from the lessons. Teachers stressed the value of the interactive
The learning machines are perceived by the children as a lessons in a cultural context where such discussions are often
mixture between mobile phones, games devices and video considered taboo. The children requested that new content be
players. This makes them attractive to the children, who added to the devices so that they could learn about a greater
quickly become familiar with them and enjoy the process of variety of topics. The most commonly requested subjects were
being congratulated by the device for answering questions Agriculture, Mathematics, English, Religious Education,
correctly. Observed learners appeared to enjoy working in Science and Technology, Physical Education and Music.
groups and taking turns in pressing the buttons. Teachers also Similarly, the teachers requested that new content would
noted a similar impact and emphasized the improved listening remain linked with the curriculum reform in order to provide
skills of the students: them with support in this transition.
They are able to explain things now – and the gadgets really The impact of the initiative on the students is intrinsically
help them with listening skills – if they do not listen then they linked to the way in which the teachers are affected, and this
miss what has been said and they cannot answer the question is now considered.
– and it makes them be fully attentive.
Increased attentiveness in class and greater motivation to
B. Impact upon Teachers
attend school also had an effect on the attainment of the
children in both curriculum and life skills based lessons. Both In assessing the overall impact upon teachers the study
students and teachers reported that the use of audio and video, considered two factors, teacher enthusiasm and teacher
as well as the continuous assessment quizzes, increased workload.
retention and affected attainment. One teacher reflected on The majority of teachers were enthusiastic about the
tests completed the previous week: introduction of the program into their school and several
We had the mid-term tests last week and more pupils did commented that they felt honored to have been selected as a
well than before – when I asked them why they said it was school for the pilot. They were pleased that MoEST officials
because of the gadgets – ‘we just remember what we have were visiting their school and showing an interest in their
learnt on them’. The increase in attainment has gone up by work. They were keen to learn about technology and could see
about 30% on average from what they normally achieve. the positive effects on themselves and the learners. They also
Despite such assertions it was difficult to assess quantitative noted how the introduction of the learning machines had
impact on student attainment due to the short period of time helped them in altering their teaching style, adopting new and
since the beginning of the intervention. A comparison of innovative approaches:
scores between the baseline and second test demonstrated that We love it – before the gadgets it was just talk and write for
certain sectors of the curriculum had considerably more some subjects – but now the kids can see the pictures. Some of
correct answers after using the learning machines. However, the children can be sitting doing one thing while others are
the majority of students had only completed 15% of available doing the gadgets.
lessons at the time of the second test and the overall Young trainee teachers particularly enjoyed using the
improvement in curriculum attainment was limited to 1.5%. technology and often engaged in using it without
Similarly, improved attainment in life skills was difficult to apprehension. This confidence had a clear effect on the
quantify as test scores related to knowledge gained rather than manner in which the children approached the devices and
necessarily to behavioral and lifestyle change. The overall made use of them in the classroom. The enthusiasm of the
impact on life skills attainment in the baseline test was a 3% students also affected the teachers:
improvement. Again, the questions which were answered …because if the learners are enthusiastic then it makes us
correctly indicated an increased understanding of the content enthusiastic. If the gadgets help the students to learn then the
for those lessons which the teachers had chosen to use most teachers have to be happy that the students are doing well.
regularly. However, this enthusiasm was not felt universally. Mature
Although quantitative change proved difficult to assess, teachers often showed less interest in the program, were
many students interviewed were able to recall accurate fearful of using the technology and reluctant to let the children
information on the life skills lessons they had used and use it independently. In one of the test schools after six
expressed satisfaction to have learnt skills useful to their daily months of usage there were only three teachers still using the
lives. Learners explained how they had put into practice what learning machines. All the other teachers in the school had
had been taught in the lessons, indicating potential behavioral become uninterested in the program, considering the
change: additional workload to be too much of a burden.
[my favorite lesson was…] Preventing malaria – because it The teachers reported a variety of different experiences
affects many children in the village. I learnt that we must regarding the impact of the program on their workload. This
sleep under treated bed nets and must not play with stagnant was dependent upon the manner of implementation in each
32
school and the degree to which the learning machines were the momentum of the program and improving teacher
incorporated within the lesson schedule. The overall feeling motivation. The limited budget of the MoEST makes it a
from the teachers was that the introduction of the gadgets had constant challenge to deliver adequate training, especially in
increased the workload in certain areas and decreased it in remote areas. The teachers requested opportunity to ask
others: questions and provide feedback, requiring effective
There are two things, lesson preparation - for this it might communication channels between the stakeholders. As
take more effort - and lesson delivery - for this it might take highlighted by one official:
less time. It has made us do more work – we have to guide the If you come back frequently then the schools will know you
students. And there are not enough gadgets – so if we had are coming and they will keep using the gadgets but if they
enough gadgets then it would be ok and then it would not think no one is coming to visit them then they will forget to use
increase our workload. them. This is because when they are monitored they will feel
Since the intervention was a pilot project the majority of that somebody is appreciating their job – they are happy that
schools chose to use the gadgets after the school day had they can be involved and they see that somebody cares.
finished so as not to interfere with the daily timetable. This The feedback from teachers demonstrated that sustained
meant that the teachers were required to stay in school for an monitoring from MoEST officials constituted a significant
extra hour every day to facilitate the lessons. Most were capacity building process. The impact was especially
willing to do this but some expressed reluctance due to pre- pronounced in marginalized rural areas which were
existing after-school commitments. It was also commented in inaccessible and rarely visited. In light of this is it clear that a
two schools that the initiative resulted in more preparation continuing focus on teacher training which both develops
work for the teachers because they needed to become fully skills and instills value is intrinsic to the success of this and
familiar with the devices themselves before making use of any ICT for education initiative. The ability of the teachers to
them in lessons with the children. In the two schools where utilize technology with confidence and adapt it to the specific
the teachers had chosen to incorporate the gadgets into the needs of the classroom is a central determining factor.
school day then a decrease in workload was reported. In these Without this integration any initiative remains an appealing
cases the gadgets helped with effective facilitation, the add-on but does not have significant effect on the culture of
management of large classes and provided support in the classroom, pedagogy, or rationale for learning.
preparing teaching and learning materials for the lessons.
In light of this, the program would benefit from the
C. Effectiveness of the Technology
development and distribution of a pedagogical guide for
teachers regarding the effective integration of the learning The success of the initiative, whilst not determined by the
machines into the learning routine of the children. Teachers technology, was dependent upon it operating effectively
should also be encouraged to utilize the full range of lessons within each school. Usability, durability and the reliability of
that are available on the devices. This would complement the the devices and charging solution were the three areas
development of a schedule for teachers to use the machines assessed in the study.
during school time, giving clear explanation how it can most The lesson observations demonstrated that the majority of
appropriately support the curriculum. children were able to operate the learning machines
The increased ability of the children to memorize what had independently, understanding the purpose of the buttons and
been taught was also recognized to make the job easier by one completing quizzes without supervision. The process of lesson
teacher from Standard 4: selection occasionally required guidance from teachers but
Gadgets have simplified the work for the teachers – when this was rarely necessary because the most capable children
the child sees something he remembers – when he just hears were seen to take a lead in the learning groups, demonstrating
he forgets. When they see it, it stays in their mind. considerable confidence with the devices. However, a
Others noted that the impact of the learning machines in significant factor limiting usability was the low number of
encouraging so many children to return to school had caused devices per classroom and subsequent high number of
class sizes to expand significantly, resulting in more work students per group, on occasion as many as ten learners. This
attempting to manage the children. A common concluding was not conducive to learning as the machine is designed for
comment from teachers was that the workload would decrease small groups, with small screen and quiet audio output. In
significantly if there was a lower pupil to gadget ratio and if such situations, dominant children were observed bringing the
the lessons on the gadgets were fully compatible with the machine to their ear in order to hear the lesson, meaning other
syllabus and incorporated into the curriculum. The program children were unable to engage and became easily distracted.
would therefore benefit from an increased allocation of 20 This was confirmed by a participating official from the
learning machines per school. This would facilitate a more MoEST who noted:
concentrated impact and allow a greater number of children in Bringing such technology is very important - it is only
smaller groups to benefit from the content on the devices. that at the moment the gadgets are very few – there needs to
In light of the challenges presented, it is clear that regular be more. It is important that you have the right number of
training sessions for the teachers would assist in maintaining gadgets for the number of children because it is difficult for so
33
many children to share. for the development of user interaction. Market influences on
Due to the nature of such a pilot program it was anticipated the development of mobile technology also ensure sustained
that there would be significant problems regarding device improvements in hardware capabilities and the subsequent
durability. However, after six months of usage only 8% had emergence of new educational possibilities.
developed problems, the majority being due to software issues However, within all such initiatives, it is necessary that the
easily resolved through reformatting. Two of the devices application of technology is conceptualized as a tool in
stopped working due to the on/off switch breaking, these were facilitating the overall aim of catalyzing a more fully effective
repaired in situ and left fully functioning. A major concern in approach to education. In reaching this goal many of the
this respect was that teachers often assumed they had broken challenges documented remain the same regardless of the
the devices when all that was actually required was for them presence of technology.
to be charged fully. It is not overly surprising that the introduction of ICT into
The solar panels operated effectively in each of the five test primary schools in Malawi caused a dramatic increase in
schools throughout the program with all of them maintaining attendance figures. The significant question for ongoing
adequate power to charge the devices as often as necessary. research remains one of long term impact on approach to
Several program schools outside the five included in the test schooling, once initial enthusiasm surrounding the program
did report problems including two faulty batteries, one loose has subsided. Whilst a significant achievement, increasing
connection and one stolen panel. In addition to these, one school attendance figures is not the most significant long-term
school adopted the practice of disconnecting the battery from challenge facing education in Africa [5]. Instead, it should
the panel to remove it for safe keeping during the night and constitute an initial building block which can be capitalized on
this prevented the battery from charging fully. through adopting an integrated approach to developing the
capacity of the education system. One dimension to this will
be the effective utilization of curriculum enhancing
VI. CONCLUSIONS technology, such as the Interactive Learning Program. For full
The analysis of impact on students, teachers and benefits to be realized, such programs require transition from
technological effectiveness has demonstrated the educational their position of peripheral curiosity to one of being integrated
potential of the Interactive Learning Program. The as a sustainable teaching tool. Future research is required to
introduction of the devices had significant positive impact assess the degree to which this is occurring in Malawi.
upon school attendance and levels of enthusiasm. Also A variety of different solutions are required to address the
noteworthy was the increased value placed on teachers and the challenge of providing equitable access to good quality
role of education within the community. However, alongside education in the developing world. The increasing availability
this, considerable challenges were encountered including the and decreasing cost of portable devices ensures that they are
negative impact on teacher workload, lack of classroom likely to play a prominent role in the future, however this
integration, significant infrastructural constraints and the should not lead to such technologies being presented as an
necessity for further teacher training. The future success, educational panacea. For programs to be of maximum
sustainability and potential for project scalability is dependent educational benefit in Malawi, as in many such initiatives, the
upon engaging with each of these identified challenges. The critical issues for consideration and action remain pedagogy,
observations made in the analysis speak directly to the classroom integration and teacher training.
program efficacy but are also applicable lessons for ICT for
education initiatives more widely. The introduction of
technology into education systems in the developing world is REFERENCES
a complex procedure with considerable potential for failure [1] MoEST, Link for Education Governance. Education Sector Performance
Analysis. Lilongwe, Malawi. 2007.
and therefore it is vital that rigorous monitoring and [2] World Bank, Data and statistics: Malawi. 2006.
evaluation is incorporated throughout all such initiatives. www.web.worldbank.org Accessed 21/08/08
The price of portable educational technology is constantly [3] MoEST, Primary Curriculum and Assessment Reform (PCAR).
Implementation Plan 2007-2010, Lilongwe, Malawi. 2008.
decreasing, making initiatives such as the Malawi Interactive
[4] UNESCO, Institute for Statistics, country data 2008.
Learning Program increasingly cost effective. The September www.uis.unesco.org Accessed 21/08/08
2008 price of each learning machine was circa US$55 and as [5] EFA, Global Monitoring Report: The quality imperative. 2005.
each device can be used by several children the per child cost www.portal.unesco.org/education Accessed 21/08/08
[6] D. Pye and J. Stephenson, Using ICT to Increase the Effectiveness of
is considerably lower than other widely publicized solutions. Community-based, Non-formal Education for Rural People in Sub-
Significant future potential also lies in the opportunity for Saharan Africa: the CERP Project Final Report (DFID Educational
integrating the educational content with new generation Paper 50). London: Department for International Development 2003.
[7] V. Tinio, ICT in Education. E-Primers for UNDP, New York 2003.
mobile phones and other handheld devices becoming [8] J. Leach, DEEP IMPACT: an investigation of the use of information and
progressively more available across the region. Linked to this, communication technologies for teacher education in the global south.
increased ability to send and receive data in standard formats Researching the Issues, 58. DFID 2005.
[9] D. Wagner, Monitoring and evaluation of ICT for Education: An
(for example Macromedia Flash, xml, mp3 or mp4) will allow Introduction. Chapter 1 in: D. Wagner et. al. 2005. Monitoring and
34
Evaluation of ICT in Education Projects – A Handbook for Developing [35] J. Burgess, The Art of Interviewing. In: A. Roger and H. A. Viles, (eds)
Countries, InfoDev 1-19, 2005. The Student Companion to Geography. Blackwell: Oxford, 242-249,
[10] D. Keats, The genesis and emergence of Education 3.0 in higher 2003.
education: the potential for Africa. First Monday 12, 3, 2007. [36] M. Q. Patton, Qualitative Research and Evaluation Methods. California:
[11] InfoDev, Knowledge Maps: ICTs in Education. World Bank 2005. Sage, 2002.
[12] C. Fadel and C. Lemke, Technology in Schools: What the research says. [37] R. Chambers, Participatory Workshops – a sourcebook of 21 sets of
Metiri Group, Commissioned by Cisco Systems 2006. ideas and activities. United Kingdom: Earthscan, 2002.
[13] L. Cuban, Oversold and Underused – Computers in the Classroom. [38] G. Valentine, Being seen and heard? The ethical complexities of
Harvard: USA 2001. working with children and young people at home and at school. Ethics,
[14] T. James and J. Miller, Developing a Monitoring and Evaluation Plan for Place and Environment. 2, 141-155, 1999.
ICT in Education. Chapter 4 in: D. Wagner et. al. 2005. Monitoring and [39] D. Stewart and P. Shamdasani, Focus Groups: Theory and Practice.
Evaluation of ICT in Education Projects – A Handbook for Developing Newbury Park: Sage, 1990.
Countries, InfoDev. 57-77 2005. [40] P, Bhattacharjee, Stepping Stones – A participatory tool to integrate
[15] D. Wagner, B. Day, J. S. Sun, Recommendations for a pro-poor ICT4D gender into HIV/AIDS work. Development in Practice, 10, 5, 691-694,
non-formal education policy. Final Report for, Imfundo: Partnership for 2000.
IT in Education, DFID, 2004. [41] Sida. The Octagon. A tool for the assessment of strengths and
[16] M. Sharples, Disruptive Devices: personal technologies and education. weaknesses in NGOs, 2002.
Educational Technology Research Paper Series 11, The University of [42] P. Sigsgaard, Monitoring without indicators: an ongoing testing of the
Birmingham, 2000. MSC approach. Evaluation Journal of Australasia. 2, 1, 2002.
[17] H. U. Hoppe, R. Joiner, M. Milrad and M. Sharples, Wireless and [43] J. Dart & R. J. Davies, A dialogical story-based evaluation tool: the most
Mobile Technologies in Education. Journal of Computer Assisted significant change technique, American Journal of Evaluation. 23, 2,
Learning, Guest editorial, 19, 3, 255-259, 2003. 137-155, 2003.
[18] F. Lehner and H. Nosekabel, The Role of Mobile Devices In E-Learning [44] M. Huberman, The Many Modes of Participatory Evaluation. Chapter 7
-First Experiences With A Wireless E-Learning Environment. WMTE in J.B. Cousins and L. M. Earl (eds) 1995. Participatory Evaluation in
103-106, 2002. Education. Studies in Evaluation use and organisational learning.
[19] J. Leach, T. Power, R. Thomas, X. Fadani and A. Mbebe, 4D London: Routledge, 103-114, 1995.
Technologies: appropriating handheld computers to serve the needs of
teachers and learners in rural African settings. Centre for Research and
Development in Teacher Education, Open University: UK, 2003.
[20] C. Pontefract, Learning to Share, Insights Education 1 Missing the
Connection – using ICTs in Education, February 2003
[21] R. Dhanarajan, Learning Technologies, where is the challenge?
Education, Communication, Information 1, 1 Spring 2002
[22] MoEST, Primary Curriculum and Assessment Reform (PCAR).
Literature Review of Primary Education in Malawi, Lilongwe, Malawi,
2002.
[23] G. Bunyi, Rethinking the place of African indigenous languages in
African education. International Journal of Educational Development 19,
4-5, 337-350, 1999.
[24] EFA, Global monitoring report: Literacy for Life, 2006.
www.portal.unesco.org/education Accessed 21/08/08
[25] R. Kozma and D. Wagner, Core Indicators for Monitoring and
Evaluation Studies for ICT in Education. Chapter 3 in: D. Wagner et al.
2005. Monitoring and Evaluation of ICT in Education Projects – A
Handbook for Developing Countries, InfoDev. 35-57, 2005.
[26] J. Taylor and S. Soal, Measurement in Development Practice: From the
Mundane to the Transformational. Chapter 5 in L. Earle (ed) 2004.
Creativity and Constraint: Grassroots Monitoring and Evaluation and the
International Aid Arena. INTRAC Policy Series 18, 95-110, 2004.
[27] P. Morgan, Measuring the Development of Capacity: Is this Still a Good
Idea? Chapter 2 in L. Earle (ed) 2004. ‘Creativity and Constraint:
Grassroots Monitoring and Evaluation and the International Aid Arena’.
INTRAC Policy Series 18, 47-62, 2004.
[28] D. Watson, Monitoring and evaluation of capacity and capacity
development. ECDPM, Discussion paper 58B. April 2006
[29] D. Cabrera, A Theory of Systems Evaluation. Systems evaluation and
evaluation systems whitepaper series. New York: Ithaca, Cornell
University, 2006.
[30] J. Baxter and J. Eyles, Evaluating qualitative research in social
geography: establishing ‘rigour’ in interview analysis. Transactions of
the Institute of British Geographers, 22, 4, 505-525, 1997.
[31] M. Bamberger, J. Rugh and L. Mabry, Real World Evaluation. Working
Under Budget, Time, Data and Political Constraints. London: Sage,
2006.
[32] D. Gray, Doing Research in the Real World. California: Sage 2004.
[33] M. C. O’Sullivan, What is happening in the classroom? A common-
sense approach to improving the quality of primary education in
developing countries. Teacher Development, 9, 3, 301-314, 2005.
[34] J. W. Creswell, Educational Research: planning, conducting, and
evaluating quantitative and qualitative research. New Jersey: Pearson
Education, 2005.
35
Claim Mobile: Engaging Conflicting Stakeholder

Requirements in Healthcare in Uganda
Melissa R. Ho, Emmanuel K. Owusu, and Paul M. Aoki
Abstract—Claim Mobile is a platform designed to support a fraud and transformed supply-chain management for the E-
project that subsidizes healthcare by reimbursing health service choupal project [6]. While health information is critical to the
providers in Uganda for treatment of patients with sexually improvement of healthcare in developing regions, financing
transmitted infections. As with many development projects, the
Uganda Output-Based Aid (OBA) project involves a number of healthcare also remains a significant unsolved problem. Can
stakeholders: the service providers, the project implementers, we take lessons from e-Choupal and apply them in the
the financiers, and the Ugandan government. Design of an healthcare sector? The design of usable, reliable, and fraud-
appropriate solution requires meeting the various and conflicting resistant tools for management of these aid flows is an area
requirements of all of these stakeholders. In this paper we detail with potential for very significant impact.
the rapid design and testing of a pilot implementation of a
mobile and web-based system for processing claims forms, based
on two prior field visits to Uganda. Based on a comparative However, in the case of healthcare, the financial models are
device study, semi-structured interviews, health clinic surveys,
and a brief deployment, we affirm the selection of the mobile very different from commercial markets – financing of health-
phone as a platform from the health clinic perspective, and care typically comes through transnational aid agencies like
further suggest that effective design for development requires the World Bank and International Monetary Fund (IMF), and is
more than addressing requirements of the the “users” of the implemented by non-governmental organizations (NGOs) and
mobile phones but also all the other stakeholders involved, who the local government. Since the NGOs are typically experts in
may have conflicting requirements.
Index Terms—mobile phone, ICTD, health, participatory de-
health, not technology, data processing is often outsourced to
sign, Africa, HCI third-party information technology (IT) vendors. Relationships
between the vendors, the NGOs, the local governments, and
the transnational aid agencies are not always smooth - and
I. I NTRODUCTION
limitations in communications infrastructure means that the
Mobile phones are frequently touted as being the appro- information flows between them are scattered at best.
priate and sustainable platform for rural healthcare in Africa.
They are relatively cheap, durable, consume less power than
In this paper we suggest that the “closed loop model”
laptops and desktops, and incorporate a battery that makes
generally used by researchers in deployments of mobile health
them more amenable to use in places with intermittent or no
applications does not map onto the financial and political
power. Commonly proposed uses are for data collection [1],
realities of the mainstream of healthcare provision in Africa,
[2] and decision support for rural health workers [3], [4]. Some
and limits the ability of pilot programs to increase their scale
projects also use mobile devices as a platform for information
and impact. We describe an innovative, IT-based, NGO-run
dissemation as well as data gathering [5]. However, these are
healthcare access program in Uganda, and our experiences
all generally “closed loop” systems in which researchers are
designing and deploying Claim Mobile, a mobile-phone based
able to control all aspects of the system design and operation,
system intended to address inefficiencies and help the program
focusing their research primarily on the rural health workers
scale to additional districts. We argue that in addition to
that will be using the mobile phones.
addressing the needs of the primary users in the system, the
Other applications have even more potential for large-scale health workers, our design must consider the requirements,
impact. In the agricultural sector, we have observed how motivations and concerns of the other stakeholders: the IT
the introduction of transparent market prices and subsequent vendors, the NGOs, the government, and the aid agencies.
hiring of “middlemen” to purchase from farmers has reduced Our designs must consider the larger order ramifications of
how we may positively and negatively impact both the “users”
Manuscript received October 1, 2009. This work was supported by the
Blum Center for Developing Economies and the U.S. National Science who will be generating the data, and the entities that will be
Foundation Summer Undergraduate Program in Engineering Research at engaged in managing and using the information in the resulting
Berkeley (SUPERB) under Grant No. 0453604. database. Just as the e-Choupal project assimilated the mid-
Melissa R. Ho is with the School of Information at the University
of California, Berkeley, 94720, USA. (phone: +256 777 723 786; email: dlemen by hiring them as kiosk operators, we propose that we
mho@ischool.berkeley.edu). can design applications structured to accommodate conflicting
Emmanuel K. Owusu is with the Computer Engineering Department at stakeholder requirements, while also alleviating information
Iowa State University, 50011, USA. (email: kwame@iastate.edu).
Paul M. Aoki is with Intel Research, Berkeley, CA, 94704, USA (email: inequalities resulting from limitations in the system prior to
aoki@acm.org) the introduction of the information technology.
36
Mbarara, Uganda
• HIV prevalence: 10% of adult population (15-49
years)
• Syphilis prevalence: about 5-7% of adult population
• 1 in 4 households had at least one phone.
• 39% reported STI symptoms
• only 1/3 sought care
• 54% of respondents who sought any STI treatment
reported using private clinics.
Fig. 1. A HealthyLife voucher. The ‘M’ in the top left is a note written TABLE I
by the health service provider to indicate that the first client voucher on the S OME BACKGROUND STI STATISTICS ABOUT M BARARA , U GANDA [7].
left was given to a male client, and that therefore the partner vouchers on the
right should be given to a female client.
can take two weeks or more to move from the providers office
II. BACKGROUND to the management agency. The current data management
system requires all claims to be submitted on paper forms
Providing effective health care in poor countries is an to the management agency. At least another two to four
essential component to economic development and poverty weeks are spent reviewing each claim, cleaning data from
reduction. Unfortunately donors supporting this endeavor often improperly-completed forms, and verifying that the service
find that resources given are not matched by desired gains. took place among suspect claims. Two months or more can
The output-based aid (OBA) model of financing seeks to ad- go by before the provider is reimbursed for service provision.
dress this by paying healthcare providers directly for services In Uganda, private providers traditionally operate on a fee-
rendered instead of paying for the service provision up front. for-service model, receive prompt payment, and do not have
However, OBA program management is information intensive, a large operating margin. In many cases, payment is provided
necessitating much paperwork to track and reimburse payment prior to service. Delays in payment result in delays in pro-
claims. Smartphones (mobile phones with advanced features curement of replacement prescriptions and medical supplies,
such as the ability to run third-party software) have the often leading to a temporary hiatus in service. Encouraging
potential to alleviate this burden. In collaboration with a local provider involvement in the OBA program requires a great
NGO and their partnering IT vendor, we have proposed to deal of confidence on the part of the providers to participate. If
deploy a number of smartphones for use in an OBA project a system to shorten claims processing could be devised, more
based in Western Uganda, with dual goals of reducing claim providers could join the scheme and more patients could be
processing time and improving communication between the provided the life-saving STI treatment voucher subsidy.
health care providers and the OBA management agency. The remainder of this paper details the system we are cur-
The project is managed by the local branch of a multi- rently piloting, in which claims are submitted via Internet from
national NGO and a for-profit health insurance company, in a mobile phone directly to all the parties in the management
collaboration with the Ugandan Ministry of Health (MoH) agency. In addition to describing our user studies and how this
and Ministry of Finance (MoF). The project is primarily has informed the design of the system, we discuss the problem
funded by an aid agency based in Europe, with additional of negotiating conflicting stakeholder requirements. We find
funding for the expansion coming from a separate transna- that in projects with multiple stakeholders, the introduction of
tional funding agency. Together, they run a voucher program a system may disrupt balances of power, particularly around
called HealthyLife, which treats sexually transmitted infec- the flow of information and money. As a result, the design of
tions (STIs), reimbursing providers for the diagnosis and full this system, in order to secure positive support from all parties
course of treatment only after the patient is seen. This program involved, must carefully balance stakeholder incentives.
was implemented in response to the high burden of sexually
transmitted infections in Uganda, and began in July 2006 III. M ETHODS
in four districts of southwestern Uganda: Mbarara, Ibanda, The research described here involved an iterative process of
Kirihura and Isingiro (See Table I). field research and prototyping. The fieldwork and deployments
Patients buy treatment vouchers in pairs, one for the client have been done over the course of three visits to Uganda:
and a second one for the client’s sexual partner (See Figure 1). an initial two-week visit in Summer 2007 to establish a
Each voucher is good for one consultation (generally including relationship with the project, in which we also conducted a
a lab test to diagnose the STI) and three follow-up visits. survey of the clinics in the program; a followup visit for
During the consultation, the provider completes a claim form three weeks in November 2007; and a five-week pre-pilot
recording the client’s demographics, the examination and deployment in August-September 2008. During all three visits
laboratory results, a diagnosis and details of the course of we conducted semi-structured interviews with the various
treatment prescribed (See Figure 5). Completed claims forms stakeholders, and directly observed claims form entry and
are sent to the voucher management office in the city of processing. When given permission, we did audio and video
Mbarara, the main urban center of Western Uganda. Forms recording of interviews and user study activities. In all, we
37
have approximately 30 hours of audio, and have done detailed

interviews in seven of the 12 participating clinics (in addition
to the initial survey of all of the clinics), as well as intensive
observation in two clinics, a rural, high-claim-volume clinic
with very little exposure to computers, and an urban low-
claim-volume clinic with its own computers. The last visit
entailed a comparative user study as well as deployment of
the mobile phones in the latter two clinics.
A. Clinic Surveys
The clinic survey was conducted in conjunction with a
larger survey of available infrastructure at contracted clinics. Fig. 2. This diagram illustrates the flow of money and information between
selected stakeholders in the OBA project, both currently, and how it will be
We asked 14 questions, assessing familiarity with computers once the Claim Mobile system is fully deployed.
and mobile phones, but primarily gathering feedback from the
health clinics on the claims process (e.g., how long it takes
them to fill out the paper forms, and what their priorities be conducted in May 2009, with mobile phone-based claims
might be for improvement of the process). We also collected processing expanded to 8-10 additional clinics.
various documents from the management agencies regarding
the performance of each clinic, including all available financial IV. S TAKEHOLDERS
reports on processed claims, and in November, we returned to In this section we describe the funding, management and
seven of the clinics to do in-depth surveys and to follow-up service provider organizations to which we alluded in Sec-
on the survey findings. tion II. Fundamentally, all stakeholders want to improve STI
treatment and reduce the prevalence of STIs. Each stakeholder
B. Rapid Development and Pre-pilot Deployment also has a financial interest in the success of the overall project
Initial prototyping occurred in early 2008, and we returned - staying afloat for the health service providers, and staying
to Uganda in Summer 2008 to do a three-stage pilot deploy- within the aid agencies’ target budgets for the management
ment, first testing the functionality of our software, second agency partners. The discussion highlights the ways in which
reviewing the proposed claims process with the management the various stakeholders have competing as well as common
agencies, and finally taking the phones to the health clinics to interests. While we will detail several stakeholders in this
test the mobile phone interface in the field. During this time section, there are six key stakeholders: the aid agencies who
we also conducted another round of semi-structured interviews fund the OBA project, the financial management agency
to gather information on changes in the claims submission (FMA) which receives the funds from the government and
process (for example, claims processing had moved from disburses them, their program management office (PMO) in
Mbarara to the national capital, Kampala). We did iterative Mbarara which runs the program and interacts directly with
development based on feedback from the various stakeholders, the health service providers, the technical management agency
trying out features as they were suggested, and developing (TMA) that manages the claims processing, the health service
new tools as seemed merited by findings in our interviews. providers (HSPs), and us, the mobile platform developers
To gain a more in-depth understanding of health clinic life, (MPDs). Figure 2 illustrates some of the relationships between
we stayed overnight for three days in the rural health clinic, these entities which we will describe in detail in the remainder
thereby supplementing the the semi-structured interviews with of this section, based on qualitative fieldwork and document
direct observation of actual practice. analysis.
The primary purpose of this last field visit was to con-
duct a pre-pilot demonstration, using the mobile phones to A. Aid Agencies
submit actual claim data to the management agency, have it As the funder of the HealthyLife STI treatment program,
reviewed, and have the management agency provide feedback the involvement of the European aid agency is more than
to the health clinics via the mobile phones. We simulated apparent. Their role in the management of the program is
the proposed process, physically following the claims forms more supervisory – a consultant goes to Uganda at irregular
from the time the patient comes into the health clinic, through intervals to help with planning of the program, and they do
the preparation of the claims summary forms, physically some monitoring. They also have commissioned another non-
transporting the forms to the management agency where we profit, affiliated with a North American university, to conduct
observed the claims approval, and data entry into the existing an evaluation of the program. Ultimately, however, they control
database. We simultaneously had the service providers submit the flow of money to the financial management agency, which
the claims form via Claim Mobile, enabling the management then pays the IT vendor to handle the technical aspects of the
agency to provide feedback to the service providers through operation.
the system. The pre-pilot is still operational, with mobile In the past year, the European aid agency has worked with
phones remaining in the two clinics, and the full pilot will an additional transnational aid agency to fund the expansion
38
of the project into additional districts. While they may not for the PMO, and one for the TMA. However, the copy that
have a direct impact on the information processes in the remains in the PMO does not have the voucher number, a
project itself, the funders’ internal actions have direct impact critical piece of information, and with stacks of hundreds of
on the project as a whole. In one example, a delay in claims per month, the information is not in a format actually
payment to the European aid agency resulted in a delay in accessible to the program office until the TMA sends back
payment to the two management agencies. As a result the IT claims summaries. However, even this is stripped-down and
vendor ceased processing of claim forms until payments were only includes the value of the claims, without any patient or
received. However, not only did the voucher program grind diagnostic information from the claims.
to a halt, but reimbursement to the participating providers for This poses a problem for the PMO staff’s interaction
patients already seen was delayed as well; the management with the health clinics. They lack sufficient information to
agencies ended up with a backlog of claims forms to process, effectively counsel and train the clinics, and often feel like
exacerbating the length of time it takes to process claims and they do not know what is going on with the program because
further delaying payment for services. they do not have access to the claim data for the long claims
processing cycles.
B. Financial Management Agency (FMA)
In the version of Claim Mobile developed in early 2008,
The local NGO partner that acts as the financial manage- we intended to make the claims process more efficient by
ment agency (FMA) is the Uganda office of a multinational enabling the mobile phones to submit claims directly to the
non-profit sexual and reproductive health organization with a database (originally co-located in the PMO, now located in
goal of reducing unintended pregnancies and unwanted births Kampala). In the Summer 2008 interviews it became clear that
through family planning and other methods. Their role in having the claims data bypass the PMO staff would deny them
this project is to receive the funds from the aid agencies even more of the information they need in their interactions
via the Ugandan government, using these funds to pay the with the clinics. Based on this realization, we discussed the
health clinics and to pay for other program costs, including possibility of an intermediary application, a website in which
the database software development and management. The main the project coordinator in the PMO would be able to view
office in Kampala runs this program (as well as others) and claims as they are submitted, as well as any status updates.
manages several clinics throughout Uganda, one of which was Furthermore, Claim Mobile could facilitate another of the
a participating clinic in the HealthyLife program until Summer project coordinator’s key roles in the OBA ecosystem: as the
2008. In addition, they have a program management office primary interface between the health service providers, the
(PMO) in Mbarara which is directly in charge of coordinating project coordinator would also be able to send messages to the
the the OBA project. In the claims process, the FMA disburses service providers through Claim Mobile, either as individual
payments to each of the service providers, based on claim messages, or broadcast announcements.
reports from the technical management agency (TMA).
C. Program Management Office (PMO) D. Technical Management Agency (TMA)

The HealthyLife PMO in Mbarara has five full-time staff, The TMA is a for-profit health insurance agency based in
in addition to two people that help with cleaning and cooking, Kampala, providing conventional employer-based health insur-
and the FMA staff that come in and out of Kampala for related ance for the formal private sector as well as conducting non-
programs. There are two computers in the office, one in the profit health management for targeted low-income informal
project coordinator’s office, and another in the finance office, sector populations. That is, their work for the HealthyLife
shared by the Behavior Change Campaign (BCC) coordinators program is in addition to their private health insurance pro-
who go out into the field to run community radio advertising gram, and is part of a company effort to help deliver quality
programs and to distribute vouchers. Their Internet connection affordable healthcare to the poor. Their business is highly
was down when we arrived, but was repaired the same day technical, and they have a wholly owned software company
and largely functional for the remainder of our four weeks based in Chennai, India. As the IT vendor, the TMA’s respon-
there. They share a 56kbps dial-up connection over a local sibility in the program is to provide the claims forms, and
area network. the Voucher Management Unit System (VMUS), the database
It became clear through our interviews in this office that, implemented by their software company to cross-check the
while the PMO is the nominal clearinghouse for information claims and to generate reports. The TMA also prints (through
between the TMA and the health clinics and is primarily another agency) the glossy color vouchers that the patients
responsible for communication with the health clinics, they purchase in exchange for subsidized care.
actually have the least information of all of the stakeholders Although the data entry clerks and the database engineer
in the OBA program. At the point in which the database were initially located in the FMA’s PMO in Mbarara, they are
processing moved from Mbarara to Kampala, all of the claims actually employees of the TMA, and moved when the TMA
information moved there as well. They have been able to shifted claims operations to their offices in Kampala in March
change the claims process such that the health clinics submit 2008. They carefully enter each claim into the database, later
two copies of each claim to the management agencies, one updating its status with information from the clinical officer
39
Fig. 3. This is a sample summary sheet prepared for one health clinic, showing a partially paid claim (QC11=Wrong consultation fees), and another claim
quarantined for having the wrong voucher. While these summary reports can be informative, most clinics are not familiar with the quarantine codes, and they
often don’t reach the health clinic for several months after the original claim has been submitted, often too late for the clinic to rectify any errors indicated
on the report.
(a doctor) who “vets” the paper claims1 . They then produce software vendor. Again, while this situation is being resolved,
two reports: a summary report for all clinics, and an itemized the TMA asserts its control over the project by processing the
report (see Figure 3) for each clinic detailing the status of each claims, but refusing to pass on the summary reports to the
claim, as well as any quarantine codes (Table II) or rejection FMA. While the project continues running, and the service
reasons for any partially paid or rejected claims. providers continue to see patients, this introduces additional
In addition, when required, the TMA produces reports delays into the claims process, and frustrates the health clinics,
(based on the information in the database) for the FMA, the whose payments are delayed without explanation.
European aid agency, and the aid agency’s evaluating partners.
E. Service Providers: Health Clinics/Hospitals
Although these reports were not part of the original specified
mandates for the operation, they have proved necessary for Service Providers are selected on the basis of a number of
the program’s external evaluation, and there has been much factors (e.g., services offered, capacity, personnel, geographi-
friction over the work involved in the creation of reports. cal location). In one respect, they are the origin of the primary
The relationship between the TMA and the FMA in this information in the claims management process, producing
program is highly contentious. While initially they were equal the claims records, which are then used to determine reim-
partners in the program, both reporting directly to the Euro- bursement. At the same time, as is perhaps typical, they are
pean aid agency, changes in funding have led to a situation in information-poor, because they are not given tools to use this
which the TMA reports to and is paid by the FMA. On top of information effectively. At the point of claims submission, they
this, the funding for the expansion of the program has been are no longer agents in the process, and must wait passively
delayed a number of times, from October 2007 to April 2008, for both payments and any feedback reports produced from
and again to September 2008. While the TMA has received the information in their claims.
some payment, both the TMA and FMA have been operating Code Description
QC01 No indication of date of treatment
without pay (but with promise of pay) since April 2008, just QC02 No indication of time of treatment
QC03 Wrong visit type: Consultation or follow-up, etc
to keep the program running while the aid agencies work out QC04 Wrong demographic information: no age and name of client
QC05 Wrong/No syndrome, no diagnosis
the details of the new grant and the expanded program. This QC06 Wrong Clinical examination / not applicable to OBA
QC07 Wrong/Poor diagnosis
is part of the reality of dealing with aid-funded projects – QC08 Wrong investigation/poor lab reporting
QC09 Wrong drugs prescribed/invalid treatment
unexpected delays in funding are common, and projects are QC10 Over prescription: more than enough
QC11 Wrong consultation fees
subject to the vagaries of arbitrary rebudgeting. While the QC12 Wrong patient status: cured or not cured
QC13 Next date of visit: wrong or not filled in
FMA is often powerless to address the issue, in this case, QC14 Wrong voucher attachment/interchanged vouchers on followup visits
QC15 Unclear claim/uncharged claim and treatment contradicts other visits
the TMA often chooses to respond by cutting off program QC16 Partner treated on client form
QC17 Exceeded VMUS ceiling limit
access to the database, ceasing claims processing and refusing QC18 Treated syndromically and asymptomatically
QC19 Unclear/wornout/blank attached vouchers
requests for reports, until their problems have been resolved. QC20 Claim without patient thumbprint
QC21 No voucher attachment
Perhaps in response to these database shutdowns, but of- QC22 No doctor’s signature
QC23 Diagnosis contradicts clinical examination
ficially as part of the aid agency’s project policy and the QC24 Used drugs not on OBA list
QC25 Undercharged/overcharged drugs, double lab charged
Ugandan government’s policy on software developed for QC26 Patient free/normal from STI or cured not allowed for next visit
QC27 Diagnosis not catered for by project
government-funded projects, there is an expectation that the QC28 Follow-up contradicts previous visits
TMA’s VMUS software should be turned over to the project. TABLE II

However, since the TMA outsources development of this soft- C ODES USED BY THE MANAGEMENT AGENCY TO INDICATE REASONS FOR
PARTIAL PAYMENTS .
ware to its partner company in India, this IT vendor considers
its software to be part of its key assets, and sees its role in
the project as a software licensor and service provider, not a
Providers are expected to follow a rigorous course of
1 Theclinical officer is employed by the FMA, and was terminated in March diagnosis and treatment — they must select a lab test based
2008 due to temporary lack of funds. on symptoms presented, and prescribe particular medications
40
many claims are approved with incomplete data (e.g., missing

demographic information for the patient). In any event, the
claim form and its contents are the object of much dispute
between the service providers and the management agency.
Often, the service providers (especially more distant providers)
feel disempowered to address this dispute and choose to accept
the given variance in payment as the cost of participation in
the program.
The health clinic survey we conducted in July 2007 explains
some of their extreme disengagement from the OBA program.
Fig. 4. “All above denied b’cos rest of P[atien]t mgt n[ot]. ethical”: this Out of the twelve clinics surveyed, three said they had not
is a sample medical advisor review of a claims summary, occurring often gotten feedback from the OBA program at the time of the
well after the original mistake has been made several times, before it could survey (July 2007), and 4 said they did not know how many
be caught and corrective measures could be made, as noted in the first line:
“Cipro pricing b4 C[ontinuing] M[edical] E[ducation]” claims had been rejected. While they were receiving payments
on a regular basis (albeit late - 9/12 clinics defined “timely
processing” as less than 15 days, half of the current processing
on the basis of the results of the test. Any deviations from this time of 30 days), there was no mapping between the claims
treatment are penalized; the service providers are not paid for they submitted and the payments they received.
medications given that are not prescribed by the program. This I don’t know. I don’t know how we are performing. I
is reasonable by public health and insurance standards, and don’t know how we are faring. And of course my staffs
necessary for the cost-effectiveness of the program. However, are also complaining. They are overworked, they dont get
despite educational illustrative posters, training sessions, and any benefit from the project, and of course it takes a lot
on-site continuing education provided by the PMO’s clinical of time. They need to be motivated as individuals. All that
officer and project coordinator, the learning curve on the exact will depend on – are we making any profits?
protocol to be used is high, and the subsequent errors are This has deeper ramifications than inability to follow up on
costly for the service provider. Figure 4 illustrates a particular quarantined, partially paid, and rejected claims. This commu-
problem in which costly errors are caught well after their first nications gap between the health clinics and the OBA program
occurrence, often after the provider’s staff has have made the leads to continued errors in adherence to treatment protocol, a
same mistake for a month. feeling of lack of control over health clinic finances, and dis-
A few months into the program, there had been so many couragement on behalf of the participating service providers.
quarantined claims (claims that had been held for review At the time of the second field visit in November 2007, we
due to discrepancies from the treatment protocol) that the were able to follow up with the PMO and the health clinics,
management agencies and the service providers were required and noted that this situation had improved. The then-informal
to do a financial settlement, in which the service providers practice of passing on copies of the itemized clinics reports to
were paid some percentage of the value of the disputed claims. the service providers was formalized, and clinics are receiving
Subsequently, the approval process was modified such that more feedback on their claims. However, there are physical
deviations from the protocol were partially paid (e.g., minus limits to a paper and in-person based communication system,
the cost of the incorrectly prescribed drug), and could be and it remains to be seen how this practice will scale as more
disputed in later reviews with the PMO’s clinical officer. Here clinics join the program.
are some figures on the value of the disputed claims for one It should also be noted that the service providers differ
of the rural service providers for a randomly selected month greatly, in number of clients, setting, and availability of
of the program: resources. While some clinics have computers and use them
regularly, in one clinic, our smartphones were greeted with
Claims submitted: 294
enthusiasm because they were the “first computers we have
Approved: 259 (88%), $1379
seen.”
Approved, but adjusted: 27 (9%), $149
Quarantined: none F. Patients
Rejected: 8 (3%), $51 Patients are the real target beneficiaries of the HealthyLife
Total Requested: $1642 program. They purchase vouchers from distributors (at a
Total Paid: $1526 (difference: $114) heavily-subsidized price) and then go to the service provider
for diagnosis and treatment. After an initial consultation with
Thus the claims submission process, while nominally a the health worker, they are directed to the lab technician,
simple information flow between cooperating organizations, who performs the requisite test and sends them back to the
becomes a site of financial contention. Claims are disapproved health worker with a slip of paper indicating the lab result.
for a number of reasons (see Table II), which could be disputed The health worker then fills out the remainder of the form,
but in practice are not. However, due to the change in policy, writes a prescription, affixes the appropriate stub from the
41
voucher, and has the patient sign and fingerprint the form, at and a revised version of the claim form. In addition, all of
which point their participation in the claims form process is the codes and tables in the web application database also
complete, until they return for a follow-up consultation. For the include mappings to their equivalents in the TMA database,
follow-up, the service provider checks recovery progress and so the information can easily be transferred between the two
prescribes additional medication if necessary. In some clinics, databases.
patients are given a copy of the claim form, which they are
directed to keep and bring back for the follow-up. However, A. Claim Mobile
most clinics do not depend on the patient copy of the claim The web application is designed primarily with three func-
form, and just go back through their time-ordered record book, tions in mind: claims submission, feedback/communications
finding the prior consultation manually. Sometimes patients process, and in-clinic claims management.
either accidentally swap vouchers with those of their sexual Claims Submission: This is the bulk of where the service
partners, fraudulently give their own voucher or the partner’s providers will spend their time. In this case, we adopt common
voucher to someone else, or simply choose to go to a different design strategies such as (1) using pre-filled checkboxes to
clinic for follow-up. Claims are quarantined or rejected if any reduce the amount of required text entry, (2) limiting answers
of these potential errors are detected, but not until the claim to valid options to reduce coding errors (see QC01,02, 04-09,
has been processed by the TMA, and the fraudulent patient has 27 in Table II), (3) downloading logistical data such as drug
already been treated. Since the original voucher is attached to prices into the application to eliminate pricing errors, and (4)
the submitted claim, the clinics do not always have a way of calculating dependent values such as expected claim amount
verifying these external aspects of voucher validity. Although to eliminate arithmetic errors and save time. However, we
their direct involvement in the claims process is minimal, it is must counter-balance potential fraud by also introducing cross-
their identity that is often contested in the vetting process. checks that are not clarified explicitly. That is, to encourage
G. Mobile Platform Developer (researchers) accurate clinical reporting (as opposed to clinical reporting
that has been “fiddled” to make electronic claim submission
As ICTD researchers, we are of course also stakeholders
more convenient or favorable), providers are allowed to submit
the claims management process – initially as outside observers,
inconsistent claims but are warned that they should clarify any
later as designers interested in using technology to measurably
discrepancies from normal OBA treatment protocol.
improve the process, and finally as researchers interested in
watching the mechanisms by which the process changes over Closing the Feedback Loop: Based on our primary finding
the course of the project. From an outside perspective, our role from the clinic surveys and follow-up interviews, we have
is most allied with TMA, the technical partner in the project; also included the ability for the clinics to send queries to the
however, since the aid agencies and FMA are interested in management agencies about particular claims and to receive
replicating the mobile device system in other OBA projects, live updates on the a claim’s status (e.g., whether it has been
there is a vested interest in the new technology from other approved, the amount for which a claim has been approved,
stakeholders as well. and explanations why the full amount may not have been
approved – see Figure 6). Any changes to a claim’s status
V. D ESIGN are included in this annotation audit trail, and anyone with
Claim Mobile is a two-part system, including a web-based access to the claim can respond to and receive queries.
PHP/MySQL application and a Java-based mobile application
running on Palm (GarnetOS) phones. For the pilot program,
the web-based application has a single level for all users, but
the final implementation will be tiered, having appropriate
access levels for service providers, management agency users,
medical advisors, etc. Both the web and the phone applications
require user login to protect patient data.
The two applications are paired, designed such that the
phone-based application uploads claims to the web site, and
downloads configuration information (drug lists, status feed-
back, claim form backups) from the web site. Eventually,
the web application will also connect to the TMA database, Fig. 6. On this screen the service provider can view the current status of their
sharing the cross-checked and validated claims form data claim, as well as any annotations or feedback from the management agency
made in response to their queries.
directly so the TMA’s staff do not have to do redundant data
entry.
To facilitate end-user training, both of the applications In-Clinic Claims Management: In the phone application,
are based on the original claim form and largely retain the the service providers can also link between consultations and
same structure, titles, and information. Figure 5 illustrates follow-up visits, as well as between client and partner visits,
some of the mappings between the phone-based user interface so they can easily check to see if the valid voucher is being
42
Fig. 5. From paper form to mobile phone: a mapping of the mobile phone interface equivalents for each section of the claim form.
Fig. 7. This is the list of claims currently entered or downloaded to the

mobile phone. The first number is the claim form number, followed by the
patient name, and then a number indicating the current status of the claim.
(1-unsubmitted, 2/3-under review, 4-preliminary approval, 5-quarantined, 6-
approved, 7-rejected)
used by a returning patient, and that treatment of a partner

or during a follow-up matches the medical history. For new
phone installations, or if the claims data is lost, the mobile
application will automatically download all prior claims data Fig. 8. Claim Mobile Web: the web-based view of the submitted claim forms
is also based largely on the original claim form, split into several sections,
from the web application. Future versions of the application and ending with annotations for the claim, and a form for adding additional
will also include financial summaries, outpatient statistics, and annotations.
other reports that may be useful to the clinics.
B. Claim Mobile Web given to the mobile application is used to authenticate with
The web application, having been commissioned in the the application when submitting claims and annotation data.
middle of the fieldwork in response to program office findings There are three primary views in the web application. The
(see Section IV.C) has two main functions: receiving claims claims list can be filtered by service provider and is modeled
and displaying them for review. on the claim summary report (Figure 3). From each claim in
Much of the claims receipt is invisible to the web appli- the claim list, the user can either (1) click on the claim number
cation user, and written as a backend for the mobile phone to access the individual claim (and annotation/status update
application. The login user for the mobile application and functions, see Figure 8) or (2) click on a patient name to view
the web application is the same – and the login information all consultations and follow-ups for the both the client and
43
partner associated with that particular voucher. This allows the inconsistency is in lab reporting – some lab tests require a
viewer to correlate treatments, lab tests, and diagnoses across value to indicate the result, and where not included, the data
visits. entry clerk just changes the lab test in the database to one
which does not require a result. This is an error, which never
C. The Modified Claims Process gets communicated back to the service providers because only
In the modified claims process, the service providers con- errors which accompany a payment change are reported in the
tinue to complete and submit the claim forms. However, in claim summary sheet.
order to receive faster payment, as well as the immediate cross- The 18 claims from the urban clinic spanned dates from
checked feedback from the phones on claims completion, they February 16, 2008 through August 25, 2008. During the
also enter the data on the mobile phones, submitting each process of simultaneously entering some of the claims into
claim to the web application as it is completed. Prior to claims Claim Mobile with the service provider, we were able to
submission the status of the claim is “Unsubmitted (1),” after identify some problems: missing personal information, missing
which it can progress through a number of stages. The service next visit date, and wrong consultation fees. However, not
provider can verify that a claim has been successfully received all claims were entered into and reviewed via Claim Mobile,
if the claim status has been updated to “Under Medical Review and, as can be seen from Figure 3, three paper claims were
(2)” or “Under Administrative Review (3)” for medical advisor submitted with wrong consultation fees, an error that would
review or database cross-checking (validation of voucher) not have occurred with an electronic submission. In addition, a
respectively. If a digital claim has been verified, the TMA will fourth claim was submitted with the wrong voucher. We were
set its status to “Preliminary Approval (4)” until the paper able to catch this while entering the claim into the mobile
form with the voucher has been received. Once preliminary phone, noting that the voucher number did not match the
approval has been received, the clinic can be paid. If no consultation type, but too late to change the submission and
voucher is received, or the wrong voucher is attached to the retrieve the correct voucher. As a result, the claim has been
form, then the preliminary payment is subtracted from the next quarantined until the correct voucher is given to the program
month’s payment until the error is resolved. In the meantime, office in Mbarara.
the service provider can view status updates as they are made
to the web application and sent to the mobile phones, and can
send annotations on each claim, which then appear in the web
application and in the status update window (Figure 6) when
they are received.
VI. P RE -P ILOT R ESULTS
Having detailed the claims submission process, our findings
with respect to the various stakeholders in the OBA project,
and the design of the Claim Mobile system, we now discuss Fig. 9. Rural clinic staff entering data from claim forms into two of the
phones.
some specific results from our user studies.
The pre-pilot demonstration was designed to last one claims
cycle, following one week’s worth of claims (submitted in With regard to the digitally-submitted claims, we spent
parallel through Claim Mobile and on paper) for two clinics about a day training the staff in the rural clinic on how to
through the claims submission process. A total of 35 claims submit the claims, and returned later to spend another day in
were submitted to the web application, including the full training. They were very enthusiastic, and although only one
complement of 18 claims from the urban clinic, 12 out of person was actually responsible for submitting claims, they
the 86 paper claims from the rural clinic, and 5 additional were all training each other (Figure 9). However after the the
claims from the urban clinic following the pre-pilot study. researchers’ departure they have still not submitted any claims.
The claims from the rural clinic spanned August 9, 2008 It is unclear whether this is from technical difficulties or lack
to August 27, 2008. We observed three patient consultations of time to enter the claims into the phone. The urban clinic
during our two visits to this clinic, as well as the preparation has continued to submit claims, with five new claims arriving
of the summary sheets for the 86 claims, taking careful note in the two weeks since the pre-pilot study.
of what the service provider verified on each form. Notably, An interesting outcome from our observation of the claims
although “syndrome” is a required field (see QC05 in Table II), review process is that there are many errors that are made
it was left blank in almost all of the claim forms. In one that affect the quality of the data, but are never communi-
case, the drug was entered correctly, but with the wrong cated to the service provider, in part because they have no
reimbursement value, and in another case, a drug was entered, attached financial consequences. The annotation feature (see
but no reimbursement was claimed either in the subtotal or Section V.A) enables attachment of quarantine codes to any
the total. At the time of the claim approval process, they claims that were in error without affecting the payment of the
were not reimbursed for the drug, because it had not been claim, providing feedback to the service providers on how to
claimed in the amount, although it had been listed. Another better complete the claims in the future.
44
One concern that emerged from this proof-of-concept study, B. Information Poverty
however, was with the reliability and the speed of Internet In addition to trying to address delays, we have also tried
access in the TMA office where the claims processing was to address information asymmetry and information poverty
occurring. Accessing individual claims took a long time, and within the system, identifying where stakeholders are disad-
the online database was completely inaccessible when the vantaged by lack of information, or lack of tools with which
Internet connection was down, which occurred infrequently to use the information.
but noticeably often. It may not be desirable for the claims This past year’s move of the data processing from Mbarara
submission process to introduce a dependency on Internet to Kampala especially has further exacerbated the gap between
connectivity where it is unreliable. the people that have the information and the people that
Unfortunately, the financial and claim review aspect of the can make use of it. While limited remedies have been made
pre-pilot was halted early due to administrative and political to rectify the situation, with a paper-based process, these
reasons, the result of which is that payments in the OBA remedies have been ineffectual, leaving the PMO without
program as a whole have been halted; so, while the technical access to necessary claims data, including voucher numbers.
feasibility of the system has been demonstrated, the logistical Through extensive stakeholder interviews, in particular with
details are still in process. We found that while the TMA’s the program office in Mbarara, we have identified the need
database entry staff were enthusiastic at the prospect of for a transparently accessible database, with the ability to
spending more time reviewing claims and less time doing just generate reports based on the submitted claims data. While
data entry, their participation in the pre-pilot was limited by control over access to the database is a key means by which the
two factors: the press of other claims that still needed to be TMA asserts itself in the OBA project, this practice is highly
processed, and pressure from the TMA to be secretive about disruptive to the OBA program, causing deep difficulties for
the data being processed until certain political issues had been the health clinics and the program office, rather than affecting
sorted out. the financiers or the FMA.
The initial design of Claim Mobile, reflected the paper-
VII. D ISCUSSION based process, and directly submitted claims from the mobile
A. Understanding Delays in the System phones in the health clinics to the TMA, bypassing the PMO
entirely. In response to our findings, we developed Claim
Delays can occur in a number of places in the claims pro- Mobile Web as a means of re-engaging the staff of the PMO
cess, not all of which can be accounted for by the introduction in the mobile claims process. The integration of the new web
of mobile phones. However, there are three key bottlenecks: application database is specifically meant facilitate resolution
1) the delay in the health clinic between when the health of information gaps, not only sharing the information with the
clinic sees the patient and when the claim is submitted, 2) people that can make use of it, but also giving them the tools
the time it takes to process the claims, entering each on into they need to make sense of the information.
the database, and 3) administration of feedback to the health Likewise, for the mobile-phone application, we also specif-
clinics, especially in case of errors. ically do not design one-way system in which claims data is
Claim Mobile is able to address all three of these cases by going out and only money returns. Instead, the claims data
1) encouraging providers to submit claims as they see patients, created in the clinic is also used within the clinic to help them
2) reducing the data entry burden through the use of digital improve patient care, as well as the accuracy of future claims.
claims, and 3) eliminating the possibility of a number of errors,
and providing a digital feedback mechanism to supplement the C. Related Work
infrequent in-person feedback. There have been a number of recent technical projects on
However, another source of delay is the administrative and the use of ICTs for healthcare in Africa [8], [9], [10], [5],
political dynamics by which program administration halts, and specifically on mobile devices for healthcare in Africa [3],
although health distributors continue to sell vouchers, and [11], [12]. However, many of these projects are design-focused
health clinics continue to see patients. During these times and technology-driven, reflecting on designing a working
payments are delayed unexpectedly for undetermined lengths technological solution to complete a particular task, rather than
of time, as can be observed from the early termination of our reflecting on the role of the technology in the system and how
pre-pilot study. Understanding this particular delay is key: the various solutions or approaches might affect social processes
TMA halts the program by withholding information, specif- within the system.
ically claim reports, from other stakeholders in the system. Braa describes two action research projects to deploy the
This is possible because the database is owned and controlled Health Information System Program (HISP) in Cuba [13]
entirely by the TMA. What happens when another outside and in South Africa [8], using Actor Network Theory (ANT)
database is introduced, with independent control? In this case, to think about how human and non-human (e.g. documents,
the data was not so much the key as the data entry staff that events, software, standards) interact. He specifically addresses
were responsible for approving the claims and validating the the challenges of designing for the multiple levels of entities
vouchers. At the same time, it is unclear where Claim Mobile involved in district health information systems, able to com-
Web fits in with the political strategy of the TMA. pare deployments across Mozambique, India, South Africa,
45
and Cuba. However, these entirely government-based contexts the fieldwork and setting up this pilot, as well as their honest
are much more hierarchical than the highly disparate multi- feedback on this paper.
organizational context described here.
R EFERENCES
VIII. C ONCLUSIONS AND F UTURE W ORK [1] Karen G. Cheng, Francisco Ernesto, and Khai N. Truong. Participant
and interviewer attitudes toward handheld computers in the context of
There is clearly much additional work to be done, in which hiv/aids programs in sub-saharan africa. In CHI ’08: Proceeding of the
the lessons learned from this pre-pilot study will be applied twenty-sixth annual SIGCHI conference on Human factors in computing
in the development of a new version of Claim Mobile for a systems, pages 763–766, New York, NY, USA, 2008. ACM.
[2] Cynthia Casas and William LaJoie. Voxiva: Case Study. December
full pilot in Spring 2009. 2003.
The outcomes from this study were three-fold. Firstly, the [3] Brian DeRenzi, Neal Lesh, Tapan Parikh, Clayton Sims, Werner Maokla,
choice of the mobile phone as a platform was affirmed by Mwajuma Chemba, Yuna Hamisi, David S Hellenberg, Marc Mitchell,
and Gaetano Borriello. E-imci: Improving pediatric health care in low-
the health clinics, for reasons of battery life, design for income countries. In CHI ’08: Proceeding of the twenty-sixth annual
readability, portability (susceptibility to theft), and ease of data SIGCHI conference on Human factors in computing systems, pages 753–
entry. Where we were concerned about introducing “qwerty” 762, New York, NY, USA, 2008. ACM.
[4] E.S. Berner and M.J. Ball, Editors. Clinical Decision Support Systems:
keyboards to novice users, our fears were alleviated, and Theory and Practice. Springer-Verlag, 1998.
all of our users assured us that “we can learn,” which they [5] Henry Lucas. Information and communications technology for future
did, quickly. Secondly, the mobile platform is not a sufficient health systems in developing countries. Social Science & Medicine,
66:2122–2132, May 2008.
solution for this program, and alone has the potential to exac- [6] ITC - e-Choupal. http://www.itcportal.com/sets/echoupal frameset.htm.
erbate information asymmetries between the stakeholders. To [7] 2006 Venture Strategies and Mbarara University population survey. http:
address this issue, we complement the mobile phone platform //www.oba-uganda.net.
[8] Jørn Braa and Calle Hedberg. The Struggle for District-Based Health
with a web application. However, Internet accessibility issues Information Systems in South Africa. The Information Society, pages
may require further development to enable local hosting and 113 — 127, 2002.
synchronization of the web application [10], [14]. Our final [9] Tessa Tan-Torres Edejer. Disseminating health information in developing
countries: the role of the internet. BMJ (British Medical Journal), pages
point is related – we consider the plethora of stakeholders 797—800, 2000.
in this project, and note that as technology providers we are [10] Rowena Luk, Melissa Ho, and Paul M. Aoki. Asynchronous remote
not coming in as naturally neutral players. Our projects are medical consultation for Ghana. In CHI ’08: Proceeding of the
twenty-sixth annual SIGCHI conference on Human factors in computing
necessarily disruptive, and equally potentially disrupted by systems, pages 743–752, New York, NY, USA, 2008. ACM.
other dynamics within the program as a whole. As a result it [11] Tapan Parikh. Position Paper: Mobile Phones may be the Right Devices
for Supporting Developing World Accessibility, but is the WWW the
is necessary for us as researchers to position ourselves and our Right Service Delivery Model? In W4A at WWW2006. ACM, 2006.
designs carefully, making sure to take into account the needs [12] Tapan Parikh and Edward D Lazowska. Designing an architecture for
of all of the stakeholders, and not just our primary users. delivering mobile information services to the rural developing world. In
Proceedings of WWW2006. ACM, 2006.
IX. ACKNOWLEDGMENT [13] Jørn Braa, Ola Hodne Titlestad, and Johan Sæbø. Participatory Health
Information Systems Development in Cuba the Challenge of Addressing
We thank Ben Bellows for his contributions to this project, Multiple Levels in a Centralized Setting. In Proceedings of Participatory
and feedback on the paper. In addition, we would like to Design Conference 2004. ACM, 2004.
[14] Michael Demmer, Bowei Du, and Eric Brewer. Tierstore: A distributed
thank all the staff at our partnering organizations and in the file-system for challenged networks. In Proceedings of File and Storage
health clinics for their assistance and patience in conducting Technologies (FAST), 2008.
46
Computer Games in the Developing World: The

Value of Non-Instrumental Engagement with
ICTs, or Taking Play Seriously
Beth E. Kolko, Cynthia Putnam
 online traffic absorbed by gaming related activity. There is no

Abstract—This paper argues that it is important to study non- ranking of popular games in each country, or listing of most
instrumental uses of ICT, including computer games. popular gaming cafes. There‟s no comparison of how much it
Specifically, the article presents the results of qualitative and costs to play Counterstrike or how many youth under the age
quantitative work spanning eight years of investigation in
of 15 are estimated to have consoles in their homes. The HDI,
Central Asia focused on computer gaming in public Internet
cafes as well as private spaces. The results presented demonstrate in short, ignores games as an element of information and
that people do indeed play games in resource constrained communication technology infrastructure within a country.
environments. The paper demonstrates that games constitute a Games, in fact, are invisible to the UN, as they are to just
significant portion of the ICT ecology in developing regions and about every non-governmental organization and multi-lateral
provide a pathway to people’s “first touch” of a computer, that organization that has worked to create information and
gamers have more frequent interaction with technology than
communication related projects in the past decade or more.
basic Internet users, that games bring more diverse users to
computers by providing a pathway to ICT use for people with The goal of this paper is not, actually, to argue that the HDI
lower levels of education, and that games can motivate innovation should include computer games as a measure of development.
in the technology space. Additionally, our findings indicate that However, the purpose is to establish that games lead to
both genders engage in game playing. The article makes the case learning computer skills and that, indeed, games are played by
that games can be a source of informal learning about ICT, and people throughout the world. Games are in fact a pivotal piece
as such, games and gaming culture in the developing world merit
of a country‟s computerization, how its population gains
further study.
information and communication technology (ICT) related
Index Terms—international development, Central Asia, ICTs, skills, and how ICTs themselves begin to diffuse in
digital games, gender, Internet cafes, technology use, developing world contexts.
entertainment
II. GAMES, LEARNING, AND CULTURE
There is a robust academic community dedicated to
I. INTRODUCTION exploring the connection between games and learning,
drawing heavily from both education and psychology. Games
The Human Development Report is published by the United are an increasingly central topic in educational research, with
Nations every year. It‟s an amalgam of facts and figures about scholars researching games as part of informal learning [28],
infrastructure, health, agriculture, policy and other key skills like collaboration that people learn while playing
characteristics of most countries in the world that provide an multi-player games [29], psychological and reaction time
overview of a country‟s stage of development. The Human skills gained from games [36], and the creation of actual
Development Index (HDI) is extracted from the report, and educational games designed to teach complex skills [6,7],
includes, among other things, measures for landlines per [9,10] [30].
capita, Internet nodes, and mobile phones. The HDI, however, As a companion piece to academic work, significant
has no column for gamers per capita. Nor is there an entry for attention and dollars have been targeted at exploring the issue
percentage of GDP spent on playing games, or amount of of games and informal learning in the US (e.g. the MacArthur
Foundation‟s 5-year, $50 million project on Digital Media and
Manuscript received September 22, 2008. This work was supported in part Learning and the Robert Wood Johnson Foundation‟s $8.25
by U.S. National Science Foundation grants #0326101 and #0219350, and by
a 2000 Fulbright Scholar award. million program in Health Games Research). There are also
B.E. Kolko is an Associate Professor with the Department of Human organizations dedicated to NGO-like activity around games,
Centered Design & Engineering at the University of Washington, University including the Serious Games initiative, Games for Change,
of Washington, Seattle, WA, 98195 USA. (e-mail:
bkolko@u.washington.edu). and Games for Health; these groups focus on developing
C. Putnam is a PhD candidate with the Department of Human Centered games with explicitly pro-social goals in mind.
Design & Engineering at the University of Washington, University of Much of the research in the games and learning area has
Washington, Seattle, WA, 98195 USA. (e-mail:
cyputnam@u.washington.edu).
focused on the learning that occurs while people engage with
47
games. Developing world initiatives include work by groups locations, and it was administered in Russian and other
such as the South Africa-based Mindset Network which has regional/local languages. In addition to the general sampling
developed mobile phone-based games to teach math skills to scheme, a three-stage process was used to select respondents
girls (Mathstermind and Fashion Network), and literacy and that included Probability Proportional to Size sample of
numeracy games for disadvantaged youth developed by Primary Sampling Units (PSUs); consecutive random
Pratham in India. These specific pro-social gaming projects sampling of households in each PSU, and selection of a
similarly focus on in-game content and what people can learn household respondent using a Kish Grid method.
as a result of playing. The survey instrument was designed by a team of
This paper, however, is more interested in how engagement researchers from the University of Washington, pilot tested in
with games serves as an ICT usage entry point. In other words, each country in conjunction with local researchers, and then
not what do people learn because of a specific game, but, revised based on analysis of pilot data. BRIF Research Group,
rather, what do they learn because of the specific activity of based in Kazakhstan, translated the survey instrument from
gaming. To that end, this paper focuses attention on the English to other languages. The University of Washington
commercial game space and how it diffuses throughout and team back translated the completed Russian translation. Likert
affects nascent computer users in developing regions. This scale measures and other question formats were developed in
question of effect can be situated within a theoretical response to initial open-ended interviews, ethnographic field
framework tied to cultural theory that investigates issues of notes, pilot tests and research on performing surveys in post-
identity and agency. Generally, cultural theory provides Soviet contexts. Several steps were taken to guarantee high
another lens through which we can view the importance of quality fieldwork including: (1) approximately 30% of
digital artifacts that transcends their literal or functional interviews were checked through a back visit to the
meaning [32]. Again, in the most broad terms, cultural respondent‟s home; (2) interviewers were trained through
theorists such as Bruno Latour [35] provide a framework workshops and practiced in a pre-testing phase; and (3)
against which we can examine technological artifacts not for statistical analysis of logical inconsistencies were double
what they are, but for what they enable. That is, games and checked with the original paper questionnaires and eliminated
non-instrumental uses of technology are important in the if necessary.
developing world not (only) because they teach people to Given the low rate of current Internet penetration in Central
collaborate or improve language skills, but, as this paper will Asia, the survey also focuses on pre-existing patterns of
demonstrate, because they provide an alternative mechanism information use, information seeking behavior, and levels of
by which many people experience their first “touch” of a trust in various producers and sources of information.
computer. Additionally, we found that games allow users with Since Internet usage rates in the region tend to be low, the
less education and English language skills to interact with survey targets the general population rather than Internet
computers, and that they foster innovation and creative users; the findings include overall attitudes towards
engagement with technology. technology from a wider audience rather than simply usage
patterns from a much smaller group.
III. METHODOLOGY
B. Ethnography
Our research is the result of longitudinal qualitative and
Ethnographic results are from the CAICT project as well as
quantitative work done as a component of a larger project on
earlier fieldwork conducted by the first author during a six-
the effect of information technology on society. The Central
month residence in the region in 2000. Fieldwork data
Asia Information and Communication Technology Project
collection methods follow standard ethnographic format for
(CAICT) is a multi-year study of ICTs. The project goals
participant-observation and include field notes and
include investigation of how ICTs diffuse within societies,
photography.
how cultural issues affect technology usage, and how patterns
In addition, interview studies have been conducted with
of trust and confidence in media and institutions change over
various groups to further investigate issues relevant to survey
time as technology diffuses in diverse communities.
findings. Qualitative data reported in this paper are drawn
A. Survey from multiple separate studies conducted in 2005 and 2006
The project includes a yearly survey of four countries in including interviews with Internet users, interviews with
Central Asia: Kazakhstan, Kyrgyzstan, Tajikistan, and mobile phone users, interviews with computer gamers, and a
Uzbekistan. The quantitative results presented here are based design ethnography. In addition, ethnographic observations
on the nationwide survey of 1000 respondents in each of these are based on field notes collected during separate data
four countries age 15 and older, administered in 2006, 2007, collection trips in 2000, 2002, 2003, 2004, 2005, and 2006.
and 2008, for a total of 1000 respondents per year per country,
and a total of 12,000 respondents overall. Survey sample was IV. BEYOND STEREOTYPES OF GAMES AND GAMERS
based on census information for age, gender, ethnicity, and Before addressing the specific ways in which games
geographic location as released by each country‟s government. provide an important pathway to ICT engagement, it is
The survey includes multiple urban and rural sampling important to deconstruct some of the stereotypes of games and
48
gamers. Recent research has begun to establish that gamers in

developed countries are not all male and not all teenagers [33-
34], that in fact the audience for games is diverse in terms of
gender and age [1][22-24]. Studies of complex multi-player
gaming environments, for example, demonstrate that although
male players outnumber female players, it is those female
players who actually spend more time online playing [25-26].
Games themselves come in many flavors, including the violent
shooting games that gain so much press attention, but they
also include puzzle games and simulations. Game systems
such as the Nintendo Wii or games with alternative input
modes like Rock Band or Dance Dance Revolution have
similarly changed the cultural activity of games and broadened
the audiences to which they appeal [27].
While not identically diverse, gamers internationally also
Figure 2: Photograph of Starcraft competition
belie some stereotypes. While gamers are still more likely to
announcement outside Kazakh game club
be male and under 30, in our research population, female
respondents and those over 30 were as likely to play games as
they were to use the Internet, See Figure 1. This finding
30%
indicates that depending on the population segment, gaming is
25%
either a more used pathway to ICT use than the Internet, or it
is at least equivalent. 20%
10% 15%
9%
8% 10%
7% 5%
6%
5% 0%
Kyrgyzstan Kazakhstan Tajikistan Uzbekistan
4%
Female 8% 18% 5% 8%
3%
Male 17% 26% 18% 19%
2%
1% Figure 3: Male and Female Gamer Populations in Central
0%
Male Female Under 30 Over 30 Asia, 2008
Gamers 8% 5% 9% 4%
Internet users 5% 5% 6% 4% Internet cafes vary in quality of equipment and furniture.
Some businesses will have recent equipment, flat panel LCD
Figure 1: Demographics of gamers vs. Internet users screens, and glossy office furniture; others have older PCs,
CRT monitors, and plastic chairs. Often the ones the feel more
V. SETTING THE STAGE: GAMING CONTEXT IN CENTRAL ASIA like a business center are located in the center of the city and
Public gaming centers in Central Asia meet many of the cater to a mixed clientele. However, businesses located in
preconceptions of computer game culture, but they are also neighborhoods similarly can serve a diverse clientele,
nuanced environments. The crowds of young boys that are although not necessarily at the same time. And games exist in
seemingly ubiquitous are reminders of gamer stereotypes, a context of ICT diffusion overall, with Internet growth
although our survey results indicate that women do indeed creeping slowly and mobile use skyrocketing, see Figure 4.
play games (see Figure 3), although women tend to play at
home rather than in public cafes. 70%
Many of the popular game titles are familiar, but the style of 60%
play and the mechanics of getting the games to work in 50%
different infrastructure environments are unfamiliar. 40%
Counterstrike (CS) and Starcraft are literally everywhere, and 30%
kiosks on the streets sell countless numbers of CS mods, the 20%
cd covers in English and Russian. Starcraft competitions can 10%
be found in many of the capitol cities in the region, see Figure 0%
6, and the World Cyber Games are a coveted destination (and Mobile Users Computer Users Gamers Internet users
Central Asian countries have placed quite high in the World 2006 20% 25% 13% 9%
Cyber Games in recent years). 2007 37% 24% 13% 8%

2008 64% 30% 14% 12%
Figure 4 Relative Rates of ICT Usage 2006-2008

49
A. A Tale of Two Tajik Internet Cafes Gigant 2 is located on the edge of Dushanbe in a residential
Plazma is an Internet café located in the center of neighborhood. It is located in a one-story strip of local
Dushanbe, the capitol of Tajikistan. Plazma is on Ryudaki businesses, There is a bread store on one side and a small dry
Street, the central boulevard in the city. It is located in a two- goods market on the other. On the sidewalk in front are
story standalone building, with a very large sign on the side several women selling seasonal fruit. Gigant 2 has streetside
advertising a bar and café on the first floor, and Internet signage advertising Internet services, prepaid IP telephone
center, computers, fax, IP telephone, and other services on the cards, DVD/VCD and VHS rentals. The café/club rents
second floor. Plazma is typical of centrally located Internet movies, burns DVDs and CDs, and offers multimedia services.
centers in the cities of the region in that it offers a wide range There are 24 stations in the main room of Gigant. There is
of services. However, it is also a particularly good example of also a back room that is attached to the main room and
a business that has specifically not bifurcated into two separate separated only spatially. There are six stations in the back
establishments in order to accommodate both gamers and a which serves as a VIP room and includes two couches and a
growing population of Internet users. Plazma has, instead, large floor fan. Gigant is nominally air conditioned, and heavy
blended the two business models and offers a central room plastic sheets hang as a barrier at the front door to keep the
with ten computers with Internet access arranged on long cooler air inside, but in the height of summer it is very warm
tables but separated by small tabletop dividers to allow inside. There are two IP telephone booths in the café, and a
privacy, 3 IP telephone booths, a central desk where items can site-wide LAN for watching movies. Dozens of games were
be photocopied, text typed, faxes sent, and – very welcome in preloaded onto the machines, but Lineage, CS, Need for
the Tajik summer – air conditioning. There is also a sign Speed, and Cossacks were reported to be most popular. The
posted in each mini workstation providing some guidance business remains open all night, and from 9pm to 6am the
about usage that reads in both Russian and English: “It is only services offered are Internet access and Lineage. CS was
categorically not allowed to visit sites with pornographic not offered during the overnight because it was too difficult to
photos and video materials. In case of visiting above find many other players online during that time frame.
mentioned sites, administration has a right to fine you from 80 Gigant has a three tired pricing structure, similar but not
up to 300 somani,” the equivalent in 2006 of US$23-86. The identical to Plazma, see Figure 8. Playing games costs one
sign serves not just as instruction not to visit porn sites, but somani per hour, being “online” also costs one somani per
also as a not-so-subtle reminder that one‟s online activity is hour, and “Internet” costs 2.5 somani per hour. The distinction
being constantly surveilled. What should be obvious, of between online and Internet is something that appears with
course, is that the specific sites or what constitutes some regularity in developing regions, and it is a further
pornography is left open to interpretation. That signage echoes motivation for individuals to develop usage patterns that
signs in Uzbekistan cafes that warn about “inappropriate downplay the importance of open-ended web browsing. In the
content” which can be either pornographic or political. case of Gigant, online activities include chatting, but Internet
On the same floor, but in a separate glass enclosed room is means web browsing since ISPs often charge by the kilobyte.
the gaming center with 28 game stations preloaded with games Under those circumstances, then, browsing can be extremely
such as Grand Theft Auto, Cossacks, Need for Speed, Call of costly and, perhaps even more of an inhibiting factor, not
Duty, FIFA, and Half Life/CS and equipped with high quality particularly transparent in terms of expense.
headsets. On the wall in the main room where the business
services are offered is a large embossed sign that runs the
length of the room that has the site‟s web address as well as
their tagline in Russian and English: All Our Life is a Game.
The manager reports that Counterstrike and Starcraft are
among the most popular games at the café. During the school
year the café is crowded, but during summer vacation students
visit less often – presumably because they spend less time in
the center of the city. The customer base for the gaming
stations ranges from schoolchildren to adults, and the adults
tend to call ahead to see if a station is free and what games are
in session before they come to the café.
Using the services in Plazma costs two somani an hour to
play games and four somani an hour to use the Internet. The
ratio – games costing half what Internet access costs – is in
line with general trends in the region. As argued above, when Figure 5: 2006 pricing structure at Gigant 2: Games 1
Internet access is metered and slow, the price differential somani, Online 1 somani, Internet 2.5 somani
between playing games and browsing makes it increasingly
attractive to see games as the open-ended entertainment Gigant is a café that melds general Internet services and
activity associated with computer technology rather than gaming into one space. In the afternoon, it was packed
aimless browsing or opportunistic link following. exclusively with schoolage boys. The manager asserted that in
50
the evening, when adults returned home from work, the somewhat counter to assumptions made about users in
clientele turned over. Indeed, around 6 pm, some of the boys resource-constrained environments, but the so-called frivolous
drifted out, but increasingly adults – including women, began uses of technology are, indeed, often what brings people
coming into the business. through the door [31].
These two profiles capture many of the business The sheer amount of game activity and gaming culture as
environments where people play games in developing world revealed by longitudinal ethnographic work in Central Asia
contexts. Their public nature, idiosyncratic policies, and tiered establishes the importance of games as a part of ICT ecology
pricing structures are common elements and evidence of in this resource-constrained region. And when schools are not
variability across venues. wired and home access rates are low, game cafes are likely
places for people‟s “first touch” with computers.
A. Nurturing public Internet access sites in early years: the
VI. FINDINGS:
GAMES AS A SIGNIFICANT PART OF AN ICT
role of games
ECOLOGY AND POTENTIAL SITE OF A USER‟S FIRST “TOUCH” OF
A COMPUTER In the past several years, a number of studies of Internet
cafes have demonstrated the importance of such public access
Throughout this project, we have maintained a focus on the
sites for users in emerging markets and other developing
study of diverse users‟ information landscapes -- their ICT
regions; many of these studies have focused on NGO-
ecology. In other words, this work takes particular interest in
sponsored telecenters rather than commercial sites [4-
how specific pieces of ICT form unique mosaics in different
5][8][[11-13][18][21]. Generally speaking, though, in the
usage contexts, giving individuals and communities access to
global south, access to the Internet is more often in public,
variable modes of interaction, communication, and
shared space than it is in private homes, or even the workplace
information-seeking. What has become clear over the years is
or school. There are several characteristics of public access
that games and gaming provide an important – if often
usage, however, that mean people‟s relationship to the Internet
overlooked – piece of this puzzle [15]. Often dismissed as
develops a particular shape. [17] Public access generally
irrelevant to capacity-building projects, overlooked as a
brings an awareness that one‟s usage is economically
measure of a nation‟s ICT sophistication, and prey to
constrained, metered by time and often by kilobyte, inhibiting
countless stereotypes about users, games are a largely invisible
the link-following that characterizes much Internet usage in
component of the ICT landscape in developing regions.
broadband contexts.
However, from the standpoint of users gaining facility with
Internet cafes, for the purposes of this discussion, are
ICTs, and allowing ICT-related industries to gain an early
publicly accessible commercial or noncommercial sites where
foothold, games play an important role as they are often the
people gain access to a variety of ICTs, including Internet, IP
first attractor – the first ICT with stickiness – for novice users.
telephone, photocopying, faxing, etc. Many commercial sites
In other words, games are an alternative pathway to some
augment their income by providing some café-like amenities,
users‟ first „touch‟ of a computer.
but often the name “Internet café” is conceptual only. But
On a functional level (as well as philosophical level), ICTs
Internet cafes are also often the first places that ICTs come to
mediate distance and time [2][3][19]. In a country such as
communities in developing world contexts. Whether
Kenya or Cambodia, nations with poor road infrastructure, the
established by an NGO or a local entrepreneur, Internet cafes
ability to transfer information quickly and over large distances
advertise novel services to a community that is largely
has a transformative effect on the kind of information
technologically illiterate, and they then face the challenge of
environment within which people live. On the other hand,
convincing people they have a need for services they do not
there is also the conceptual mediation ICTs play as they bring
know how to use. Indeed, the capabilities of the Internet are
fractured components of the globalization narrative to small
themselves somewhat opaque to individuals who have not
communities that might be otherwise have very limited
been exposed to a new media infused environment, and games
exposure to outside media influences. In Central Asia, ICTs
are often a way to first draw people through the door.
are a linchpin of economic development as well as a symbol of
The research conducted in Central Asia from 2000-2007
modernization. Computers in schools projects bring
pointed to an evolutionary pattern for commercial public
technology literacy to youth, e-government initiatives slowly
access sites. To illustrate, take the example of Tashkent, the
bring elements of government online and into some version of
capitol of Uzbekistan. Tashkent is a city of approximately
transparency, and within certain professional sectors
three million; it was the fourth largest city in the Soviet Union.
computers streamline practice and make international
In December of 2000, there were twelve operating Internet
collaborations more effective.
access points in Tashkent. This list included two sponsored by
However, when such top-down initiatives are bracketed,
NGOs or large multi-lateral organizations, one funded by a
and instead focus is placed on bottom-up patterns of
local cultural center, and nine commercial cafes – most of
technology diffusion, ICT usage follows a different narrative.
which did not actually serve food or drink. That number has
Adoption and usage patterns that emphasize ICTs for
since grown to the hundreds, and the actual number is
communication, gaming, and entertainment emerge. From
impossible to estimate given the fact that many operate
movies to games to ringtones, ICTs become attractive to
without being fully licensed by the government, signage can
everyday people for non-work purposes. This pattern runs
51
be missing or misleading, and there is no reliable central a community member did need occasional access to the
directory of all functioning businesses in the city. However, it Internet, there was in fact a local site to visit. However,
is safe to state that the Internet business in Tashkent as of without games and other non-instrumental uses of the
2007 was thriving. Home access is growing, numerous ISPs computers, the businesses would have had a much smaller
offer local dial-up service, others offer DSL, and schools and regular customer base on which to draw. We would argue,
businesses are increasingly getting Internet access. then, based on our interviews and observations that games
In 2000 (the early years of Internet diffusion), it was not played a pivotal role in the ICT diffusion within that country.
immediately clear to individuals why they should spend about And once people started playing games, particularly LAN-
US$1 per hour to use something that had linguistic and based games, they began chatting and utilizing other
technological barriers, particularly in a country where the communication tools. Their usage of ICTs expanded beyond
average monthly salary was US$20-30. Interviews with novice games, but it was the gaming activity that allowed them to
users over the course of five months in 2000 pointed to such enter the so-called information society.
confusion and illustrated the cultural gaps that acted as
B. The appeal of games as a pathway to ICTs compared to
additional barriers to entry for new ICT users. There was an
other technologies
absence of external information resources in people‟s lives, so
using analogies to describe the Internet as, for example, like While initially games might have been appealing because
yellow pages or encyclopedias was not particularly resonant. they were an inexpensive way for users to begin
Keyboards were often in English yet people spoke Uzbek or experimenting with computer technology, there was also a
Russian and needed the Cyrillic alphabet. Operating systems functionality argument in favor of them. In many developing
and software were often in Russian which was an additional world contexts, including Central Asia, Internet access tends
hurdle to Uzbek speakers (Uzbek is a Turkic language, not to be fairly slow. Often an Internet café will have a dozen or
Slavic). Printer drivers didn‟t have the Uzbek character set, more computers sharing one dial-up or DSL line. Such
there was very limited content online that was relevant to local bandwidth limitations play out, for example, in a
inhabitants, and keyboarding or typing skills were not contemporary web page from a more developed country
common. Literacy rates themselves were quite high, although taking two or three or five minutes to load. In and of itself that
the multi-lingual nature of life could make reading and writing may not seem a hurdle, but the time adds up.
a bit more complicated depending on primary and secondary Using a basic free webmail application like Yahoo or
language ability. [14][16] Hotmail becomes an exercise in patience. To write an email in
However, because copyright laws were loosely if at all Yahoo mail, for example, takes about six page loads: one to
enforced, there was no shortage of entertainment media load mail.yahoo.com; a second to enter a username and
available. The latest games could be bought for about US$1 in password and wait for login authentication, a third to click on
kiosks on the street, and VCD movies were about the same the inbox and check mail; a fourth to click on compose to
price. At the same time, games could be played either write a message; a fifth to click on send and wait for the
standalone on individual computers or over a café-based LAN. message to go out; a sixth to click on check mail again and be
Alternatively, movies could be watched on the computers by returned to the inbox. Do the math, and suddenly it‟s 20-30
an individual or a group. Cafés developed a two-tiered pricing minutes to write one email – and that‟s without having to
structure where it cost roughly half as much per hour to play check someone‟s email address by loading the addressbook.
games. What emerged, then, was a pattern of usage in these Because these are public Internet sites, pop mail or SMTP is
Internet cafes where the majority of customers came in to of limited utility, and so webmail quickly became the mail
either use computers to type documents (also much cheaper application of choice. However, people were forced to use
than Internet access), or to play games. During the day, webpages created in 2000 with access speeds more
schoolchildren would come to play a variety of games, and approximate to 1995. The Internet, then, was simply not
often café owners would let the youngest children hang around perceived as terribly efficient; indeed, it was not at all
and watch movies or play games if the café was otherwise efficient. Opportunistic web browsing was not a habit easily
empty. Playing games provided young people with their first adopted when web pages took so long to load. Expensive and
experiences touching computers. Playing demystified the slow, the Internet was a window to frustration rather than a
technology, allowed them to gain keyboarding skills, taught window to the world.
them how to navigate operating systems and GUI menus, and Added to this limited utility, however, was a slightly more
generally build comfort and facility with computers. In those acute problem of surveillance and censorship.. Many Internet
early days, in 2000, it appeared that revenue from activities cafes had signs posted warning that accessing inappropriate
other than Internet usage allowed Internet cafes to remain content would result in fines or arrest. The managers or
open in their communities. administrators would sit at a desk and monitor traffic; users
Although it was impossible to get owners to disclose actual were crowded elbow to elbow with very little privacy for the
revenue figures, in conversation and observation over the items on their screen. Inappropriate content was rarely if ever
course of six years of field visits, it became clear that for many defined. Sometimes it was specified as political information,
businesses, the revenue generated by game players helped or inappropriate political information; other times
sustained these early adopter businesses. Consequently, when pornography was banned. But what constituted inappropriate
52
political content could change; a website with acceptable we wanted to identify those users who overcame the barriers
regional news one day might be overly critical of the president to entry for either activity. Our goal in this exploration was to
the next and suddenly be on the banned list. The habit of investigate if the two groups differed, and if so, how they
censorship and surveillance for one‟s Internet usage provided differed and what the differences would suggest about
another tamping mechanism for Internet use at the Internet potential barriers.
cafes, and another motivation for those who wanted to learn Two direct logistic regression analyses were performed
about computers to gravitate toward games. using SPSS to predict gaming and Internet use outcome from
Amidst all this gaming in the region – activity that was eight predictors: (1) age; (2) gender; (3) years of schooling;
economically as well as socially appealing, gaming-specific (4) living in a rural or urban environment; (5) mobile phone
centers gradually emerged. While game playing allowed use; (6) ability to speak and read English; (7) ability to speak
Internet cafes to remain open in the early days of ICT and read Russian; and (8) socio-economic status reported on a
diffusion, eventually the industry matured substantially scale of one to three.
enough that gaming centers split off from Internet cafes. There
A. Comparing predictors of game playing to Internet use
were enough Internet users producing Internet-only revenue
that many businesses could focus on one or the other. Many First, we analyzed game playing. A test of the full model
crossover businesses remain, especially in countries with with the set of predictors against the null model with no
lower percentages of Internet users (Uzbekistan, Tajikistan), predictors was significant, 2(9, N=12000) = 2268.52, p<.001,
but one can see that as the sector matures and a business can Nagelkerke R2 = .317, indicating that the set of predictors
take in enough revenue from Internet access to remain reliably distinguishes between individuals who play games
sustainable, that game cafes separate themselves from Internet and those who do not. The approximate variance in predicting
centers. game playing accounted for by the set of predictors is 32%.
It is not a zero sum game, though, and as the numbers of According to the Wald criterion all eight variables reliably
Internet users grows, gaming remains a vibrant activity. predicted computer gaming--listed here in order of influence:
Indeed, the presence of multiple gaming centers throughout (1) age; (2) ability to speak and read Russian; (3) owning a
the region – particularly in the capitol cities -- makes it clear mobile phone; (4) living in an urban environment, (5) gender;
that games are being played, both players and games are (6) ability to speak and read English; (7) years of education;
diverse, and gaming provides a social interaction platform. and (8) SES.
Next, we analyzed Internet use. A test of the full model with
70%
the set of predictors against the null model with no predictors
60% was significant, 2(9, N=12000) = 2852.51, p<.001,
50%
40% Nagelkerke R2 = .447, indicating that the set of predictors
30%
20% reliably distinguishes between individuals who use the Internet
10%
0% and those who do not. Internet use accounted for by the set of
General Inter- predictors is a striking 45%
Home Friend's School
Access net cafe Work
***
Point ** ***
house ** According to the Wald criterion, all eight variables reliably
Kyrgyzstan N = 121 49% 6% 27% 15% 8% 15%
predicted Internet use--listed here in order of influence: (1)
Kazakhstan N = 215 62% 2% 5% 22% 10% 12%
ability to speak and read English; (2) age; (3) years of
Uzbekistan N = 115 27% 10% 30% 17% 20% 11%
education; (4) owning a mobile phone; (5) living in an urban
Tajikistan N = 122 29% 3% 30% 17% 18% 11%
environment; (6) ability to speak and read Russian; (7) gender;
and (8) SES.
Figure 6: Sites of Gaming in Central Asia, 2008
*** = p <.001; ** = p <. 05 Statistics based on Chi-square test
600.00
500.00
VII. SURVEY FINDINGS: LESS ADVANTAGED USERS MORE
400.00
LIKELY TO GAIN INTRODUCTION TO ICTS THROUGH GAMES
300.00
As discussed in the previous section, games can provide a
200.00
mechanism by which youth can be drawn into ICT training
centers, and games also provide a revenue stream for owners 100.00
of public access ICT sites. Additionally, our survey results 0.00

Years Living
demonstrate that games offer tangible assistance in of in an
Mobile
Age Gender phone English Russian SES
overcoming barriers to entry for novice users by allowing school- urban
use
ing /rural
users with less education and English language skill to gain Gaming 562.34 97.72 56.36 112.10 113.51 75.40 122.21 44.02
experience with ICTs. Internet Use 248.22 12.41 196.40 142.06 171.21 339.05 135.45 11.19
In order to explore potential barriers to entry for novice
users we wanted to first identify what types of people play Figure 7: Predictor effects on the models according to the
Wald Statistic
games compared to those that use the Internet, In other words,
53
While the sets of predictors are the same, there are notable very few guilds, and the auction house is not a particularly
differences. First, the set of variables does a much better job at effective mechanism for trading goods. While economically
predicting Internet use than game players, indicating there is viable in this context, Azeroth, the world of WoW, is also a
more homogeneity in the Internet using population; however, largely empty world for these players. However, clearly the
both models were very good at predicting the outcome. online game has a draw, as do many other game genres. The
Second, the importance of individual variables is different. local hacks that people use to be able to play games are a
Whereas the ability to speak English is very important to testament to the appeal that games of diverse genres have for
predicting Internet use, it was less influential on game playing. users.
Additionally, education level is very important to predicting
Internet use, but not as much for computer gaming. B. Games as motivation to gain technical expertise
Conversely, age is a stronger predictor of gaming. Together, Similar local adaptation can been seen among local gamers
these items suggest that entry into the game playing realm is in Bishkek. A group of, neighbors, residents of the same large
more accessible to a different segment of the population. If the apartment building, discovered that they all liked to play the
goal, then, is to broaden participation in ICTs, games provide same games, but they preferred to play at home because they
a pathway to ICT usage for users who would not necessarily saw it as cheaper and they had “more freedom.” So they ran a
have the education or English language ability to become LAN down the outside of their nine story Soviet style flat,
Internet users. See Figure 9.
connecting eight neighbors together so they could play
This finding is significant in that it clearly establishes that
together. Gamers display a fair amount of creativity in
gaming and Internet use attract different segments of the
population, and that an individual‟s likelihood to gravitate to ensuring they are able to remain connected with global gaming
one version of ICT over the other is not only due to culture. Their narratives of themselves as gamers, however,
availability or interest, but is also due to societal factors such emphasize the public and social aspects of their gaming which
as education. is not necessarily consistent with gaming in the US, but does
resonate with gaming culture in Korea and China.
VIII. FINDINGS: GAMES AS A GATEWAY TO AND MOTIVATION One of eight Kyrgyz gamers interviewed in 2006, Yuri
FOR INNOVATION echoed the pattern of games as a motivator for learning more
about technology. He started playing games in the 4th grade,
The strong attractor force of games can also be a motivator
and was introduced to games by playing Flight Sim at a
for innovation. Many games encourage users to become active
friend‟s house, long before computers had been introduced to
participants in a digital environment, and the enthusiasm many
his public school. After Flight Sim, he began going to
players feel for games can motivate them to learn new
computer clubs where he fell into the world of Counterstrike.
technical skills in order to facilitate their playing.
Now 19, he plays games over a LAN only in the clubs
A. World of Warstan because, as he says, Internet is too expensive. After playing
Blizzard‟s World of Warcraft (WoW) is a subscription- for years, he gained a variety of computer-related skills and
based game that requires Internet access. It is a massively- now works as an administrator at a game café. He calls his
multiplayer game that has thousands of players on each server. friends and arranges for them to come after hours and gather
Much of the gameplay relies on collaboration, and players together to play. Although none of the other gamers
form guilds; trade happens in auction houses, and the game interviewed in Kyrgyzstan in 2006 had direct stories of their
requires a certain critical mass of players to run effectively. gaming leading to jobs, they did all emphasize the collective,
The technology and banking infrastructure required by WoW shared knowledge of their gaming circles, the enjoyment of
would seem to relegate it to an impossible game in the context gathering in public places together to compete, share, and
of Central Asia, a region with relatively slow Internet access improve skills.
and essentially no credit cards. However, in at least two of the Their stories also stand in contrast to the usual ICTD
capitol cities in the region, local ISPs have devised a way to narratives about how newcomers become first acquainted with
offer WoW to their subscribers. ICTs and the associated information society. Often,
Locally run servers allow DSL subscribers to play in- development projects that incorporate technology emphasize
network, so, for example, there is a Tashkent-wide version of instrumental uses of technology. Games, however, are part of
World of Warcraft. The game is in Russian but with a server an open-ended and alternative exploration of technology.
attached to the provider. Most of the people are from the
capitol given the pattern of DSL diffusion. Players recruit IX. CONCLUSION
friends and classmates, but because of the low in-game Much of the research on game culture concerns itself
populations, the basic mechanics of the games are different primarily with gaming in extremely developed contexts –
from standard play in developed world contexts. where resources are abundant, and where gamers‟ relationship
For about UDS$12-13 per month, players get unlimited with technology has evolved in the context of that abundance.
hours online with both WoW and ICQ. All other Internet Game culture is just as interesting, and just as vibrant, and just
activity is metered by kilobyte, and MSN and Yahoo as reflective of complex social dynamics when it takes place
Messenger are metered as well. As for WoW, there are only a in resource constrained environments and when the gamers
few hundred users online in the world, which means there are who adopt these technologies have very different relationships
54
with the technology. Central Asia is one such resource [10] Gee, J. P. (2003) What video games have to teach us about learning and
literacy. Palgrave/Macmillan: New York, NY
constrained environment, and while the specifics of the region [11] Hudson, H. E. (2006). From Rural Village to Global Village. Mahwah,
give it particular political and economic constraints, the stories New Jersey: Lawrence Erlbaum Associates, Publishers.
that emerge of games‟ importance in allowing ICTs to take [12] Jensen, M., & Esterhuysen, A. (2001). The Community Telecentre
Cookbook For Africa - Recipes For Self-Sustainability. Paris: United
hold during early stages of technology diffusion, as well as the
Nations Educational Scientific and Cultural Organization.
persistent public nature of gaming activity help to broaden the [13] Khelladi, Y. (2001). What Works:The Infocentros Telecenter Model.
picture of what games can be and what purpose they can serve Washington DC: The World Resources Institute.
within an overall ICT ecology. [14] Kolko, B.E. (2002). “International IT Implementation Projects: Policy
and Cultural Considerations.” Proceedings from the Annual IEEE IPCC
It may also be that gaming culture is the non-instrumental Conference, Portland, OR, September 2002. 352-359.
use of ICT that can parallel some of the adoption stories for [15] Kolko. B.E., Thayer, A. (2003).“Games as Technological Entry Point: A
mobile phones. Games are fun, and there‟s nothing wrong Case Study of Uzbekistan.” Proceedings of the Digital Games Research
Association. Utrecht University. 19 pages.
with fun, even in a development context. If talking to friends [16] Kolko, B.E (2006) “Cultural Considerations in Internet Policy and
and family is part of the leverage that gets people to adopt Design: A Case Study from Central Asia.” Critical Cyberculture
mobile phones, then playing with friends may be part of the Studies: Current Terrains, Future Directions. Ed. David Silver and
Adrienne Massanari. New York: NYU Press. 145-157.
leverage that gets young people to develop further skills with [17] Kolko, B.E., Rose, E.J., Johnson E. (2007). “Communication as
computers. Information-Seeking: The Case for Mobile Social Software for
As we have argued in this paper, gaming culture is vibrant Developing Regions.” Proceedings of ACM World Wide Web
Consortium Conference. 863-872.
and thriving, and it provides a potential “first touch” with [18] Mercer, C. (2006). Telecentres and transformations: Modernizing
ICTs , especially for people who may not have access to Tanzania through the Internet. African Affairs, 105(419), 243-264.
computers at home, work, or school. Additionally, our survey [19] Poster, M. (1990). The Mode of Information. University of Chicago
Press.
demonstrates that games provide different barriers to access
[20] Proenza, F. J. (2001). Telecenter Sustainability - Myths and
than Internet usage, broadening the reach of ICTs. Finally, Opportunities. Journal of Development Communications, 12.
games and the enthusiasm that gamers have, can motivate [21] Proenza, F. J., Bastidas-Buch, R., & Montero, G. (2001). Telecenters for
innovation and technological skill acquisition. Socioeconomic and Rural Development in Latin America and the
Caribbean. Washington DC: Inter-American Development Bank.
Ultimately, the goal of this paper has been to bring attention [22] Schott, Gareth R. and Horrell, Kirsty R. (2000). “Girl Gamers and Their
to games and some of their possibilities as an element of ICT Relationship with the Gaming Culture,” Convergence, v. 6, n.4, 36-53.
ecology in resource constrained environments. Games have [23] Taylor, T.L., Kolko, B.E. (2003). “Boundary Spaces: Majestic and the
Uncertain Status of Knowledge, Community and Self in a Digital Age.”
been almost completely ignored by the ICTD discourse, and as Information, Communication & Society. 6:4, 497-522.
the community seeks to better understand how and why people [24] Yates, Simeon J. and Littleton, Karen. (1999). "Understanding
adopt ICTs into their lives, it seems important to include Computer Game Cultures: A Situated Approach," Information,
Communication, & Society, 2:4.
entertainment-related uses of ICT as part of that adoption [25] Williams, D., M. Consalvo, S. Caplan & N. Yee. (2009, in press).
pathway if we hope to truly comprehend the cultural and Looking for gender (LFG): Gender roles and behaviors among online
economic changes associated with technology. gamers. Journal of Communication.
[26] Williams, D., N. Yee & S. Caplan (2008). Who Plays, How Much, and
Why? A Behavioral Player Census of Virtual World. Journal of
ACKNOWLEDGMENTS Computer Mediated Communication.
[27] Nadia Bianchi-Berthouze, Whan Woong Kim, Patel Darshak, "Does
Our thanks to Mark Chen, members of the CAICT research body movement engage you more in digital game play? And Why?",
team, including Erica Johnson, Emma Rose, and Odina Proceedings of the Int. Conf. of Affective Computing and Intelligent
Salikhbaeva. Thanks also to our local researchers, research Interaction, LNCS 4738, 102-113, Lisboa (Portugal), September 2007
[28] Stevens, R., Satwicz, T., & McCarthy, L. (2007). In-Game, In-Room, In-
participants, and blind reviewers.
World: Reconnecting Video Game Play to the Rest of Kids' Lives. In
Katie Salen (Ed.), The Ecology of Games: Connecting Youth, Games,
REFERENCES and Learning (pp. 41 - 66). Cambridge: The MIT Press.
[29] Nardi, B. and Harris, J. (2006). Strangers and friends: collaborative play
[1] Cassell J. and H. Jenkins (eds.), (1998). From Barbie to Mortal Kombat:
in World of Warcraft. In CSCW '06: Proceedings of the 2006 20th
Gender and Computer Games, Cambridge: The MIT Press.
anniversary conference on Computer supported cooperative work, pages
[2] Castells, M (1996). The Rise of the Network Society, The Information
149-158, New York, NY, USA. ACM Press.
Age: Economy, Society and Culture Vol. I. Blackwell: Cambridge, MA.
[30] Holland, W., Jenkins, H. & Squire, K. Theory by Design (2003). In
[3] Castells, Manuel (1997). The Power of Identity, The Information Age:
Perron, B., and Wolf, M. (Eds). Video game theory reader. (pp. 25-46).
Economy, Society and Culture Blackwell: Cambridge, MA
London: Routledge.
[4] Colle, R. (2005). Memo to Telecenter Planners. Electronic Journal of
[31] Sandvig, C. (2006). The Internet at Play: Child Users of Public Internet
Information Systems in Developing Countries, 21(1), 1-13.
Connections. Journal of Computer-Mediated Communication, 11 (4).
[5] Dagron, A. G. (2001). Prometheus Riding a Cadillac? Telecentres as the
[32] Wise, John MacGregor. (1997). Exploring Technology and Social
promised flame of knowledge. Journal of Development Communication:
Space. Thousand Oaks, CA: Sage.
Special Issue on Telecentres 12(2).
[33] Lenhart, Amanda, Kahne; Middaugh, Ellen, Rankin Macgill, Alexandra;
[6] Dubbels, B. (2003) Video Games as Metaphor for Learning and
Evans, Chris; Vitak, Jessica. (2008). “Teens, Video Games and Civics:
Curriculum Design
Teens' gaming experiences are diverse and include significant social
[7] Egenfeldt-Nielsen, S. (2003). Review of the research on educational
interaction and civic engagement”. Pew Internet and American Life
usage of games
report. September 16, 2008. Available at
[8] Fuchs, R. (1998). Little Engines that Did: Case Histories from the
http://www.pewInternet.org/pdfs/PIP_Teens_Games_and_Civics_Repor
Global Telecentre Movement. Ottawa: IDRC.
t_FINAL.pdf
[9] Garris , R. , & Ahlers, R. (2003). Games, motivation, and learning: A
[34] Lenhart, Amanda; Jones, Sydney; Rankin Macgill, Alexandra. (2008).
research and practice model
“Adults and Video Games” Pew Internet and American Life Project
55
Data Memo. December 7, 2008. Available at

http://www.pewInternet.org/pdfs/PIP_Adult_gaming_memo.pdf.
[35] Bruno Latour. (2005). Reassembling the social: an introduction to
Actor-network theory, Oxford, New York: Oxford UP.
[36] C. Shawn Green andDaphne Bavelier. (2003). “Action video game
modifies visual selective attention.” Nature. Vol 423. pp. 534-537.
56
Content Creation and Dissemination by-and-for

Users in Rural Areas
Sheetal K Agarwal, Arun Kumar, Amit Anil Nanavati, Nitendra Rajput
IBM Research, India Research Lab
4, Block C, ISID Campus, Vasant Kunj, New Delhi, India.
Email: sheetaga,kkarun,namit,rnitendra@in.ibm.com
Abstract—83% of the world population does not have access information that caters to a wider section of the population,
to Internet. Therefore there is a need for a simple and affordable thus leaving out the details for specific villages. One of the
interaction technology that can enable easy content creation and common methods of providing local information (such as a
dissemination for this population. In this paper, we present the
design, development and usage pattern of a VoiKiosk system health-related warning) is done through announcements using
that provides a voice-based kiosk solution for people in rural loud speakers in vehicles. Figure 1 shows one such vehicle
areas. This system is accessible by phone and thus meets the being used by a government healthcare official to announce
affordability and low literacy requirements. We present usability an upcoming health camp in the village.
results gathered from usage by more than 900 villagers during
four month of the on-field deployment of the system.
The on-field experiments suggest the importance of locally
created content in their own language for this population. The
system provides interesting insights about the manner in which
this community can create and manage information. Based on the
use of the system in the four months, the VoiKiosk also suggests a
mechanism to enable social networking for the rural population.
I. I NTRODUCTION
The total rural population in 2005 is about 51% worldwide.
If we look at less-developed nations, then this number is
even larger, at 57%, which translates to more than 3 billion
people. For India, the percentage of rural population is even
higher, 71%. Based on the projections in [1], this number is
not likely to change fast in the coming years (the percentage
yearly change from rural to urban is projected to be around
TriC
1 Pyr.mea.IT IBM India Research Lab June 24, 2008
1%). Even so, most of the HCI research has been focussed Fig. 1. Announcement of a government organised health camp.
on the interaction with computers, to which this population
does not have access. Therefore it is extremely interesting to There is a lack of enough information sources that can create
look at sections of this large population base and identify the sufficient locally relevant content such as doctor visit organised
information and communication needs, the means to achieve by the government, change in school timings, delay of the train
these needs, and study the interaction modalities of this service, list of movies in the village community, etc. If there
society. is a system that can enable this community to create their own
A majority of this 71% of rural population in India is content, the information and communication ecosystem can be
illiterate or at least textually-illiterate [2]. Therefore though significantly improved.
they can count currency and make phone calls using numbers, Since not much is studied about the rural population, to
the ability to read or write is very limited. The means of build an information system that can be used by rural popu-
interaction of this society is either completely voice-based or lation to create and access content, it is extremely important
paper-based. Some people (esp. the money lenders, shopkeep- to follow the following design steps:
ers) make paper notes of their business. Most people (such 1) Perform a needs-study to identify the information needs
as farmers) interact using voice, and get information through of this population. It is important to understand what
radio, television and other public announcement methods does a community consider as information. To enable
in villages. News papers continue to be another source of easy access of information, a meaningful categorisation
accessing information. is also needed.
However the common problem across these information 2) Follow a participatory design process to develop a low-
sources for these villagers is the lack of locally relevant con- cost, low-literacy solution. Since a majority of this rural
tent. Most news papers, radio and television programs provide population in India lives below USD 2 per day [3], it
57
is important that the solution is affordable. Moreover, also been done on improving the usability for mobile Internet
the low literacy considerations also need to be taken interaction. In [14], the authors highlight the various usability
in developing the end-user interface of this system. issues by studying the different critical comments by mobile
Participatory design process is important to study a Internet users. Despite all these solutions, the usefulness of
population that has not been exposed to interaction Internet information for people in developing countries is very
modalities beyond paper, radio and TV. limited due to the fact that the content that is relevant for
3) Observe usage pattern of the solution for the rural a villager is not available on the Internet [15]. This paper
population. Since not much research has been performed positions itself in the space of creation and dissemination of
in enabling local content creation, the usage pattern of locally relevant content and is supportive of the statement:
this population will be interesting to observe and can access to relevant information is key to the development of a
provide important insights. society.
In this paper, we follow these three principals to build
B. Alternate Models for Local Information Systems
VoiKiosk – a system for increasing reachability of information
kiosks in developing regions. A VoiKiosk is a voice based Community Radio systems have been studied as a alterna-
service available on the Telecom network [4]. As opposed to tive solution to provide information in rural areas. In [16],
PCs, the phone penetration in rural India has been significant the authors present a study that evaluates the effectiveness
and continues to grow. Also based on field studies and litera- of 10 community radio stations in different rural communi-
ture reviews semi-literate and illiterate people are more com- ties in South Africa and report the popularity of programs.
fortable with speech-based interfaces to access information A sustainability study of the community radio service is
services [5]. Leveraging the increased mobile penetration and presented in [17] for Nepalese radio stations. Extensions to
comfort with speech based interfaces, we present this alternate community radio service to involve audience participation
model to create and host voice-based kiosks (VoiKiosks). have been presented in [18]. Among other things, the success
We identify the information needs of villagers (Section III) of community radio lies in the availability of relevant content
in south of India by visiting several villages and talking for the rural communities.
to villagers and officials of an NGO that operates in these The CAM system [19] provides an interface to capture
villages for more than seven years. Based on the needs, we local data using the mobile phone camera. It has been applied
implement the VoiKiosk system (Section IV) and involve a for microfinance data capture [20] and for procurement and
limited number of users and the NGO (Section V) in the quality control in a rural cooperative [21]. Using camera phone
participatory design phase. We deploy the system live in one to capture paper content is another approach to create local
village and gather important usage statistics by four months content that can be used in several application domains. We
of live field deployment of the system (Section VI). We derive the importance of relevant content presented in the cited
conclude the paper by discussing important insights gained work and use these concepts in this paper.
during the entire process of needs-gathering, participatory
C. Speech and Mobile based Systems for Low-literacy Users
design, prototype testing and usage analysis.
Since textual literacy is a major issue in developing coun-
II. R ELATED W ORK tries, especially in rural areas, several interesting information
Several studies have suggested that rural communities have systems for this community are speech based. The community
very different information and communication needs and radio systems mentioned above are completely based on audio
patterns [6] [7]. Moreover, it has also been studied that information. Studies from a spoken dialog system for rural
technological innovation opportunities in rural areas have not India [22] show that users are able to navigate the dialog
been studied in great depth [8]. In [8], the authors present field system irrespective of their literacy levels. A speech-based
studies suggesting that use of a local stakeholder contributes system for providing health information to low-literate users
significantly to a better design of the technologies for rural has been presented in [23]. This paper reports that there is a
communities. While there are significant design challenges need to educate the users by creating a video to explain and
for the rural community, at the same time, the potential of demonstrate the system for a new user.
technology to improve their lifestyle is huge [9]. This forms Information systems based on low-end mobile devices
the motivation for the work presented in this paper. provide an affordable alternative to PC-based systems. A
system for providing information to farmers using mobiles
A. Information Access and Dissemination has been presented in [24]. The authors in [7] conclude that
In the current world of globalization, access to information exploring mobile social software holds significant potential as
is key to the development of a society [10]. The increasing an information tool in rural areas. The MobilED [25] and the
amount of information on the WWW [11] is a reflection of this AudioWiki [26] are recent solutions that address this space
fact. Not surprisingly, therefore, there have been several efforts by providing information to this population through mobile
to provide this information to people in rural areas. There are devices.
solutions in the network space to provide low-cost Internet The work presented in this section leads us to believe that
access access for rural areas [12], [13]. Significant work has creation and dissemination of locally relevant content through
58
speech interface using mobile devices can be used as an The NGO has a Village Coordination Officer (VCO) for
effective system to provide information needs of the rural every village. A VCO is usually a woman who has had 12
population in developing countries. System level technology years of formal education. The VCO can only understand
development work in this domain has been presented in [27] local language, Telugu. For every 3-5 villages, the NGO has
and [28] where the authors present technologies for content positioned a Nodal Coordination Officer (NCO) to oversee
creation and linking. A solution based on these technolo- activities in these villages. The NCO can understand English
gies [29] proposes to organize the unorganized urban poor and is a degree holder in any subject. The NCO can operate
businesses. In this paper, we extend the concepts of the the computer. Every village has a Gram Vikas Samitee (GVS)
World Wide Telecom Web [30] and apply them for providing that is a committee of 18 members from the village who
information solutions to the rural population in developing have volunteered to work with the NGO on the 18 different
countries. modules.
III. N EEDS S URVEY B. Needs Gathering on the Field

We visited four villages accompanied with the NGO field
In order to understand the information needs of a commu-
staff to understand the information needs of this community.
nity, it is important to know the community for a significantly
The NGO staff provided the development history and details
long time. Selective interviews are often restrictive and may
of the community for these villages. These four villages were
not provide deep insights that can be gained with time. The
identified based on the varying demographics to get a broader
understanding of the working and living methods can either be
perspective. The first village (Vandram) has agriculture as the
learned through an extensive ethnography study or by learning
main source of livelihood and paddy is the main crop. Juvvala
from experience of organizations who have done such study
Palam is one of the larger villages and is geographically
and have a presence in the community [8]. We relied on the
located as the central hub for nearly 4 lakh population in
latter approach. This was one of the reasons for us choose
nearby villages. This village has a large population base that
villages in the Andhra Pradesh state in southern part of India.
deals with renting of transport vehicles and so the associated
A. The NGO Operations in the Field business of mechanics, drivers is also very common. The third
village (Cherukumilli) is primarily aquaculture based and has
Byrraju Foundation [31] is an NGO that has been operating a number of fish ponds where aqua farmers cultivate fishes.
in 198 villages in 6 districts of Andhra Pradesh. The NGO Ibhimavaram is a very wealthy village where farmers have
has performed detailed studies to identify the needs of the large lands and they usually employ labourers to cultivate their
villagers from a much broader perspective. Their operations farms. Except for Cherukumilli, all the other three villages had
cover basic amenities such as sanitation, drinking-water and a computer center (called as Ashwini Center). The chart in
healthcare at one end to advanced services such as distance- Table I summarises the important inputs gathered in the field.
education, remote-healthcare and rural BPOs. Following are Each meeting in a village was with the GVS committee
the main activities of the NGO in these villages: head, the NCO, the VCO and some members of the GVS
• Education: The NGO adopt one primary and one sec- committee of that village, all in one room. Most villagers share
ondary government school in every village. They provide a phone within their family. Thus the cell-phone penetration
training to teachers, conducts distance education courses, rates are more than 50% considering the number of families
and conduct health check-ups for students. that have a cell-phone. Ibhimavaram has 100% cell-phone
• Healthcare: There is a health center for every village penetration. Privacy of contacts or calls is not a concern in
where a doctor visits for 2 hours daily. The Village this open community. This culture enables easy sharing of
Coordination Officer (explained later) acts as an assistant cell-phones between family and friends. Almost all villagers
to the doctor. There is an ambulance service for all the mentioned that the mode of interaction with the cell-phone
villages. device is voice. People do not use text messaging, even though
• Sanitation: The NGO has facilitated garbage collection this is relatively cheaper. All cell-phones have Roman letter
system for all villages. The garbage collector generates inscriptions and there is no local language support either on the
and sells manure from the garbage to make the system keypad or on screen. All cell-phone users are number-literate
self-sustainable. Drinking water plants have been set in and can recognize English language digits. However very few
some villages, they provide door-delivery of water at 10 can understand the roman script for characters.
paise per liter (USD x per gallon). The cell-phone signal is sufficiently strong in all remote
• Ashwini Centers: 20 villages have computer centers. areas. Although there are lot of electricity blackouts, but these
These are used to provide distance education to students intervals are short in duration, thus charging of cell-phones is
at school times. Later they are used to provide painting not an issue. There are two major service providers (Airtel,
and embroidery classes to women. BSNL), however there are about 5 different service providers
• Agriculture/Aquaculture Advice: Experts in distant loca- available.
tions provide personal advice to agri/aqua farmers based Most officials and village committee members mentioned
on the crop/fish conditions. that people will be interested in knowing several schedule
59
TABLE I
S UMMARY OF INFORMATION NEEDS IN FOUR VILLAGES .
Village Population Families Mobiles Main DTMF Will pay? Information Required Remarks
Oper-
ators
Vandaram 2292 450 150 Airtel/ Depends Not Program schedule of Ashwini, In- Yes for
BSNL/ on initially. formation categorised on Modules, pilot
Idea/ value Pesticides, Medicines, Aqua infor-
Voda- of mation, sports updates, railway en-
fone infor- quiry.
mation.
Juvvala Palam 3800 850 600 Airtel/ Voice is same as Community news, agriculture info, Yes for
Tata/ better above NO to train, no to Electricity, doc- Pilot
BSNL/ tor visit, Teleconferencing infor-
Voda- mation announcement, panchayat
fone meetings, ankur channel program
schedule. Hub village for 4 lakh
population in nearby villages.
Cherukumilli 4047 1000 500 Airtel/ Voice After information about village, Pan-
Idea realising chayat meetings, GVS meetings
the value,
they
may. Not
initially.
Ibhimvaram 4600 3000 Idea/ Voice Yes Tractor renting, electricians, etc. GVS
Airtel panchayat meeting, national fish very in-
rates, V-agri, V-agri, v-agri terested
for
pilot.
related information such as distance education class schedule, utes, sports achievements, etc.
village committee meting schedule and minutes. Other infor- It was clear from the on-field interactions that a phone-
mation needs were based on agriculture information, commu- based information portal that enables interaction through voice
nity news and crop rates. For the Juvvala Palam village, people will be ideal from the end-user perspective. Such a system
were interested in knowing about electricians, mechanics and will ensure that the end-users will not need to learn any new
different rental options for farming equipments. Except for interaction modality nor a new device. Given the significant
Ibhimavaram, other villagers mentioned that the villagers may cell-phone penetration, a cell-phone based system can be used
not pay for the service initially, but once they can see value to fulfill the information needs of these villages.
from the information, they have the capacity to pay for the
information. IV. P ROTOTYPE IMPLEMENTATION AND TESTING
Based on these meetings, we identified the following cate- Of the 10 different categories of information identified in the
gories for which the information is required by villagers: previous section, we built a prototype of the VoiKiosk system
for four main categories. A VoiKiosk is a specific instance of
• V-Agri: Provide agriculture and aquaculture advice to a VoiceSite that can be used by villagers to create and access
farmers on cell-phones. locally relevant content. A VoiceSite can be thought of as a
• Job work: items and type required by merchants to that parallel to a Website, but which can be accessed by dialing
villagers can get business. a phone number and information can be listened rather than
• Availability of transport. being read or seen. Creation of a VoiceSite is made easy by
• Health: Mostly static info such as doctor timings, date the VoiGen system [27] through which anyone can call up the
for the eye camp, student camp VoiGen system and interact with it through voice. This can
• Aqua prices: Traders can upload prices of fish, prawns at enable any illiterate person to create her VoiceSite. Such a
which they will purchase. system enables easy local-content creation. All information in
• People: Static information about electricians, mechanics, the VoiceSite is stored as audio messages that are recorded by
carpenters of that area with phone numbers making a phone call to the system.
• Entertainment: Name of movies in the nearby theater. Since a VoiKiosk is a VoiceSite for the entire village,
• Ashwini program schedule different type of users can update content in different sections.
• General info about the village: Population, availability of We explain the type of users and their interaction with the
health center, schools, famous local food, etc. VoiKiosk in the next section. The kiosk operator (typically a
• Community news: GVS meeting timings, meeting min- NCO or a VCO) navigates through the VoiKiosk application
60
Create: Operator
to configure it to offer various services for the village. Figure Welcome
Message
Listen: All
4 shows a sample operational scenario for the VoiKiosk.
Advertisement
Health
Message
Health
VoiKiosk Message Agri/Aqua-culture
Ashwini Advertisement
Health Advice Ad
Center
Message
Kiosk-operator uploading VoiGen Health
Schedule
Message Ad
information on VoiKiosk
Create: Operator Operator Expert All

Listen: All All Farmer All
VoiGen allows creation and
Villagers calling the VoiKiosk configuration of VoiKiosk Fig. 3. Information on VoiKiosk: categories and access
for information
Fig. 2. VoiKiosk Usage Scenario

As shown in Figure 3, the information in these four
categories can be modified by different kind of users. All
A VoiKiosk act as information and service portal for a information is accessible to everyone, except for the expert
village. It can be a central point of access for a community advice, which is specific to a particular farmer. We discuss
where information relevant to the community can be posted about the different types of users in more detail in the next
and accessed directly by the users themselves. This solution section.
doesn’t rely on Internet connectivity which is most often not The call-flow for a specific interaction of the Kiosk operator
available in the rural areas and most importantly, it allows with the VoiKiosk is shown in Figure 4. The Kiosk operator
end-users to directly interact with the services, thus removing can choose to modify the welcome message for the VoiKiosk
the dependence on the kiosk operator. system, or modify information in any of the four different
Following are the four main categories of information categories. Within a category, the operator is allowed to create
available on the deployed VoiKiosk system for the village: a new information message, delete any existing message or re-
• V-Agri: Farmers use this service to consult agriculture record an existing message.
experts regarding their crop related problems. Currently a
Kiosk
picture of the crop is taken and sent to an expert, who then Operator
sends a reply back to the farmer through the foundation.
The turn-around time for this process is 24 hours. With
Record welcome
VoiKiosk, the expert will be able to post his advice for the message
farmer on the VoiKiosk, reducing the turn-around time to
4 hours. VoiKiosk identifies farmers based on their caller Edit
IDs. Edit to edit the welcome message
Services to configure services
• Health Information: Information related to different
health advisory and health camps is posted in the
VoiKiosk. The schedule of the doctor visit to the health Select a service to add:
Health, Agriculture,
center is also posted on the VoiKiosk. The VCO can Weather, Transport
change the message if there is a change in the doctor Agriculture
visit timing.
Record service
• Ashwini Center Schedule: Information regarding new Back
description
programs, schedule of daily classes or changes in the
schedule are advertised by word of mouth, local news- Add to add information, Edit to edit the
papers or posters that are pasted at various spots in the service description, Configure to configure the
service, Back to go to service menu
village. Often people go to the the ashwini centers to get
this information in person. The VoiKiosk has a Ashwini Configure
Center section where a kiosk operator can post the latest Add an agri expert
Browse registered experts
news related to the distance education program. Back to go to service menu
• Professional Services: In this section users will be able Add Browse

to record their personal advertisements. In the current
practice, micro-businesses such as mechanics, drivers, Fig. 4. VoiKiosk Usage Scenario
daily wage skilled laborers use word of mouth advertising
to reach out to clients. VoiKiosk provides an opportunity
to increase their client base and increase business op- V. PARTICIPATORY D ESIGN OF VOI K IOSK S YSTEM
portunities for them. Users call the VoiKiosk and record Given the four categories of information, based on the
their advertisement which other villagers can browse by creator and the consumer of this information, there are four
calling the VoiKiosk. types of users of the VoiKiosk system:
61
• The Kiosk Operator has the permission to modify any cognitive load on the users is less if they are told what all they
content on the VoiKiosk. He can call the VoiKiosk from can say to the system.
his cell-phone and the system identifies his caller-id and
provides the right permissions.
• Experts are allowed to post expert comments on agricul-
ture and aquaculture for specific farmers. They dial the
phone number of the farmer for whom they have to post
a message. Experts can not post comments in any other
category.
• Farmers call in the system to listen to any message from
experts. They can listen to information in other categories
too.
• Villagers can post an advertisements, or listen to any other
general information in the VoiKiosk.
A kiosk operator is usually identified for the VoiKiosk. The
kiosk operator has permissions to add different categories, and
also to add and delete information in any category. We have
developed a kiosk template for a group of villages in Andhra
Fig. 5. Initial prototype testing with the farmers.
Pradesh.
A. Participatory Design Modifications From the experts perspective, they wanted that any advice
We developed the prototype system in consultation with the they post to the farmers should also be archived in the NGO
NGO and then took this to the field to get feedback from the IT infrastructure. The NGO believes that this database of
four different type of end-users. Since the Kiosk Operator has specific advices can act as a knowledge resource in the future.
the maximum number of options available on the VoiKiosk, we We created a FTP site so that all the content in the V-Agri
started our testing with 2 Kiosk Operators. The initial design section would be pushed to the NGO server at midnight. The
of the system was such that when the system expected a user duration of any message to be posted on the VoiKiosk was
input, there would be a beep sound at the end of the system restricted to 10 seconds in the initial design of the system.
utterance. However the beep sound was present only when The system automatically detects silence and identifies the
the system used to record user voice and not at the time when end of the utterance. However if there is some noise in the
the system would do a speech recognition on the voice. Since background, then the system is not able to detect the silence
this fact was not known to the users, they did not know when and the message recording continues. This can result in user
to start speaking. Sometimes they would respond too early frustration, both during message recording as well at the time
and sometimes they were too late. This resulted in increased of listening. The other option was to ask the users to press
frustration and so to make sure that they are speaking in the a specific key to explicitly let the system know the end of
interval when the system accepts their input, they kept on utterance. But this would involves a key input, which we
repeating the same utterance, such as “exit, exit, exit”. We wanted to minimize, given the profile of target users. So we
modified the VoiKiosk interface so that after every system had fixed the duration to 10 seconds. However this duration
input, there was a beep sound. The users could then figure was too less to record some of the messages from the experts.
out that they are always supposed to wait for a beep sound Since experts were more advanced users in terms of the device
before they respond. This improved the system performance familiarity, they could use the keypad while talking over the
significantly. phone. So we increased the duration to 30 secs for the Experts.
There was a system prompt that asked the following from They now have to press the # key to indicate the end of the
the users: “Do you want information on Health or Agriculture utterance in case of background noise.
or Ashwini Center Schedule or you want to about the people The participatory design process improved the system sig-
in this village?”. We expected that people will respond by nificantly. With the first version of the system, the Kiosk
saying either of Health, Agriculture, Ashwini Center or People. operator used to take about 10 minutes to reach the Ashwini
However the users used to answer Yes or No. So we had Center part of the VoiKiosk and then update a message. Now it
to change the prompt to “Please say Health to know about takes him (and any novice user) less than 2 minutes to update
health related information, or say Agriculture to ...”. Although a message.
these descriptive prompts are too lengthy and not so natural,
VI. U SABILITY OF VOI K IOSK
but this community preferred to listen to prompts that clearly
tell the user what to say. This was a very good insight from The VoiKiosk was tested with the villagers for a period of
these participatory design experiments. Users in village have four months. We present the usage statistics of the VoiKiosk
more patience and prefer simple descriptive system prompts system by analysing the data through two approaches. The
as compared to more natural sounding prompts. Perhaps the first approach presents the analysis by looking at the data as a
62
TABLE II
whole. This analysis is presented to demonstrate the increasing S UMMARY OF DIFFERENT SECTIONS VISITED BY USERS .
acceptability of the VoiKiosk system in the village. In the
second approach, we present the results on the backdrop of Description Number Percentage
Agriculture 1640 8%
the changes that were made in the system over time. This will Health 2383 11.6%
provide more details about the effects of the usability of the Ashwini Center 1749 8.5%
system with improved interactions. In this section, we will Advertisements 7492 36.5%
present the two results in more detail.
A. User Background and Village Demographics while there used to be about 50 calls per day at the start of
The pilot was deployed for villagers in the Juvvala Palam the pilot, the last week of the pilot has witnessed more than
village in South India. This village has a population of about 300 calls every day. The interesting thing to note is that the
4000 people that form about 850 families. About 70% of these NGO did not actively train the villagers to use the system nor
families have a cell-phone. The male/female ratio is 100/94 in did they advertise this service aggressively. We had started
this village. The main occupation of this village is agriculture, with a group of about 30 users who were initially contacted
especially paddy crop. Transportation is the main business to start using the system and provide us feedback. The caller
for this village. People rent their trucks, tractors and smaller base increased from these 30 users just by word of mouth
vehicles for a few days to people in nearby villages. These are and has reached even the neighbouring villages. We have had
used by farmers in their field and for a specific social function villagers from the neighboring villages calling and creating
such as a wedding. The village a health center where the doctor their advertisements on the VoiKiosk.
visits for two hours every day. There is one bank in the village 700
where about 10% of the families have a bank account. The
average monthly income for a family is roughly about 2000 600
rupees (USD 50). All cellphone users are comfortable dialing 500
Number of calls per day
numbers and talking with other people. However the use of

text-message is very low. As a rough estimate, only about 400
10% of villagers use text-message for communication on cell- 300

phone. This is due to the fact that the local language (Telugu)
is not supported on the device and since only about 20% of 200
people know very limited English, text-messaging is almost

100
non-existent.
The village has a secondary school and a primary school. 0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116
The NGO facilitates English and Math classes through a Days since the start of pilot
distance education program in these schools. The NGO also

facilitates distance education of MS Office, spoken English, Fig. 6. Number of calls per day, starting from the launch of pilot.
embroidery and painting classes for adults. Villagers from the
Juvvala Palam village participated in the pilot. While most calls were during the afternoon and late evening
times, we were surprised to see calls arriving at 2:00 am.
B. Overall System Usage There were a total of 99 calls received between odd hours
The VoiKiosk system was live, 24 hours a day and all seven of 11:00 pm and 5:00 am. Figure 7 shows the calling pattern
days of the week for four months. In these four months, the during different times of the day. This is an aggregate number
system received a total of 20499 calls from 976 villagers. summed over the number of calls in the four months.
We assume that a unique phone number would map to a 2500
specific villager because the trace to users is based on the
caller identification in the VoiKiosk system. Table II shows 2000
the number of calls that went to the different services in the
Number of calls per hour
VoiKiosk. The advertisements section of the VoiKiosk was

1500
accessed maximum number of times. The third column in this
paper shows the percent of the total number of calls that access
1000
this service. Since this number does not add to 100, not every
call accesses a service.
500
Users spent a total of 477 hours on the VoiKiosk system.
The average call time was 83 seconds, with a maximum of 49
0
minutes and minimum of zero seconds. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
24 Hours of day
The number of calls have been increasing since the pilot (starting from 12:00 midnight)
was launched. The graph in Figure 6 hints at the increased
acceptability of the VoiKiosk system over time. On an average, Fig. 7. Hourly breakup of number of calls in the day.
63
There were a total of 2528 advertisements recorded by behind this was that by listening to the advertisements, a
463 different users. Many people recorded their advertisement novice user will know what to speak in her advertisement.
multiple times over the period of four months. Although one When the number of advertisements increased beyond 20, we
person can record only one advertisement, people often call noticed that the users were still patiently listening to them
again either to improve their advertisement, or to record a and then would post their advertisement. This usage pattern
fresh advertisement. One caller had recorded his ad 62 times. provides (at least) one of the following two insights about
Though most callers had recorded it only once. The graph in the users:
Figure 8 shows the different callers and the number of times
they have recorded their advertisements.
This population is more patient and could listen to a long
70 list of information over phone as compared to what has been
observed with the western population’s experience with
60
spoken dialog systems.
Number of advertisement recordings
50 And/OR
The villagers ascertain certain value to posting an
40
advertisement and so are ready to wait for a long time in
30 order to get a chance to post.
20 When the number of advertisements increased beyond about
40, we provided option to jump to the next advertisement
10
by pressing a key on the phone keypad. Perhaps the users
0 were comfortable with navigating the VoiKiosk system by
1 18 35 52 69 86 103 120 137 154 171 188 205 222 239 256 273 290 307 324 341 358 375 392 409 426 443 460
Unique callers
then, and therefore they were able to learn the keypad
navigation. It should be noted at this point that during the
Fig. 8. Advertisement update frequency of callers. needs-gathering phase, the users had clearly mentioned that
navigation through keypad will not be easy and so a speech-
input method should be provided to navigate the VoiKiosk.
C. Improvements with Changes in Interface
Even then, the average time spent per-call was higher (104
During the four months of the live pilot, we made two seconds) after Modification-1, it reduced to 75 seconds when
changes to the system: we provided option to record without users having to listen to
1) Enabled easy navigation of advertisements by providing all advertisements. This is a clear insight that users were able
options to jump to next advertisement through keypad to interact more efficiently with the VoiKiosk system when we
input. provided keypad-based shortcuts for faster navigation.
2) Enabled easy recording of advertisement by allowing
users to record without them having to listen to any VII. D ISCUSSIONS
advertisement. The advertisement section attracted the most attention, and
The initial system was live for 70 days. After the first change, was accessed the most. Initially, we had a couple of users
the system was live for 21 days and then after the second (a mechanic and a truck owner) upload personal (classified-
change, it was live for 30 days. Table III shows the number of type) advertisements to the VoiKiosk. Inspite of these initial
calls, average call time, number of advertisements and unique straightforward examples, the villagers soon found some very
callers in the three situations. interesting and innovative uses for this section:
• An eighth grade student announced himself and his
TABLE III
C HANGE IN V OI K IOSK USAGE WITH SYSTEM CHANGE . I NITIAL S YSTEM mobile number.
IS THE FIRST VERSION THAT WAS DEPLOYED . M ODIFICATION -1 IS THE • A man uploaded his profile for matrimonial purposes (to
SYSTEM THAT ENABLED NAVIGATION OF ADVERTISEMENTS .
M ODIFICATION -2 ALLOWED EASY RECORDING OF ADVERTISEMENTS .
invite marriage proposals). He made several attempts till
he was satisfied that all the details he wanted to capture
System State No. of calls Average time Unique callers were recorded.
Initial System 6239 79 335
Modification-1 4437 104 227 – Another person created a ”response advertisement”
Modification-2 9820 75 412 commenting on the above profile!
• A pair of young parents recorded a message in their
It is interesting to note that over a period of these four child’s voice for the child’s grandparents to hear.
months, the number of advertisements have been increasing. • A politician posted a thank you note after winning a local
Even then, due to improved navigation, the time spent by election.

users has not increased significantly. In the first deployment These examples are very demonstrative of several things:
of the system, a caller had to listen to all advertisements (a) The villagers really understood the technology enough to
before she could post her own advertisement. The reason play with it, and find innovative uses for it, and (b) Even
64
though they may have never heard of social networking on the the system over a period of four months. Users made about
Web, the need for social networking is natural and compelling, 20000 calls to this VoiKiosk system. We present the usage
and (c) they independently discovered the message/response- pattern obtained by different modifications in the VoiKiosk
to-message interface as well as voice mail (as obvious as these over these four months. We provide a detailed description of
might sound to us). how some of the services were used beyond what they were
Thus the advertisements section evolved to a message-board designed for. This leads to an interesting insight into the social
by innovative use of the interface by the villagers. networking applicability of the VoiKiosk system.
As was mentioned in the previous section, in an earlier The increasing use of the VoiKiosk system for the different
incarnation of the VoiKiosk, a villager had to listen to all the purposes leads us to believe that a voice-based mechanism for
advertisements before posting his own, but that did not seem local content creation is a very powerful interaction modality
to act as a deterrent. The villagers appear to be extremely to provide information and communication technologies in
patient in listening to all the information and spending a lot rural areas.
of time with the system, which may be a reflection of culture
or the lack of alternatives, most likely both. ACKNOWLEDGMENT
Of the 20499 calls received on the VoiKiosk system, only We thank Byrraju Foundation for supporting VoiKiosk in
2532 were for creating the content. Creating the content in Andhra Pradesh villages and for providing valuable insights.
the VoiKiosk is either by recording of an advertisement by
a villager or by recording of other messages by the Kiosk R EFERENCES
Operator or the agriculture Expert. Therefore, more than 87%
of the calls were for accesing the content on the VioKiosk [1] “World Urbanization Prospects: The 2007 Revision Population
system while only about 13% calls were made for creating Database,” http://esa.un.org/unup/.
[2] Ministry of Finance, Government of India, “Economic
the content. This indicates the usefulness of the content that Survey 2001-2002,” http://indiabudget.nic.in/es2001-
was being created on the VoiceSite. 02/chapt2002/chap106.pdf.
In the previous section, we have provided more details [3] Population Reference Bureau, “World Population Data Sheet,”
about the advertisements rather than the agriculture, health and no. ISSN 0085-8315, Aug 2006.
ashwini center. This is because the other three categories were [4] S. Agarwal, A. Kumar, A. A. Nanavati, and N. Rajput,
“Voikiosk: Increasing reachability of kiosks in developing re-
mostly information upload from one specific person (either
gions,” in Proc. Intl. Conf. on World Wide Web (WWW), China,
the Kiosk Operator or the Expert), and the access was to the April 2008.
villagers. Therefore these services, although important, were [5] T. S. Parikh and E. D. Lazowska, “Designing an Architecture for
different from the advertisements because the latter had user Delivering Mobile Information Services to the Rural Developing
generated content. By providing the ability for all villagers to World,” in Proc. Intl. Conf. on World Wide Web (WWW), May
2006.
create content, VoiKiosk becomes a much more participatory
[6] S. Seshagiri, A. Sagar, and D. Joshi, “Connecting the bottom
platform for rural areas. of the pyramid: An exploratory case study of india’s rural
Although the statistical analysis proves the acceptance of communication environment,” in Proc. Intl. Conf. on World Wide
the VoiKiosk system by the villagers, the next step is to Web (WWW), China, April 2007.
get usability feedback directly from the users. We plan to [7] B. Kolko, E. Rose, and E. Johnson, “Communication as
talk to a sample of VoiKiosk users and seek their feedback information-seeking: The case for mobile social software for
developing regions,” in Proc. Intl. Conf. on World Wide Web
in terms of the usability of the system and the value of (WWW), Canada, May 2007.
the content. We also plan to provide more critical services [8] D. Ramachandran, M. Kam, J. Chiu, J. Canny, and J. L.
such as daily crop prices to registered farmers and medication Frankel, “Social dynamics of early stage co-design in developing
reminders to patients through the VoiKiosk system. It will also regions,” in Proc. CHI, USA, April-May 2007.
be interesting to observe the usage pattern of a different village [9] C. K. Prahalad, “The Fortune at the Bottom of the Pyramid:
Eradicating Poverty Through Profits,” in Wharton School Pub-
and identify the similarities and differences and question lishing, 2004.
whether they reflect the demographics of that village. [10] E. Brynjolfsson and L. Hitt., “Paradox lost: Firm-level evidence
on the returns to information systems spending.” Management
VIII. C ONCLUSION Science, vol. 42, April 1996.
In this paper, we present a information kiosk system for [11] “Netcraft Web Server Survey,”
http://news.netcraft.com/archives/web server survey.html.
use in rural areas to create and access locally relevant content.
[12] K. Matthee, G. Mweemba, A. Pais, G. V. Stam, and M. Ri-
We identify the information needs in rural areas by talking to jken, “Bringing internet connectivity to rural zambia using a
villagers and the NGO. We develop VoiKiosk – a voice-based collaborative approach,” in Proc. Intl. Conf. on Information and
system that can be accessed by a phone to provide a cost Communication Technologies and Development (WWW), India,
effective solution that is easy to use by the less-literate people December 2007.
in rural areas. We test the initial prototype with the villagers to [13] U. Saif, A. L. Chudhary, S. Butt, N. F. Butt, and G. Murtaza,
“Internet for the developing world: Offline internet access at
evolve a participatory design of the final system. The system modem-speed dialup connections,” in Proc. Intl. Conf. on In-
was then deployed live in a village in South India and we formation and Communication Technologies and Development
captured user statistics for 4 months. More than 900 users used (ICTD), India, December 2007.
65
[14] G. Buchanan, S. Farrant, M. Jones, H. Thimbleby, G. Marsden, [23] J. Sherwani, N. Ali, S. Mirza, A. Fatma, Y. Memon, M. Karim,
and M. Pazzani, “Improving mobile internet usability,” in Proc. R. Tongia, and R. Rosenfeld, “Healthline: Speech-based access
Intl. Conf. on World Wide Web (WWW), Hong Kong, May 2001. to health information by low-literate users,” in Proc. Intl. Conf.
[15] R. Heeks, “Information Systems and Developing Countries: on Information and Communication Technologies and Develop-
Failure, Success, and Local Improvisations,” The Information ment (WWW), India, December 2007.
Society, vol. 18, March 2002. [24] R. Veeraraghavan, N. Yasodhar, and K. Toyama, “Warana
[16] E. R. Megwa, “Bridging the Digital Divide: Community Radio’s Unwired: Replacing PCs with Mobile Phones in a Rural Sug-
Potential for Extending Information and Communication Tech- arcane Cooperative,” in Proc. Intl. Conf. on Information and
nology Benefits to Poor Rural Communities in South Africa,” Communication Technologies and Development (WWW), India,
Howard Journal of Communications, vol. 18, October 2007. December 2007.
[17] F. Hussain and R. Tongia, “Community radio for development [25] “MobilED Initiative,” http://mobiled.uiah.fi/.
in south asia: A sustainability study,” in Proc. Intl. Conf. on [26] T. Leinonen, F. Aucamp, and E. Sari, “Audio Wiki for mobile
Information and Communication Technologies and Development communities: information system for the rest of Us,” in Work-
(WWW), India, December 2007. shop on speech in mobile and pervasive environments, Mobile
[18] S. R. Sterling, J. O’Brien, and J. K. Bennett, “Advancement HCI 06, September 2006.
through Interactive Radio,” in Proc. Intl. Conf. on Information [27] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. A.
and Communication Technologies and Development (WWW), Nanavati, “Voiserv: Creation and delivery of converged services
India, December 2007. through voice for emerging economies,” in WoWMoM’07 Pro-
[19] T. S. Parikh, “Designing an Architecture for Delivering Mobile ceedings of the 2007 International Symposium on a World of
Information Services to the Rural Developing World,” Ph.D. Wireless, Mobile and Multimedia Networks, Finland, June 2007.
Dissertation, University of Washington, 2007. [28] S. Agarwal, D. Chakraborty, A. Kumar, A. A. Nanavati, and
[20] T. S. Parikh, P. Javid, S. K., K. Ghosh, and K. Toyama, “Mobile N. Rajput, “HSTP: Hyperspeech Transfer Protocol,” in ACM
Phones and Paper Documents: Evaluating a New Approach for Hypertext 2007, UK, September 2007.
Capturing Microfinance Data in Rural India,” in Proc. CHI, [29] A. Kumar, N. Rajput, S. Agarwal, D. Chakraborty, and A. A.
Canada, April 2006. Nanavati, “Removed for blind review,” in Proceedings of the
[21] Y. Schwartzman and T. S. Parikh, “Using CAM-equipped Mo- World Wide Web, April 2008.
bile Phones for Procurement and Quality Control at a Rural [30] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. Nana-
Coffee Cooperative,” in MobEA V, Mobile Web in the Developing vati, “Removed for blind review,” in SIGCOMM Workshop on
World, Canada, May 2007. Networked Systems for Developing Regions, Japan, Nov 2007.
[22] M. Plauche and M. Prabaker, “Tamil market: a spoken dialog [31] “Byrraju Foundation,” http://www.byrrajufoundation.org.
system for rural India,” in Proc. CHI, Canada, April 2006.
66
E for Express1: “Seeing” the Indian State

through ICTD
Renee Kuriyan and Isha Ray
electricity bill payment and issuance of government

Abstract— This paper examines how, in their attempts to certificates and forms. The article drew attention to Andhra
liberalize and modernize their operations, Indian states are using Pradesh (AP) in particular as the “hot-bed” of e-governance
ICTD e-governance services to represent themselves in a new since it started e-Seva 2 , a project which was upheld as a
way to their citizens. It reveals how states come to be seen by model of efficient
their citizens through their everyday interactions at ICTD
telecenters. The research finds that, with its e-governance government service delivery [1]. The cartoon accompanying
services, the state is trying to recast its image to fit market- the article, entitled “E for Express” symbolized this new form
friendly principles such as economic efficiency, accountability of service delivery and the accompanying changes in the
and effectiveness. Citizens simultaneously trust the government Indian bureaucracy (Figure 1). It depicted a sleek Indian
as credible and are disillusioned with it as inefficient. Telecenter-
express train carrying passengers on top of its roof and inside
provided e-governance services are partially re-shaping the
boundaries between state, civil society and markets.
its cars, holding laptops and cell phones. The Indian railways
were once emblematic of the country’s bureaucratic and slow-
Index Terms—e-governance, India, state, public-private moving ways. At the same time, often associated with Gandhi
partnerships traveling on their rooftops, the always-overflowing railways
were symbols of accessibility to the Indian masses. The
passengers in this cartoon, however, wearing expensive
I. INTRODUCTION business suits and holding laptops, were a far cry from usual
images of the “common man.” The cartoon also highlighted
“Paying an electricity bill [in India] could easily involve a
the links between modern technologies and an Indian
day’s wait at a government office where a cross official
bureaucracy that was changing from being slow and “neither
would demand a bribe for doing his job. The same was
equitable nor efficient” 3 , to one that is fast-moving, capable,
true for phone bills, water bills, taxes and all other
and accessible to all. It hinted that with the spread of e-
interactions with government. Often the customer would
governance services and state of the art technologies the
first have to go to a bank to get a banker’s draft and then
public’s experience of the government was being transformed.
take it to a queue at the payment office. Even a small firm
would need an employee whose sole task was to pay bills
and deal with other aspects of officialdom. Now all of this
can be done online.” (The Economist, Special report on
technology and government, February 14, 2008)
The Economist report cited here, entitled the “Electronic

Bureaucrat,” highlighted India as a nation whose government
was on the path to technologically “leapfrog the rich world”
by putting its services online. It argued that the Indian Figure 1: E-governance in India, “E for Express” (Source: Economist,
government was able to serve its citizens more effectively February 14, 2008)
through its pioneering efforts in e-governance, with online
This material is based upon work supported by the National Science

2
Foundation under Grant No. 0326582, the Technology and Infrastructure for E-seva was launched in Hyderabad in 1999 with goals of ”looking at
Emerging Regions group, Microsoft Research India, and Intel Research. 'service' from the citizens' point of view” and redefining citizen services using
Renee Kuriyan is a Research Scientist with Intel Research, Beaverton, OR, state-of-the-art technologies. It is a public private partnership between the
USA (phone: 503 329 0688; e-mail: renee.kuriyan.wittemyer@intel.com). Government of AP and private service providers. It provides services such as:
Isha Ray is an Assistant Professor with the University of California, payment of utilities bills, government certificates, licenses, permits,
Berkeley, Energy and Resources Group (e-mail: isharay@berkeley.edu). transportation department services, bus reservations, passport services, and
business services based on agreements with private businesses such as cell
1
The Economist, Special report on technology and government, February phone providers and banks (http://esevaonline.com/).
3
14 2008 The Economist, Special report on technology and government, 2008
67
This paper examines the implied hypothesis that, through 3 years from 2004 to 2006. We examined four projects in the
Information and Communication Technology (ICT) projects, states of Kerala and Andhra Pradesh, all of which provided e-
and in particular those that provide e-governance services, the governance services to citizens via entrepreneur-mediated
nature and boundaries of the Indian state are being redefined. telecenters. These were Akshaya in Kerala, and three separate
We do this through an analysis of telecenters that provide e- projects in AP called Rural Eseva, Rural Service Delivery
governance (and other) services in two states in India: Kerala Points (RSDP), and Rajiv Internet Village Centers (Rajiv).
and AP. Both states have been enthusiastic adopters of such The telecenter projects offered a glimpse into the realities and
telecenters, and both states have formed public-private perceptions of good governance principles and government
partnerships (PPPs) in order to promote their e-governance policies on ICTD. The telecenters in AP and Kerala focused
agendas. We ask: Do citizens “see the state” [2] differently in on e-governance services that 1) provided entitlements
light of governance and development services provided (including certificates, licenses), 2) provided information
through these entrepreneur-mediated telecenters? (sectoral, agricultural, or health), 3) provided redress for
grievances, and 4) enabled government bill payments
The term “see the state” is a twist on the pioneering work of
(electricity, taxes, utilities). In Kerala the telecenter project
James Scott, who, in Seeing Like a State (1998), examines the
offered computer education services as well.
ways in which the state sees (and thus controls) its citizens.
We conducted 31 interviews with state actors within the
He argues that the state often attempts to make its populations
Government of India (GOI) and within the AP and Kerala
“legible” through simplified, yet strategic and technical,
state governments, using a semi-structured interview protocol.
processes (such as mapping, censuses, and various other
These interviews explored each state’s strategies for
standardized modes of representing the population).
delivering e-governance services, and how states perceived
Conversely, Corbridge et al in Seeing the State (2005)
their roles and those of the private sector in the process. We
examine how citizens see government agencies, through an
conducted open ended, key-informant interviews with 16 local
ethnographic analysis of state-citizen interactions in eastern
entrepreneurs in Kerala and AP from the four different
India. We build on Corbridge et al to further explore the
projects. Through these interviews and shorter conversations
shifting dynamics of state-citizen relationships through ICTD
with other entrepreneurs, we explored the role of
telecenters in southern India.
entrepreneurs in delivering governance services and the
position of the entrepreneur with respect to the state. In the
We begin by highlighting some of the important ideas in the telecenters of each project, we observed entrepreneurial
literature on governance, e-governance and their associated behavior, and engaged in informal conversations with the
reforms in India. Then, using two vantage points -- the view users. We interviewed telecenter users and non-users in each
of the state and view of the citizens -- we argue that in both district, using an open-ended interview protocol, exploring
Kerala and AP, the state governments are turning to the perceptions of both the project and the state with respect to
private sector as partners not only for the provision of e- their roles in development. Sixty-five interviews with
governance services, but to associate themselves with a liberal household members were conducted in Kerala, and 70 in AP.
market order and the modernization process more broadly. All interviews were recorded, transcribed and coded using
The state is trying to recast its image to fit market-friendly interview analysis software. Finally, we analyzed literature
and policy documents on good governance, ICTD, and the
principles such as economic efficiency, accountability and
modern Indian state.
effectiveness, all of which embody the “good governance”
agenda in India and elsewhere. These efforts reflect the state’s
III. GOVERNANCE REFORMS AND E-GOVERNANCE IN INDIA
attempt to reposition itself in the context of a liberalizing
economy and to alter the way in which it is perceived by its The expression “governance” is ambiguous and has become
citizens. We find that citizens simultaneously trust the a catch-all term in the study and practice of development. The
government as credible and are disillusioned with it as academic literature on governance is varied, containing
perspectives from institutional economics [3], sociology [4,
inefficient. We find that e-governance through decentralized
5], development studies [6] and governmentality-oriented
entrepreneur-mediated telecenters are partially redefining
theory [7]. From a purely technocratic perspective,
perceptions and expectations of the state, the lay citizenry and
governance can be understood as a set of administrative or
the private sector. This hybrid version of government is managerial tools, which, properly applied, lead to “good
gradually reworking both the way the state sees itself and how governance.” Administrative reforms of Western governments
citizens see the state. in the 1980s and 1990s, implemented under the market-
oriented framework of new public management (NPM,[8, 9]
underscored concepts such as efficiency, open markets,
II. METHODS accountability, customer service and decentralization. NPM
Using a combination of methods such as interviews, models indicated that market-based mechanisms for service
participant observation, and literature and document review, delivery were more competitive and therefore more efficient
we explored the reworking of these relationships through the than traditional government-based provision. The overall
case of ICT for development (ICTD) telecenters. Primary data philosophy was that governments should be encouraged to
collection took place on several trips to India over a period of perform like incentive-driven private businesses and
68
entrepreneurs. Similar principles were later invoked in “good for help with financial contributions and the modernization
governance” reform programs in developing countries. process overall. During our research, several government
Development practitioners in recent years have defined officials talked about the financial constraints of the
good governance principles as transparent policymaking, government and its inability to scale service delivery without
professionalism and accountability in government actions, and private sector participation. For example:
civil society participation. International development agencies
“The government is not having many funds. The private
have sought to operationalize good governance by
sector has funds and if the government supports [them]
restructuring and privatizing state bureaucracies,
they can invest. In Kerala, [the] government is running
decentralizing state power through local government or
out of money. We go for heavy loans from the Asian
nonstate actors, reforming legal systems, and implementing
Development Bank... For the last ten years it’s been like
public private partnerships [10]. The rationale behind these
that… There is a lot of work [to be done], but not much
reforms was that by “combating corruption, nepotism,
money.” (Interview, IT Mission, Kerala)
bureaucracy, and mismanagement … aid would be effectively
used to achieve the objective of reducing poverty” [11: 270].
Critics of good governance argued that it had a neoliberal bias
Officials in Kerala indicated that inadequate technical capacity
towards market-led development due to the influence of
was another reason why the government needed to move
donors. Critics also charged that the words ‘governance’ and
towards a PPP model for service delivery:
“government” were being used interchangeably because:
“There was no capacity with the government. That had to
“Governance appears to be used in place of government as if
be built up and that is not something you can do
“government” was a difficult word to sell in a privatized,
overnight. In e-governance you find a much higher
market-oriented society. Governance is about a ‘reinvented’
acceptance of the private sector as a player. The
form of government, which is better managed” [12: 18].
government acknowledges that they don’t have the
In India, the good governance agenda developed against the
technical or financial capacity in the government and they
background of a long history of concern over state-society
need to look to the private sector”
relationships rooting back to India’s anti-colonial struggles
[2]. Debates about good government with respect to
appropriate power for India’s citizens, corruption, and abuse These constraints and incentives led several Indian states to
of executive powers have been part of the national agenda introduce decentralized e-governance delivery projects
from the 1960s. The reforms of the 1990s, which liberalized through PPPs. E-governance measures aimed to improve
the economy to a great extent, were introduced largely to administrative processes by using ICTs, and to build
address India’s balance of payments problem [2], But a connections to promote socioeconomic activity [16]. In
broader program of administrative and economic reforms was accordance with the principles of good governance, e-
also supposed to tackle political challenges, or what Kohli governance was meant to emulate the private sector’s qualities
called India’s “growing crisis of governability” [13: 23]. By of reliability, transparency, scalability and treating citizens
this time, the centralized state had lost a great deal of more like customers. The mainstream press was generally
legitimacy, and devolution through administrative reforms supportive of the promise of fewer lines, administrative
promised a range of benefits [14]. Prime Ministers from the efficiency and the electronic bureaucrat (Figure 2). For
1990s onwards turned to a mix of deregulation, privatization, example, the Indian Express newspaper in a 2005 article
civil service reform, decentralization and PPPs to address a credited e-governance with bringing citizens to the center of
range of administrative “failures” in India. service delivery:
Decentralization in particular was an integral part of the “E-governance has, in many cases, restored the
governance reforms and was viewed as a way to reduce the choice to the citizen as to the quality and adequacy of
role of the state by fragmenting its authority and making it services he is entitled to expect from public
more responsive and efficient. Local governments were organizations. Citizen-centric governance meant
supposed to have better information on local needs, and were government was for the people and the services were
considered closer to their constituencies and thus more tailored to meet their requirements. 4 ”
politically accountable, than centralized government agencies.
Decentralization would thus “expand service deliveries as
authority goes to those more responsive to user needs”[15:
173]. However, even local government institutions could lack
accountability mechanisms and be vulnerable to corruption or
to being captured by local elites. This frequently frustrated the
goals of equitable public provision of services to the general
population (Ibid).
In addition to decentralization, therefore, the administrative
reforms included PPPs in service delivery. Since Indian states
were often in fiscal trouble, with limited budgets to deliver 44
“Giving to people their new right: The right to good governance”
government services, they also needed private sector partners (http://www.indianexpress.com/india-news/full_story.php?content_id=76910)
69
everyday flows of power, money, commodities and

information. The poor in rural India form their understandings
of the state through financial and legal transactions at local
government offices [2]. They note that a poor person most
directly experiences the state when he or she registers for birth
or death certificates, receives a registration form, or picks up
an entitlement. These are many of the e-governance services
now offered by entrepreneurs in telecenters. With telecenter
entrepreneurs in effect replacing the functions of local
bureaucrats, we examined the ways in which the state is being
“seen” or experienced by different groups of people accessing
e-governance services through ICT telecenters. Our research
indicates that at least some segments of the rural and peri-
urban population have developed new images of the state
through their encounters with telecenter entrepreneurs. They
evaluate these experiences against their previous and often
negative encounters with the local arms of the state.
Figure 2: The Electronic Bureaucrat (Source: Economist, February 14, 2008)
We argue that both Kerala and AP, through changing their
governance initiatives, are deliberately re-working images of
The focus on e-governance led many state governments to
the state versus the private sector. The growing acceptance of
adopt “integrated citizen service portals” or ICTD telecenters.
service delivery through ICTs, the political reality of
Telecenters (or kiosks) were meant to bring government and
economic liberalization policies, and discourses of India as a
other services closer to citizens and businesses through
technological leader in e-governance have all influenced the
partnerships with local entrepreneurs [17]. Telecenter
types of images the state tries to portray to its citizens. We
entrepreneurs had market-based incentives to perform
find that how the state is then perceived by civil society
efficiently; were situated at a village level; and in most cases
depends on a combination of factors -- the ambience of
were members of the communities that they served.
telecenters compared to government offices; how
Consequently, it was felt, they were best suited to make e-
entrepreneurs interact with the public; how the state brands
governance services responsive to the felt needs of the local
the telecenters to citizens (whether it is considered a private or
population.
a government office); and the extent to which citizens both
Despite the Indian states’ embrace of e-governance through trust, and are disillusioned with, the government.
telecenters, there have been a number of critiques of these
Our observations in Kerala and AP support neither the
projects. It has been argued that telecenter services were not
reaching the poor [18], were unsustainable [19, 20], and faced strong proponents of the good governance agenda who
political challenges in actually delivering e-governance advocate a minimized role for the state, or the strong
services [5]. We now examine the influence of this new form opponents who fear that service through PPPs will, in effect,
of governance from the perspectives of the state and of the ‘privatize’ the state. We find that, rather than removing the
citizens that these initiatives are meant to serve. state and supplanting it with the private sector, telecenter
projects under the PPP model have created a space for the
state to construct a better image of itself with respect to its
IV. CONSTRUCTING THE STATE AND CHANGING ITS IMAGES citizens. These telecenters and their electronic delivery of
Our research aims to understand how the state represents services are being used as, and to some extent are becoming,
itself to its citizens and how it comes to be seen by them symbols of responsiveness and of accessibility to all. The
through their everyday interactions at ICT telecenters. Several partnering of the state with private entrepreneurs is a key
works of political anthropology depart from the conventional mechanism through which the state is trying to recast its
treatment of the state as an institution defined by its powers image and be re-imagined by its citizens.
and operations, and instead examine the state as it is
‘imagined’, constructed and conceptualized through the A. Representations of the State in Kerala and AP
experiences of its citizens [21-24]. These works analyze how Throughout India, at e-governance and telecenter
the state becomes “socially effective through particular conferences, and in our discussions with state officials at the
imaginative and symbolic devices”[25: 981]. Gupta (1995) central government and AP and Kerala state government
argues that through ethnographic research on the state, for levels, we found a similar discourse about the state’s need to
example with lower level officials and politicians, it is change the way it treats, and how it was perceived by,
possible to “illuminate the quotidian practices of bureaucrats citizens. There was a general consensus among government
that tell us about the effects of the state” on the lives of officials that the state needed to behave more like the private
ordinary people [24: 376]. sector in service delivery. Officials in Kerala and AP indicated
What quotidian practices “construct” the state? Corbridge et that one way to accomplish the change was to partner with
al. (2005) argue that state-citizen interactions are based on private entrepreneurs who had an economic incentive to
provide good customer service.
70
Entrepreneur-run telecenters represented a new concept of Our conclusion that both AP and Kerala were quite
government service delivery that was fast, efficient, hassle concerned about their public images is consistent with
free, and accessible to the “common man” (Figure 3). These Madon’s (2005) study of image formation in the state of
attributes allowed states to renegotiate their standing with Kerala. She found that, given the historical resistance to and
their citizens, a standing that was clearly low on account of suspicion of the private sector in Kerala, the government was
previous encounters between the state and its citizens. A careful to manage the image of the Akshaya project. Rather
former AP official was clear on the expected differences: than advertising the Akshaya project to the public as a PPP at
first, it initially portrayed the project as social development
“The appearance is totally unlike a government office.
bringing e-literacy to all. After a few years the government
The ambience created is more of a private sector feeling-
began to emphasize the private sector’s role in delivering
it has a uniform ambience built across all the service
services to citizens through state-entrepreneur partnerships.
centers across the state- same design, same color. You
Madon attributed this change to the growing optimism and
have a feeling that you have not walked into a
acceptance of IT in governance reforms and increasing
government office, but you’ve walked into something the
confidence in the private sector in Kerala. The Kerala
private service sector would run. The beauty is that it
government in this case judiciously managed the image of the
starts at 8 in the morning and closes at 8 at night.”
role of public-private partnerships and governance reform.
Despite their acceptance of private sector participation,
state officials in Kerala and AP asserted that the credibility of
the government brand was highly important for the PPP-based
implementation of government services. One official we
interviewed in AP stated,
“The image is that these private telecenters are the
government. If you look at the transactions- there are
millions a month. Probably 99% are government to
citizen transactions… they are branded as government.
Figure 3: Old versus new images associated with government offices and That is why the credibility is much better. Because
services. (Photographs shown in presentation at National e-governance people will think twice if somebody else [who is private]
conference in Kochi (2006) to show old and new forms of government service
delivery) wants to collect your electricity bill and pay taxes. People
will think twice!” (Interview, 2006).
Frequent reference to the telecenters’ appearance or
ambience revealed that state employees were at least as State officials asserted that despite its reputation for poor
concerned with their image as they were with making quality and slow services, some level of government
government services more streamlined. Several officials association with the entrepreneurs created credibility for the
acknowledged that government offices had historically been telecenters. The images of the state were constantly being
very different in appearance than something that the private constructed and renegotiated not only by state and project
sector would run. They recognized that, unlike government officials, but by citizens themselves, as we show in the next
offices in which officials often closed early, were not present section.
at their offices, and were on the whole unaccountable, these
new telecenters had much more responsiveness and flexibility. B. “Seeing the state” through entrepreneurs
State employees explained that with telecenters, they were
Citizens “see the state” through the individuals who
aiming for transparency, speed, and convenience for the
represent it. Because the personal characteristics of key
citizen, and that wanted the government to be seen as more
individuals, such as a pro-poor officer or a corrupt official,
“common man friendly”. Our research showed that the
often appear large in the minds of citizens who have
business-like appearance of the telecenter actually represented
experienced them, the views of government are likely to be
a type of streamlining to users and also to the officials.
fractured [2] The local bureaucrat has traditionally embodied
Both states emphasized the business-like aspect of these
the “state” in India and the state is then “seen” through the
telecenter projects, a feature that is at the core of the good
everyday interactions between ordinary people and this
governance agenda. State actors, especially in AP, insisted
person-cum-state:
that the credibility of the telecenter and e-governance projects
could only be achieved when they were managed by the “The manner in which these officials negotiate the
private sector because of their good services, longer hours, tensions inherent in their location in their daily
lack of resource constraints and flexibility. But both state practices both helps to create certain representations
governments wanted to make sure that the government’s name of the state and powerfully shape assessments of it,
and brand was associated with these “private” telecenters. By thereby affecting its legitimacy.” [24:388]
having the private sector deliver the services but branding the Corbridge et al argue that “encounters with the
centers with the government name, the state had found a developmental state build up a dynamic picture of “it”, both as
concrete way to represent itself as protecting the public good an idealized set of values and practices (the state as it should
and doing so in a business-like manner. work) and also as its flawed but more commonly experienced
71
counterpart (the state as it does work)” (2005: 119). Today, “At government offices, people are not helping me. Even
the local bureaucrat is no longer the exclusive embodiment of when we go to a government office, they won’t help us -
the state. When accessing ICT telecenters, citizens’ everyday people like us. Even if you have a small job, they will not
interactions with the state are now being mediated by small- give any amenities, like ration card. They will have to
scale entrepreneurs as they, and not local officials, provide a give, but they won’t give it. We have to do everything on
host of government services to the populace 5 . These our own. Here [at the telecenter] there is more respect for
entrepreneurs, even though they are not state employees, come people. “
to represent the state at least in part, and also affect its
While poorer citizens seemed concern with the need for
legitimacy.
more respect from the government, middle-class individuals
complained that traditional service delivery methods were
During our research many entrepreneurs emphasized their
slow and inefficient. A middle-income elderly man described
ability to deliver government services through the PPP
going to a government office compared to a private telecenter
strategy better than the government could deliver them.
to pay his electricity bills:
However, they relied on being “seen” as the government,
“At a Government office, a person has to wait 5-6
particularly when they collected electricity bill payments or
hours. A person uses Rural e-seva now. Before they
issued government certificates. Entrepreneurs confirmed what
had to travel for many hours to many offices. [Now]
state officials had claimed, that being thought of as the
instead of 4 hours, I use 10 minutes. We use [Rural]
government or having some government endorsement gives
e-seva for electricity bills and telephone.”
people the confidence to pay their bills at the telecenters. The
Speed and convenience were not the only advantages
association with the government also prevented people from
offered by the telecenters. Household members seemed to
thinking the telecenters were corrupt. An entrepreneur in AP
think that dealing with private providers was easier overall
admitted:
compared to dealing with employees at a government office.
“If people think we are working under private people
In the words of a middle aged man:
they won’t come. I think and tell villagers that we got
“It is easier to deal with these private people than the
a loan from the government to establish this center.
government - if you have 2-3 private centers, definitely
So it is a government center.
the private person will grab people by providing offers,
private institutions, providing amenities. They give you
For the most part, entrepreneurs providing e-government
water. You won’t find water in a government office. They
services to citizens benefited from the association with the
[the private centers] provide good amenities to you.”
government’s name. Entrepreneurs faced challenges, however,
branding their centers as “government” institutions when they In this sense, it seemed that the government had outsourced
provided more than just e-governance services to citizens, the provision not only of services but also of customer
such as computer education courses[19]. Rajalekshmi’s study satisfaction to the private sector.
of the Akshaya project also found that the most important
reason for its acceptance was “the trust that people had in
government as an institution and the fact that this project was Thus we found that citizens had mixed feelings towards the
spearheaded by the government” [17: 29]. Our study in both government with respect to basic governance services.
AP and Kerala corroborated this observation; citizens’ trust in Ordinary citizens trusted the credibility of the government
government institutions allowed these privately-run payment name but were dissatisfied with the quality of government
and other e-governance systems to function. services, and with having to put up with rude and bureaucratic
treatment in an often-corrupt system. Citizens had similarly
divided feelings towards the private sector. Civil society in
C. Citizens’ perspectives
India may be disillusioned with government provided
We now turn to the perspectives of the households. Several services, but it was equally skeptical of the private sector as
of our respondents said that the private sector, meaning the the protector of the poor.
telecenter entrepreneurs, treated all customers with a level of
respect that they did not receive from the government. Our Hansen’s study of the Indian state as a guarantor of order
interviews with households revealed that ordinary citizens, stated that, “It may be well that ordinary Indians are less in
especially the poorer ones, resented the bureaucratic and rude awe of the state than a few decades ago, but it is still regarded
manner in which government officials often treated them. One as indispensable for public order and for recognizing
woman, living below the poverty line, complained: communities, leaders or claims as legitimate.”[26: 37]. The
resistance to private sector participation in service delivery
among some segments of civil society was rooted in the belief
5
Undoubtedly, the range of people’s encounters with the state extends that this sector was solely concerned with profit making and
much beyond their interactions with telecenter entrepreneurs. Nor do we wish was frequently corrupt [27]. Thus, just as the state officials
to imply that telecenter entrepreneurs are the only intermediaries between the
state and its citizens: field-level government employees and NGO and entrepreneurs indicated, the citizens we interviewed
representatives, for example, also play this role. But telecenter entrepreneurs agreed that, despite a reputation for poor services, the
are a novel form of intermediary in that they are simultaneously the state and government brand was nevertheless accountable and credible,
not the state, at least when they are providing government sanctioned e- particularly for the delivery of governance services. These
governance services.
72
divided feelings explain why the state sees the private Figure 4: Signboard outside Rajiv Internet Village Center (2006)
entrepreneurs as indispensable to its new image and why the
entrepreneurs see the state as equally indispensable to their E. Conclusion
ability to provide e-governance services. This paper analyzed how, in their attempts to liberalize and
modernize their operations, Indian states are using ICTD to
D. The hybrid state represent themselves in a new way to their citizens. Equally, it
We found that through these telecenter projects a new form reveals how modernizing states come to be seen by the
of government is being experimented with that attempts to citizens through their everyday interactions at ICTD
combine the accountability of an elected government with the telecenters. We find that e-governance initiatives in Kerala
efficiency and customer service associated with the private and Andhra Pradesh are not only about delivering services
sector. The state’s new vision of itself contrasts sharply with efficiently (“e for express”) but also about updating the state’s
the images that its population might have had from previous image to one that is modern and market-friendly. This goal
encounters at conventional state offices. This hybrid version has partly been achieved via the delivery of government
of government is gradually reworking both the way in which services through telecenters, implemented in partnership with
citizens with access to ICTD telecenters now perceive it and local entrepreneurs. We find that citizens with access to these
also the state’s perception of itself in relation to its citizens. telecenters generally prefer them to conventional government
offices. In particular, the poorer citizens feel they are treated
Scholars of the Indian state have argued that, at the most with some respect at the centers, while middle class citizens
local levels, the boundaries between state and civil society are appreciate the conveniences and amenities that these new
often blurred [24]. Through neighborhood ICT telecenters, centers offer.
the lines between states and civil society appear to merge as
telecenter entrepreneurs hold multiple positions as local As the good governance agenda has taken hold in India,
community members, private actors, and also representatives Kerala and particularly Andhra Pradesh have embraced the
of the state in the delivery of government services. By language and the ideal of the modern state conducting its
negotiating their multiple roles in their daily practices, business in a business-like manner. This embrace has been
entrepreneurs create representations of the state that affect only partial, however. With service delivery via privately-
both its legitimacy and their own. managed telecenters, we do not find that the governments in
The running together of the public and the private realms is Kerala or AP have been privatized or have withdrawn, as the
especially evident where profit-seeking entrepreneurs offer critics of the “good governance” agenda frequently aver (or as
public services in a market setting. When walking into an ICT the proponents of the agenda tend to promise). Rather, they
telecenter, citizens encounter painted signboards that mark the play a critical role in managing and constructing their image
telecenters as both private and public. In the case of the Rajiv to their citizens. Citizens’ simultaneous trust in and
Internet centers, the signboards might advertise the brands of disillusionment with their governments, combined with their
more than 7 different public and private entities (See Figure simultaneous admiration for and suspicion of private
4). The state’s name and authority still define the terrain on enterprise, have kept the states practically and discursively
which private entrepreneurs have to operate. At the same time, important to the success of e-governance.
entrepreneurs create new experiences that alter the way in
which the state is seen by citizens. Everyday interactions with Given that changes in perceptions of the state (for the most
telecenter entrepreneurs contribute to citizens’ sense of how, part in a positive way) is one outcome of these projects, our
and for whom, government operates. Rather than thinking of research suggests that policymakers could explore how, and if,
the state and market as distinct spheres, our research found these changed perceptions can be used in evaluating projects.
that, in these part-public part-private telecenters, the very Given the widespread implementation of telecenters
definitions of the state and the market were mutually throughout India and the government’s commitment to
constituted. implement 100,000 telecenters in the country, our research
also raises the question: Will citizens in other parts of India
encounter the state via telecenters in a similar way as the
citizens of AP and Kerala?
We note that, as of now, most Indians still do not have access

to e-governance, and most government business is still not
conducted electronically. We agree with those who argue that
the transparency and efficiency expected from public-private
partnerships do not always materialize, and with critics of the
rush towards telecenters when the results of such projects
have been mixed. But our research does indicate that, in
Kerala and AP, these public-private telecenters have created a
space for the state to renegotiate its role and image in public
service delivery.
73
[24] A. Gupta, "Blurred Boundaries: The Discourse of Corruption, the

Culture of Politics, and the Imagined State," American Ethnologist,
vol. 22, pp. 375-402, 1995.
[25] J. Ferguson and A. Gupta, "Spatializing states: towards an
V. REFERENCES ethnography of neoliberal governmentality," American
Ethnologist, vol. 29, pp. 981-1002, 2002.
[1] E. Lucas, "The Electronic Bureaucrat A Special Report on [26] T. Hansen, "Governance and myths of state in Mumbai," in The
Technology and Government," The Economist, vol. February Everyday State and Society in Modern India, C. J. Fuller and V.
16,2008, 2008. Benei, Eds. New Delhi: Social Science Press, 2000.
[2] S. Corbridge, R. Wiliams, M. Srivastava, and R. Veron, Seeing the [27] R. Vyas, P. Small, and K. De Riemer, "The private public
State: Governance and Governmentality in India. Cambridge: divide:impact of conflicting perceptions between the private and
Cambridge University Press, 2005. public health care sectors in India," Int Journal Tuberc lung Dis,
[3] O. Williamson, "The New Institutional Economics: Taking Stock, vol. 7, pp. 543-549, 2003.
Looking Ahead," Journal of Economic Literature, vol. 38, pp. 595-
613, 2000.
[4] J. Kooiman, Governing as governance. London: Sage Publications,
2003.
[5] S. Madon, "Governance lessons from the telecenters in Kerala,"
European Journal of Information Systems, vol. 14, pp. 401-416,
2005.
[6] R. Jenkins, "The emergence of the governance agenda:
sovereignty, neoliberal bias, and the politics of international
development," in The Companion to Development Studies, V.
Desai and R. Potter, Eds. NY: Oxford University Press, 2002.
[7] N. Rose, Powers of Freedom: Reframing Political Thought.
Cambridge: Cambridge University Press, 1999.
[8] A. Cordello, "E-government: towards the e-bureaucratic form,"
Journal of Information Technology, vol. 22, 2007.
[9] L. Kaboolian, "The New Public Management: Challenging the
Boundaries of the Management vs.
Administration Debate," Public Administration Review, vol. 58, pp. 189-193,
1998.
[10] S. Batterbury and J. Fernando, "Rescaling Governance and the
Impacts of Political and Environmental Decentralization: An
Introduction," World Development, vol. 34, pp. 1851-1863, 2006.
[11] V. Nanda, "The Good Governance Concept Revisited," ANNALS,
AAPSS, vol. 603 January 2006, pp. 269-283, 2006.
[12] G. Stokker, "Governance as theory: five propositions,"
International Social Science Journal, vol. 50, pp. 17-28, 1998.
[13] A. Kohli, Democracy and Discontent: India's Growing Crisis of
Governability. Cambridge: Cambridge University Press, 1990.
[14] P. Bardhan, "Decentralization of Governance and Development,"
The Journal of Economic Perspectives, vol. 16, pp. 185-205, 2002.
[15] P. Bardhan, "Governance issues in delivery of public services,"
Journal of African Economies, vol. 13, pp. 167-182, 2004.
[16] R. Heeks, "Understanding E-governance for Development,"
Institute for Development Policy and Management, vol. Working
Paper No.11, 2001.
[17] K. Rajalekshmi, "E-governance services through telecenters: The
role of human intermediary and issues of trust," Information
Technologies and International Development, vol. 4, pp. 19-35,
2008.
[18] J. Thomas and G. Parayil, "Bridging the Social and Digital Divides
in Andhra Pradesh and Kerala: A Capabilities Approach,"
Development and Change, vol. 39, pp. 409-435, 2008.
[19] R. Kuriyan, I. Ray, and K. Toyama, "Information and
Communication Technologies for Development: The Bottom of the
Pyramid Model in Practice," The Information Society, vol. 24, pp.
93-104, 2008.
[20] J. Pal, S. Nedevschi, R. Patra, and E. Brewer, "A Multi-
Disciplinary Approach to Studying Internet Kiosk Initiatives: The
Case of Akshaya," Proceedings of the Global e-development
Conference, 2004.
[21] J. Scott, Seeing Like a State: How Certain Schemes to Improve the
Human Condition have Failed. New Haven: Yale Agrarian Studies
Series, 1998.
[22] G. Joseph and D. Nugent, Everyday Forms of State Formation:
Revolution and the Negotiation of Rule in Modern Mexico.
Durham: Duke University Press, 1994.
[23] J. Comaroff, "Reflections on the Colonial State in South Africa and
Elsewhere: Factions, Fragments, Facts, and Fictions," Social
Identities, vol. 4, pp. 321-361, 1998.
74
Evaluating the Accuracy of Data Collection on

Mobile Phones: A Study of Forms, SMS, and Voice
Somani Patnaik, Emma Brunskill and William Thies
Abstract—While mobile phones have found broad application In migrating mobile data collection from PDAs to cell
in reporting health, financial, and environmental data, there has phones, a critical issue is ensuring the accuracy of data entry.
been little study of the possible errors incurred during mobile In the context of healthcare, an errant entry may prevent life-
data collection. This paper provides the first (to our knowledge)
quantitative evaluation of data entry accuracy on mobile phones saving treatments from reaching patients, or may cause the
in a resource-poor setting. Via a study of 13 users in Gujarat, prescription of unnecessary treatment that is costly and dan-
India, we evaluated three user interfaces: 1) electronic forms, gerous. In financial applications, entry errors may jeopardize
containing numeric fields and multiple-choice menus, 2) SMS, the economic standing of communities that are already very
where users enter delimited text messages according to printed poor. Due to the importance of this issue, several researchers
cue cards, and 3) voice, where users call an operator and dictate
the data in real-time. have studied the error rates incurred as PDAs are deployed
Our results indicate error rates (per datum entered) of 4.2% in developing regions. As detailed later (in Table II), the
for electronic forms, 4.8% for SMS, and 0.45% for voice. These error rates are generally less than 2% (i.e., 2 errors per 100
results caused us to migrate our own initiative (a tuberculosis entries) in programs where users received at least an hour
treatment program in rural India) from electronic forms to voice, of training [12], [24], [4]. However, in the context of mobile
in order to avoid errors on critical health data. While our study
has some limitations, including varied backgrounds and training phones, studies of data accuracy are distinctly lacking. The
of participants, it suggests that some care is needed in deploying closest work is by Parikh et al., where a hybrid system of paper
electronic interfaces in resource-poor settings. Further, it raises forms and camera-equipped mobile phones has demonstrated
the possibility of using voice as a low-tech, high-accuracy, and error rates of less than 1% [28]. For standalone data collection
cost-effective interface for mobile data collection. on low-end phones, we are unaware of any previous study
I. I NTRODUCTION with a rigorous evaluation of data accuracy. This research
Mobile devices have shown great promise for improving the opportunity is highlighted in Table I.
efficiency and effectiveness of data collection in resource-poor In this paper, we provide a quantitative evaluation of data
environments. Compared to a traditional process that relies entry accuracy using low-cost mobile phones in a resource-
on paper-and-pencil forms with subsequent transcription to a constrained environment. We evaluate three practical user
computer system, mobile devices offer immediate digitization interfaces for entering data on a mobile phone: electronic
of collected data at the point of survey. This allows for fast forms, SMS, and voice. Electronic forms consist of numeric
and automated data aggregation. It also improves adherence fields and multiple-choice menus, and can be implemented in
to complex or context-dependent questionnaires, as the device Java or a native phone platform. The SMS interface requires
determines which questions should be answered or skipped. users to send a structured SMS messages to a server, with
The benefits of mobile data collection have been demon- logical fields separated by delimiters in the message. The voice
strated mostly in the context of personal digital assistants interface represents a normal telephone call, with a live human
(or PDAs) [31], [10], [8], [2], [32], [12], [24], [9], [16], operator that enters the data into a centralized spreadsheet.
[4], [3], [15]. Given the recent explosion of mobile phones We evaluated these interfaces in a study of 13 health workers
around the world, there is growing excitement in extending and paramedical staff over a month-long period in Gujarat,
the successes achieved on PDAs to a phone-based platform. India. Each participant was trained and evaluated on all of the
While high-end phones provide the same capabilities as PDAs, interfaces. We focus on the collection of health data relevant
low-end phones lack features such as high-resolution displays to tuberculosis (TB), as we anticipate deploying an electronic
and touch-screen capabilities. To empower the full population system in a real TB treatment program. The data in this paper
of nearly 4 billion mobile phone subscribers [26] with the represent only simulated patient interactions.
capabilities of mobile data reporting, it will be important to Our results indicate an error rate of 4.2% for electonic
establish usable interfaces that are portable to inexpensive forms, 4.5% for SMS, and 0.45% for voice. These represent
phones, and there have been a number of recent efforts in the fraction of questions that were answered incorrectly; as
this space (see for example [13], [1], [22], [25], [7]). each patient interaction consisted of eleven questions, the
probability of error somewhere in a patient report is much
Manuscript received September 22, 2008. higher. For both electronic forms and SMS, 10 out of 26
Somani Patnaik and Emma Brunskill are with the Massachusetts Institute reports (38%) contained an error; for voice, only 1 out
of Technology (email: somanip@mit.edu, emma@csail.mit.edu). During this
research, William Thies was affiliated with both the Massachusetts Institute of 20 reports (5%) contained an error (which was due to
of Technology and Microsoft Research India (email: thies@microsoft.com). operator transcription). As detailed in Section VI, error rates
75
PDAs Cell Phones

Malaria monitoring in Gambia [12]
Clinical study in Gabon [24]
Published error rates Tuberculosis records in Peru [4] None?
Sexual behavior surveys in Peru [3]
SATELLIFE [15]
DataDyne EpiSurveyor [31]
Cell-Life in South Africa [13]
EpiHandy [10]
Jiva TeleDoc in India [1, p.42]
Infant health in Tanzania [32]
Other programs Pesinet in Mali [22]
e-IMCI project in Tanzania [8]
Malaria monitoring in Kenya [25]
Respiratory health in Kenya [9]
Voxiva Cell-PREVEN in Peru [7]
Tobacco survey in India [16]
Ca:sh project in India [2]
TABLE I
P REVIOUS WORK IN EVALUATING THE ACCURACY OF MOBILE DATA COLLECTION IN THE DEVELOPING WORLD .
are distinctly higher for health workers than for hospital staff, of each interface (Section IV). We describe the setup of our
though this difference may also be influenced by variations in user study (Section V) and the results obtained (Section VI),
our training environment. and we discuss the implications (Section VII). We conclude
We were surprised and alarmed by these results. In our own in Section VIII.
treatment program, our original intent was to utilize electronic II. R ELATED W ORK
forms. However, we consider it to be an unacceptable risk
As summarized in Table I, there have been several initiatives
that 38% of submitted forms – containing critical health
to apply PDAs and cell phones for mobile data collection in
information – may contain errors. For this reason, we have
the developing world. While a fraction of the studies on PDAs
overhauled our plans and will implement a treatment program
includes an experimental analysis of the error rate incurred, we
using voice rather than forms or SMS. While the cost of a
are unaware of any study which systematically measures the
live operator may be prohibitive in many countries, in India it
accuracy of data entry on a cell phone. This is the principal
proves to be very cost-effective. The increased cost of a human
novelty of our work.
operator is more than compensated by the decreased cost
Lane et al. provides a review of nine randomized controlled
of voice-only handsets, voice-only cellular plans, decreased
trials that compare the effectiveness of PDAs and paper
training time, and decreased literacy requirements for health
forms for data collection [21]. Six of the trials reported entry
workers. We offer a more detailed analysis in Section VII.
accuracy, with varying results: two studies found PDAs to be
While the results of this study have changed our own ap- more accurate than paper [20], [29], three studies found the
proach to implementing mobile data collection, we caution the accuracy to be similar with both methods [17], [23], [36], and
reader in extending the results of the study beyond its original one study found that paper was more accurate [35]. None of
context. In particular, we are focused on the scenario in which the trials were in the context of the developing world (they
users have limited cell phone familiarity and there is limited took place in North America and Europe).
time to perform training. If either of these variables changes, Previous studies of PDA entry accuracy in the developing
it may be possible to implement high-accuracy mobile data world are summarized in Table II. In cases where workers
collection with electronic forms or SMS. Also, while the error received at least an hour of training, error rates are under 2%
rates that we report on mobile phones are 3-8x higher than (i.e., 2 errors per 100 questions). As early as 1991, Forster et.
those previously reported for PDAs, our data are unable to al evaluated the use of PDAs for a malaria morbidity study in
distinguish whether this difference is due to the devices, or the Gambia [12]. Employing secondary-educated workers who
due to other aspects of the study demographics, training, and received five days of training, they report error rates between
evaluation. A future study could address this question directly 0.1-0.6% and argue that the PDAs offer improved accuracy
by evaluating both phones and PDAs in the same context. and efficiency over paper forms. Missinou et al. employed
Despite these limitations, our study is the first (to our PDAs in a clinical study in Gabon, employing four clinicians
knowledge) that evaluates data entry accuracy on mobile who had no prior PDA experience and received 8 hours of
phones. Based on our results, we submit only that electronic training [24]. They report a 1.7% rate of discrepancy between
forms and SMS may need further validation before gaining PDAs and paper forms, and note that clinicians preferred
widespread deployment in accuracy-critical applications, and the PDAs. Blaya et al. found that error rates improved from
that voice may deserve more attention as a high-accuracy and 1.3% (with paper forms) to 0.37% (with PDAs) in reporting
low-cost means of data collection. tuberculosis bacteriology data in Peru1 [4]. The authors also
The rest of this paper is organized as follows. We start by argue that PDAs are cost-effective [5].
reviewing related work on mobile data collection (Section II). 1 Blaya et al. reports errors per form, rather than errors per entry [4]. Via
Then we consider the tradeoffs between electronic forms, personal communication with the author, we determined that there were an
SMS, and voice (Section III) and detail our implementation average of 7.5 entries per form, yielding the error rates quoted here.
76
Application Location PDA Education Level Training Error Rate

Malaria morbidity [12] Gambia Psion Organizer II XP Secondary 5 days 0.1%-0.6%
Clinical study [24] Gabon Palm m500 3 M.D.s, one clinical officer 8 hours 1.7%
Bacteriology data [4] Peru Palm Zire Post-secondary (2-3 years) 16 hours 0.37%
Sexual behavior [3] Peru Palm Zire Secondary or less 2-3 mins 14%
TABLE II
E RROR RATES MEASURED BY PREVIOUS RESEARCHERS IN APPLYING PDA S FOR MOBILE DATA COLLECTION IN THE DEVELOPING WORLD .
Higher error rates have been reported in the case of prior to data entry. Users that lacked prior camera or computer
self-administered surveys, when limited training is possible. experience were trained to a level of comfort within 5 to 15
Bernabe-Ortiz et al. evaluate the use of PDAs for surveys of minutes. A separate study measures error rates of 1% or below
sexual behavior in Peru [3]. To protect patient privacy, the using the CAM system [28]. This represents an interesting and
PDAs were intended for use by actual subjects, rather than by useful design point, especially in cases where paper forms are
health workers. As only some subjects had finished secondary already ingrained into the workflow. We focus on solutions
education, and subjects received only 2-3 minutes of training, that are independent of any paper workflow, and which do not
the authors observed a 14% discrepancy between electronic necessarily require a camera-phone (while Java-phones often
and paper forms. However, the error rate was substantially have cameras, our SMS and voice solutions are suitable to the
lower for subjects who had finished secondary schooling. most inexpensive phones).
Additional programs have applied PDAs for data collection While electronic forms have been widely deployed, there
in the developing world, but have not provided a rigorous are fewer solutions that rely on user-constructed SMS mes-
analysis of entry accuracy. SATELLIFE uses PDAs for dis- sages for mobile data collection. One example is a system
seminating and collecting medical information in numerous from Dimagi, Inc. which monitors water treatment plants in
countries [15]. There are anecdotal reports that the PDAs India [34]. We are unaware of other systems which rely on
improved data quality [19], and the benefits of decreased error a cue card (as we do in our evaluation) for submitting a
rates were estimated on a five-point scale [6]. Users of the structured SMS message to a server.
system have also rated its usability [11]. However, we are un- Others have considered broader issues in the contextual
aware of a quantitative assessment of the error rates incurred. design of user interfaces for data collection in the developing
DataDyne EpiSurveyor [31] has been widely deployed for data world. Examples include interface design for Auxiliary Nurse
collection in Sub-Saharan Africa; while it has been argued that Midwives in India [14] and a methodological framework for
the system is more accurate than paper forms [30], we are evaluating health devices [18]. Our focus is on assessing the
unaware of a controlled study. EpiHandy also provides tools entry accuracy for a range of standard interfaces.
for deploying electronic forms on PDAs and has been deployed
in South Africa, Uganda, and elsewhere [10]. PDAs have III. U SER I NTERFACES
also found application for gathering infant mortality data in Three of the central modes on a cell phone that can be used
Tanzania [32], for pediatric care (as part of the e-IMCI project) to perform data collection are voice, SMS and an electronic
in Tanzania [8], for assessing respiratory health in Kenya [9], forms application. Data collection performed by voice can be
for surveying tobacco use in India [16] and for maternal and further split into systems that link the data collector with a live
child health (as part of the Ca:sh project) in India [2]. These operator, those that connect to an automated interactive voice
studies lack formal evaluations of entry accuracy. response system, and those that allow the user to record a
Cell phones have also found broad application for mobile message. We focus our discussion around live voice operators,
data collection in the developing world. Cell-Life employs SMS and electronic form based systems, and examine some
electronic forms on mobile phones to improve TB and HIV of the strengths and weaknesses of these various approaches.
treatment in South Africa [13], [33]. Electronic forms are We use SMS to refer to data collection systems that involve
also used by Jiva TeleDoc for improving rural healthcare in information entered by a structured text message: in particular
India [1, p.42], and by Pesinet for monitoring infant health we assume that the information is entered by following a small
in Mali [22]. Mobile phones with forms are also being used cue sheet with a flowchart that directs the collector how to
to monitor malaria in Kenya [25]; while PDAs were also enter the data. To our knowledge, using cue cards to guide
piloted, the authors note that phones are more intuitive due data entry by text message has not been done previously.
to worker familiarity. Voxiva’s Cell-PREVEN uses interactive In contrast, electronic forms (particularly on personal digital
voice response and voice recording to monitor adverse events assistants) have been widely used. In this paper, we use the
amongst sex workers in Peru [7]. We are unaware of any term “electronic forms” to denote any external application that
quantitative evaluation of entry accuracy in these projects. can be placed on a phone, and that automatically guides the
To avoid the complexities of navigating electronic forms, user how to enter data, through the use of text, menus or other
the CAM framework offers a hybrid system in which paper tools. In a voice operator interface, the user simply calls a live
forms are used for organization while phones are used for operator, who asks the user a series of questions to elicit the
data entry [27]. Each field on the paper form is annotated information needed. Figure 1 illustrates each interface as used
with a barcode, which is recognized by a camera on the phone in our particular experiment.
77
1 2 3
Electronic Forms Interface
General Strengths
Easy patient identification
Ongoing cost is low (SMS or data plan)
Can store visits when connectivity is poor
a) General Weaknesses
Requires programmable phones
Requires basic literacy skills
Hard to alter survey questions
Hard to enter in free-form notes Electronic form screenshots
Application can be deleted by user 1. Start application
2. Select the patient
Our Results: Accuracy & Efficiency 3. Select "new visit"
We measured 4.2 errors per 100 entries 4. Enter the patient's temperature
The average interaction was 99 seconds 5. Select severity of patient’s cough
4 5
SMS Cue Card

SMS + Cue Card Interface
1. Create a new SMS Message 4. Enter a Space
General Strengths Press Center Button Press *
Select “Messages”
Can be used with any phone
Select “Create Message”
Ongoing cost is low (SMS) Select “New Short Message”
Many workers familiar with SMS 11. Enter the Patient’s Cough
No Cough - Press 1
General Weaknesses Rare Cough - Press 2
2. Switch to Numeric Input Mode Mild Cough - Press 3
Requires basic literacy skills
b) Changing survey requires new cue card Press Menu Button Heavy Cough - Press 4
Hard to enter in free-form notes Select “Entry Mode” Severe Cough - Press 5
Select “Numeric” (with blood)
No confirmed receipt of data delivery
Worker can forget or lose cue card
Quite easy to fake visits (copy old SMS)
3. Enter the ID of the Current Patient
Aamir Khan - Press 1 21. Check Yourself
Our Results: Accuracy & Efficiency Abhishek Bachchan - Press 2 Your finished message should be
We measured 4.5 errors per 100 entries Aishwarya Rai - Press 3 formatted similarly to the following:
The average interaction was 97 seconds … 10 372 62 68 4 1030007
Sample Voice Interaction

Voice Interface
Operator: Hello. What is your name?
Patient Worker (to operator): My name is Lipika.
General Strengths
Can be used with any phone I am calling to record a patient visit.
No literacy required of workers Operator: What patient are you visiting?
Easy to change survey questions Worker (to patient): What is your name?
Easy to add in free-form notes Patient: Pavathi (reading from note
Hard to fake a visit: operator can ask new note sheet)
questions Worker (to operator): Pavathi.
c) Worker Operator: That’s Pavathi, right?
General Weaknesses Worker (to Operator): Yes
Ongoing cost of operator salary (operator records name)
Voice plans often higher cost than SMS Operator: What is her temperature?
Awkward 3-way social interaction Worker (to patient): What is your
temperature?
Our Results: Accuracy & efficiency Patient: 97.1 (reading from note sheet )
We measured 0.45 errors per 100 entries Operator Worker (to operator): 97.1 degrees.
The average interaction was 140 seconds Operator: 97.1 deg.
(operator records temperature) …
Fig. 1. The three user interfaces evaluated in this paper: a) electronic forms, b) SMS + cue card, and c) voice.
78
In general, there are a variety of factors that affect the data. High quality data can perhaps be characterized by two
choice of a data collection interface. These may be loosely simple criteria: whether or not the data is intentionally faked
categorized into operation, effectiveness and cost. Figure 1 by the user, and the accuracy of data that is not intentionally
supplements the below discussion by summarizing some of faked (which is the focus of this paper).
the strengths and weaknesses of each interface. Intentionally faked data can lead to incorrect conclusions
and potentially lead to significant misallocation of resources
A. Operation when interventions are based on false data. There may be an
We use “operation” to refer to factors involved with the incentive to fake data when users are busy and collecting real
general infrastructure of the data collection system. Initially data is time consuming, due to the data recording itself or
there is the investment of time to set up the system, and transportation time to reach the source of the data (such as
then train the workers who will be performing data collection visiting remote patients). Unfortunately in SMS systems it is
on the system. For voice or SMS interfaces, the set up quite easy to fake data, particularly for cell phone savvy users
time for workers is minimal: each worker must simply be that can copy and paste prior SMS messages. Faking electronic
provided with a phone, if he or she does not currently own forms is slightly harder as it requires the user to sequentially
one. However, electronic forms require that the application be fabricate data across an entire form. It requires the most effort
downloaded onto the phone, which requires either an Internet- for users to fake data while speaking on the phone, as the
enabled phone in an area of good connectivity, or specialized operator can always ask a new question to try to ascertain if
development tools and an external computer. the user is fabricating the data.
Training time for each application is an open issue, and is Voice also has the benefit that it is easy for users to convey
one of the factors we investigate further in this study. Worker additional information (not included in the original survey),
education and worker cell phone familiarity are likely to affect whereas it is more challenging to spell out text using the
how easy it is to set up each user with an interface, and train keypad, particularly in other languages which may or may not
them how to use it. We expect that a voice interface requires be supported on a given phone. Voice is also likely to have
the least amount of education and background to get users fewer operational risks: users may accidentally delete the form
equipped to start performing data collection. In particular, a application, or forget their SMS cue card, but since an operator
voice interface does not require that its users be literate. can always call a worker directly, the voice system is fairly
System coverage and reliability are also critical factors to robust. Voice also makes it easy for users to correct previous
ensure good data collection. Voice calls have priority over visits, by simply calling back the operator. This is also easy
SMS, and there is the possibility of lost SMS messages. to do by modifying and resubmitting a saved electronic form.
The delivery mechanism with electronic forms can vary: both However, it is also important to consider the speed of data
GPRS and SMS can be used. GPRS has the advantage that entry, how much the user likes the interface, and the accuracy
there is an acknowledgment of whether the data was sent; of data entry. To our knowledge there are no prior studies
however not all locations have coverage. From the user side, comparing the accuracy and speed of data entry using SMS,
voice appears to be the most reliable and has the most far electronic forms and voice. Since we regard these as some of
reaching coverage; however, this also requires that there exists the most critical factors in choosing an interface, this is a large
a sufficient number of operators so that users can always reach motivation for our current study.
a person when they call. If this is not always possible then
there may be a reliability penalty as users may have to call C. Cost
back later (or wait for the operator to return their call).
In addition to reliability, a good system should enable One of the other important considerations is cost: the most
some degree of flexibility. Despite good initial prototyping, beautiful, user-friendly, accurate interface may still not be
it may sometimes be important to be able to modify the practical if the cost overhead is too high for the particular
data collection interface, fix an error, improve usability, or problem. Costs consist of fixed one time costs as well as
add or remove information to be collected. If users have an ongoing marginal costs.
Internet enabled phone and are always working in areas of For all three interfaces users must have a cell phone. An
high connectivity, then updating an electronic form system is electronic form requires a programmable phone (such as a
quite feasible. However, if this is not the case, then users must Java-enabled phone or Windows phone) but both SMS and
reprogram their phone using the same specialized tools needed voice applications can be used with any phone. The ongoing
for initial set up. SMS is similarly challenging to update since cost for an SMS phone depends on the rate per message
a new cue card must be distributed to direct the user to enter which is typically quite low. An electronic form can send data
the data. In contrast, voice is trivial to update, as the operator using SMS or through a data plan; typically SMS is cheaper
can simply ask a new set of questions. depending on the amount of data that is being collected.
Voice minutes are frequently more expensive than SMS. But
B. Effectiveness most importantly, voice has the ongoing cost of the salary of
In any data collection effort, one of the key considerations the operator, which is an additional overhead not shared by
is the effectiveness of the program at obtaining high quality electronic forms or SMS.
79
IV. U SER I NTERFACE I NSTANTIATION The electronic forms underwent several design iterations,
including gathering feedback from a 3-day session with 22
The prior section discussed some of the general factors
health workers in Bihar, India, prior to this study. Based on
important to consider when designing and selecting a data
feedback from the workers in Bihar, we choose to employ
collection interface. We now discuss the context for our data
hybrid English/Hindi menus for some of the forms, since some
collection effort and the interfaces we evaluated.
medical terms are easier to understand in English, but others
A. Domain context are easier to understand in Hindi. We also changed from using
multi-select lists (with a checkbox per symptom) to using
Soon the authors, along with other collaborators, intend to individual yes/no questions.
conduct a trial that examines whether increased information Figure 1a shows a series of screenshots of the form interface
and monitoring can improve heath outcomes and adherence used for the present study. The Java application can be set
during tuberculosis treatment in Bihar, India. Treatment will up to either relay this information via SMS or GPRS. This
be conducted by having tuberculosis patients regularly visit distinction is important for cost considerations but does not
health workers and receive drugs as part of a directly observed affect the interface testing considered here.
therapy (DOT) strategy. During these visits, health workers
will collect data about their patients and report this information C. SMS implementation
by mobile phone back to a central office. This information For the SMS interface we designed a cue card that instructs
will be aggregated and analyzed to inform doctors and the the worker how to record information about the patient into a
trial manager about which patients may need to be visited, text message; Figure 1b displays a subset of the cue card used.
for example, if a patient is not improving or is experiencing All information is coded numerically; this is done to reduce
adverse side effects. To support this effort we need a user the amount of cell phone familiarity necessary, as well as to
interface that enables fast and accurate data collection. increase the speed of data entry. Participants enter in data as
The data collected during a patient visit will include both prompted by the cue card and then send the text message at
identification and health status information. The worker will the end of the interaction. The final part of the cue card as
enter in information to identify both the worker name (done displayed in Figure 1b shows a sample text message.
only once at the start of treatment, in the case of forms
and SMS) as well as the patient name. In addition, the D. Voice implementation
health worker will record the patient’s current temperature, For the voice interface the worker calls a live operator.
weight and pulse, as well as the presence or absence of The operator asks the worker a series of questions about the
seven symptoms: night sweats, chest pain, loss of appetite, patient’s health, which prompts the worker to ask the patient
nausea, coughing with blood, yellow eyes and fatigue. These that question. This means that workers interact simultaneously
symptoms were chosen based on advice gathered from tuber- with an operator and a patient; we are unaware of previous pro-
culosis health experts. The worker will also record whether the grams that have taken a similar approach. Figure 1c displays
patient’s current cough is absent, rare, mild, heavy or severe a sample interaction. The live operator confirms answers with
with blood. the worker; this adds to the length of each call but is done
The trial intervention is centered around the hypothesis that to increase accuracy. This can be particularly important when
better, more frequent data collected about patients can improve the phone connection is poor or there is background noise.
tuberculosis health outcomes and therefore high quality data
collection is critical. However, even if an interface encourages V. S TUDY M ETHODOLOGY
high quality data, it is still essential that such a data collection The user study took place in the Surat and Bharuch districts
method also be easy to use and affordable in order for such of the Indian state of Gujarat during July and August of 2008.
an intervention to have widespread applicability. Originally
we were planing to use electronic forms for data collection. A. Participants
However, since there appeared to be a dearth of literature As detailed in Table III, the study participants consisted of
in evaluating mobile data collection accuracy, we decided to six community health workers and seven hospital paramedical
evaluate the accuracy, speed and usability of three mobile staff. The community health workers were associated with the
phone interfaces. The results of this evaluation influenced our Dahej public health center; five of the paramedical staff were
choice of an interface for use in the treatment program. at the Reliance Tuberculosis hospital; and the remaining two
paramedical staff were at the dispensary of the Sardar Vallab-
B. Electronic forms implementation hbhai National Institute of Technology. The study participants
We created a Java application which provides a sequence of were recruited through contacts of the first author.
electronic forms that guide the worker to request information Initially, we had hoped to perform the study entirely with
from the patient. The worker identification number is encoded community health workers, as they are often the primary
once into the phone and is included with each recorded agents of remote data collection (including in our upcoming
visit. The worker has to either enter numeric data or make tuberculosis treatment program). However, this turned out to
a selection from a multiple-choice menu to encode symptoms. be infeasible because some community health workers were
80
unable to travel to the Dahej public health center for training six community health workers completed only one interaction,
and testing, and it was not feasible for us to travel to each while others completed two interactions (we did not anticipate
worker’s home. This prompted us to recruit participants from that voice would become a focal point of this study until
two other centers. There were also some logistical challenges halfway through our experiments).
in performing the studies due to adverse weather conditions The lag time between training and testing was exactly
and the bomb blasts occurring in July 2008 in the Surat area. one day for seven of the participants, and ranged between
The education level of the health workers ranged from 10 half a day and two days for the remaining participants. All
to 12 years, while the education of the hospital staff ranged participants received a brief refresher and supervised entry
from 10 years to a B.A. degree. The average age of the study session immediately prior to testing.
participants was 26.4 years (range 19-35). Seven participants
owned a cell phone, four participants had used but did not own VI. R ESULTS
a cell phone, and two participants had never used a cell phone The results of the user study are detailed in Table III. We
previously. Eleven of the participants were native Gujarati present both the accuracy of data entry, as well as the time
speakers and all spoke Hindi. needed to interview patients and report the data.
On average, electronic forms and SMS offered comparable
B. Training error rates of 4.2% and 4.5% per entry, respectively. The
Participants were trained by at least two trainers in small voice interface proved to be approximately 10x more accurate,
groups of at least two. Initially, examples were presented with an error rate of 0.45% per entry. While only one out of
on a whiteboard and participants were instructed to practice thirteen participants performed perfectly on both the forms and
entering in the data on either electronic forms or as an SMS SMS interfaces, twelve out of thirteen participants performed
using the cue card. After this stage, a paper with a set perfectly on voice. A Student’s two-tailed, unpaired t-test
of example patients was handed out, and participants were revealed that voice had a significantly lower error rate than
instructed to practice entering in this data. In the final stage, electronic forms (p < 0.01) and SMS (p < 0.01); no significant
participants were instructed to practice role playing patient– difference was found between the error rates of electronic
worker interactions with each other. forms and SMS (p = 0.84).
Participants received variable amounts of training, ranging It is important to note that our results indicate a bimodal
from 45 minutes to 8 hours, depending on their experience and distribution of error rates: participants 7-13 performed notably
availability. The longer training sessions were not necessarily better than participants 1-6. While there are many compound-
more effective, as they were performed in larger groups. While ing differences between these participants, including the man-
it would have been desirable to achieve more uniform training, ner in which we conducted training, we refer to them by their
this was difficult given the logistics of transportation and occupation in order to simplify the discussion; participants 1-6
worker schedules. Prior to the completion of training, all are health workers while participants 7-13 are hospital staff.
participants had completed at least two perfect interactions As summarized in Table III, health workers exhibited an error
on both electronic forms and SMS, and at least one perfect rate of 7.6% for forms and 6.1% for SMS, while hospital staff
interaction on the live operator mode. exhibited an error rate of 1.3% for forms and 3.2% for SMS.
Throughout the user study, we employed Motorola L6i In addition, the only voice error occurred with health workers.
cell phones for training and testing. This is the cheapest Unfortunately, our data are insufficient to explain the dif-
Java-enabled phone from Motorola (the source of our current ferences observed between these two groups of participants.
development tools) that is available in India; see Appendix On average, the hospital staff were older, more educated, and
A-1 for a cost analysis. All interfaces and related tools (cue more likely to own a cell phone than the health workers. It is
cards, etc.) were presented in Hindi, and the mobile phones plausible to suspect that these factors contributed to the higher
used had dual Hindi menus. accuracy achieved by hospital staff. However, due to logistical
reasons, our training procedure also differed between the two
C. Testing groups: health workers were trained in a large group for 6-8
Participants were tested in pairs, alternating who was being hours, while hospital staff were trained in small groups for
tested on data entry, and who was playing the fake patient for 1-2 hours. Our trainers were also somewhat more experienced
that data point. The order of the interfaces was randomized: when working with hospital staff, as health workers were
for a given participant pairing, the order of voice, SMS, and trained first. We re-iterate, however, that training continued
electronic forms was alternated. For the voice interface, the until all participants were able to complete two perfect trials
first author acted as the operator and was located outside of on forms and SMS, and one perfect trial on voice.
the room testing was being conducted in; however, there was To better understand the error rates observed using each
always an additional person associated with the experiment interface, we tabulate the exact sources of error in Appendix
inside the room at all times with the participants. A-2. We classify errors by their entry type (numeric, multiple-
During testing, each participant performed two complete choice, yes/no). We also inspect whether each error could be
patient–worker interactions (in the role of the worker) for each detected, by a trained eye, using the submitted data only;
of the forms and SMS interfaces. For the voice interface, the in the future, such errors could potentially be flagged or
81
Owns Used Total Accuracy of Entries Time per Interaction

Cell Cell Training (Wrong / Total) (Average)
ID Occupation Education Level Age Phone? Phone? (Hours) Forms SMS Voice Forms SMS Voice
1 Health worker pre-secondary (class 10) 25 X 8 1 / 22 3 / 22 0 / 11 2:00 1:45 3:07
4 Health worker secondary (class 12) 19 8 2 / 22 1 / 22 0 / 11 1:33 1:27 2:34
5 Health worker secondary (class 12) 19 X 6 2 / 22 0 / 22 1 / 11 1:45 1:27 2:12
6 Health worker secondary (class 12) 20 X X 6 2 / 22 2 / 22 0 / 11 1:35 2:10 2:00
7 Hospital staff pre-secondary (class 10) 30 2.5 0 / 22 2 / 22 0 / 22 2:25 1:40 2:05
8 Hospital staff secondary (class 12) 32 X X 2 0 / 22 1 / 22 0 / 22 1:42 1:17 2:35
9 Hospital staff secondary (class 12) 28 X X 0.75 0 / 22 1 / 22 0 / 22 1:30 1:17 1:55
10 Hospital staff post-secondary (B.A.) 35 X X 1.5 1 / 22 0 / 22 0 / 22 1:25 3:15 2:00
11 Hospital staff post-secondary (D. Pharm.) 26 X X 1 0 / 22 0 / 22 0 / 22 1:05 0:55 2:10
12 Hospital staff post-secondary (D. Pharm.) 24 X X 1 0 / 22 1 / 22 0 / 22 1:07 1:25 1:52
13 Hospital staff post-secondary (M.S.W.) 30 X X 0.75 1 / 22 0 / 22 0 / 22 1:10 1:15 3:15
Average (health workers only) 7.6% 6.1% 1.5% 1:50 1:41 2:32
g (hospital
Average ( p y)
staff only) 1.3% 3.2% 0% 1:29 1:35 2:16
Average (across all interactions) 4.2% 4.5% 0.45% 1:39 1:37 2:20
Std. Dev. (across all interactions) 5.9% 6.4% 2.0% 0:28 0:45 0:28
TABLE III
R ESULTS OF THE USER STUDY. A LL PARTICIPANTS WERE EVALUATED ON TWO INTERACTIONS WITH THE FORMS INTERFACE AND TWO INTERACTIONS
WITH THE SMS INTERFACE . T HE COMMUNITY HEALTH WORKERS (1-6) WERE TESTED ON ONE INTERACTION WITH THE VOICE INTERFACE , WHILE THE
PARAMEDIC HOSPITAL STAFF (7-13) WERE TESTED ON TWO INTERACTIONS . AVERAGES AND STANDARD DEVIATIONS ARE SHOWN AT BOTTOM .
automatically fixed using self-correcting forms. Finally, we forms and SMS). One factor that contributed to the slower
tabulate whether each error is potentially dangerous (e.g., entry rates using voice was the cellular coverage in our study
a severe cough reported as a mild cough would prevent a area; the connection between participants and the operator was
physician from delivering needed care). highly unreliable. The audio quality was frequently degraded
Electronic forms witnessed errors in each entry type; only beyond recognition, and calls were occasionally dropped and
three of the twelve errors are evident from the values sub- re-started. While many resource-poor environments have ex-
mitted, while five errors may be dangerous. Surprisingly, cellent cellular coverage (including the area of Bihar that
eight of the errors were due to numeric entry problems on we are planning to target with our treatment program), the
the electronic forms. Two errors were due to a mis-placed weak coverage in our study area nonetheless reflects a realistic
decimal point in the temperature entry; while our interface hazard of voice in some environments.
automatically places the decimal point if needed, the user In addition to quantitative results, we also solicited qual-
failed to enter the right number of digits in the temperature. itative feedback from each participant, asking them to rank
The SMS interface also witnessed errors in each entry type; the interfaces by their order of personal preference. The
out of thirteen errors, eight are detectable and seven may be forms and SMS interfaces were most popular amongst the
serious. Three of the errors could perhaps be averted with a participants, with each receiving six votes as the most popular
revision of the SMS cue card: to indicate the absence of a interface. Only one participant preferred the voice interface
patient cough, many participants entered the code “0” rather to the others. This feedback is indicative of the poor phone
than the desired (though perhaps less intuitive) value of “1”. connections experienced during the trial; many found voice to
Unlike the forms interface, workers sometimes entered the be frustrating due to the bad call quality. We were surprised
wrong patient identity when using SMS. that any participants preferred the SMS interface, given the
The voice interface witnessed only a single error for the relatively cryptic message that is produced in the end; however,
entire duration of the trial. We consulted a videotaped record participants noted that fewer keys are required under SMS
of the interaction in question (we taped one interaction for each than under electronic forms (which requires scrolling and
participant), and found that the error was incurred by the op- selection). We also note that 8 of the 13 participants preferred
erator in translating the participant’s report into a spreadsheet. the interface on which they demonstrated the fastest entry time.
While such transcription errors could indeed occur in practice,
it is encouraging that the participants were not responsible for VII. D ISCUSSION
any errors on the voice interface. In addition to the factors examined in our experiment, cost
While the voice interface offered the lowest error rates, it is a critical variable for selecting a data collection interface.
also led to the longest entry times. Electronic forms and SMS For the purposes of our own decision making with regards to
averaged 1:39 and 1:37 per interaction, respectively, while the selecting an interface for our tuberculosis treatment program,
voice interface required 2:20 on average (1.43x higher than we performed a simple cost analysis. Details are provided in
82
Appendix A-1, but in summary, the expected cost for data worker to a live operator if the patient symptoms entered are
collection for each patient during his/her treatment is US worrisome. We look forward to exploring solutions for han-
$7.89 using electronic forms, US $4.59 using voice, and US dling these different tradeoffs, and considering IVR solutions,
$2.99 using SMS2 . These results show the cost of voice is as part of our future work.
competitive with the cost of the other two interfaces. Though
SMS is slightly cheaper, in order for tracking patient symptom VIII. C ONCLUSION
status to be helpful, it is essential that the reported data be Given the widespread excitement in using mobile phones
close to error-free. This data will be used to guide doctor for collecting and analyzing data in the developing world, it
intervention, and faulty data may lead to unnecessary visits is important to establish that the data entered on these devices
or worse, missed visits when a patient is sick. The voice meets the strict accuracy requirements of health, finance, and
interface had close to perfect accuracy and was significantly other applications. In this study, we provide a quantitative
more accurate than SMS or electronic forms. Voice also allows evaluation of data entry accuracy on mobile phones using
for additional, unscripted information to be easily collected, electronic forms, SMS, and voice interfaces in a resource-poor
and provides a social dimension to the health worker’s job. setting.
We anticipate that this social dimension could potentially lead Our results indicate that, within the context of our study, the
to higher performance and a lower turnover rate amongst error rates for electronic forms (4.2% of entries wrong) and
workers, since talking to an operator is likely to increase the SMS (4.5% of entries wrong) may be too high to deploy these
worker’s feeling of being supported and integrated in a larger solutions in a critical application. In contrast, the accuracy of
project. Voice also allows for verification to be performed the voice interface was an order of magnitude better (0.45%
easily: operators can simply request the worker to verify the of entries wrong), with only a single error observed across all
data entry just given, which can be particularly useful for trials. This result has influenced us to overhaul our plans for
unusual entries. In addition, a voice interface can be replicated an upcoming tuberculosis program in Bihar, India, to switch
very easily in other contexts– no special software or cue cards to a voice-only interface. Employing a voice interface requires
need to be developed, and any cell phones can be used. While the employment of an operator, and may not be cost-effective
voice requires longer entry times for workers, this represents a in all countries. However, in India, the cost of this operator
very small fraction of their overall working day. For all these is more than compensated by the lower cost of voice-only
reasons, we have now decided to use a voice interface for our handsets, voice-only cellular plans, decreased training time,
upcoming tuberculosis treatment program. and decreased literacy requirements on health workers.
Despite the many advantages of voice, there are still several While this study provides an initial data point for the
challenges that must be addressed in practice. In our upcoming accuracy of data collection on mobile phones, further research
treatment program, workers will be actively examining and is needed to distinguish the factors that are responsible for the
collecting data from patients and must report this information errors observed. In the case of electronic forms, we observed
back to an operator. Calling the operator and keeping him error rates that are 3-8x higher than previously measured
on the line as the worker examines the patient may lead to on PDAs. Our data are insufficient to diagnose whether this
a slightly awkward social interaction. Another more general difference is due to the devices themselves (screen resolution,
challenge for voice interfaces is how to handle scenarios in touch screen vs. keypad, etc.) or due to other aspects of the
which a user calls and the operator line is busy. One potential evaluation (worker education, training duration, etc.). A future
solution for these two challenges is to have the worker write study could address this question directly by evaluating PDAs
down the data on paper and then call the operator. This and mobile phones in the same focus group. However, it is
introduces an additional opportunity for transcription errors not our goal in this paper to prescribe the optimum device for
but has the side benefit of creating a paper trail that may mobile data collection. Rather, we aim only to highlight that
be used for later verification. To handle missing calls the there exists at least one context in which electronic forms and
operator could be responsible for calling back workers, or SMS may be too error-prone for large-scale deployment in
workers could leave a message that would be transcribed by an accuracy-critical application. In this same context, there
the operator. is evidence that a low-tech alternative (voice) provides an
An alternative solution to these challenges would be to use accurate and cost-effective solution.
an interactive voice recognition (IVR) system. IVR could also
IX. ACKNOWLEDGMENTS
be useful when there is very frequent data collection or when
each survey questionnaire is very long. Hybrid live-operator- We are very grateful to the many people who graciously
IVR systems are also possible, such as directing the worker facilitated and participated in our user study. We thank the
initially to an IVR system, but automatically transferring the hospital staff at the Sardar Vallabhbhai National Institute of
Technology in Surat, and also at the Reliance TB Hospital in
2 We use Motorola phones for the electronic forms due to our current set Hazira. We thank Joshnaben Godia of the Taluka Development
of development tools. Moving to the cheapest available Java-enabled phone office in Vagra, Gujarat, and especially Suprava Patnaik for all
would decrease the forms cost to $5.39. However, in practice the cost of voice
phones can also be reduced by leveraging existing phones in the community. her help with our study. This work was supported in part by
The cost of voice remains competitive with forms in most practical scenarios. the MIT Public Service Center.
83
R EFERENCES [22] Mali: Mobile service helps bring down infant mortality, 2007.
Balancing Act News, Issue No. 364, http://www.balancingact-
africa.com/news/back/balancing-act 364.html.
[1] Achieving the Health Millennium Development Goals in Asia and the [23] J. S. McBride, R. T. Anderson, and J. L. Bahnson. Using a hand-held
Pacific: Policies and Actions Within Health Systems and Beyond. United computer to collect data in an orthopedic outpatient clinic: A randomized
Nations Publications, Dec. 2007. trial of two survey methods. Medical Care, 37(7):647–51, July 1999.
[2] V. Anantraman, T. Mikkelsen, R. Khilnani, V. S. Kumar, N. R. Machi- [24] M. A. Missinou, C. H. O. Olola, S. Issifou, P.-B. Matsiegui, A. A.
raju, A. Pentland, and L. Ohno-Machado. Handheld computers for rural Adegnika, S. Borrmann, D. Wypij, T. E. Taylor, and P. G. Kremsner.
healthcare: Experiences from research concept to global operations. In Short report: Piloting paperless data entry for clinical research in Africa.
Workshop on Development by Design, 2002. Amer. Journ. of Trop. Medicine and Hygiene, 72(3):301–303, Mar. 2005.
[3] A. Bernabe-Ortiz, W. H. Curioso, M. A. Gonzales, W. Evangelista, J. M. [25] Mobilizing against malaria. Expanding Horizons (by Nokia), pages 12–
Castagnetto, C. P. Carcamo, J. P. Hughes, P. J. Garcia, G. P. Garnett, 13, Feb. 2007.
and K. K. Holmes. Handheld computers for self-administered sensitive [26] Nokia CEO message, 2008. http://www.nokia.com/A4942317.
data collection: A comparative study in Peru. BMC Medical Informatics [27] T. S. Parikh. Using mobile phones for secure, distributed document
and Decision Making, 8:11, 2008. processing in the developing world. Pervasive Computing, 4(2), 2005.
[4] J. Blaya and H. S. F. Fraser. Development, implementation and prelimi- [28] T. S. Parikh, P. Javid, S. K, K. Ghosh, and K. Toyama. Mobile
nary study of a PDA-based bacteriology collection system. Proceedings phones and paper documents: Evaluating a new approach for capturing
of the AMIA Annual Symposium, 2006:4145, 2006. microfinance data in rural India. In CHI, 2006.
[5] J. A. Blaya, W. Gomez, P. Rodriguez, and H. Fraser. Cost and imple- [29] P. Quinn, J. Goka, and H. Richardson. Assessment of an electronic daily
mentation analysis of a personal digital assistant system for laboratory diary in patients with overactive bladder. BJU International, 91(7):647–
data collection. The International Journal of Tuberculosis and Lung 52, May 2003.
Disease, 12(8):921–7, Aug. 2008. [30] J. Selanikio. Advantages of using handheld computers for
[6] Cost effectiveness study report for the PDA data capture and transmis- data collection. DataDyne. http://www.datadyne.org/files/Brief-
sion. SATELLIFE, Sept. 2004. http://www.healthnet.org/coststudy.php. HandheldAdvantages.pdf.
[7] W. H. Curioso, B. T. Karras, P. E. Campos, C. Buenda, K. K. Holmes, [31] J. Selanikio and R. Donna. Datadyne brief. DataDyne.
and A. M. Kimball. Design and implementation of Cell-PREVEN: A http://www.datadyne.org/files/DataDyne brief.pdf.
real-time surveillance system for adverse events using cell phones in [32] K. Shirima, O. Mukasa, J. Schellenberg, F. Manzi, D. John, A. Mushi,
Peru. Proceedings of the AMIA Annual Symposium, 2005:176180, 2005. M. Mrisho, M. Tanner, H. Mshinda, and D. Schellenberg. The use of
[8] B. DeRenzi, N. Lesh, T. Parikh, C. Sims, W. Maokla, M. Chemba, personal digital assistants for data entry at the point of collection in
Y. Hamisi, D. S. Hellenberg, M. Mitchell, and G. Borriello. E-IMCI: a large household survey in southern Tanzania. Emerging Themes in
Improving pediatric health care in low-income countries. In CHI, 2008. Epidemiology, 4(1):5, 2007.
[9] L. Diero, J. Rotich, J. Bii, B. Mamlin, R. Einterz, I. Kalamai, and [33] D. Skinner, U. Rivette, and C. Bloomberg. Evaluation of use of
W. Tierney. A computer-based medical record system and personal cellphones to aid compliance with drug therapy for HIV patients. AIDS
digital assistants to assess and follow patients with respiratory tract in- Care, 19(5):605–607, 2007.
fections visiting a rural Kenyan health centre. BMC Medical Informatics [34] SMS data management. Dimagi, Inc., 2008. http://www.dimagi.com/
and Decision Making, 6(1):21, 2006. content/sms-data-management.html.
[10] T. Engebretsen. Acceptance of information technology by health re- [35] B. Tiplady, G. Crompton, M. Dewar, F. Boellert, S. Matusiewicz,
search projects in low-income countries: Intention to use and acceptance L. Campbell, and D. Brackenridge. The use of electronic diaries in
of using EpiHandy (IUAUE). Master’s thesis, Agder University College, respiratory studies. Drug Information Journal, 31(3):759–764, 1997.
Dec. 2005. [36] I. Walker, C. Sigouin, J. Sek, T. Almonte, J. Carruthers, A. Chan,
[11] Evaluation of the SATELLIFE PDA project, Feb. 2003. Bridges.org. M. Pai, and N. Heddle. Comparing hand-held computers and paper
[12] D. Forster, R. H. Behrens, H. Campbell, and P. Byass. Evaluation of a diaries for haemophilia home therapy: A randomized trial. Haemophilia,
computerized field data collection system for health surveys. Bulletin 10(6):698–704, Nov. 2004.
of the World Health Organization, 69(1):107–11, 1991.
[13] R. Fynn. Remote HIV/AIDS patient monitoring tool using 3G/GPRS A PPENDIX A-1. C OST ANALYSIS
packet-switched mobile technology. In Appropriate Healthcare Tech-
nologies for Developing Countries, pages 129–138, 2006. In our basic cost analysis we first assume that the treatment
[14] S. Grisedale, M. Graves, and A. Grünsteidl. Designing a graphical user pool is 1000 patients. In our treatment program each worker is
interface for healthcare workers in rural India. In CHI, 1997.
[15] T. Groves. SatelLife: Getting relevant information to the developing
responsible for 10 patients, so there is a total of 100 workers.
world. British Medical Journal, 313(7072):1606–9, 1996. Each worker must be equipped with a cell phone. Our current
[16] P. C. Gupta. Survey of sociodemographic characteristics of tobacco use development tools for electronic forms are tied to Motorola,
among 99,598 individuals in Bombay, India using handheld computers. and require a Java-enabled phone. The cheapest such phone in
Tobacco Control, 5(2):114–120, June 1996.
[17] R. N. Jamison, R. H. Gracely, S. A. Raymond, J. G. Levine, B. Marino, India is the Motorola L6i which is 75 US dollars. In contrast,
T. J. Herrmann, M. Daly, D. Fram, and N. P. Katz. Comparative study both the SMS interface and voice interface can be used on
of electronic vs. paper VAS ratings: A randomized, crossover trial using any cell phone, one of the cheapest of which is the Motorola
healthy volunteers. Pain, 99(1-2):341–7, Sept. 2002.
[18] D. R. Kaufman and J. B. Starren. A methodological framework for Motofone F3 ($26). Voice calls are slightly more expensive
evaluating mobile health devices. Proceedings of the AMIA Annual than text messages: a 3-minute voice call, which is longer than
Symposium, 2006:978, 2006. the average time in our experimental study, is about 3 rupees
[19] H. Ladd and N. Sewankambo. Uganda health information network:
Technical report (submitted to the International Development Research
(0.065 US dollars, Airtel carrier). SMS messages using Airtel
Centre). SATELLIFE & Uganda Chartered HealthNet, Dec. 2004. are 1.5 rupees per message (0.0327 US dollars). The average
http://www.healthnet.org/idrcreport.html. call length in our user study is 2 minutes and 20 seconds;
[20] S. O. Lal, F. W. Smith, J. P. Davis, H. Y. Castro, D. W. Smith, D. L. therefore conducting 100 calls would require slightly under 4
Chinkes, and R. E. Barrow. Palm computer demonstrates a fast and
accurate means of burn data collection. The Journal of Burn Care & hours. We therefore anticipate that a 100 call load would be
Rehabilitation, 21(6):559–61; discussion 558, 2000. reasonable for an operator working 8–9 hours per day, in order
[21] S. J. Lane, N. M. Heddle, E. Arnold, and I. Walker. A review of to include a liberal number of breaks. Our program design
randomized controlled trials comparing the effectiveness of hand held
computers with paper methods for data collection. BMC Medical involves each worker visiting each patient to record symptom
Informatics and Decision Making, 6:23, 2006. information every two weeks. At this rate a single operator
84
Interface Fixed Cost Marginal (Ongoing) Cost Total cost

working five days per week could handle the 1000 calls over Forms $7.50 $0.39 $7.89
the two week period. Based on our experience in hiring a Voice $2.60 $1.99 $4.59
qualified operator in Bihar for $100 per month, we choose a SMS $2.60 $0.39 $2.99
conservative estimate of an operator salary of $200 per month. TABLE IV
The length of treatment is six months. A PPROXIMATE COST PER PATIENT INCURRED BY EACH USER INTERFACE
AS PART OF A 6- MONTH TUBERCULOSIS TREATMENT PROGRAM IN I NDIA .
Given the assumptions above, we calculate the total cost per F IXED COSTS COVER THE PHONE , WHILE MARGINAL COSTS COVER
patient over the course of the treatment for each interface. Note TRANSMISSION VIA VOICE OR SMS, AND , WHERE APPLICABLE , THE CALL
that we are only focusing here on the aspects of the interfaces OPERATOR SALARY. H EALTH WORKER SALARIES DO NOT DEPEND ON THE
INTERFACE AND ARE EXCLUDED .
that lead to different costs and we are not considering the
salary of the workers or additional program overhead. The cost
of the phone per patient is simply the cost per phone multiplied This cost analysis assumes that we continue to use the
by the number of workers divided by the number of patients. Motorola L6i Java-enabled phone for the electronic forms
The cost of an operator per patient is the salary of the operator interface. There are some cheaper Java-enabled phones that we
per month ($200), multiplied by the 6 month treatment length, may be able to use in the future, such as the $50 Nokia 2626,
divided by the number of patients, yielding a cost of $1.20 per but this would require us to obtain new development tools.
patient. Workers will upload health information approximately This would change the cost per patient for electronic forms
12 times per patient (once every two weeks). Therefore the to be $5.39. This still means that voice is less expensive than
cost of communication per patient is equal to the cost for forms in terms of cost per patient. Also, the cost of voice could
each data entry (either SMS or a voice call) multiplied by 12. be further reduced by leveraging existing phones belonging to
Table IV displays the cost breakdown per patient. the health workers.
Due to the high cost of phones that can support external While the above analysis is conducted for a specific program
applications, such as Java-enabled phones, voice is cheaper in India, informal data suggests that in some other countries
than electronic forms over a single 1000-patient program, even voice may also be worth considering. For example, the average
given the ongoing cost of an operator salary. SMS is the salary of call center operators in Peru is approximately 150
cheapest since it requires no operator and can be used with US dollars per month. The biggest cost considerations when
any phone. Perhaps most important is that the cost for each comparing interfaces in new locations are likely to be the
interfaces is less than $10, a small sum compared to the total operator salary, the cost of voice calls compared to SMS, and
cost of approximately $90–100 needed to treat a tuberculosis the expected frequency and duration of conversations between
patient in India. workers and the operator.
A PPENDIX A-2. D ETAILED L OG OF A LL DATA E NTRY E RRORS

Error Error Error
Number Interface Mode Entry Type Entry Name Correct Entry Actual Entry Detectable? Dangerous?
1 Forms Multiple-choice Cough "mild" "none" X
2 Forms Multiple-choice Cough "heavy" "mild" X
3 Forms Numeric Temperature 100.3 103.0 X
7 Forms Numeric Temperature 100.8 108.0 X
4 Forms Numeric Temperature 98.5 98
5 Forms Numeric Temperature 98.7 98.687
6 Forms Numeric Temperature 100.2 100.0
8 Forms Numeric Weight 62 empty X
9 Forms Numeric Weight 68 67
10 Forms Numeric Weight 68 93 X
11 Forms Yes/No Fatigue Yes No X
12 Forms Yes/No Nausea No Yes X
13 SMS Multiple-choice Cough "1" (none) "0" (disallowed) X
16 SMS Multiple-choice Cough "3" (mild) "0" (disallowed) X
17 SMS Multiple-choice Cough "5" (severe) missing X X
18 SMS Multiple-choice Patient ID "6" (Akshaye Khanna) "5" (Akshay Kumar) X
19 SMS Multiple-choice Patient ID "7" (Anil Kapoor) "1" (Aamir Khan) X
20 SMS Numeric Temperature 1003 103 X
21 SMS Numeric Weight 54 45 X
24 SMS Yes/No Yellow eyes "6" "2" X
25 SMS Yes/No Fatigue "0000007" "000007" X
26 Voice Numeric Weight 69 59 X
85
FOLKSOMAPS - Towards Community Driven

Intelligent Maps for Developing Regions
Arun Kumar, Dipanjan Chakraborty, Himanshu Chauhan,
Sheetal K. Agarwal, Nitendra Rajput
IBM India Research Lab.,
4, Block C, Institutional Area Phase II,
Vasant Kunj, New Delhi - 110070
Abstract— were offered in the manner done today. Furthermore, lack of

Many services taken for granted in the developed world are stable infrastructure including electricity, internet connectivity
often missing from the developing countries. One typical example and the lower purchase power of people, also complicate the
is that of map systems that form the basis of several location
driven services. Its heavy reliance on content, provides a huge situation further.
barrier towards building such systems. Further, in developing The urban metropolitan cities, however, are beginning to ex-
countries like India, the infrastructure typically has a history of perience such services as the factors affecting semi-urban/rural
unplanned development, leading to unorganized addresses and
absence of standard naming conventions for roads. Detailed map areas are not applicable there. Examples in India include
systems such as online maps have only recently started becoming MapMyIndia.com and MapsOfIndia.com. These websites pro-
available but for major cities. Remote towns and villages remain vide detailed mapping information for metropolitan cities but
out of reach till date. contain only high level content for other areas. Even in
In this paper, we investigate a community-driven approach for metropolitan cities (about 100 million people stay in Indian
creating maps in developing regions - following Web 2.0 princi-
ples, but not entirely relying on the existing Web. Our system, metropolitan cities [2]), the (online) maps often do not contain
dubbed F OLKSOMAPS is an intelligent, community constructed enough detail/content to get driving directions from door-to-
map system, particularly architected with developing regions in door. Secondly, lack of structured addressing conventions and
mind. We present the design of F OLKSOMAPS, followed by an poor road signs makes it difficult to follow the maps4 . So,
implementation of our proof-of-concept system. We present user even people comfortable with maps, often need to ask people
studies aimed at understanding the uptake, usability and utility
of F OLKSOMAPS. The results indicate a strong need for such a on the streets to find their way.
community-generated mapping ecosystem. To overcome the prohibitive cost of developing and main-
taining such map services for semi-urban areas, as well as
I. I NTRODUCTION
to address the limitations of using maps in urban cities, we
As defined in Wikipedia: “A map is a visual representation propose utilizing the collective efforts of the community who
of an area’s symbolic depiction highlighting relationships would be motivated to populate, maintain and access content
between elements of that space such as objects, regions, and for their benefit. In this paper, we present F OLKSOMAPS a
themes. Maps may represent any space, real or imagined, community driven map system that leverages Semantic Web 5
without regard to context or scale” technologies to create and manage a community generated
The map systems in developed countries have advanced knowledge base and makes use of web and voice applica-
to a state where users can view street level information tions [3] to provide access to its services.
in 3D and annotate the maps with their own personalized It is non-trivial to build such a system since several issues
content1 . Such systems include Google Earth2 and MSN’s Live crop up. For instance, due to unplanned, historic development
Maps3 etc. Due to the prohibitive development cost involved, over centuries, cities, towns and villages in developing coun-
offering such systems becomes viable if profitable services tries typically do not have well structured naming of streets,
can be offered on top of the core content base. The existing roads and houses. For example, postmen in villages often need
map systems generate revenue from services such as driving to know the inmates by name in order to reach their houses.
directions, finding local businesses and advertising. This leads to imprecise directions and key landmarks become
In contrast, sparsely populated semi-urban and vast rural very important in specifying locations and directions. We make
areas of developing countries such as India do not have use of such insights among other obtained from a user study
detailed map systems built for most locations. Further, the for the design of our system.
semi-literate, low income, Non-ITsavvy population [1] re-
siding in these areas cannot use such services even if they
1 http://www.wikimapia.org 4 A tiny segment using GPS navigation might get along in the cities, however
2 http://earth.google.com they become handicapped outside urban areas due to lack of maps.
3 http://maps.live.com 5 http://en.wikipedia.org/wiki/Semantic Web
86
II. S URVEY OF C URRENT M ODELS B. Results of Survey with Non-ITsavvy Subjects

In this section, we present results of survey done with end- As expected none of the subjects in Non-ITsavvy category
users who depend upon location/map-based information for used maps to find information about locations. In fact, many
their daily business and/or personal needs. We investigate their were not even aware of the concept of maps. A majority of
current models to identify how they manage and use location them mentioned that they currently rely on asking people on
information. the street to know the location of a shop in an area. They
ask people nearby or ask their friends or colleagues about the
location. The others mostly find a shop on their own. 3 subjects
A. Survey Process
mentioned that the information that they receive from people
We selected the subjects considering their technical back- on the street are often not correct and not accurate, and this
ground, so that we get a good variety. We targeted two leads to frustration. About 87.5% of these subjects rely on
categories of people: (1) ITsavvy: People for whom computers other people to provide them travel directions. However, only
and Internet is part of life (2) Non-ITsavvy: People who do a small section of these felt the need for precise directions
not use computers and Internet (reasons range from them (needed only in congested residential areas).
being less literate to being economically challenged) but use About 75% people preferred the choice of a phone based
low-end cellphones (primarily for communication). We believe interface and were willing to upload content. About 56% also
that such a mixed set would be able to provide us the right mentioned that they would be willing to pay for such a service.
insights to evaluate the need for a community-driven map for Key Insights : Use of maps for Non-ITsavvy segment is nil
developing regions. and they rely primarily on others for travel directions. Most
With the Non-ITsavvy class, we conducted the survey in prefer a voice based interface and many are willing to pay
a face-to-face interview mode, where specific questions were for the call. Also, precise directions are not necessary for this
asked, albeit in a very informal and interactive manner. On user segment.
the other hand, we circulated our questionnaire to the ITsavvy
C. Results of Survey with ITsavvy Subjects
class and requested them to fill it up independently. The
questions were targeted to understand the subject’s current As shown in Table I, a majority of ITsavvy subjects make
model for finding landmarks and directions in the city. There use of maps for finding locations and directions to locations.
were 21 questions in total. At a high level, our aim was to The interesting part is that even with maps, a majority of these
understand the following: people rely on other people for location related information.
This is primarily because the unstructured nature of city layout
1) How do people find out points of interest (ranging from
coupled with broken, missing, faded, hidden (behind posters,
very small mom-and-pop shops to popular landmarks) ?
graffiti) or even inconsistent sign boards make it almost
2) How much do they rely on maps or people on the streets?
impossible to rely entirely on a map for travel directions. Many
Are all their information needs satisfied by maps?
a times, the maps do not contain fine-grained information in
3) How do they provide location information (information
the first place.
about landmarks, directions) to other people ?
2 out of 24 subjects mention that maps help them to find
4) Would they be interested in consuming and producing
exact destination, while 19 mention that they rely on maps,
information for a community-driven map system ?
only for a rough idea of the direction or rely on public
We surveyed a total of 40 subjects, with 24 from the ITsavvy transportation and mostly ask people (if road signs are not
category and 16 from the Non-ITsavvy category. The average enough) for the exact location once they reach nearby their
age of the ITsavvy subjects was 26.37 years, ranging from destination.
21 to 34 years. The average age of the Non-ITsavvy subjects Most subjects mentioned it would be helpful to have a
was 32.68 years, ranging from 21 to 62 years. Most of the phone-based location and direction finding system in addition
Non-ITsavvy subjects either did not have a formal education to the web based interface, and about 79% expressed interest
or primary education (10th standard) and were working in in contributing to the service by uploading content either over
the city (security guards, car mechanics, cab drivers etc). The phone or through a web-based portal.
ITsavvy set consisted of engineers, editors, reporters, business Key Insights : Even though this segment makes heavy use
owners etc. We summarize the key findings in Table I. of maps, they still are forced to rely on other people due to
TABLE I various factors. Many people prefer rough directions in the
C URRENT M ODEL : S UMMARY OF U SER R ESPONSES beginning (as they probably know the city) of their journey,
and want detailed directions only towards the end. People rely
Responses Non-ITsavvy ITsavvy
Use maps 0% 66.6%
a lot on community information, and asking people on the
Rely on people to find shops 68.75% 75% streets for directions and location is a common practice.
Rely on people for directions 87.5% 67%
Want precise directions 35.7% 58.3% III. U NDERSTANDING T ECHNOLOGY R EQUIREMENTS
Will upload content 87.5% 79%
F OLKSOMAPS harnesses user-generated content about lo-
cations and aims to provide map-based services that repre-
87
sent user’s intuitive way of finding locations and directions to guide the person. This came as a surprise to us, as we
in developing regions. We conducted an additional survey were expecting this community to be more reliant on maps.
with the aim of understanding the technological requirements 10 out of 24 subjects usually are able to give exact direction
for F OLKSOMAPS. In this survey, we primarily focused on to a known landmark using relationships as shown in table
understanding the ways in which users express directions and II. 13 subjects mentioned that they guide the person to the
location information. In addition, to provide the subjects with nearest big landmark and then guide the person by phone or
an idea of the system, we created a paper prototype that shows pick the person from the landmark. 14 out of 24 subjects felt
how a user would typically access the services and populate confident that people should be able to follow the way they
content to it. guide. Table II shows the most commonly used relationships
The paper prototype was an audio recording of system to describe proximity of a landmark to another one. Typically,
prompts and user responses. We created audio prototypes for while giving directions, such relations are used to relate a
two scenarios 1) User calls F OLKSOMAPS to find directions to less visible or a less known landmark with a more visible or
a particular location 2) User calls to add information about a popular one.
location These prototypes were created using audio recording Interestingly, 21 out of 24 subjects either use both time
and editing tools and were used to illustrate the concept of and kilometers as a measure of distance and only 3 subjects
F OLKSOMAPS to some users. claimed to use only kilometers to represent distance. 17 out of
We conducted this survey along with the working model 24 participants never use zip codes while 6 use them rarely.
survey with a total of 40 subjects - 16 from Non-ITsavvy and This study, coupled with insights obtained from the survey
24 from ITsavvy community. Here is a sample of the questions of current models, has been useful to design the ontology as
we asked: well as the output of F OLKSOMAPS- differentiating it from
• How do you give directions to people on the road or traditional map-based systems prevalent today.
friends? IV. T HE T ECHNOLOGY
• How do you describe proximity of a landmark to another
one? Based upon insights gained from the surveys we devel-
• How do you describe distance? Kilometers or using time-
oped F OLKSOMAPS – a community generated map system.
to-travel? F OLKSOMAPS is designed to be populated by end users for
their own consumption. This section presents the architecture
A. Non-ITsavvy Community and design of F OLKSOMAPS system while highlighting design
When asked about how they personally give directions to choices that differentiate it from the established notion of map
anyone who asks them, 12 out of 16 subjects said they make systems. They key differences are listed below.
use of landmarks to explain the direction to the destination. • It relies primarily on user generated content rather than
They use names of big roads to describe a location, and use data populated by professionals.
“near to”, “adjacent to”, “opposite to” relations with respect • It strives for spatial integrity in the logical sense and
to visible or popular landmarks to point the destination. 5 does not consider spatial integrity in the physical sense
subjects said they can provide exact directions within one as essential. For instance, information such as “Building
kilometer of the destination. 4 said that they usually give A is located near to circle J after taking first turn on the
directions up to the nearest landmark thereafter which people circle while arriving from location B.” is treated complete
will need to ask again. 9 subjects felt confident about guiding and correct for tracing path from B to A. In other words,
a person to the exact landmark. the direction and distance parameters are not specified
Interestingly, 6 of 16 subjects said that they use time (only) in precise terms. This is because, as evident from the
as metric to measure the distance between any two locations. surveys, the end users are not likely to specify physical
12 out of 16 subjects mentioned that they use either time data while populating geographical landmarks.
or kilometers. 3 subjects mention that they sometimes use • A visual representation is not essential to F OLKSOMAPS
“rough” distance measures in terms of kilometers. One subject which is important considering the fact that a large
did not have the notion of kilometer as a measure at all. segment of users in developing countries do not have
TABLE II access to Internet.
C OMMON R ELATIONSHIPS USED TO EXPRESS RELATIVE LOCATION • F OLKSOMAPS is non-static and intelligent in the sense
that it infers new information from what is entered by
Relationships Non-ITsavvy ITsavvy
A isNear B 93.75% 67% the users.
A adjacent to B 81.25% 45.8% • The user input is not verified by the system and it
A opposite B 75% 75% is possible that pieces of incorrect information in the
knowledgebase may be present at different points of time.
F OLKSOMAPS adopts the Wiki model and allows all users
B. ITsavvy Community to add, edit and remove content freely. From the estab-
Only 2 out of 24 subjects tell people to use maps to guide lished Wikis on the Web we expect that the community
them to their house. The rest either use landmarks on roads would actively remove or edit invalid content and keep
88
the maps up-to-date. However, to limit malicious intent, to query the system either when they are stuck on road
the system places two minor restrictions described in the looking for directions or before starting on a trip and would
next section. be pressed for time. Further, the user set also consists of
people who might be illiterate or semi-literate or not very
A. Conceptual Design ITsavvy. Considering these, we discuss three different modes
We use the notion of a landmark as the basic unit of of interaction that the system needs to support to cater to
representing nodes in F OLKSOMAPS. A location represents different user segments for different tasks.
more coarse grained geographical area such as a village, city, There are three main tasks that a user can perform with
country etc., in addition to also representing a landmark. The the system. First is to find a landmark/location by specifying
core knowledgebase of the system needs to capture few key its name possibly including some related information such
logical characteristics of locations that users are interested in as nearby places or enclosing area. Second, users can ask
specifying and making use of. These include the following: for tracing a path between two locations. Third, users could
• Direction i.e. the positioning of a location relative to add to the knowledgebase by adding information about a
another one. From the surveys we found out that users location/landmark that they know of. In addition, some or all
are comfortable with providing relative information such users may also be given the facility to edit or remove entries
as ‘towards left of’, ‘on the right side of’ etc. instead of from the knowledgebase.
absolute direction in the form of north, east, west, south We also consider three sets of users who would interact with
compass points. the system. On one extreme, the users of F OLKSOMAPS are
• Distance i.e. the measure of amount of space between ITsavvy people who can access it over the Web. F OLKSOMAPS
two locations. This can be represented as numbers along provides a web interface to these users for submitting queries
with units in which the distance is expressed. From the as well as to update its knowledgebase by adding new locations
surveys, we learnt that F OLKSOMAPS needs to consider and related information. On the other extreme, we have users
both time and metric units to represent distance. who are illiterate or semi-literate and cannot afford to have
• Proximity and Reachability i.e. representation of infor- high end devices but can use an ordinary low end phone
mation stating that one location is in close proximity to for voice communication. Studies done earlier [3], [4], [5]
another or is reachable from another respectively. suggest that a voice-based interaction works well for this
• Layer i.e. granularity of geographic area that a location user segment and for them F OLKSOMAPS supports a voice
name represents. It could be a division as big as a whole based interface for querying the system. The third segment
country or as small as a village. The notion of direction of our users lies between the two extremes and consists of
and distance from a location, are interpreted with respect mobile people with low end devices who are familiar with
to the layer that the location represents. In other words, SMS. F OLKSOMAPS allows SMS based querying and location
direction and distance could be viewed as binary operator updates in a constrained form for these users.
over locations of the same level. For instance, ‘is towards
C. System Architecture
left of’ would be appropriate if the location pair being
considered is <Libya, Egypt> or say <South Korea, Figure 1 shows the architecture of the F OLKSOMAPS sys-
Japan> but not if the pair is <Sheraton waikiki hotel, tem. As shown, users can upload content into the knowl-
Mexico> where Sheraton Waikiki hotel is in Honolulu, edgebase through an SMS interface, a web based interface
Hawaii. or through a voice interface. Similarly, the content delivery to
We model the knowledgebase for representing and storing the consumers also happens through these multiple interfaces.
these concepts in two parts. The first one makes use of Web The knowledgebase consists of an ontology and a graph
Ontology Language (OWL) 6 to model the categorical charac- database. An ontology is used as the primary repository of
teristics of a landmark, i.e. direction, proximity, reachability the location information. This is because the user generated
and layer. Use of a Semantic Web language to represent content cannot be expected to be complete. It is essential to
relationships between locations brings in the advantage that be able to infer facts not explicitly populated by users in
the system can reason on those and infer newer relationships order to have a pragmatic map system. The graph portion
not explicitly specified by users of the system. The second part of the database captures additional information that either
makes use of a graph database to represent distance between cannot be expressed appropriately in the ontology or needs
landmarks which is numerical data. The two modules are used to be processed differently. This includes numeric data such
in conjunction to generate answers to queries submitted by as distances between locations.
users to the system. The central block of the figure forms the core of the
runtime system of F OLKSOMAPS, acting as an intermediary
B. User Interaction between the consumers of the service and the knowledgebase.
The user interaction aspect is critical for the success of It consists of a module each corresponding to the tasks listed
F OLKSOMAPS. This is especially true since users would tend above, namely location insertion/removal, location finding and
path finding. It provides a similar interface to the users across
6 http://www.w3.org/TR/owl-ref/ interaction modalities for information upload and retrieval.
89
Not shown in the figure is another concept labeled Space. It
Browser
Location Finding
Voice
Voice
Module
is defined as complementary and disjoint to Location. This be-
Ontology
Ontology
API
Path Finder
SMS Module came necessary since OWL ontologies follow the open world
SMS
G/W
Location
Insertion Module
assumption8 which means that a relation not explicitly asserted
Graph
in the ontology being reasoned upon cannot be concluded to
API
Graph
Web
App
Location DB
Removal Module be false since it may be specified elsewhere.
<owl:ObjectProperty rdf:about="&commgis;partOf">
Fig. 1. The System Architecture <rdf:type rdf:resource="&owl;TransitiveProperty" />
<rdfs:domain rdf:resource="&commgis;Location" />
<rdfs:range rdf:resource="&commgis;Location" />
</owl:ObjectProperty>
In the next section, we provide some details of the design <owl:Class rdf:about="&commgis;Landmark">
<rdfs:subClassOf rdf:resource="&commgis;Location"/>
of F OLKSOMAPS knowledgebase. <rdfs:subClassOf>
<owl:Restriction>
<owl:onProperty rdf:resource="&commgis;partOf"/>
<owl:allValuesFrom rdf:resource="&commgis;SubArea" />
V. K NOWLEDGEBASE D ESIGN <owl:allValuesFrom rdf:resource="&commgis;Area" />
<owl:allValuesFrom rdf:resource="&commgis;Village" />
The F OLKSOMAPS knowledgebase consists of two parts.

</owl:Restriction>
</rdfs:subClassOf>
A graph database and an ontology of locations. The graph

</owl:Class>
database is primarily a graph data structure based representa-

Fig. 3. OWL definitions of landmark class and partOf property
tion of the locations. The locations are represented by nodes
and the edges between two nodes of the graph are labeled
with the distance between the corresponding locations. Given • Relationships
the insights gained from user surveys, precise distances (and Each location can be related to other locations to log-
exact directions) are not key components of a map for our ically represent the geographical relationship that exists
target users. Therefore, Graph DB is an optional component between them in the world. As can be observed from
and we do not discuss it in this paper. the figure, apart from isA relation of all locations with
The other, more important, part that makes F OLKSOMAPS the Location concept, each location concept is related to
intelligent, is the ontology of locations that helps construct one or few location concepts through a partOf relation.
paths and retrieve information that no user may have explicitly This relation helps establish the layering between various
entered. This is what we describe next. geographic locations, as identified in Section IV.
Landmarks lie at the lowest strata of the ontology. Since
A. The Location Ontology the administrative structure of a country remains rela-
tively static and is generally well known, the top part of
Location Country
the ontology can be bootstrapped in advance and users
can be allowed to contribute instances from the lower half
State Union Territory CountryCapital consisting of landmarks and sub areas. Also, some key
landmarks such as historic sites of national importance
Division
and key government offices such as the parliament could
StateCapital District
Bootstrapped also be pre-populated into the ontology.
The ontology supports several relationships in order to be
Pargana
able to specify the four key characteristics of a location
Tehsil City Area Town defined in Section IV. In the interest of space we present
here details of only two key relationships – nearTo and
Legend PostOffice
connectedTo that model the proximity and reachability
PartOf characteristic respectively.
isA SubArea Village User-generated Relation nearTo is a symmetric relation defined between
two locations to express the fact that they are in close
Landmark proximity to each other. This could intuitively mean a
few hundred meters or a kilometer. The logical integrity
Fig. 2. The Folksomaps Ontology Design of nearness can be applicable to locations other than land-
marks. For example, two towns in the same district can
Figure 2 depicts the location ontology that we created. be considered nearTo each other, compared to two towns
As shown, all concepts in the location ontology derive from in different districts. To capture these, for SubAreas a
concept Location. The highest level concept that the ontology sameTown relation and for Areas a sameDistrict relation
currently represents is Country. The rest of the concepts are and so on are defined. However, for the purpose of this
defined specific to India keeping in view the administrative paper, we stick to the basic nearTo relation between
structure of the country 7 . landmarks. All the modules make use of this relation
7 http://en.wikipedia.org/wiki/Subdivisions of India 8 http://en.wikipedia.org/wiki/Open World Assumption
90
to search for or add/update locations specified in user’s new entry is rejected. Also, this landmark is added only
query. within the context of its parent, i.e. it’s SubArea. If the
The connectedTo relation is a symmetric as well as a SubArea supplied does not exist, then also the landmark
transitive relation. It expresses the fact that two locations is rejected. In addition to the landmark name, the user
are reachable from each other via one or more paths can optionally specify other meta information such as
that can obtained from the ontology facts. Users may which all landmarks are located nearby and to which all
add connectedTo relation between location instances that landmarks is this new one is connected to.
they are familiar with. F OLKSOMAPS makes use of the 5) editLandmark() : This method allows you to search for a
ontology to infer new connectedTo relations based upon landmark or a relation instance and allows you to delete
these individual assertions. The Path Finder module relies it in the fashion similar to addLandmark().
on this information to compute paths from the individual In the next section, we describe our prototype implementa-
connections supplied by users. tion that we used to conduct user studies.
• Bootstrap Process
The F OLKSOMAPS system could be bootstrapped from VI. P ROOF - OF -C ONCEPT I MPLEMENTATION
existing databases to populate instances of location types
in the upper part of the ontology. Two such sources We have implemented a prototype of F OLKSOMAPS and
of data in the absence of a full-fledged Geographical deployed it at our lab. The prototype’s knowledgebase includes
Information System (GIS) system come from the Telecom the ontology module and does not have the optional Graph
Industry and the Postal Department. While postal depart- module. We used OWL to implement the ontology. We im-
ment is obvious, the telecom companies also maintain plemented the F OLKSOMAPS modules for finding a location,
database of various circles that they operate in. Given that finding a path and adding a landmark. The ontology API
mobile phones have penetrated into the remote rural areas used is JENA9 with Pellet reasoner10. We bootstrapped the
as well, the upper ontology can be populated from their system with data about New Delhi, the capital of India and
data as well. While the actual GIS data benchmarking initialized it with SubAreas located in a couple of Areas under
require significant efforts and cost on the field to map the South Delhi district. We implemented a Web based interface
spatial data, the telecom and post office data give a very as well as a Voice based interface for this prototype. The Web
good logical view of the locations. This complements our based interface supports all the implemented modules and is
system’s design goal of providing a logical view rather developed using Java Server Pages (JSP)11 . On the other hand,
than a spatial view to the users. the Voice based interface is accessible over a phone call and
supports find location and find path modules. It is developed
B. Knowledgebase API using JSPs and VoiceXML12 .
This subsection describes the API that we built for accessing Users are allowed to populate the F OLKSOMAPS system
the knowledgebase for finding path, location or for adding a with new landmarks and associate them to the SubArea
landmark. which they belong to. Additionally, users can also provide
1) findLocation() : This method allows a user to search for information about the landmark. This includes other landmarks
a location specified by its name. Optionally, extra infor- located near to the one being added, and other landmarks that
mation can be supplied which includes the landmark’s are connected to this by road etc. Figure 4 shows a partial
relationship with another landmark or its attributes. snapshot of the populated Folksomaps knowledgebase.
2) findPath() : This method takes source location name India New Delhi
and destination location name along with a filtering (Capital)
criteria and returns a list of locations that represent the South Delhi
(District)
Bootstrapped Data
path traversal from source to destination. The location

names can optionally be augmented with a list of nodes Munirka
(Area)
VasantKunj
(Area)
Malvianagar
(Area)
Saket
(Area)
that represent meta information about the position of VasantVihar

(Area)
Mahipalpur
(Area)
the source or destination node in the ontology. The JNU PVRPriya

Sector B
(SubArea)
Sector C
(SubArea) PVR
filtering criteria specifies additional restrictions (e.g. VasantLok
Market
Badam
Market Anchal
Saket
Plaza
User Populated Landmarks
only traverse paths connected by a nearTo relation) on GoldenDragon

Hotel
partOf
the path traversal algorithm. connectedTo

GDGoenka
ISID
Sarvapriya
Vihar
3) doesExist() : This method determines whether the sup-

isNear
Country/Capital/District School GrandHotel
/Area/subarea
plied landmark name already exists in the knowledge- Landmark
base.
4) addLandmark() : This method allows the user to insert a Fig. 4. A partial snapshot of populated ontology
new landmark into the knowledgebase. It takes the name 9 http://jena.sourceforge.net/
of the new landmark and also its immediate parent, i.e. 10 http://pellet.owldl.com
its SubArea name. Optionally, its next parent, i.e. PostOf- 11 https://java.sun.com/products/jsp
fice is also supplied. If this landmark already exists, this 12 http://www.w3.org/TR/voicexml20/

91
Users are also allowed to query F OLKSOMAPS for get- website. In addition to the bootstrapped data, the knowl-
ting location information and directions. Figure 5 shows the edgebase also contained some pre-populated locations. We
screenshot of webpage for getting directions from a source to requested them to populate content into F OLKSOMAPS while
destination location. Figure 6 shows the flowchart for querying restricting the locations to a set of 6 Areas in South Delhi
through the Voice based interface. Error steps are not shown District. This was done so that the content populated does
in this flowchart. The results from F OLKSOMAPS reflect the not get thinly spread out and is relatively rich for querying
14
ways in which people would essentially give directions. . After populating some landmarks known to them, users
then queried the system for finding information about other
landmarks and travel directions to those. After this, we asked
them a few questions about their view of F OLKSOMAPS
system.
For the Non-ITsavvy subjects, we briefed them about the
purpose of the proposed system and gave an explanation of
the prototype. We then walked them through the voice based
interface, by querying for some location and requesting for
a path to that location from another. This was followed by a
question answering session. All subjects grasped the concept
fairly quickly and were able to see the benefits they could
derive from such a system.
B. Results of Survey for Non-ITsavvy subjects
We conducted a total of 22 surveys with Non-ITsavvy
participants using the voice interface. The set of interviewed
Fig. 5. Web UI showing the results of querying for directions
people consisted of porters, security guards, elderly people,
draughtsmen, waiters and service staff. The results are tabu-
Welcome to
lated below.
Folksomaps
Source Path Find Location

Location
'Location'
Location? Name?
or 'Path' ?
Play
Get Get
Location
Src_Name Loc_Name
Details
Sample User Interaction

Destination System: Welcome to Folksomaps. Using this service you can find information
Location? about locations in Delhi. To find information about a location say
Information, to find a route between two locations say Route
User : Route
Get System: Please speak the source location

Dest_Name User : IIT
System: Please speak the destination location

Play User : ParkBalluchiRestaurant
Path
System: You can go from IIT to SDA.
Details
From SDA you can go to GreenParkMarket
From GreenParkMarket you can go to ParkBalluchiRestaurant
Fig. 6. Voice UI flow for querying F OLKSOMAPS. Red dots indicate voice
recognition steps.
VII. S OLUTION S URVEY Fig. 7. Interviewing the Non-ITsavvy users
In this section, we present results of user studies we con- TABLE III

ducted to verify the benefit and acceptability of the proposed S UMMARY OF N ON -IT SAVVY U SER R ESPONSES TO F OLKSOMAPS
system. We further present insights that we obtained from
users while conducting the survey. For the ITsavvy segment, Questions Yes No
we let the subjects try F OLKSOMAPS through the web based Would you call to get directions? 100% 0%
Prefer calling over asking people on the street? 82% 18%
user interface of the system. For the Non-ITsavvy segment, Ready to pay for call (else want ads)? 45% 55%
we conducted the survey with the voice based interface of Will upload content? 73% 27%
F OLKSOMAPS13 . Voice Interface preferred over SMS? 91% 9%
Results need to be very accurate? 86% 14%
A. Survey Process
For ITsavvy survey participants, we started with a small As can be seen from Table III, all the subjects surveyed were
introduction before giving them access to the F OLKSOMAPS interested in using F OLKSOMAPS system. Most of them pre-
ferred the option of calling up a number for directions rather
13 The voice based interface was in Hindi language and allowed a restricted
set of landmarks to keep the speech recognition accuracy high for the 14 We envision that the actual content in a deployed system will be much
prototype more richer than the content populated by the survey participants.
92
than asking someone on the street. Apart from convenience, cities, more than the road route from a source to destination,
this overwhelming response can be attributed to a couple of people from the underprivileged segment were more interested
reasons. First, asking on the street does not always work. On a in knowing the bus route numbers that could take them to
secluded road, there may not be anyone to ask. Several times their destination. Road routes serve well those people who
people end up giving wrong directions (possibly to hide their travel by their own vehicles but the underprivileged rely
lack of awareness or to not appear rude) leading to precious primarily on public transportation. Similarly, for rural areas
time spent traveling on a wrong route and recovering from that consist of remotely located towns and villages, what helps
it. Also, the point of view of street vendors who often get the underprivileged people is information regarding modes of
enquiries from passersby about directions is quite interesting. transportation (train, bus, boat, cycle-rickshaw, taxi etc.) to
They candidly admitted that during the course of the day take from source to destination, where to make a switch and
they get so many such queries for detailed travel instructions estimated travel time. Time tables of these public transport
that it is easy to get irritated and shrug them off. Second, an mechanisms are another important feature for this segment
important insight provided to us by our subjects was the sense that can become an essential part of F OLKSOMAPS.
of security that they would get with such a system. We were Key Insights : We realized that accuracy of the information
informed that even though asking for travel directions from is a key requirement and more the meta information available,
strangers on the street is an option, it exposes the enquirer merrier it would be for these consumers. Also, voice based
to criminal elements, who often take advantage of their lack interface is indeed a preferred mode for this user segment
of knowledge and use it to rob them, the activity sometimes over SMS and the calls to the system should to be free of cost.
resulting into a worse situation such as a murder. This is
C. Results of Survey of ITsavvy subjects
especially true for first time visitors to the city from remote
towns and villages or female citizens traveling at odd hours For ITsavvy segment, we conducted the survey with a total
of the day. of 15 subjects using the web based interface. We also told
Many people were willing to pay for the call even a them that the system has a voice based interface available
small premium over normal charges as they saw value to over a phone call and supporting similar API. The user list
having this information available to them at all times. Yet consisted primarily of software professionals apart from a
a majority of the subjects preferred the advertisement model couple of businessmen. As expected, the ITsavvy community
where an advertisement played in the beginning of the call had significant experience in using the current online maps of
pays for the entire call. This is understandable, given this cities in Indian metropolitan cities and were able to carefully
segment’s high sensitivity to cost. A few users suggested that evaluate our approach, considering the map services that are
the advertisement model was better since most of these users already operating in metropolitan cities. We try to capture
primarily have pre-paid SIM cards and often they do not have learnings from their feedback.
sufficient balance to make outgoing calls. TABLE IV
Most people were willing to upload content, though a few S UMMARY OF IT SAVVY U SER R ESPONSES TO F OLKSOMAPS
refused as they were hesitant due to not owning a phone. Questions Yes No
Almost everyone preferred the voice based interface over Would you access it for directions? 93% 7%
SMS even though we demonstrated speech recognition errors Prefer F OLKSOMAPS over asking people? 87% 13%
Ready to pay for call (else want ads)? 67% 33%
during the study interviews. The primary reason for this cited
Will upload content? 87% 13%
by them was that many people are either not comfortable using Prefer Web for upload? 92% 8%
SMS or not comfortable using a mobile phone itself. However, Results need to be very accurate? 53% 47%
some users who were well versed with SMS preferred it over
voice. As is evident from the results (Table IV), most survey
In terms of accuracy of returned results, most people asked participants mentioned that they would like to use this service
for full accuracy while a very few were okay with minor and that it would certainly be more convenient than asking
mistakes. The need for strong accuracy is driven by the fact people around in the streets. Interestingly, a bulk of the
that most of these people either use public transport, or use a ITsavvy community did not stress on getting fine-grained
bicycle or even walk to reach their destination. The cost of a direction all the time. They were fine with getting high level
wrong input for them is huge compared to a person driving directions involving major landmarks.
in his own or rented vehicle. In fact, one of the main reasons Most people were fine with paying for the service when
for preferring a voice call over asking people for directions offered on phone. Most were also willing to upload content
was to avoid wrong directions. This is an important feedback into F OLKSOMAPS but preferred to do so over a web based
since we started with the assumption that we do not need strict interface as opposed to SMS or a voice based interface.
controls over the content and the wiki model would work. But Accuracy of responses was important for this segment as
the tolerance for incorrect information is low and we need to well, though not as strongly as for the Non-ITsavvy segment.
factor this in. As discussed earlier, this segment typically used their own
We also learnt that meta information is as important to Non- vehicles and need high level directions rather than precise
ITsavvy users as the landmarks themselves. For instance, in route.
93
A few participants pointed out that while voice-based access destination. Voice interface also takes care of the language
is good for interactive session, they would still prefer the barrier since content can be delivered in local language as
content is sent to them via SMS so that they can store it demonstrated in other systems [3]. Voice based interfaces are,
for future access, pointing out that it is easy to forget the however, constrained with the capability of speech recognition
directions if you just hear it. technology which is under slow but constant improvement.
However, this segment had other expectations from the sys- Our surveys indicate that most people would like to con-
tem. Some subjects mentioned that the system should adapt to tribute to F OLKSOMAPS knowledge base. However, in real-
the user’s request and have the ability to produce fine-grained life, there are several factors that provide impedence for a
details depending on whether the destination is a popular user to be an effective information producer. Reasons range
landmark, or an area or a house in a colony. Few others men- from users becoming busy, loosing interest after an initial
tioned that the directions provided by F OLKSOMAPS should surge, etc. In a live deployment of F OLKSOMAPS, one needs
take into consideration the amount of knowledge the subject to also consider pragmatic business models (such as bartering
already has about the area, i.e. it should be personalized based models, advertisements or incentives) using which an in-flow
upon user profile. One subject mentioned that the current of information can be sustained to keep improving the quality
prototype appeared more suitable for driving directions but of system responses.
not for pedestrians. Accuracy of results being of primary concern universally,
Just like the Non-ITsavvy community, the ITsavvy com- further research is needed to ensure that various modules of
munity also reflects the need for meta information on such a F OLKSOMAPS would guarantee correct and precise results
community-driven map. One subject mentioned that he would given that the data input by the users is correct in the first
be interested in uploading traffic information on routes and place. Also, as the knowledgebase as well as the userbase
would benefit from community uploading such information on grows, established scale-up techniques would have to be
the system. A few subjects mentioned that frequent changes applied for real life environments.
in road plans due to constructions should be captured by
such a system - thus making it more usable than just getting IX. R ELATED W ORK
directions. Two research areas that are very relevant for F OLKSOMAPS
Key Insights : While accuracy and convenience score with are the areas of research in use of IT for underprivileged
IT-Savvy population as well, this segment turned out to be in developing regions and semantic tools for geographic in-
more ambitious in terms of deriving benefits from such a formation systems. Apart from these, work in the area of
system. Based upon the feedback listed above, we learnt that intelligent user interfaces for masses in developing regions is
the user interfaces of F OLKSOMAPS needs to be rich and also relevant.
adaptive to the information needs of the user when considering There is a lot of literature on means to harness available
this community. It also appears to the authors that dynamic information and user generated content [11], [12], [13] to
and real-time information augmented with traditional services deliver useful services to underprivileged in developing re-
like finding directions and locations would certainly add value gions. [14], [15] talks about voice-driven technologies (e.g.
to F OLKSOMAPS. audio wiki) to capture user content from Non-ITsavvy masses.
The Neighbourhood Mapping [16]15 initiative proposes the
VIII. D ISCUSSION
involvement of school students to gather community input
F OLKSOMAPS we believe, presents a novel approach to- in the context of building maps. The project used PDAs
wards developing a self-sustaining map system, harnessing coupled with GPS to build an information repository that
community input, particularly targeted towards developing could be used of planning purposes. F OLKSOMAPS builds
countries, where there is a need for such a system. further along this direction and proposes to create alternatives
Reasoners using ontologies consume space and compute to well established solutions in developed countries, solely
power. [6], [7], [8] reports ways through which spatial reason- through user generated content. Specifically, it focuses on
ing can be made faster. This is an issue with F OLKSOMAPS creating a framework (exploiting ontological reasoning), where
as well. However, F OLKSOMAPS as a system compliments by geographic information can be captured, enriched, and
this body of work as its focus is not on improving reasoning funneled back to the masses - customized to the needs of
capabilities or address scalability needs of underlying ontology developing regions. To the best of our knowledge, we are
reasoners. Rather F OLKSOMAPS can benefit from this body not aware of any such community-driven map system for
of work by adopting the solutions suggested to improve the developing regions.
computational and reasoning efficiency. There has been considerable work on Place ontologies,
Given their preference towards voice based interface over retrieval and storage of geographical information using on-
SMS [9], [10], designing an efficient and user-friendly voice- tologies [8], [17], [18], [6], [7]. For example, [8] talks about
based user interface for the masses is important for F OLK - limitations of OWL to support spatial reasoning, integrity
SOMAPS . For example, while finding directions, user interface rules, and proposes a combination of spatial data-based store
should be designed in a way that facilitates users to specify the
level of detail they are looking for, varying from source to the 15 http://www.csdms.in/NM/
94
and ontology-based reasoning to better represent geographic [3] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. A. Nanavati,
information and resources. [6] further critically evaluates “WWTW: A World Wide Telecom Web,” in ACM SIGCOMM Workshop
on Networked Systems For Developing Regions, Aug 2007.
ontology-based approaches towards geographic information [4] M. Plauche and M. Prabaker, “Tamil Market: A Spoken Dialog System
retrieval while [7] presents a spatially aware search engine, for for Rural India,” in Working Papers in Computer-Human Interfaces
semantic interoperability of distributed and heterogeneous GIS (CHI), 2006.
[5] T. S. Parikh, “Mobile Phones may be the Right Devices for Supporting
on the Internet. In contrast to this literature, F OLKSOMAPS Developing World Accessibility, but is the WWW the Right Service
does not focus on improving reasoning capabilities of OWL or Delivery Model?” in International Cross-Disciplinary Workshop on Web
address spatial and logical integrity issues of Place ontologies. Accessibility (W4A), Scotland, May 2006.
[6] A. Abdelmoty, P. Smart, C. Jones, G. Fu, and D. Finch, “A critical
Rather, we focus on designing concepts of an ontology that evaluation of ontology languages for geographic information retrieval
is suitable for capturing map information from communities, on the Internet,” Journal of Visual Languages and Computing., pp. 331–
keeping only logical integrity in mind, and by exploiting 358, August 2005.
[7] M. Kun and B. Fuling, “An Ontology-Based Approach for Geographic
currently standardized semantics of OWL. F OLKSOMAPS in Information Retrieval on the Web,” in International Conference on
a way is hence complimentary to this body of work and can Wireless Communications, Networking and Mobile Computing, Sept
benefit from it. 2007, pp. 5959–5962.
[8] A. Abdelmoty, P. Smart, and C. Jones, “Building Place Ontologies
A body of research focuses on intelligent user interfaces for the Semantic Web: Issues and Approaches,” in Proceedings of the
for developing regions [19], [10], [20] and infrastructures to 4th ACM workshop on Geographical Information Retrieval, Lisbon,
take IT services to the economically challenged and Non- Portugal, Nov 2007.
[9] “Africa:Impact of Mobile Phones,” Vodafone Policy Paper Series, Mar
ITsavvy masses in developing regions [3], [21]. F OLKSOMAPS 2005.
at its core can certainly benefit from user interface designs [10] A. Kumar, N. Rajput, D. Chakraborty, S. Agarwal, and A. A. Nanavati,
to facilitate upload and download of map information. The “Voiserv: Creation and delivery of converged services through voice for
emerging economies,” in In Proceedings of the WoWMoM, Finland, June
system, by having multi-modal front-ends fits well with the 2007.
architectural principles outlined in [3], [21]. [11] R. Abraham, “Mobile Phones and Economic Development: Evidence
from the Fishing Industry in India,” in IEEE/ACM International Confer-
ence on Information and Communication Technologies and Development
X. C ONCLUSION (ICTD), Berkeley, USA, May 2006.
[12] K. Ramamritham, A. Bahuman, C. B. S. Duttagupta, and S. Balasun-
In this paper, we investigated feasibility of a community- daram, “Innovative ICT Tools for Information Provision in Agricultural
driven approach towards creating maps for developing re- Extension,” in IEEE/ACM International Conference on Information and
gions. Our system dubbed F OLKSOMAPS, has the potential of Communication Technologies and Development (ICTD), Dec 2006.
[13] P. K. Reddy, G. Ramaraju, and G. Reddy, “eSaguTM: A Data Ware-
providing an effective alternative to expensive map solutions house Enabled Personalized Agricultural Advisory System,” in ACM
using community input, making map-based services (finding International Conference on Management of Data, China, June 2007.
directions, finding locations and landmarks) available to peo- [14] P. Kotkar, W. Thies, and S. Amarasinghe, “An Audio Wiki for Pub-
lishing User-Generated Content in the Developing World,” in HCI for
ple in developing regions where such services are currently Community and International Development (Workshop at CHI 2008),
missing. F OLKSOMAPS builds on the current models adopted Florence, Italy, April 2008.
by users in developing regions and leverages their collective [15] L. Wang, P. Roe, and B. Pham, “An Audio Wiki Supporting Mobile
Collaboration,” in Proceedings of the ACM Symposium on Applied
knowledgebase thus overcoming the huge cost barrier in Computing (SAC), Brazil, March 2008.
developing such a system. It is specifically designed to provide [16] R. Mallick, H. Kalra, and D. Banerjee, “Infusing map
content that is intuitive for the users. culture through participatory mapping,” GIS@development.
http://www.gisdevelopment.net/magazine/years/2005/feb/infusing.htm,
We conducted a total of about 77 interviews in the process Feb 2005.
of evaluating a need for such a system and testing our proto- [17] J. Kay, W. Niu, and D. Carmichael, “ONCOR: Ontology and Evidence
type for verifying usability and utility of F OLKSOMAPS. Our based Context Reasoner,” in Proceedings of International Conference
on Intelligent User Interfaces (IUI), Honolulu, Hawaii, Jan 2007.
surveys suggest that the community is very receptive towards [18] G. Look and H. Shrobe, “Towards Intelligent Mapping Applications:
the concept of a community-driven map as that alleviates A Study Of Elements Found In Cognitive Maps,” in Proceedings of
some of the problems (reliance on people, security, inaccurate International Conference on Intelligent User Interfaces (IUI), Honolulu,
Hawaii, Jan 2007.
directions, etc) they face in day-to-day life. We intend to build [19] I. Medhi, A. Sagar, and K. Toyama, “Text-Free User Interfaces for
further upon this system and overcome its current limitations Illiterate and Semi-Literate Users,” in IEEE/ACM International Confer-
to bring it even closer to users’ expectation. ence on Information and Communication Technologies and Development
(ICTD), Berkeley, USA, May 2006.
[20] J. Sherwani, S. Tomko, and R. Rosenfeld, “Sublime: A Speech- and
ACKNOWLEDGMENT Language-based Information Management Environment,” in In Proc.
ICASSP, May 2006.
The authors would like to thank Amit Nanavati for suggest- [21] A. Kumar, N. Rajput, S. Agarwal, D. Chakraborty, and A. A. Nanavati,
ing the ’Folksomaps’ name. “Organizing the Unorganized - Employing IT to Empower the Under-
privileged,” in Proceedings of 17th ACM International Conference on
R EFERENCES World Wide Web(WWW), Beijing, China, April 2008.
[1] “The Use of ICTs by Small and Informal Businesses,”

http://research.microsoft.com/ jdonner/ Papers/donner micros slides.pdf.
[2] “World Gazetteer. India - Metropolitan Areas,” http://www.world-
gazetteer.com/wg.php?x=&men=gcis&lng=en&dat=80&geo=-
104&srt=pnan&col=aohdq&msz=1500&va=&pt=a.
95
HIV Health Information Access using Spoken

Dialogue Systems: Touchtone vs. Speech
Aditi Sharma Grover, Madelaine Plauché, Etienne Barnard, Christiaan Kuun
Based on this perceived value of SDSs in the developing

world, a number of exploratory studies have been performed
Abstract— This paper presents our work in the design of a in recent years. Barnard et al. [5] report on preliminary
SDS for the provision of health information to caregivers of HIV experiments performed to assess the usability of a telephone-
positive children. We specifically address the frequently debated based information service for access to government
question of input modality in speech systems; touchtone versus information in South Africa. A kiosk-based SDS for
speech input, in a new context of low literacy users and a health agricultural information was developed by Plauche et al. [6],
information service. We discuss our experiences and fieldwork
and evaluated with semi-literate users in rural Tamil Nadu,
which includes needs assessment interviews, focus group sessions,
and user studies in Botswana with semi and low-literate users. India. Nasfors [7] also developed an agricultural information
Our results indicate user preference for touchtone over speech service, aimed at mobile telephone users and deployed in
input although both systems were comparable in performance Kenya. The most sophisticated speech technology in this
based on objective metrics. category was employed in the telephone-based information
service for community health workers developed by Sherwani
Index Terms— Spoken dialogue systems, DTMF, Touchtone, et al. [8]; this was piloted in Sindh, Pakistan. Agarwal et al.
Speech interfaces, Health information, HIV, Illiterate users, [9] have implemented a telephone-based kiosk system, which
Semi-literate users, Low literate users, ICT, Developing regions, they call “VoiKiosk”; this is being trialled in rural villages in
Information access, Africa. Andhra Pradesh, India.
I. INTRODUCTION Each of these pioneering studies was primarily aimed at
assessing the feasibility of using speech technology in various
There is a widespread belief that spoken dialogue systems settings in the developing world. However, in the process of
(SDSs) will have a significant impact in the developing world determining feasibility, a number of practical lessons were
[1]. This belief is based on a number of factors. Firstly, also learnt. For example, it was found that user acceptance of
illiteracy is predominantly a problem of the developing world such systems is proportional to the difficulty that users would
(according to a recent estimate, about 98% of the illiterate have to access the same information through other
people on earth live in the developing world [2]), and speech- mechanisms [6] (thereby confirming the concept of the
based access to information may enable illiterate or semi- “motivated user”), and two studies [5, 8] found that it may be
literate people to participate in the information age. Also, the preferable to use more verbose, less efficient user interfaces to
availability of traditional computer infrastructure is low in the guide inexperienced users for whom time pressure is not a
developing world, but telephone networks (especially primary concern.
mobile/cellular networks) are spreading rapidly [3]. (For
example, a recent community survey in South Africa [4] The current contribution similarly has a twofold aim: both to
found that 73% of households owned at least one mobile explore the use of an SDS in a new environment (namely, by
phone, but only 7% of homes had internet access.) A further caregivers of HIV positive children in Botswana, Southern
factor is the strong oral culture that exists in many traditional Africa), and to contrast different input modalities used in such
societies, which is likely to render such systems more a system. In particular, we compare systems using key-presses
acceptable than text-based or graphical information sources. (“DTMF or Touchtone”) with those that use automatic speech
Finally, the availability of relevant services and alternative recognition (ASR) for user input in this application. To this
information sources is often low in the developing world. end, we start with the most basic variant of speech systems
(key-press replacement), and compare such systems with
Manuscript received September 22 2008. This research was supported by DTMF input. The motivation for this choice is twofold: on the
OSI, OSISA and the NRF under the Key International Research Capacity one hand, key-press replacement is likely to be more
(KISC) programme, UID no. 63676. acceptable in the developing world, where general numeracy
(Corresponding author) A. Sharma, E. Barnard and C. Kuun are with the
Human Language Technologies (HLT) Group in the Meraka Institute, CSIR, is less common; on the other, such systems are much easier to
Pretoria, South Africa. (phone: +27128413028, email: <asharma1, ebarnard, develop than natural-language systems, and are therefore more
ckuun>@csir.co.za) attainable in the resource-constrained environments that
M Plauché was a visiting researcher at the Meraka Institute during the project. typically characterize the developing world.
(email: mad@brainhotel.org)
To our current knowledge DTMF and ASR input modalities

have not been compared systematically in the developing
world. However, in a review of developed-world applications
96
comparing DTMF to ASR input by Lee and Lai [10], both primary and specialty medical care for HIV/Aids, to catering
user preference and performance are found to depend on the for the psychosocial needs of HIV patients and their families
nature of the task, the personality of the user, and the [18]. Baylor provides treatment to over 2100 children infected
capabilities of the speech-recognition system. With the with HIV and 260 families across Botswana. The centre is
exception of only one call-routing system [11], all studies involved in a number of support activities such as community
found that simply replacing key-presses with speech does not outreach programmes for patients in rural areas, servicing 20
improve user performance or perception. Conversely, well- communities outside Gaborone and a “Teen Club” to provide
engineered speech recognition systems with natural-language moral support and counselling to teenagers living with HIV
input are often preferred to DTMF for tasks that are not easily [18].
accomplished with DTMF. In terms of user preference, studies
A child is typically brought to Baylor to be tested for
in laboratory settings [10], [12] report that users find speech
HIV/Aids by a caregiver. A caregiver is any individual who
input more interesting and enjoyable to use, but contrastingly
takes care of an HIV positive child; it may be the child’s
in a recent informal poll of over a thousand users of real-
parents (who themselves might be HIV positive), other family
world information-access systems [13], almost half of the
members or an unrelated community member. Children who
users responded that they would prefer to use a speech input
test as positive receive free treatment from Baylor for the
modality “as little as possible” and only 8% would do so
remainder of their infancy and adolescence. Caregivers of
“most of the time”. Despite these reservations, numerous
such children are also counselled and trained. Baylor provides
applications that are completely reliant on speech input are
free lectures for caregivers three times a week, where many
currently in use in the developed world - examples are the
aspects of HIV/Aids, antiretroviral (ARV) medication are
health-management systems described by Migneault et al. [14]
explained and advice is given on how to live with the
and commercial voice portal systems, such as the Tellme
condition. Each caregiver on average attends two lecture
portal (which serves 40 million phone calls per month,
sessions. The primary focus of the lecture sessions in Baylor
according to its providers [15]).
is on adherence to ARV medication, with topics such as the
principles of HIV, universal precautions, basics of ARV
Below, we first provide background on the health-care
therapy, medication dosage, side effects and storage, and
application selected for our study and describe the system that
importance and strategies for adherence, being covered.
was developed as well as the experimental protocol employed
(Section II). Section III contains our experimental results,
including user profiles, usability measurements and task
completion rates. Finally, we discuss the scope and
generalizability of our results, and conclude with thoughts on
next steps to be taken along this research trajectory.
II. OPEN PHONE: HIV/AIDS HEALTH INFORMATION LINE IN

BOTSWANA
A. Background
HIV/Aids is perhaps the gravest health pandemic to face the

world, and Southern Africa has been the worst hit region. Of
the 33 million people infected with HIV worldwide,
approximately two-thirds are inhabitants of sub-Saharan Fig. 1. Baylor in Gaborone, Botswana.
Africa [16]. Within sub-Saharan Africa, Botswana has one of
highest HIV prevalence rates, with 1 in every 4 adults being B. Open Phone Development
HIV positive [16]. The hardest hit in Botswana are women;
nearly 40% of pregnant women (ages 25-39) are living with
HIV and infection levels are increasing amongst pregnant Open Phone is a pilot HIV/Aids community-oriented SDS
women aged 30-34 years, with nearly one in two living with service that makes use of language technologies to address
HIV [17]. Aids deaths have orphaned approximately 120 000 acute informational needs of caregivers of children with HIV.
children (ages 0-17) and another 14 000 children (aged 0-14) 1) Preliminary Investigations
are living with HIV in Botswana.
During our study a partnership was established with The Our initial investigations started in April 2007, where we
Botswana-Baylor Children’s Clinical Centre of Excellence conducted interviews and discussions with 2 doctors, 4 nurses,
(hence forth referred to as Baylor). Baylor is a specialised 9 caregivers, a social worker and the technical manager. We
paediatric institute serving the Botswana capital, Gaborone also accompanied 3 community outreach workers on visits to
and its neighbouring areas since June 2003. The centre is 2 caregivers’ homes. The intention of this investigation was to
staffed collaboratively by U.S. and Botswana health identify specific needs of the various user groups and their
professionals. The services provided by Baylor range from day-to-day tasks.
97
This trip also served to acquaint us with HIV domain- Although caregivers are encouraged to call Baylor with
specific terminology used by the interviewees to describe their any questions they may have, most are reluctant (and
situations, tasks and other notable work processes. These unable) due to the high costs of mobile phone calls.
‘hidden’ pieces of information allowed us to build a profile of
These challenges and issues formed the basis of our
the individuals interviewed and make informed choices later
motivation for the design of a health information SDS that
in the process of application design.
provided not only adherence education from Baylor lectures
This initial field trip highlighted a number of challenges and but also general health information tailored to the needs of
issues for consideration in providing health information to caring for HIV positive children. An SDS in the local
caregivers: language Setswana, toll-free and accessible at any time
through a simple telephone call, could greatly support Baylor's
Caregivers often struggle to recall material covered in
services to caregivers of children with HIV.
the lecture sessions.
Caregivers often have questions regarding general 2) Content Development
health information, for example, how to deal with
infections, nutritional needs, hygiene requirements and Our design process started with the identification of relevant
commonly held misconceptions about HIV. content and development of a framework (Fig.2) which
detailed the broad health topics that needed to be covered by
Travelling to Baylor to address health information
the SDS. Interviews with Baylor staff (nurses, doctors,
queries is not always possible for caregivers due to
nutritionists) and the following printed sources were used to
family/work responsibilities, transportation costs and
create locally relevant accessible HIV health related content:
time constraints.
The majority of caregivers are semi and low literate Baylor Adherence lecture materials
populations. No written material is used and thus there HIV Aids Care & Counselling, A Multidisciplinary
is a lack of reinforcement and support for remembering Approach [19]
material learnt in the Baylor lectures. Where there is no Doctor: A Village Care Handbook
[20]
Most caregivers are uncomfortable with English, thus
HIV, Health & your Community, A Guide for Action
Baylor lectures and all interactions with caregivers are
[ 21]
in Setswana. Baylor staff explains complex health
information in accessible terms in local language.
Facts on
Cereals/Rice/ How H IV is W hen to use
Cure Nutrition
Bread spread gloves
Answers to
Traditional
C om m on Fruits and
Practices Food Groups W hat to do if
Questions about Vegetables gloves are not
HIV M yths
M eat & Beans Dealing with available
body fluids &
Infection How to clean up
M ilk and Dairy infected waste
How HIV works products blood properly.
Learn about HIV
Safe preparation H ow to dispose
How HIV is Preparing and
of food infected gloves
spread? Storing food
The spread of HIV & waste
Safe storage of
from M other to During food
Child pregnancy W hen to wash
2. Nutrition
1. H ygiene & your hands
During
Breastfeeding Cleanliness W ashing your
Childcare
hands with an
antiseptic
G eneral Health W ashing your
5. Facts about HIV hands W ater shortage
H elpline & hand washing
3. Com m on Keeping a clean
Sicknesses Danger Signs 4. ARV household
Cleaning
Hom e-based care M edication H ow AR Vs work ? household item s
Diarrhoea
Learn about M edication side
Danger Signs
ARVs effects How to wash
Vom iting Hom e-based care your linen.
How to store ARVs.
Danger Signs
W hen to give
Breathing Hom e-based care W hen and How to
m edication?
Problem s give m edication ?
How to give
Danger Signs W hat to do if you m edication ?
m iss an ARV
Fever Hom e-based care dose? Planning ahead
Danger Signs Traditional Vom ited dose

m edication
Constipation Hom e-based care Forgotten dose
Drugs & Alcohol
Fig. 2. OpenPhone content framework.
98
We also held focus group sessions with 27 caregivers at

Baylor on a second visit. The discussions were led by a A sample system-user interaction follows.
Setswana facilitator and were aimed at soliciting input from
caregivers on what kind of health information they typically User: [Dials number...]
require under the broad topics of the framework. The System (Introduction): Hello and Welcome to the Health
caregivers were first given a demonstration of a health SDS Helpline, I am Nurse Lerato and I know you probably have
prototype and then asked what kind of information they would many questions about caring for someone with HIV.
like the system to provide to them. For instance, the facilitator System (Overview): I can tell you about Hygiene &
would introduce the topic of “Nutrition” and then ask e.g. Cleanliness, Nutrition, Common Sicknesses, ARV
“what kind of food does a child with HIV require? Medication, and Facts about HIV. If at any time during
Caregivers’ responses were used to fuel discussion and direct your call you want to start our conversation again, you can
further questions on sub-topics of the framework. Thereafter press 0.
caregivers were asked to prioritize (rank) topics they had System (Main Menu): For Hygiene & Cleanliness, please
listed in order of need in a group consensus setting. In many press 1, for Nutrition, press 2, for Common Sicknesses,
cases caregivers debated and gave reasons for their choices. press 3, for ARV medication, press 4 or for Facts about
HIV, please press 5.
We found that the topics in our content were mostly inline User: [Presses 2.]
with what caregivers proposed in the focus groups. However, System: Eating a balanced diet of different foods
the topics’ ranking done by the caregivers varied from ours, helps people with HIV stay healthy and strong. A healthy
for example, they ranked Hygiene & Cleanliness as most diet does not have to be costly and contains food from all
important which we had thought was secondary to ARV the different food groups. Healthy food is always prepared
medication. The rankings allowed us to determine the topics and stored in a clean environment...
the SDS should focus on providing first and foremost and
were later also used to determine to order of main menu As a final step, the design team decided on the specific
options in the SDS. The focus group sessions allowed us to wording of all audio prompts in Setswana. This required
revise the content and framework to build a SDS that caters careful consideration for the dialect, the register (informal or
for the health information needs expressed by the caregivers formal), the cognitive load (short audio prompts, but long
themselves. enough to provide contextual “anchoring”), and use of
appropriate user interface metaphors. For instance, a system
3) System Design that suggests to the user “say Main Menu” is unlikely to make
sense to users who have never interacted with visual or audio
Once the initial content had been created and approved, the interfaces before. Instead, we chose the metaphor of a
dialogue flow and audio output of the SDS had to be designed “conversation” and asked the user to “Start our conversation
in a user-friendly manner. We carefully crafted the primarily again”. The Setswana-speaking linguist on the design team
text-based content for spoken speech and simple language. played an essential role in the selection of ASR keywords that
For instance, words like “lower cranium” will have little were locally relevant and logical to users, yet acoustically
meaning for semi & low literate users; such references were dissimilar.
translated into more accessible language, using terminology
and metaphors used by Baylor staff in their lecture sessions. 4) Final SDS Development
For example, the white blood cells are the ‘soldiers’ of the
body, and ARV medication is the ‘ammunition’ for these All the prompts and the health content were translated using
soldiers. a registered translation service into the dialect of Setswana
spoken in Botswana. We specified that the content would be
Since the SDS was aimed at low literacy users and users spoken aloud and the intended audience would likely be low-
who may never have used a SDS before, the design goal was literate. Since user interaction with an SDS is based on audio
to create an easy-to-use interface that placed a low cognitive modality, the voice of the system plays a crucial role. With
load on the user. We grouped content under logical menus and our target audience we felt that this was an essential element
sub-menus, so that the amount of audio output under each in creating a persona that would not only make users
node was balanced with the number of menu options at any comfortable in their interactions with the system but also
given point in time. make the user experience enjoyable. Thus, our ideal voice
talent would:
The design team set the following constraints based on Sound like a caring nurse willing to answer
guidelines for speech user interface design [22, 23]: questions.
Be a mother-tongue Setswana speaker.
The maximum depth of menus in the SDS should not Have a full, mature female, well-articulated voice.
be more than 3 levels. Instil a sense of confidence and trust.
The maximum breath or number of options at any
menu should be no more than 5.
99
Our recruited voice talent was a well-regarded local soap due to either user’s time constraints or a technology failure
celebrity, which meant her voice would be familiar to many of experienced in system set-up. The experimental set-up
the target users. All the system prompts and content included a facilitator, an observer and a WOZ operator. The
recordings were done in a professional recording studio. facilitators were local graduate students who were trained by
The SDS was built using the Asterisk telephony platform, the authors to facilitate in the local language, Setswana. The
the current set up in Botswana is a free standing PC, observers took notes on user behaviour, number of verbal
connected through an Asterisk card to an ISDN line. The prompts needed by a user, any user comments, and general
ISDN line allows up to two calls simultaneously and users dial body language of the user.
a local telephone number to access the system. While the
system is operational, all aspects of the calls are logged. From Each user was introduced to the system and asked to sign a
start to finish all key-presses are monitored and all audio is consent form. Emphasis was placed on communicating that
recorded. users were not being tested but rather the system for purposes
of improvement. Thereafter, each user watched a five minute
video showing how a caregiver could use the system to find
C. User Study
typical health information that they might need (Fig. 4).
The goal of the study was to compare the most basic variant
of speech systems (key-press replacement) with DTMF input.
Thus, we built two identical systems that differed only at the
menu prompts in choice of input modality i.e. in one system
the user would press a key to chose a menu option and in the
other they would say a keyword or phrase. For example, a
DTMF menu option would be; “to hear about Nutrition, press
1,” whereas the ASR menu option would say, “to hear about
Nutrition, say Nutrition.”
The ASR was simulated using the Wizard-of-Oz (WOZ)

methodology [24] where a researcher played the role of the
speech recogniser. The ‘wizard’ listened to the speech input
of the user and chose the next state of the system on this basis.
The ‘wizard’ only accepted the exact keyword or phrase that
the user was allowed to say at a particular menu option, any
other input was directed to a No-Match state. In the case of
the DTMF system, key presses were handled by the back-end
telephony platform. Both systems ran from a PC laptop which Fig. 4. User watching a full-context video.
was connected to standard telephone through a voice over
internet protocol (VOIP) gateway. The context of the video was carefully matched to depict
typical scenarios where a caregiver could use the SDS to
obtain health information. For example, in the first part of the
video, a caregiver with a sick child who has just thrown up his
ARV medication is unsure of whether to give the medication
again. A friend of the caregiver arrives and tells her about a
telephone information system that she can call and learn when
and how the next dose of ARV medication can be given if a
child throws up his medication. Research [25, 8] on designing
interfaces for low-literacy users has shown that such a full-
context video greatly improves user studies by offering not
only a demonstration of how to use the interface, but also the
source (and therefore trustworthiness) of the content, the
context in which you might use such a service, and the
potential impact in your day-to-day life.
After this video, the facilitator gave a short demonstration to

the caregiver on how to retrieve “Nutrition” information as
Fig. 3. DTMF vs. ASR experiment set-up. shown in the second part of the video. For each task, the
caregiver was asked to dial a number themselves to access the
The user study was held over a period of five days in April SDS. This simple action served as a quick check to verify the
2008 at the Baylor premises. A total of 33 caregivers were caregiver’s ability to recognise numbers and use a phone. In
part of the study, of which 27 tried both the DTMF and ASR order to show the difference between the DTMF and ASR
systems. The remaining 6 caregivers did not try both systems systems we used separate telephone sets for each system (Fig.
100
3). After the caregiver finished the tasks using the DTMF in Fig. 5). For example, if a caregiver started with DTMF
system, for example, the facilitator explained that they would using Task Set A he/she would then proceed to do ASR with
now try a similar system (on the other handset) but one where Task Set B (Quadrant 1). Each caregiver thus did a total of 4
they would now say a keyword to obtain information. An tasks and users were approximately evenly assigned between
example of a task explanation is shown on the following page. Set A (15 caregivers) or Set B (18 caregivers). The
Facilitator: What’s the name of any good friend in your permutations in the order of trials also help to counter the
neighbourhood? impact of subject fatigue and learning effects within the small
Caregiver: Kabelo sample size.
Facilitator: Your friend Kabelo says she must wash her hands
frequently to keep her family safe from disease
but she has very little water at home so she
doesn’t know what to do. You’ve heard that
Baylor has a phone number (help line) that can
answer many health questions. So you decide to
call the phone number, can you help your friend?
Since the study was a within-subject comparison, we

refrained from using the same tasks in both DTMF and ASR
to prevent bias of previous knowledge. Thus, we created two
sets of tasks; Set A and Set B with two tasks each (an easy and
a difficult task; Task 1 and 2). The tasks were designed to be
similar across Sets and to require a user to make three correct
menu selections (Menu levels 1, 2, 3) to reach the specified Fig. 5. Order of Trials and Task Sets in experiment.
information. Recall that our HIV health helpline is only three
menu levels deep. The correct paths for each task are shown in After completion of each modality’s trial, a post
Table I. questionnaire was administered verbally to the caregiver. It
consisted of ten questions adapted from the PARADISE
TABLE I
TASK DESCRIPTIONS WITH CORRECT SDS MENU OPTIONS.
evaluation framework [26]. The facilitator recorded the
response in a 5-point Likert scale format based on the strength
of the user response.
Task Description Menu Menu Menu Level
Level 1 Level 2 3 We then verbally interviewed the caregivers to gather
Task 1, Set A Hygiene Washing Water demographic data on education levels, language, and
and your shortage and occupation, telephone usage (mobile and landline). Caregivers
Find out how to wash
cleanliness hands hand washing were also interviewed on their familiarity with technology
hands during water
(computer, mobile phone, TV, radio, video/DVD machine)
shortage.
based on factors such as use, ownership, frequency of use,
Task 1, Set B Hygiene Body What to do if place of use, and reason for use. Data was also gathered on the
and fluids & gloves are not number of children they take care of, how often they visit
Find out how to protect
cleanliness infected available Baylor and how they usually resolve their information queries.
hands when gloves are not
waste At the end of each session caregivers were provided with
available.
small non-monetary incentives (juice, potato chips and fruit
Task 2, Set A Common Fever Home Care * for the child and gloves & household disinfectant for the
Sicknesses caregiver) to thank them for their participation.
Find out how to care for a
child with fever.
Task 2, Set B Common Diarrhoea Home Care *

Sicknesses
Find out how to care for a
child with diarrhoea.
* Home Care can only be reached after the user has heard the Danger Signs of
the selected illness. For this reason, Task 2 is slightly more difficult than Task
1.
Additionally, to minimize any possible bias due to the order

of trial of the DTMF and ASR systems or the use of a Task
Set (A or B) with a particular modality, we systematically
ensured that our data covered all possible combinations of
Order of Modality and Task Set (illustrated as Quadrants 1-4
101
Fig. 7. Methods of Health Resolution by caregivers.
In terms of technology familiarity, 30 out of 33 (91%)

caregivers owned mobile phones and 85% of these knew how
to load their mobile phones with ‘airtime’ (pre-paid phones
which require users to load money by calling the network
provider’s service number and entering a sequence of digits
from the pre-paid calling card). Average mobile phone costs
Fig. 6. A user taking part in the study. per month were 68 Pulas ($10.5 USD) with an average cost
per call being reported as 4.5 Pulas ($0.75 USD). Only 30%
III. RESULTS of caregivers reported having access to a landline telephone
In this section we present results from our user study, and of these only 9% had the landline at home (Table II).
including a description of our users (Section A), their task
completion scores, other usability metrics from the study and
a comparison of their performances on the DTMF and ASR
parallel systems (Section B).
A. User Profile
During our user study, we had 33 caregivers who
participated, of which 27 tried both DTMF and ASR systems.
The caregivers were all female with the exception of one male
caregiver. The age of our users ranged between 22 and 61
years old with the average age of 34 years. The average
number of years of schooling amongst our users was 9 years
but 2 users had 0 years of schooling. All of the users could
read and write the local language Setswana and approximately
Fig. 8. Time since last Health Question by caregivers.
79% of them knew some English. In terms of occupation, 47%
were unemployed and of the 53% who were employed, the
majority were in low-income occupations such as cook, TABLE II
cleaner, house maid, hair dresser, or security guard. SUMMARY OF TECHNOLOGY OWNERSHIP AND USAGE BY
Caregivers reported that they visit Baylor between 1-3 times a CAREGIVERS.
month, with average travelling distance and time at 28 km and Technology Use Ownership
1 hour respectively, some travelling from as far as 130 km,
with average cost of travel at 18 Pulas (approx. $3 USD). The Mobile Phone 91% 91%
average waiting time at Baylor reported by the caregivers was
2.5 hrs, also travel time to Baylor ranges from 30 minutes to 3 Landline 30% 9%
hours; together, these represent a significant portion of a Computer 15% 3%
working day and are a substantial burden to the caregivers.
Caregivers were also asked what they usually do if they TV 76% 71%
have questions regarding the child’s health and when the last Radio 91% 91%
time was that they had such a question. One third of the
caregivers usually go the local clinic to resolve their health DVD/Video Machine 41% 38%
queries and another quarter go to Baylor for this purpose (Fig.
7). Forty percent of the caregivers had a query regarding the ATM 35% N/A
child’s health within the last 6 months, and another 21% had
more recent queries; in contrast, only 12 % could not
remember specific queries (Fig. 8).
102
B. Broad Usability Metrics and Observations
System usability was determined using objective (Task

completion rate, response time, routing time) and subjective
(user preference, Likert ratings) metrics [27].
1) Response Time
Response Time measures the time it takes someone to

respond to the system for the first time. It is usually measured
not from the system start, but from the end of the Main Menu
prompt, which in our case was 29 seconds long. This results in
three basic categories of responses: (1) people who barge-in
will have a negative response time, (2) people who respond in
the 4 seconds of silence after the end of the Main Menu
prompt will have a response time between 0 and 4 seconds,
and (3) people who listen to one or several timeouts before
responding will have a response time greater than 4 seconds.
The Mean Response Times of our users (Fig. 9) indicates Fig. 10. Task Completion Rates for DTMF and ASR.
many barge-ins, roughly corresponding to the time the correct
option at the first menu level was played: Hygiene and For this study, we provide the task completion rate at each
Cleanliness for Task 1, Common Sicknesses for Task 2 (Table menu level of the SDS (Fig. 10). Across both tasks and both
I). Although we expected that our users would exhibit some input systems, 60% to 84.38% of caregivers selected the
short term learning effects, we found that response time did correct first menu option (Menu Level 1). Approximately half
not decrease from the first call to subsequent calls. of those people were able to select the appropriate sub-menu
option (Menu level 2) when the task was to find out about
“Common Illnesses” (Task 2), with the ASR system yielding
the highest task completion rate for this menu level (56.67%).
Caregivers seemed to have had more trouble correctly
selecting Menu level 2 options for Task 1, with ASR again
yielding the highest task completion rate of the two systems
(40.63%). Task completion rate for level 3 ranged from 2.85
% to 23.68%, with Task 2 causing the most difficulty. Rates
did not vary significantly by input mode (DTMF vs. ASR) or
by task set (A vs. B).
3) Routing Time
Routing Time measures the time it takes a user to reach the

beginning of the node which contains the correct information
for the assigned task. Fig. 11 shows the mean routing time
Fig. 9. Average Response Time (time from end of the Main Menu prompt from our user study. Users’ routing time was similar across
until the user's first input). both tasks and both systems for both Level 1 and Level 2. As
expected, due to the forced loop through Danger Signs for
2) Task Completion Rate Common Sickness, caregivers took longer at Level 3 for Task
2.
Task completion rate measures how frequently users were
able to reach the node that provided the correct information
for their assigned task.
103
Fig. 11. Routing Time for completed tasks. Recall from Task Completion
Rates (above) that the Level 1 mean response is averaged over approximately Fig. 12. Usability Metrics for DTMF and ASR systems.
75% of our users, whereas Level 2 and Level 3 times are averaged over fewer
users, in some cases only one or two.
5) User Preference
Only two users correctly completed Task 2 with the ASR Systems were rated separately using the PARADISE [26]
system; both users took over 4 minutes to do so. With the framework after use. The Likert scores were found to be
DTMF system only one person completed Task 2, doing so in unreliable, however. Despite efforts to elicit honest, critical
less than 3 minutes. All users that correctly completed Task 1 feedback to the system (for example, we had a different
using the ASR system did so in about 60 seconds. For Task 1 person in a separate room conduct the post-study evaluations),
on the DTMF system, most users completed the task in just all caregivers gave the system the highest marks possible
over 60 seconds, but one user took over 6 minutes. across all categories and were hesitant to provide any
criticism. For those 27 (out of 33 total) caregivers who tried
both the DTMF and the ASR system, we were able to elicit
4) Other Usability Metrics feelings of preference for one system over the other (Table
III). Most caregivers (59%) preferred the DTMF system over
Results from our user study show that user interaction with the ASR one (19%) and 22% indicated no preference. Both
the ASR and DTMF system are very similar across a wide the DTMF and the ASR systems were judged to be the faster
range of additional usability metrics, illustrated in Fig. 12. Use system by those who preferred it. Our measurements of
of the Main Menu global (press '0' or say 'Simolola') and the routing time and task completion, however, show the systems
Exit global (press '9' or say "Fetsa"), were similar for both are comparable.
systems. There were almost the same average number of TABLE III
timeouts (when 4 seconds elapsed with no user response) and USER PREFERENCE FOR DTMF OR SPEECH INPUT
repeats (when user chooses to repeat an information node) for
the ASR and the DTMF system. On average, caregivers used Preferred Num Reasons Given Example Remark
the barge-in function one more time when using the ASR System ber
system but the total number of turns taken by users was
similar for both systems. Note, all the usability metrics are DTMF 16 Clearer "Its quick and the
means for each call, whether the task was completed or not. instructions (7) doctor gives you the
Faster to use (4) instructions, you
More private (2) just have to follow
them."
ASR 5 More accessible "Its faster and old

(2) people can also use
Faster (1) it. The button
Clearer system takes too
instructions (1) long. "
Hands free (1)
Both or 6 Similar (2) "They are similar.

None With one you press
buttons, with the
other, you say
104
things." would not help the caregiver any further. In some cases, they
would ask to try again, which we allowed if time permitted.
6) Social Factor Correlations Most were very interested in the content and many referred to
the voice they heard as ‘the doctor’. Only one caregiver
We examined several social factors based on our users' recognized the voice of the SDS as the celebrity soap star
responses to questionnaires to see if they correlated with although most commented that “the ‘doctor’ explained very
either their interaction with the system, based on objective nicely”.
metrics, or their reported system preference (DTMF vs ASR).
During the interviews, all of our users enthusiastically
Employment and experience loading airtime on a mobile indicated that they would like to use the service again; many
phone were significant factors in the overall task completion said that it would be very valuable for educating themselves
rate (p=0.09 and 0.02, respectively). Those users who were and their family/friends on caregiving aspects for children
employed or who had experience loading airtime completed with HIV. A SDS such as OpenPhone could also serve as
more tasks during the study. Age, education, use of landlines, persuasion tool for caregivers trying to educate others, as
mobile phones, or ATM machines were not significant factors. explicitly reaffirmed by a caregiver “now I can tell them at
Only previous use of a landline was a significant factor in use home that the doctor (SDS voice) says the same thing
of system globals (p=0.01). Use of landlines was also a factor, (referring to a HIV related topic) that I’m telling them”.
along with loading airtime, in whether a user barged-in during
system use (p=0.1 and 0.1 respectively). Caregivers who use a
landline and load airtime were more likely to barge-in during IV. DISCUSSION
system use. No questionnaire responses were found to be
significant factors for system repeats or overall response time.
From our pilot study, we found that there were no
Employment, previous use of an ATM, and experience significant differences between task completion rates (ASR
loading airtime on a mobile phone were significant factors in only performed slightly better) or other usability metrics for
the overall correct response (p= 0.07, 0.09, and 0.03) 1 . both systems. This agrees with a number of previous studies
Caregivers who were employed, used ATM machines and in the developed world [12, 28, 29] where no major
loaded airtime had more correct responses to tasks during the differences were found in terms of performance. However,
user study. Age, education, and amount of mobile phone use subjectively the majority of our users preferred DTMF (59%)
were not factors in correct response. over ASR (19%), which is in contrast to formal studies in the
developed world [9]-[11] (where user preferences largely
Loading airtime was the sole significant factor in user
favour speech), but correlates with the observation that simple
preference of DTMF over the ASR system (p=0.1). Those
key-press replacement with keywords is generally not viewed
people who load airtime regularly preferred DTMF over ASR.
favourably. The users who did prefer ASR did not as in
Employment significantly correlated (p<0.1) with Overall
developed world studies comment on the aesthetics of speech
Task Completion and Overall Correct Response. Age and
input, that “speech is more entertaining or enjoyable” but
monthly mobile phone costs were not found to correlate
rather on the utility of speech “more accessible for older
significantly with user performance or user preference for
people or faster”.
either the DTMF or ASR system.
7) Other Observations Our finding that users’ employment and experience loading
airtime correlated with higher task completion but that
We observed based on body language and explicit remarks education level does not, indicates that technological literacy
that several caregivers were nervous at first, and then became is a more important factor in adopting new technology than
more relaxed during the first few minutes of the study. Many literacy itself. This may also contribute to the finding that
caregivers (and their children) showed signs of fatigue while ‘loading of airtime’ was the sole significant factor found in
trying the second system (sometimes an hour later). DTMF preference over ASR.
Caregivers sometimes had trouble understanding the task.
It is also interesting to note that our users who had minimal
They would often try to find information about ‘Nutrition’, for
exposure to SDSs (except loading airtime), were relatively
example, which was the topic of the demonstration, instead of
comfortable using our system for the first time, as indicated by
searching for information on their assigned task. Caregivers
their frequent, timely barge-ins. Also, our users noticed the
also often clarified with the interpreter what the task was,
value of speech (allowing hands-free operation,
what the keywords were, and what they should press or say
innumerate/older people being able to use it) and DTMF as
during their trial. The interpreter would nod or say 'yes' but
well (provides privacy). This highlights that even if users may
be technically inexperienced and unfamiliar with an ICT
1
The following categorical groups (yes and no) were divided into two application, they have valuable and sound judgment on the
groups and the resulting dependent values analysed: Employment, previous utility of the interface. Also, both DTMF and the ASR
use of an ATM, experience loading airtime, use of landlines, use of ATM
machines. A p-value of 0.1 and less indicates a significant difference between systems were judged (subjectively) to be the faster system by
the two groups, whereas a p-value higher than 0.1 indicate no significant those who preferred it, which indicates that time constraints,
difference. may also be of importance for low literacy users, in contrast to
105
earlier observations. The DTMF vs. speech input comparison interactive (perhaps with audio cues, getting intermittent user
could be improved in terms determining task performance by feedback) and conversational.
using a between-subjects experiment design; however this In addition, the nature of educational information services
approach would not reveal user preference. tends to be rather exploratory; where a user may peruse
various topics related to his/her general query (e.g. a user may
Our preliminary investigations indicated that a good fraction
want to, in general know about “Dealing with body fluids and
of caregivers were in the low and non–literate range. In our
infected waste” which has 4 topics all related to that option).
user study though, we encountered that many of our recruits
This in our case translated to some users struggling to find the
were semi- to low literate. Also, whilst our user numbers
exact menu option related to the task and exploring related
represent a significant sample size for a developing world user
sub-topics. One user even singled out that she would like to
study, it may in comparison to studies in the developed world
know the mapping of the menu options beforehand to help her
be on the ‘small’ end of the spectrum. The above-mentioned
locate the information she is looking for. This experience
issues emphasize some of the challenges faced in research for
highlighted challenges of using hierarchical menus [8] and the
the developing world; that the very users for whom an SDS
importance of paying attention to the taxonomy and
could be most useful may be the hardest for us to reach and
vocabulary of the system to enable easy navigation for the
also that user recruitment in developing regions can present
user.
significant obstacles [30].
SDS design for smaller languages also introduces challenges
Whilst this pilot study illustrated that telephony services
on other dimensions including prompting and persona. The
could in fact be easily used by semi and low literacy users,
prompts and content of a SDS application will typically be
and that an SDS in local language can be a powerful health
translated from a language such as English to the local
education tool, the decisive factor in widespread uptake is
language. Thus, great care has to taken in the prompt writing
likely to be the cost incurred by the caller for the service. The
phase to ensure that intended meaning of the original prompt
majority of caregivers said that even though the service would
(English) is still preserved in the translated prompt (local
be useful to them they would only be able to make use of it, if
language) and conveyed in the simplest and shortest way
the service is toll-free. An average phone call in Botswana of
possible. Often, a concept described by a single word in
5-10 minutes to the SDS would cost a mobile phone user $1-2
English has no direct translation in another language. For
USD. From our questionnaires, the average cost per month for
instance, whereas a keyword in the English version of our
mobile phone usage was $10.5 USD. Thus, a single phone call
application was “Safe food”, it became the phrase, “Dijo tse di
to the SDS would consume, 10-20% of a caregiver’s monthly
siameng” after translation, in order to adequately describe the
mobile phone budget, making the case for a toll-free number
concept.
all the more imperative.
Moreover, not only should the translated SDS prompts
Notwithstanding our findings that show promise for SDSs
convey the intended meaning but the designer should ensure
for low literacy users, we did encounter challenges in
that the persona of the local language system is in line with
introducing the concept of a SDS to our users, for instance a
cultural and contextual expectations of the intended audience.
few users did not fully realise that the system was automated.
For instance we ensured our prompts not only had the right
For example, at the end of one call, a user proceeded to ask
balance of formality and gravity appropriate for the message
the ‘nurse’ (system persona) a question when prompted by the
(HIV info) but were also understandable and conversational.
system to leave a comment (and waited for the answer).
Another user repeatedly acknowledged what the ‘nurse’ was
Our experience in employing multiple data-gathering
saying by responding with “Yes, yes” or “I agree with you”.
techniques in the needs investigation phase (interviews and
An obvious solution here might be to use a text-to-speech
discussions, observations, field visits, and focus groups) better
(TTS) voice for the prompts. However, we run the risk of a
equipped us in trying to comprehend the needs of our users.
mismatch between the target audience and the context (e.g.
The interviews and discussions helped us establish a rapport
emotionally sensitive in terms of health care) or culturally-
with our users and stakeholders (Baylor staff), whilst also
based communication norms of the community. This in turn
providing the flexibility of follow-up questioning and
may affect the willingness of users to interact with the system.
recalibration of interviewer terminology when needed.
Observations on the other hand helped to further reveal the
A general design challenge for SDSs is to ensure a minimal
issues that users are unable or unwilling to articulate or
cognitive load on the user. This magnifies more so in the case
express. Field work allowed us to pick up on the cultural
of information access applications where lengthy pieces of
nuances and social-economic context of the users and enabled
information need to be provided (e.g. in a health care context).
us to gain a deeper understanding of the sensitive environment
In our case, all the users indicated in the subjective
(HIV/health) we were working in. Finally, through focus
questionnaire that the length of the content was not too long;
groups we were able to observe the interaction amongst
however, during the experiment some users did mention the
caregivers and most importantly it enabled us to obtain
need to concentrate in order “not to miss all the things being
specific targeted design information in a group setting and
said by the ‘nurse’ ”. There is a need to address the above-
correct our earlier assumptions on the ranking of SDS health
mentioned issues by developing the dialog to be more
topics.
106
REFERENCES
[1] R. Tucker and K. Shalonova. "The Local Language Speech Technology
V. CONCLUSION Initiative", SCALLA Conference, Nepal, 2004.
[2] SIL International. (2001). Facts about illiteracy. [Online] Available at::
http://www.sil.org/literacy/litfacts.htm (last accessed 15 Sept 2008).
This paper has addressed the frequently debated question of [3] ITU: International Telecommunications Union. (2003). Mobile
input modalities of touchtone vs. speech in a completely new overtakes fixed: Implications for policy and regulation. [Online].
Available:
context; low literacy users and a domain (health) different
http://www.itu.int/osg/spu/ni/mobileovertakes/Resources/Mobileovertak
from the usual business centric applications (call routing, es_Paper.pdf, (last accessed 5 September 2008).
voice mail systems or banking). Our pilot study also served to [4] Statistics South Africa. (2007). Community Survey 2007. [Online]
confirm the feasibility of SDS applications for semi and low Available at: http://www.statssa.gov.za/publications/P0301/P0301.pdf
literate populations. (last accessed 15 Sept 2008).
[5] E. Barnard, L. Cloete, and H. Patel. “Language and Technology Literacy
Barriers to Accessing Government Services,” Lecture Notes in Computer
It is interesting to speculate on the applicability of our Science, vol. 2739, pp. 37-42, 2003.
findings to other resource-poor countries. Our intuitive [6] M. Plauche, U. Nallasamy, J. Pal, C. Wooters, and D. Ramachandran.
judgment is that the only characteristic of our user population “Speech Recognition for Illiterate Access to Information and
Technology,” in Proc. IEEE International Conference on Information
that strongly influenced our results was their relative and Communications Technologies and Development ‘06, pp. 83-92,
familiarity with mobile telephones; however, the matter May 2006.
requires detailed investigation. In future work we would [7] P. Nasfors. "Efficient Voice Information Services for Developing
therefore like to further explore the space of non-literate users Countries", Master Thesis, Department of Information technology,
Uppsala University, Sweden, 2007.
and the suitability of SDS applications for them, as well as [8] J. Sherwani, N Ali, S. Mirza, A. Fatma, Y. Memon, M. Karim, R.
investigate the interaction of task types (linear vs. non-linear) TRongia and R. Rosenfeld, “Healthline: Speech-based Access to Health
and application domain – informational (Openphone) vs. Information by low-literate users”, in Proc. IEEE International
transactional (tracking a social services payment) with the Conference on Information and Communications Technologies and
Development ‘07, Bangalore, India, Dec. 2007.
input modality. Once the system is free of charge and has been [9] S. Agarwal, A Kumar, AA Nanavati and N. Rajput, "VoiKiosk:
in use for several months, we would like to study whether Increasing Reachability of Kiosks in Developing Regions", in Proc. of
OpenPhone leads to a change in health habits and improves the 17th international conference on World Wide Web, pp. 1123-1124,
the ability of a caregiver in providing care to children, like 2008.
[10] K. M. Lee and J. Lai “Speech Versus Touch: A Comparative Study of
change of hygiene habits, better nutrition, and fewer the Use of Speech and DTMF Keypad for Navigation” International
misconceptions about HIV. Journal of Human–Computer Interaction, vol. 19, no. 3, pp.343–360,
2005.
Throughout the design and development process, we [11] B. Suhm, J. Bers, D. McCarthy, B. Freeman, D Getty, K Godfrey and P
Peterson, "A comparative study of speech in the call center: natural
experienced that beyond usability and creating simple,
language call routing vs. touch-tone menus". In: Terveen, Loren (ed.)
accessible user interfaces for low literacy users, factors such Proc. of the ACM CHI 2002 Conference on Human Factors in
as cultural and social context, establishing relationships with Computing Systems Conference, Minneapolis, Minnesota. pp. 283-290,
stakeholders and user communities, and localizing content, April 2002.
play a vital role in the success of an ICT intervention in the [12] C. Delogu, A. Di Carlo, P Rotundi and D. Sartori, "Usability evaluation
of IVR systems with DTMF and ASR", in Proc. of the 5th International
developing world. We intend to draw on these valuable Conference on Spoken Language Processing (ICSLP 98), paper 0320,
experiences and carry our work forward in serving the Australia,1998.
information needs of citizens of developing regions with [13] T. Pearce and M Bergelson. (2008). Alignment index for speech self-
accessible and usable telephony based services. service. Dimension Data Technical Report [online]. Available at
http://www.dimensiondata.com/NR/rdonlyres/9191A848-5F35-459F-
8239-
8D9D2248414E/8791/mainstreamspeechalignmentindexreport2.pdf
ACKNOWLEDGEMENTS (Last accessed 15 Sept 2008).
[14] J.P. Migneault, R. Farzanfar, J.A. Wright, and R.H. Friedman. “How to
The authors wish to thank the staff of Botswana Baylor write Health Dialog for a Talking Computer,” Journal of Biomedical
Children’s Clinical Centre of Excellence for the generous Informatics, vol. 39, no. 5, pp. 468 – 481, Oct. 2006.
manner in which we were received, in particular Dr Paul [15] Tellme Networks Inc. Who uses TellMe? [Online]. Available at:
http://www.tellme.com/about (Last accessed 15 Sept 2008).
Mullan, for the amount of help and time given to us. We [16] UNAIDS (August, 2008). Report on the Global AIDS Epidemic. Joint
would also like to thank members of the HLT group at United Nations Programme on HIV/AIDS. [Online] Available at:
Meraka Institute who provided valuable contributions http://www.unaids.org/en/KnowledgeCentre/HIVData/GlobalReport/200
throughout the project; Victor Zimu, Jama Ndwe, Mpho 8/2008_Global_report.asp (last accessed 7 Sep 2008).
[17] K. Seipone, W. Jimbo, F.d.l.H. Gomez, K. Ampomah, J. Othwolo, O.
Kgampe, Louis Joubert, Richard Carlson, Bryan Mcalister,
Kaluwa, M. Busisiwe. “Trends in HIV Prevalence Among
Mark Zsilavecz, Marelie Davel and Alta de Waal. We also
Pregnant Women in Botswana 2001-2005” 16th Intl. Conf
greatly appreciate the time given to us by Connie Ferguson, Aids. Toronto, Canada, Aug. 2006.
our celebrity voice talent. This research would not have been [18] Botswana-Baylor Children’s Clinical Centre of Excellence, Annual
possible without the joint support of OSI and OSISA. The Report 2006, Gaborone, Botswana, 2006.
project was also in part supported by the NRF under the Key [19] A. Van Dyk, HIV Aids Care & Counselling, a Multidisciplinary
Approach. 2005, Third Edition: Pearson Education, South Africa.
International Research Capacity (KISC) programme, UID no.
63676.
107
[20] D. Werner, C. Thuman and J. Maxwell. Where there is no Doctor, a

Village Care Handbook.. 2007, New revised edition: Hesperian, CA,
USA.
[21] R. Granich and J. Mermin. HIV, Health & your Community, A Guide for
Action. 2001, Hesperian, CA, USA.
[22] G.A. Miller. "The Magical Number Seven, Plus or Minus Two: Some
Limits on Our Capacity for Processing Information," The Psychological
Review. vol. 63, no. 2 pp. 81-97, 1956.
[23] Suhm B. “IVR Usability Engineering using Guidelines and Analyses of
end-to-end calls” in D. Gardener-Bonneau and H.E. Blanchard (Eds).
Human Factors and Voice Interactive Systems. 2008, pp. 1-41, Second
Edition, Springer Science: NY, USA.
[24] N.M. Fraser and G.N. Gilbert, “Simulating speech systems”, Computer
Speech and Language, vol. 5, no. 2, pp. 81-99, 1991.
[25] I. Medhi and K. Toyama, “Full-Context Videos for First-Time, Non-
Literate PC Users” ”, in Proc. IEEE International Conference on
Information and Communications Technologies and Development ‘07,
Bangalore, India, Dec. 2007.
[26] M.A. Walker, D. Litman, C.A. Kamm and A. Abella. “PARADISE: A
general framework for evaluating spoken dialogue agents” in Proc. of
the 35th Annual Meeting of the Association of Computational
Linguistics. ACL/EACL 97, 1997.
[27] Nielsen, J., (1993). Usability Engineering. AP Professional, Boston,
MA, USA.
[28] Foster, J. C., McInnes, F. R., Jack, M. A., Love, S., Dutton, R. T., Nairn,
I. A., et al. (1998). An experimental evaluation of preference for data
entry method in automated telephone services. Behaviour & Information
Technology, 17, 82–92.
[29] Goldstein, M., Bretan, I., Sallnas, E.-L.,&Bjork, H. (1999). Navigational
abilities in voice-controlled dialogue structures. Behaviour &
Information Technology, 18, 83–95.
[30] E. Brewer, M. Demmer, M. Ho, R.J. Honicky, J. Pal, M. Plauche, and
S. Surana. “The Challenges of Technology Research for Developing
Regions,” IEEE Pervasive Computing, vol. 5, no. 2, pp. 15-23, April-
June 2006.
108
ICT4What? – Using the Choice Framework to

operationalise the Capability Approach to
Development
Dorothea Kleine
development outcomes top-down and a-priori is unsuitable in

Abstract— Identifying the specific contribution of the use of the context of multi-purpose technologies which could
ICTs to specific development goals has proven to be extremely empower individuals to attain development outcomes of their
difficult. This paper argues that instead of trying to make ICTs own choosing.
fit with a linear conceptualisation of impacts and an often
economistic view of development, ICT4D should be used as a
prime example of a development process which has to be The paper is structured in six parts. In section one I briefly
analysed in a systemic and holistic way. Amartya Sen’s capability introduce Sen’s capability approach before presenting, in
approach offers a way of thinking about development not as section two, some important steps towards operationalising it.
economic growth, but as individual freedom. The Choice Building on this body of work, section three develops the
Framework is presented as a way of operationalising this Choice Framework as a further way of operationalising Sen’s
approach and visualising the elements of a systemic
conceptualisation of the development process. An individual case
approach. Methodological implications are explored in section
study, related to telecentres in rural Chile, is used to demonstrate four before in section five the Choice Framework is applied in
the way the Choice Framework can be applied as a guide to a a case study of one particular individual’s usage of the
systemic and holistic analysis. Internet in a telecentre in rural Chile. The final part points out
limitations of the model, directions for further empirical
Index Terms— ICT4D, Amartya Sen, capability approach, research and calls for more theoretical work on the nature of
Choice Framework, telecentres, Chile the “development” element in ICT4D. The paper concludes
by highlighting some concrete implications this theoretical
work may have for practitioners.
I. INTRODUCTION
T he paradox is this: ICTs and particularly the internet are

widely regarded as groundbreaking inventions that have
changed the way millions of people live their lives, and yet
II. EVOLVING THEORIES OF DEVELOPMENT
Research located in the contested intellectual space that is
researchers and practitioners in the field of ICT and
‘development’ needs to be able to answer the fundamental
development often struggle to prove specific impacts of the
question of what is understood as development. Broadly
technology to funders. There may be specific reasons why
speaking, debates in development studies range from positions
particular projects fail, even some generalisable patterns of
which equate development with economic growth (e.g. [2],
failure [1], but the overall degree to which the ICT4D
[3], [4], [5]) through to critical perspectives stressing that
community has to struggle when trying to legitimise its work
uneven development, dependency and inequality are inherent
to funders is astonishing in the context of a general discourse
in capitalist development (e.g. [6], [7], [8], [9]) to ideas of
about how much these technologies have changed our lives.
alternative, bottom-up development recognising social and
This paper tries to unravel the reasons behind this paradox by
ecological as well as economic goals (e.g. [10], [11]), and
arguing two fundamental points: On a theoretical level, while
radical “post-developmentalist” critiques that often dismiss the
there have been interesting alternative theoretical approaches
entire “development project” altogether (e.g. [12], [13]).
to development, including Amartya Sen’s capability approach,
the mainstream discourse’s conceptualisation remains heavily
The most influential challenge to the mainstream growth-
focused on economic growth, which is too narrow to capture
focused view of development has come from Amartya Sen’s
the impacts of ICT. Secondly, and on a practical level, the
capability approach (also known as the capabilities approach)
common way of measuring impact by defining the intended
in which development is defined as “a process of expanding
the real freedoms that people enjoy to lead the lives they have
Manuscript received September 22, 2008. This work was supported in part
by the Dr. Heinz Dürr Fellowship Programme. Author’s Details: Dorothea reason to value” [14]. His understanding focuses on
Kleine is Lecturer in Human Geography at Royal Holloway, University of development as freedom of choice. While this understanding
London, Egham, TW20 0EX and a member of the UNESCO Chair/ICT4D of development is a minority position within institutions such
Collective (email: dorothea.kleine@rhul.ac.uk).
109
as the World Bank [15], Sen’s approach has influenced the A. Empowerment
development discourse and it provides a means of building a One of the most interesting attempts to operationalise Sen’s
bridge between those working in international development ideas is offered by Alsop and Heinsohn [31]. Writing for the
organisations and researchers in academia. World Bank, they link choice with their definition of
empowerment 2 . They define empowerment as “enhancing an
The approach developed by Amartya Sen ([16], [17], [18], individual’s or group’s capacity to make effective choices and
[19]) argues that development is about the freedom of choice translate these choices into desired actions and outcomes”
in the personal, the social, the economic and the political [39]. ICTs could be seen as useful tools in such processes of
sphere. In Sen’s approach, “functionings” are the various empowerment
things a person may value doing or being, such as being
adequately nourished, being healthy and being able to take Alsop and Heinsohn see material and non-material assets,
part in the life of a community. In Sen’s terminology a or resources, as the basis of individual agency which, together
person’s “capability” refers to the alternative combinations of with the structural conditions frame empowerment processes.
functionings that are feasible for her/him to achieve [20]. The In their attempt to use empowerment as a middle-range
focus of development thus becomes increasing a person’s theoretical concept to convert the development paradigm of
capability set, or her/his substantive freedom to lead the life choice into a construct that is of use to practitioners, Alsop
she/he values. Functionings represent the “outcome” and Heinsohn build a crude framework which connects
component, while capabilities are the “freedom” component in “individual agency” with an “opportunity structure” from
this approach. 1 In Sen’s more holistic view of development, which follow the degree of empowerment an individual has to
economic growth plays an important, but not exclusive, role. achieve development outcomes. The different “degrees of
Sen suggests ontologically focusing on human wellbeing and empowerment” are: existence of choice, use of choice and
methodologically focusing on capabilities. achievement of choice [40]. Individual agency is measured by
an individual’s asset endowment, consisting of
While welcoming Sen’s approach at offering a more “psychological, informational, organisational, material, social,
holistic view of development, scholars have been struggling to financial or human” assets [41]. These assets are listed, but
find a balance between its conceptual richness and its not defined. An actor’s opportunity structure is said to be
potential to be operationalised. Several scholars [21], [22], shaped by the “presence and operation of the formal and
[23], [24] have attempted to operationalise the approach. informal institutions” [42] and measured by the presence and
Within this field the majority of studies use capabilities as a operation of laws, social norms and customs. Alsop and
normative basis for the research while measuring functionings Heinsohn have applied their framework in the evaluation of
as a proxy [25], owing to the practical difficulty of measuring World Bank projects with women, on rural water supply and
capabilities. Some authors (e.g. [26], [27]) have challenged sanitation, on school decentralisation and with school drop-
Sen to draw up a general list of capabilities, but Sen has outs.
refused to do so, claiming that specific lists of capabilities
ought to be drawn up for a given research or policy context
[28] and, crucially, that the process of choosing capabilities B. The sustainable livelihood framework
should be left to the individual [29]. The dilemma which Another literature which can be linked to the capability
emerges is how to apply the capability approach to specific approach is the literature on livelihoods. Based on earlier
areas or sectors in a meaningful way while retaining open- work on livelihoods ([43], [44], [45] the Sustainable
ended development outcomes that do not presuppose Livelihood Framework (SLF) used by the UK Department for
individuals’ choices. International Development (DFID) [46] offers an analytical
tool to understand in a systemic way the elements influencing
the lives of the poor. Duncombe has demonstrated how the
II. OPERATIONALISING SEN’S APPROACH SLF can be applied to ICT4D research with microenterprises
Sen intended his approach to be combined with other [47], while retaining the focus on poverty reduction through
theoretical approaches [30]. The following section explains economic growth. The SLF includes the concept of an
how, in order to operationalise the approach for ICT4D and individual’s “capital portfolio” made up of five “capitals”:
other areas of development, elements have been drawn human capital, natural capital, financial capital, physical
together from the literature on empowerment and on capital and social capital.
sustainable livelihoods to enhance the application of Sen’s
approach. In operationalising the SLF, human capital is measured by
formal education and health indicators, but there has been a
2
The concept of empowerment originated in work on gender relations and
community participation (e.g. [33], [34]) and has been increasingly discussed
in development studies (e.g. [35],[36], [37], [38]. There are several competing
1
For a more in-depth discussion of the capability approach, see also [32]. definitions of the term.
110
struggle to quantify “social capital” [48]. As a result, critics A. Outcomes

have argued that “everything social” gets packed into the True to Sen’s statement that choice is both the aim and the
social capital variable [49]. principal means of development [51], the primary
development outcome is choice itself. Secondary development
Individuals own or have access to this portfolio of capitals, outcomes depend on the individual’s choice as to what lives
their “livelihood assets” with which they negotiate “policies, they value. These may include, for example, easier
institutions and processes”. They operate within a communication, increased knowledge, more income or time
“vulnerability context” and develop livelihood strategies saved. Information and communication technologies might
which then result in livelihood outcomes. Livelihood prove useful tools in achieving these outcomes. Just like other
outcomes are defined a priori – In the DFID version, “more attempts to operationalise Sen’s work, here capabilities are not
income” is listed at the top, even before “increased measured directly, though participatory research with
wellbeing”. The SLF offers a broad and systemic view of individuals and groups may reveal them to some degree.
development processes, but its Mainly, the outcome component will map or measure the
STRUCTURE
Key:
ER = Educational Resources SR = Social Resources
• institutions and organisations PsR = Psychological Resources NR = Natural Resources
In = Information MR = Material Resources
• discourses FR = Financial Resources GR = Geographical Resources
CR = Cultural Resources He = Health
• policies and programmes
• formal and informal laws
including:
- norms on usage of space DEVELOPMENT
- norms on usage of time OUTCOMES
• access to ICTs
- availability of ICTs
- affordability of ICTs Principal: Choice
- necessary skills for ICTs
DIMENSIONS OF
CHOICE
Secondary:
• existence of choice • easier communication
• sense of choice • increased knowledge
• access to markets
AGENCY • use of choice
• business ideas
SR
• achievement of • increased income
He CR choice
• more voice
Age
ER NR • time saved
Gender
Ethnicity • higher job satisfaction
PsR MR
... •...
In GR
FR
Fig. 1: The Choice Framework
set of capitals is limited and in it the development goals are achieved functionings resulting from an individual’s choices
predetermined and not up to the individual to choose. In this as a proxy to the capabilities. 3 An analysis based on the
respect, the SLF fails to mirror the thinking behind Sen’s Choice Framework would then work backwards, from the
approach. outcomes, into the systemic relationships between agency,
structure and choice, thus analysing how the outcomes were
arrived at.
III. THE CHOICE FRAMEWORK
Based on Sen’s capability approach, inspired by Alsop and B. Dimensions of Choice
Heinsohn’s work on operationalising Sen’s work, taking
Alsop and Heinsohn’s dimensions of choice, which they
elements from the SLF and informed by an in-depth research
call “degrees of empowerment” include, firstly, the existence
project with microentrepreneurs’ use of ICTs in Chile [50],
the Choice Framework was developed. After presenting it in 3
Two disadvantages of this method are that some of the individual’s
diagrammatic form (Fig 1), the following sections will in turn capabilities are not captured in the achieved functionings and that it is difficult
explain each of the key components of the framework. to trace the choices related to apparently negative outcomes. However, so far,
capturing functionings is methodologically easier and more precise – in
regards to both quantitative and qualitative methods - than capturing
capabilities.
111
of choice – whether the different possibilities exist and are, in face to face contact in the urban economy [55]).
principle, attainable for the individual if the combination of
their resource portfolio and the structural conditions would Human Resources: The term “human resources” has been
allow it. The second dimension, a sense of choice, not used for decades in the economics and industrial relations
originally included by Alsop & Heinsohn, was added as a literature. 4 In the Choice Framework, this term needs to be
result of fieldwork experiences relating to ICT and disaggregated into Health and Education and Skills
development. Individuals were aware of some possibilities the (educational resources). Within Sen’s paradigm of
new technology offered them, like email and online chat, but development, good health is a prerequisite for a person’s
not of others, like Voice over IP. This was precisely because ability to choose the life she/he values. Educational resources
their educational resources (including computer skills) and the represent education and skills acquired through formal and
dominant discourse in the Chilean media stressed some usages informal means.
over others. For any piece of research focused on a technology
which is new to the respondents, the dimension of “sense of Psychological resources: Alsop and Heinsohn [56]
choice” will play a significant role. The “use of choice” recognise the significance of “psychological assets” and give
dimension refers to whether or not an individual actually as an example “capacity to envision”. More broadly,
makes the choice and the “achievement of choice” refers to psychological assets may include self-confidence, tenacity,
whether the outcome matches the choice expressed. optimism, creativity and resilience. Spiritual or religious
beliefs stand in complex interrelation with psychological
resources – they can strengthen or weaken an individual’s
psychological resources.
C. Agency
Instead of using a terminology of capitals and asset or Information: Alsop and Heinsohn list informational assets
capital portfolios, Sen uses the term “resources” within the as a key resource. Heeks [58] calls for putting information at
capability approach [52]. Resources can be interpreted as the centre for analysis of ICTs and Development, and Gigler
individual agency-based capability inputs which, together [59], adds “informational capital” to the capital portfolio.
with structure-based capability inputs, can be converted into Access to information is the first step to knowledge
capabilities [53]. acquisition, the process of filtering and transforming
information into meaningful knowledge.
In the Choice Framework age, gender, ethnicity etc. are
conceptualised as personal characteristics of an individual Cultural resources: “Cultural capital” – which in the Choice
which may in a given social context become related to socially Framework is called cultural resources – exists, according to
constructed axes of exclusion and influence the scope and Bourdieu [60], in three states: an embodied state (the habitus a
scale of the resource portfolio. The resource portfolio consists particular person lives in); an objectified state (objects like
of: paintings, instruments and monuments which only the
initiated can use or appreciate); and an institutionalised state
Material resources: These sum up the material objects (prestige attached to, for example, academic titles).
owned, such as machinery, computer hardware and other
equipment. They are also essential inputs in the production Social resources: “Social capital” – or social resources – is
process. included in both the SLF and Alsop and Heinsohn’s work. It
has been both immensely influential and highly contested in
Financial resources: These stand for financial capital in all development discourse. For the Choice Framework,
its forms (cash, savings, shares etc.). The ability to obtain Bourdieu’s definition of social capital is used:
credit is a combination of the structural character of the
banking rules and individual collateral. “the aggregate of the actual and potential resources which
are linked to possession of a durable network of more or less
Natural resources: This includes issues such as institutionalized relationships of mutual acquaintance and
geomorphological and climatic conditions in a locality and recognition – or in other words, to membership in a group –
related aspects such as soil quality and the availability of or which provides each of its members with the backing of the
access to water as well as the attractiveness of the surrounding collectivity-owned capital, a “credential” which entitles them
nature. to credit, in the various senses of the word.”[61]
Geographical resources: Covers the practical implications Membership of these groups can be defined by kinship,
of location and relative distances, and also includes the friendship, shared ethnicity or class, friendship or informal
intangible qualities of a location alluded to by writers from commonality ties.
Marshall (who refers to the mysteries of the trade “in the air”
[54]) to Storper and Venables (who describe the “buzz” of 4
For examples, see the collection by Fitzgerald and Rowley [57]
112
diagram of the Choice Framework.

Thus, these ten types of resources – material, financial,
natural, geographical, psychological, cultural, social, and The Choice Framework is an attempt to operationalise the
educational (education and skills) resources; health; and capability approach in a holistic and systemic way, thus
information – represent an attempt to holistically map aspects maintaining much of its conceptual richness. While it may
of the agency element of the systemic framework. prove particularly useful in the area of ICT4D, the framework
could also be applied in other areas of development work.
However, it is important to recognise that this resource-
based agency can only be realised within the confines of and
in systemic interaction with a given structure. This aspect of
IV. IMPLICATIONS FOR METHODOLOGY
the Choice Framework will be analysed in the following
section.
The basic challenge that the capability approach offers to
the orthodox methodologies of development research, and
D. Structure
ICT4D in particular is that, on a fundamental level, it
Both the empowerment framework suggested by Alsop and questions the validity of outcomes that are defined a priori and
Heinsohn and the SLF take into account not only individual without consulting the individual in question. Both the
agency, but also structures which aid or constrain this agency. inclusion of a development goal and its position within a set
Alsop and Heinsohn list “formal and informal laws, of development priorities, however, relate to the question
regulations, norms and customs” [62] as elements of this which kind of life people would choose to live and this,
structure, while the DFID SLF includes these as laws and according to Sen, is what development is about. A funding
“culture” – the latter running the risk of being used as a kind institution or government may set, say, economic prosperity as
of black box into which all locally specific aspects can be the top priority. Once basic needs such as food and shelter
subsumed. The SLF includes not only laws, but also policies, have been met, however, an individual may value being close
institutions and processes. Rules, laws, norms and policies are to family members above earning more money, or may value a
embedded in, and often emanate from discourses, and healthy environment for themselves and their children over
hegemonic discourses can define the thinkspace in which economic growth. In the practice of development projects, this
policies, including ICT policies, can be conceived. Thus means that before undertaking an intervention designed to
discourses are included as part of the structure element of the improve people’s lives and later measuring its effectiveness,
Choice Framework. practitioners and researchers would have to ask individuals
about their own development priorities and let these guide the
In particular with respect to ICTs, relevant elements of the planning, implementation, monitoring and evaluation of
structure which influence an individual’s agency include development projects and programmes. Work done in this
dimensions of access, such as availability, affordability and area includes participatory monitoring and evaluation [64]
capabilities needed for using different ICTs [63]. To avoid and, in a broader context, initiatives around participatory
confusion with Sen’s use of the word, the term “skills” is used budget planning [65].
instead of capabilities. These dimensions of access are
nationally and often locally specific, path dependent and Setting development priorities in a participatory way may
embedded with other elements of the structure. make the process of development planning more complicated,
but it comes with major benefits: Firstly, morally it is the right
Structural factors such as these stand in a complex thing to do to engage the people themselves in the decisions
relationship with an individual’s resource portfolio. For that will affect their lives. Secondly, if the outcomes have
example, with the help of social resources an individual might been agreed upon in a participatory way, they are more likely
have access to the internet (at a neighbour’s house) which to be locally and culturally appropriate and may reduce the
might lead to frequent email contact with a distant family rate of failure. Thirdly, a participatory process will lead to
member, thus increasing occasions of, in Bourdieu’s terms, greater local buy-in to measures and therefore higher future
legitimate exchange with both the neighbour and the distant institutional sustainability. Fourthly, such a process harbours
relative, in turn potentially increasing social resources. the chance that the current overly economistic focus of
Similarly, a person with higher educational resources (skills development work can be broadened to include
and education) and information might find it easier to use the environmental, social and cultural aspects and thus better
existing access facilities to enhance their skills and gain mirror the diversity of the kinds of things people value in their
information. The interface between the opportunity structure lives. Last, and perhaps not least, ICT and development
and individual agency thus includes a host of reciprocal and practitioners work with multi-purpose technologies which
cumulative processes. Structural constraints need to be offer far more significant changes to people’s lives than the
recognized as being as important an element as individual economic impact they have been proven to have. Moving
agency. To reflect this, structure is placed above agency in the away from an a priori, top-down and often overly economistic
113
set of development priorities offers the chance to recognise here in the telecentre.
the diversity of the contributions ICTs can make to the social,
cultural, environmental and economic aspirations individuals If one were to apply a typical questionnaire on telecentre
may have for their lives. usage to this case, this woman’s usage experience might be
subsumed in the category “personal usage” or “other”. Yet the
The Choice Framework is one way of conceptualising such following section will offer a careful application of the Choice
choices by the individual. There are some specific Framework to this case:
implications for research design which result from the model:
A focus on the individual’s own development outcomes
A. Outcome
means that the research needs to start from these wished-for
outcomes, measure the degree to which they have been The primary outcome was that the respondent had improved
attained and work systemically backwards through structure, choice, in this case, between “no visit to Kaiserslautern” and
agency and choice to understand how these outcome have “virtual visit to Kaiserslautern”. The secondary outcome
come about. The extensive list of resources covers six less achieved was defined by the individual: “to see more of the
tangible resources (social, cultural, educational, psychological world” – which in this case, translated into “virtual visit to
resources, health and information) which pose challenges to Kaiserslautern” - in Sen’s terms an “achieved functioning”.
measurement but need to be taken into account. On the The aspiration was “visit to Kaiserslautern”, which since it is
structure side, access to ICTs is conceptualised as availability, feasible, could be seen as a “capability” in Sen’s terms. The
affordability and necessary skills. ICTs are embedded in the achieved functioning ”virtual visit” is not equal to the
wider set-up of institutions, policies, programmes, norms and capability “visit in person” but it is an improvement in
discourses. As such they need to be analysed as firmly and outcome over no visit at all.
historically engrained in the societies they affect.
Studies of development outcomes, or more commonly of
impacts, often operate with a set of impacts as defined by the
V. APPLYING THE FRAMEWORK: IMPACTS OF TELECENTRES IN funding body, government, international organisation or
RURAL CHILE commercial sponsor. This set of impacts then acts as a
checklist informing the construction of questionnaires and
The following example is part of an extensive ethnographic
interview guides, possibly with some scope for “other
study of how state ICT policies affected microentrepreneurs in
activities”. Sen’s approach, with the individual’s choice as the
rural Chile [66]. When interviewing microentrepreneurs and
primary outcome, however, would suggest that the analysis
their partners who were using a telecentre located in a public
needs to start from the ground up, asking people about what
library in rural Chile, open-ended questions revealed that apart
lives they value and what outcomes they want to see. For this
from business-related usages such as looking up prices of
individual, one of the greatest impacts the telecentre had made
machinery on the internet (carpenters), looking up photos of
was that it had given her the chance to virtually visit
furniture models (carpenters), communicating via email with a
Kaiserslautern, something few policymakers or researchers
supplier (carpenters, spice vendor) or buyer (spice vendor),
would have predicted. Indeed, some might question whether
looking up recipes (cake vendor), and looking up guidelines
this is a valid “development outcome” or “impact” for a
for government business assistance (carpenters), there were
telecentre. In Sen’s approach, expressed via the Choice
several answers which reflected what in many studies of
Framework, it is.
telecentre usage is described as “personal usage”.
Respondents’ faces lit up when they described how they now
could exchange emails or chat with relatives who were living B. Agency
abroad, children who were studying in a larger town or family The individual in question was a Chilean-mestizo 5 woman
members who were working as temporary labour on fruit in her 50s, married with four children who were all grown up
farms or in the mines in the north of Chile. One respondent, a now and had left the home. Her material resources did not
woman in her 50s whose household income was around 440 include a computer and internet access at home and her
USD per month and who together with her husband ran a financial resources made it difficult for her to spend money on
carpentry business, described how while the Word Cup was using a computer in the local cybercafés. However her social
on in Germany in 2006 she visited the world cup site to find resources (contacts with friends) had helped her gain the
links and take virtual tours of some of the German cities she information that there was free access to the internet available
was not able to visit in person. As a young woman, she had at the telecentre in the local library. Her geographical
had a pen friend from Kaiserslautern, and while he had come resources (the location of her house) and her state of health
to visit her in Chile, her dream of visiting him had never been were such that she could easily reach the telecentre on foot.
possible because, she said, the money she saved had been She had the cultural resources to not feel intimidated when
spent on her children’s education. Eventually, they had lost
touch, but now, she told me with tears in her eyes, over 25 5
i.e. not considered part of the indigenous minority which had historically
years later, she was finally able to “visit” Kaiserslautern, right been discriminated against.
114
entering a space like a library and to know the behavioural information, and psychological resources, knew that the
code there. With the help of her social resources (knowing the telecentre offered a computer and a good connection, had
librarian who was now also the director of the telecentre), her acquired the skills to navigate the Internet and run an
educational resources (literacy, rudimentary English) and her application in the free digital literacy courses offered at the
psychological resources (extrovert, willingness to ask telecentre. She felt that informal, gendered social norms
questions) she quickly learnt how to use the computers. The allowed her to go to the telecentre during the morning before
information she gained online, together with her psychological having to prepare lunch. Thus she developed a sense of
resources (curiosity, tenacity) allowed her to understand the choice, was able to choose (use of choice) and achieved her
choices she had and find the site which offered the virtual tour desired outcome (achievement of choice).
of Kaiserslautern, thus achieving her chosen development
outcome.
VI. CONCLUSION
C. Structure Applying the Choice Framework to this particular case
The agency of the individual is a shaper of, and is shaped allows us to firstly, theorise the use of ICT in a systemic and
by the structure in which it operates. In this case, as part of the procedural way which reflects the systemic and pervasive
national ICT policy, the Agenda Digital, the state of Chile had impact of ICT. The “impact of ICT” is not conceptualised in a
signed an agreement with the Bill and Melinda Gates cause- and effect chain, instead effects are carefully
Foundation for them to provide 9.2 million USD worth of disaggregated and their systemic interrelatedness and co-
hardware to be installed in public telecentres based in libraries causality is demonstrated. Secondly, the Choice Framework
around the country, running Microsoft software. The local offers a way to operationalise Sen’s capability approach in the
library was an existing institution which was able to context of ICT. Sen’s approach is currently the most well
accommodate the hardware, delivered as part of the known heterodox alternative to orthodox, growth-focused and
Biblioredes telecentre programme (availability of ICTs). The often economistic conceptualisations of development. Given
Chilean digital literacy campaign provided free IT courses to the enormous potential of ICTs to give individuals choices,
adults, and public discourse in Chile stressed the importance and indeed a greater sense of choice, Sen’s approach is of
to become “digitally literate”, so the woman had taken the particular interest to the ICT and development research
course (necessary skills). The formal rules for users of the community.
telecentre stipulated that access was free (affordability of
ICTs) but limited to 30 minutes per person at busy times, and There are three obvious limitations to the application of the
the informal rules were that people were left to use the Choice Framework, and this is where more theoretical work
computers on their own unless they asked for help from the needs to be done:
telecentre director. Norms on the usage of space made it easy
for her to go to a library as a mestizo woman with a completed Firstly, the Choice Framework aims to be comprehensive in
school education who was known in town. However, norms its modelling of the complex relationships between agency,
on the usage of time meant that she could only use the structure, degree of empowerment and outcome, and this
telecentre when she was not supposed to be home preparing automatically entails a trade-off with the depth of theorisation
meals (gendered norms on time) or when she was expected to of each element. Behind each of the terms included in the
attend to customers (business norms on time). framework lies a wealth of theoretical literatures which may
need to be synthesised for different research purposes and key
issues brought to the attention of researchers in the
D. Dimensions of choice development field. While for example, social resources can be
In the Choice Framework, an individual’s resource-based theorised by linking to the wider debate on social capital (see
agency can operate within a given structure to achieve degrees [66]), which has been received in the development studies
of empowerment, such as existence of choice, sense of choice, discourses, work on cultural capital (in Bourdieu’s sense) is
use of choice and achievement of choice. In this case, both the hardly ever linked to development discourses in the South.
choice “travel to Kaiserslautern in person” and “take a virtual
tour of Kaiserslautern” existed, the latter only since the links Secondly, the Choice Framework is relatively easily applied
were offered via the World Cup website in 2006. In a in qualitative work on the micro-level of the individual. A
capitalist market system, however, the former choice required further challenge will be how to apply the framework to
an amount of financial resources which the individual felt groups of individuals, communities, or even nations. Within
unable to dedicate to this idea. The choice “take a virtual tour this and related to a theoretical tension evident in Sen’s
of Kaiserslautern” however, required a good internet original approach, there is a complex relationship between
connection, a computer, the knowledge that the tour was individual and collective choice which will have to be
available via the website, the skills to find and run it, and time. conceptualised carefully.
The individual, thanks to among others, her social resources,
115
The example used here to illustrate the potential of the Secondly, there are some macro-methodologies which
Choice Framework is part of a far more extensive reflect the ethos of giving people the power to choose.
ethnographic and longitudinal study of how the state ICT Voucher schemes are a good pragmatic way to monitor, in a
policies affected microentrepreneurs in rural Chile [66]. In heavily supply-driven development field such as ICT4D, what
this study, connections are made between individual and products (hardware, software etc), services (trainings,
collective choice. For example, the local authority’s public computer repair, communication etc) and content (economic,
procurement policy was analysed as an expression of social, political, cultural etc) people would, after considering
collective choice, but related to individual’s views as to how their options, actually choose. From the field of participatory
their tax money should be used to create the community they urban planning come methodologies for participatory budget
wanted to live in. However, further empirical work is design, where communities get to debate and decide which of
necessary to gather experiences in the applicability of the their desired outcomes to prioritise and pursue. This is a
Choice Framework in other cultural and socio-economic practical and democratic way to aggregate individual
settings. The relationship between individual choice and capabilities in order to enable collective decision making, and
collective choice needs to be conceptualised carefully in these could also be used for ICT4D.
local contexts in order to allow for the empirical application of
the Choice Framework at the more aggregate level such as so- Thirdly, practitioners may deduce that if the ideal is for
called target groups and communities. development projects’ intended outcomes to reflect the
individual’s choices, then the more individuals are aggregated
The third limitation is a very practical one: funders prefer to a group, the less probable it is that they can agree on a
predefined and clearly measurable impacts. The Choice similar set of capabilities. From this follows that the further
Framework, however, suggests that impacts of ICTs occur in a down the directional control continuum an ICT4D project is,
systemic, pervasive and transversal way, and that outcomes the more sensitive/locally customised it has to be to the
should be defined, in line with Sen’s approach, by the choices of a smaller number of people. Big, uni-directional
individual, based on their choices as to what kind of life they development programmes with specific, a priori defined
value. There are, however, some funders who are open to desired outcomes designed for a large number of people are
methods such as participatory evaluation and monitoring, and most likely to be in contradiction to a people-centred holistic
this means there might be hope for genuinely people-centred development process as proposed by Sen and expressed in the
development work – and development theory. Choice Framework.
There are some key implications emerging from this Ultimately, researchers working on ICT, particularly the
theoretical work for practitioners of ICT. Firstly, while no internet, need to consider the question: Should we try and fit a
technology is ever completely politically neutral [67][68], groundbreaking, multi-purpose and potentially liberating
ICT4D projects can be placed on a continuum of “directional technology into orthodox notions of development – such as
control”. At one end there are projects and programmes more ICT for higher GDP, more ICT for better school results
which focus on providing people with access to a technology et cetera – impacts which we may struggle to prove? Or can
which is recognised as multi-purpose, like some telecentre the field of ICT and Development serve as a test case and
projects. On the other end of the continuum are projects and breeding ground for thinking about development in a more
programmes which carry a much more narrow set of holistic way, putting the individual and their own choices at
intentions, for example training microentrepreneurs to use a the centre of development? If the latter is the case, then we
specific e-procurement system in order to “train them” to have plenty of work to do, but the Choice Framework may
operate in a more competitive market environment under a serve as one part of the big puzzle we have to begin putting
specific set of rules [67]. The further down the directional together.
control continuum a particular project and programme is
located, the more risk there is that the intended outcomes of
an ICT4D project diverge from the capabilities, or desired ACKNOWLEDGMENT
outcomes individuals in the so-called target group would I would like to thank Diane Perrons, Robin Mansell,
choose. Thus, the more directional control is involved in the Alexandra Norrish, Macarena Vivent, Rodrigo Garrido, Tim
project or programme, the more participation of the set of Unwin, and four anonymous reviewers for their comments on
individuals who are the intended group will be needed to earlier versions of this paper.
reduce this gap. This would include conceptualising the
development process as open-ended and the so-called target
group as individuals empowered to choose the lives they
themselves value. Participatory project design and
participatory monitoring and evaluation techniques would be
most appropriate.
116
Chile, unpublished PhD thesis, Department of Geography and

Environment, London, London School of Economics, 2007
[33] C. Moser, Gender Planning and Development, London, Routledge, 1991
REFERENCES [34] W. Van Eyken, The Concept and Process of Empowerment. The Hague,
[1] R. Heeks, “Information systems and developing countries: Failure, Bernard van Leer Foundation, 1991
success and local improvisations” in The Information Society, vol. 18, [35] J. Friedmann, Empowerment: The Politics of Alternative Development,
no. 2, pp. 101-112, 2002 Oxford, Blackwell, 1992
[2] W.A. Lewis, “Economic development with unlimited supplies of [36] N. Kabeer, “Resources, agency, achievements: reflections on the
labour”, The Manchester School of Economic and Social Studies, vol.22, measurement of women’s empowerment”. Development and Change,
no. 2, pp.139-91, 1954 vol. 30, no.3, pp.435-464, 1999
[3] G. Myrdal, Economic Theory and Underdeveloped Regions, London, [37] P. Oakley (ed.), Evaluating Empowerment, Reviewing the Concept and
Duckworth, 1957 the Practice, Oxford, INTRAC, 2001
[4] A. O. Hirschman, The Strategy of Economic Development, Yale, Yale [38] A. J. Bebbington, M. Woolcock, M. Guggenheim and E.A. Olson (eds),
University Press, 1958 The Search for Empowerment. Social Capital as Idea and Practice at
[5] W. Rostow, The Stages of Economic Growth: A Non-Communist the World Bank, Bloomfield, CT, Kumarian Press, 2006
Manifesto, Cambridge, Cambridge University Press, 1960 [39] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice –
[6] A.G. Frank, Capitalism and Underdevelopment in Latin America, Structuring Analysis and Framing Indicators, Washington D.C.: World
London, New York, Monthly Review Press, 1967 Bank, p.5, 2005
[7] T. Dos Santos, Dependencia y Cambio Social, Santiago, Universidad de [40] Ibid, p.6
Chile, 1970 [41] Ibid, p.8
[8] M. Blomström and B. Hettne, Development Theory in Transition. The [42] Ibid, p.9
Dependency Debate and Beyond: Third World Responses, London, Zed [43] R. Chambers and G. R. Conway, “Sustainable Rural Livelihoods:
Books, 1984 Practical Concepts for the 21st Century”, Discussion Paper 296, Institute
[9] C. Kay, Latin American Theories of Development and of Development Studies, University of Sussex, UK, 1992
Underdevelopment, London, New York, Routledge, 1989 [44] A. J. Bebbington, “Capitals and Capabilities: A Framework for
[10] M. Nerfin (ed.), Another development: Approaches and Strategies, Analyzing Peasant Viability, Rural Livelihoods and Poverty”, World
Uppsala, Dag Hammarsköld Foundation, 1977 Development, vol. 27, no.12, pp. 2021-2044, 1999
[11] R. Chambers, Rural Development: Putting the Last First, Harlow, [45] D. Carney, Sustainable Livelihood Approaches Compared, London,
Longman, 1983 Department for International Development
[12] A. Escobar, Encountering Development: The Making and Unmaking of [46] DFID, Sustainable Livelihoods Guidance Sheets, Department for
the Third World, London, Longman, 1995 International Development, London, 1999
[13] J. Nederveen Pieterse, “After post-development”, Third World [47] R. Duncombe, “Using the Livelihoods Framework to Analyze ICT
Quarterly, vol. 21, no.2, pp.175-191, 2000 Applications for Poverty Reduction through Microenterprise”,
[14] A. Sen, Development as Freedom, Oxford, Oxford University Press, Information Technologies and International Development, vol. 3, no. 3,
1999, p.3 pp.81-100, 2006
[15] J. Harriss, Depoliticizing Development – The World Bank and Social [48] DFID, Sustainable Livelihoods Guidance Sheets, Department for
Capital, London, Anthem Press, 2001 International Development, London, 1999
[16] A. Sen, “Equality of what?”, in S. McMurrin (ed.) The Tanner Lectures [49] A. Munasib, Social Capital at the Individual Level: A Reduced Form
on Human Values, Salt Lake City, 1980 Analysis. Available:
[17] A. Sen, Resources, Values and Development, Cambridge, MA, Harvard http://munasib.myweb.uga.edu/papers/SocialCapitalDemography.pdf,
University Press, 1984 2004
[18] A. Sen, “Capability and well-being”, in M. Nussbaum and A. Sen (eds.) [50] D. Kleine, Empowerment and the Limits of Choice: Microentrepreneurs,
The Quality of Life, Oxford, Clarendon Press, 1993 Information and Communication Technologies and State Policies in
[19] A. Sen, Development as Freedom, Oxford, Oxford University Press, Chile, unpublished PhD thesis, Department of Geography and
1999 Environment, London, London School of Economics, 2007
[20] A. Sen, Development as Freedom, Oxford, Oxford University Press, [51] A. Sen, Development as Freedom, Oxford, Oxford University Press,
1999, p.3 1999
[21] M. Nussbaum, Women and Human Development: The Capabilities [52] A. Sen, Resources, Values and Development, Cambridge, MA, Harvard
Approach, Cambridge, Cambridge University Press, 2000 University Press, 1984
[22] S. Alkire, Valuing Freedoms, Oxford, Oxford University Press, 2002 [53] I. Robeyns, “The Capabilities Approach: An Interdisciplinary
[23] D. Clark, Visions of Development, Cheltenham, Edward Elgar, 2002 Introduction”, Department of Political Science and Amsterdam School of
[24] I. Robeyns, “Sen’s capabilities approach and gender inequality: selecting Social Sciences Research Working Paper, University of Amsterdam,
relevant capabilities, Feminist Economics, vol. 9, no 2-3, pp. 61-92, Amsterdam, 2003
2003 [54] A. Marshall, Principles of Economics, London, Macmillan, 1920, reprint
[25] I. Robeyns, “The Capabilities Approach: An Interdisciplinary edition 1961
Introduction”, Department of Political Science and Amsterdam School of [55] M. Storper and A.J. Venables, “Buzz: face-to-face contact and the urban
Social Sciences Research Working Paper, University of Amsterdam, economy”, Journal of Economic Geography, vol. 4, no. 4, pp. 351-370,
Amsterdam, 2003 2004
[26] R. Sugden, “Welfare, resources and capabilities: a review of Inequality [56] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice –
Re-examined by Amartya Sen”, Journal of Economic Literature, vol. Structuring Analysis and Framing Indicators, Washington D.C., World
XXXVI, pp. 1947-1962, 1993 Bank, 2005
[27] J. Roemer, Theories of Distributive Justice, Cambridge, Cambridge [57] R. Fitzgerald and C. Rowley (eds.), Human Resources and the Firm in
University Press, 1996 International Perspective, Cheltenham, Edward Elgar, 1997
[28] A. Sen, “Capability and well-being”, in M. Nussbaum and A. Sen (eds.) [58] R. Heeks, “Information and communication technologies, poverty and
The Quality of Life, Oxford, Clarendon Press, 1993 development”. Development Informatics Working Paper No.5,
[29] A. Sen, “Maximisation and the act of choice”, Econometrica, vol. 65, Manchester, IDPM, 1999
no.4, pp.745-779, 1997 [59] B.S. Gigler, “Including the excluded – can ICTs empower poor
[30] A. Sen, Inequality Re-Examined, Oxford, Clarendon Press, 1992 communities? Towards an alternative evaluation framework based on
[31] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice – the capability approach,” presented at the Fourth International
Structuring Analysis and Framing Indicators, Washington D.C.: World Conference on the Capability Approach, Pavia, Italy, Sept. 5-7, 2004
Bank, 2005 [60] P. Bourdieu, “The forms of capital”, in: J. Richardson (ed.), Handbook
[32] D. Kleine, Empowerment and the Limits of Choice: Microentrepreneurs, of Theory and Research for the Sociology of Education, New York,
Information and Communication Technologies and State Policies in Greenwood Press, 1986
117
[61] Ibid, p.249

[62] R. Alsop and N. Heinsohn, Measuring Empowerment in Practice –
Structuring Analysis and Framing Indicators, Washington D.C., World
Bank, 2005, p.9
[63] R. Gerster and S. Zimmermann, Information and Communication
Technologies (ICTs) for Poverty Reduction?, Bern, Swiss Agency for
Development Cooperation, 2003
[64] I. Guijt and J. Gavena, Participatory Monitoring and Evaluation:
Learning from Change, Brighton, Institute of Development Studies,
1998
[65] A. Shah (ed.) Participatory Budgeting. Washington D.C., The World
Bank, 2007. Available:
http://siteresources.worldbank.org/PSGLP/Resources/ParticipatoryBudg
eting.pdf
[66] D. Kleine, Empowerment and the Limits of Choice: Microentrepreneurs,
Information and Communication Technologies and State Policies in
Chile, unpublished PhD thesis, Department of Geography and
Environment, London, London School of Economics, 2007
[67] D. Kleine, “The ideology behind the technology – Chilean
microentrepreneurs and public ICT policies”, Geoforum, vol. 40, no.2.,
2009
[68] D. Kleine, “Striking a Balance”, Engineering and Technology, February
2007, p. 30-33
118
ICTD for Healthcare in Ghana:

Two Parallel Case Studies
Rowena Luk, Matei Zaharia, Melissa Ho, Brian Levine, and Paul M. Aoki
by-side analysis of two completely independent projects which

Abstract — This paper examines two parallel case studies to arose from similar objectives but resulted in two very different
promote remote medical consultation in Ghana. These projects, strategies. Two sets of researchers came to the same country to
initiated independently by different researchers in different
work with the same community, identified the same problems,
organizations, both deployed ICT solutions in the same medical
community in the same year. The Ghana Consultation Network and proposed two different solutions. Both projects were
currently has over 125 users running a Web-based application implemented over the course of 2007-2008, both aimed to
over a delay-tolerant network of servers. OneTouch tackle remote medical consultation among Ghanaian doctors,
MedicareLine is currently providing 1700 doctors in Ghana with and both sought nationwide deployment over the course of the
free mobile phone calls and text messages to other members of project lifetime. However, the Ghana Consultation Network
the medical community. We present the consequences of (1) the
(GCN) was initiated by a group of technologically-oriented
institutional context and identity of the investigators, as well as
specific decisions made with respect to (2) partnerships formed, researchers, while the OneTouch MedicareLine (ML) was
(3) perceptions of technological infrastructure, and (4) high-level initiated by a public health researcher and social entrepreneur.
design decisions. In concluding, we discuss lessons learned and GCN’s solution consisted of a Web application hosted on a
high-level implications for future ICTD research agendas. delay-tolerant network of computers running in each hospital
Index Terms — Remote medical consultation. and on the open internet. ML focused instead on a mobile
phone program involving a combination of technological
I. INTRODUCTION services and business innovation. We discuss the rationale
ICTD research focuses primarily on matters relating to its behind decisions made by each party over the course of the
target populations and their conditions, such as how project and the resulting outcomes.
individuals, communities and institutions interact with The main contribution of this paper is to reflect on some
technology. For the most part, investigators only become part ways in which the context and background of researchers
of the scene as objects of retrospective critique, usually in the affect the structure of ICTD projects. A second contribution is
context of having failed to see some vital point and achieve in identifying relevant factors which will assist in the design
some critical goal. As with any kind of research, however, and execution of ICTD projects in the future. In particular, we
ICTD research is a product of social, cultural and professional provide specific examples of how an investigator’s
influences on the investigators, influences that affect every institutional context and identity affect not only the
decision made over a project’s lifetime. It is extremely methodology used but also the interpretation of findings; how
difficult to reflect upon the effects of these influences because the partnerships chosen for co-development and co-
each project is so uniquely framed by the contexts of its deployment have a fundamental role in the development and
investigators and its target setting that it is hard for analysts to deployment of the technology; how ‘objective’ assessments of
imagine how “it could have been otherwise” [19]. existing technology infrastructure are influenced by personal
In this paper, we address this reflective gap through a side- areas of expertise; and how projects can be designed with
various degrees of technological ‘specificity’ and the resulting
Manuscript received September 22, 2008. This material is based in part implications for their impact on various development
upon work supported by the U.S. National Science Foundation under Grant indicators. We expect these findings to be of interest to the
No. 0326582. M. Zaharia is supported by the Natural Sciences and
Engineering Research Council of Canada community of ICTD researchers and practitioners as a whole.
R. Luk is with AMITA Telemedicine Inc., Toronto, ON, Canada (e-mail: The paper is organized as follows. We first present a
rowena@amitatelemedicine.org). background overview of some of the underlying issues of
M. Zaharia is with the Department of Electrical Engineering and
Computer Sciences, University of California, Berkeley, CA, USA (e-mail:
healthcare in Ghana and briefly introduce both projects. We
matei@berkeley.edu). then discuss in turn the framing of the research problem and
M. Ho is with the School of Information, University of California, how it affected the partnerships formed and resulting
Berkeley, CA, USA (e-mail: mho@ischool.berkeley.edu).
deployment strategies; the assessments of technological
B. Levine was with New York University, New York, NY. He is now
with the Department of Obstetrics and Gynecology, New York Presbyterian infrastructure by the project members; and the usage and
Hospital, Columbia University Medical Center, New York, NY, USA. appropriation of the two solutions. Finally, we discuss high-
(email: DrBrianLevine@gmail.com). level implications for ICTD research and present works
P.M. Aoki is with Intel Research, Berkeley, CA, USA (e-mail:
aoki@acm.org). related to each of these sections.
119
II. BACKGROUND
In this section, we highlight some of the background issues
of healthcare in Ghana and provide an overview of the two
projects which are the focus of our case studies.
A. Healthcare in Ghana
Like many other countries, Ghana has a tiered healthcare
system in which cases that cannot be handled by an institution
at a given tier are referred to institutions above it in the
hierarchy. All referral chains culminate in one of two teaching
hospitals, both located in the South, where specialists have the
training and resources necessary to carry out more complex
procedures. Doctors, and in particular specialists in areas such
as internal medicine, are highly concentrated in the urban Figure 1. GCN distributed, delay-tolerant server network
South. This sometimes leaves only two or three doctors to
serve in district hospitals in the rural North; predictably, rural (c)
(c) new
new thread
thread (a)
(a) primary
primary case
case list
list
doctors are confronted with heavy workloads (meaning they
can spend only a few minutes face-to-face with a given patient
per day) and isolated working environments (which prevent
them from taking advantage of many of the educational and
collaborative programs available to doctors in the South).
Formal and informal consultation is highly integrated into
the life of all doctors. Between hospitals, many doctors call
(d)
(d) navigation
navigation menu
menu
personal contacts – friends, colleagues, classmates – to seek
advice, and within hospitals this behavior is even more
frequent. Such “curb-side” consultation has long been a
common observed characteristic of medical practice [16] but it
is of critical importance in environments where specialist
expertise is spread thinly.
Continuity of care is difficult to ensure. Patient records are
almost universally paper-based. Despite attempts by some
major hospitals to transition towards an electronic system,
none of the 18 hospitals visited had a working system. Within
a major hospital, there is nothing to guarantee complete
patient records besides a doctor’s own discretion and multiple
(b)
(b) ““other
other cases”
cases” list
cases” list
reminders from administration. It is not uncommon for records
to be lost, confused, or incomplete. Figure 2. Key elements of GCN’s welcome screen
The communication infrastructure in the country is severely
limited. Hospital landlines are frequently out of service, into a local server (hosted at some participating hospitals) or
forcing many doctors to rely on personally-purchased mobile by logging into one of the two public servers (hosted with
phones. Likewise, broadband Internet infrastructure is Internet service providers (ISPs) in both Ghana and the U.S.).
unevenly available, frequently unreliable where available, and Providing local servers ensures availability and responsiveness
dependent on hospital budget allocations. Other options are to the users in the face of unreliable network connectivity and
available throughout the country, such as dial-up, satellite, and makes the task of synchronizing data between servers
mobile data plans, but these are generally expensive, transparent to them. Synchronization is automated and carried
unreliable, and/or slow as well. out over a disconnection-tolerant messaging layer called
OCMP [29] which draws inspiration from the research into
B. The Ghana Consultation Network (GCN)
delay-tolerant networking (DTN) [7],[12].
Solution Because the doctors already view consultation as a matter of
GCN is an end-to-end, computer-based system providing reaching out to personal contacts, the system is presented as a
doctor-to-doctor medical consultation on a network of servers social networking platform – a forum for medical consultation
implementing a distributed, asynchronously-synchronized with social and professional colleagues – and leverages social
database (Figure 1) [20]. The goal is to allow doctors incentives using principles drawn from the HCI and CSCW
throughout Ghana to consult with each other as well as communities [20]. The system supports two types of
medical professionals (in particular, the large Ghanaian ‘conversations’: highly structured ‘consultations’ for specific
medical diaspora) around the world. Doctors access the patients (which work much like an electronic case history) and
system through a Web-based UI (Figure 2), either by logging unstructured ‘discussions’ (which work much like online
120
forums). So, for example, a doctor unsure of how to treat a

particularly resistant case of tuberculosis might fill out various
fields of the ‘consultation,’ address it to a colleague, and wait
for advice to be appended to that consultation, whereas
another interested in general updates on malaria treatment
might create a question under ‘discussions,’ addressed to any
interested doctor. Recognizing the ubiquity of mobile phones,
GCN also incorporates text message (as well as email)
notifications of new content; yet the core of the interaction is
Figure 3. ML Phase 1 calls for free phone calls and text messages; the
designed for computers. This is based on feedback during the planned ML Phase 2 calls for MMS and data reports over SMS, and ML
design process that the wealth of data required for patient Phase 3 calls for free smartphones, reference tools, and custom applications.
management would demand a screen size larger than those
present in mobile phones. C. The OneTouch MedicareLine (ML)
To date, over 125 doctors have been enrolled from Ghana,
the U.S., Mali, Nigeria, South Africa, and the U.K. and 39 Solution
consultations have been submitted. MedicareLine is a program currently offering free calls and
text messages between any registered physician and/or
Methodology surgeon within Ghana. Its current focus has been on reducing
GCN is the product of a conventional user-centered design logistical and economic barriers to mobile phone use rather
process. There have been four rounds of iterative design and than on technological innovation. After submitting the
fieldwork, starting with exploratory needs assessment (2005) required paperwork, a doctor registered with the Ghana
and continuing in conjunction with design exercises (2006- Medical Association (GMA) receives a OneTouch GSM SIM
2007), a pilot deployment (2007), and an ongoing deployment which can be used with a privately-purchased mobile phone.
(2008). Overall, interviews or focus groups were conducted Using this SIM, the physician can now call other program
with 132 doctors in 15 hospitals throughout Ghana [20]. participants free of personal charge. For example, a physician
The most recent fieldwork in Ghana lasted six weeks in can call a specialist in the capital or a friend in a rural town to
mid-2008. Exploratory interviews on mobile phones and ask or provide medical consultation. This can be a significant
computer usage were conducted with 35 internal medicine cost saving, especially given that airtime in Ghana is relatively
(IM) doctors at a major teaching hospital. Further, six expensive compared to many developed countries. The GMA
evaluation interviews at one regional hospital and two also has a computer terminal that can send “blast” texts to all
interviews at one district hospital were conducted (both sites participants for updates and notifications.
of the original pilot). Limitations of the methodology include Future phases of the program envision new technological
an overrepresentation of internal medicine doctors from urban interventions (Figure 3). Phase 2, as yet uncompleted, calls
hospitals. for physicians to receive free MMS service so that they can
While the initial rounds of fieldwork focused on rural augment their phone consultations with photos (e.g., of a skin
hospitals which are more numerous in the North, the current condition or X-ray). Phase 2 will also allow the GMA and
live deployment is centered in the more accessible South, with other government organizations to collect data from
plans to extend later. We have established public servers in physicians via SMS. In Phase 3, ML anticipates partnering
both Ghana and the US, as well as local servers in 3 major with hardware vendors to provide each physician with a
hospitals and 2 rural district hospitals in the South. These smartphone preloaded with medical reference software.
district hospitals suffer from some of the same problems as in Phase 1 of this program has already experienced a very high
the North – a shortage of doctors and poor travel and rate of adoption. Approximately 1700 of 2000 doctors in the
communication infrastructure – although the severity of these GMA have enrolled, with over 2 million calls made to date.
challenges is much less. Our current strategy is first to recruit
specialists from the major hospitals and to test our system in Methodology
the district hospitals of the South before reaching out to more ML arises from a five-week visit to Ghana in October 2007
challenged environments in the North. and a two-week visit in March 2008. The purpose of the first
Table 1 – Comparison of GCN and ML

Ghana Consultation Network (GCN) OneTouch MedicareLine (ML)
Primary platform Computers Mobile phones
Adoption 89 doctors in 5 hospitals in Ghana >1700 doctors in Ghana
>125 doctors total from around the world
Problem addressed Network connectivity Cost of cellular airtime
Target user Doctors in Ghana and Ghanaian medical diaspora Doctors in Ghana
Partners (see Table 2) Ministry of Health, GPSF, and KNet (ISP) GMA and OneTouch (mobile operator)
Assumptions Adequate internet connection quality and coverage Adequate mobile phone quality and coverage
punctuated by regular power and network outages
Specificity (see Section VI) High – custom system software and Web application Low – only generic mobile phone service
Deployment strategy Incremental (hospital by hospital) National (available country-wide at launch)
121
A. GCN
GCN began as part of an ongoing collaboration between a
U.S. corporate research laboratory and a U.S. research
university. Specifically, it was a project of a joint research
group which had worked extensively in the area of low-cost
connectivity and delay-tolerant networking. Project funding
came from U.S. government grants as well as the corporate
sponsor. In the first brainstorming/conception phase of this
project, the axes by which ideas were evaluated were defined
as: (1) direct social impact (e.g., improvements in healthcare
delivery), (2) medium-term impact on ICT adoption in
developing regions (e.g., finding novel ways to make ICT
Figure 4. Timeline of GCN and ML Projects, including fieldwork. more relevant in addressing local problems), and (3) long-term
contribution towards HCI research. The primary short-term
visit was to introduce and promote usage of a social deliverables were software, real deployments, and research
networking Web system for international medical papers. As for the individual investigators, all came from a
collaboration that had already been developed in the States; technology and research background, collectively with
however, within two weeks of arriving in Ghana, the experience in systems, networks, and HCI.
investigator decided that access to both computers and the
Internet was a fundamental problem. Discarding the Web- B. ML
based project entirely, he moved on to exploratory interviews ML began as a project funded by a social entrepreneurship
with all stakeholders, including over 30 doctors, politicians, program and the international health program of the School of
local businessmen, and administrators, focusing on the issue Medicine at a U.S. research university. As part of the
of barriers to communication, innovation, and current international research and education component of a medical
technology usage. He conducted site visits to five regional program, there were no short-term deliverables. The individual
hospitals and one polyclinic in the urban South, and visited investigator comes from a medical program with prior
Tamale, a northern regional capital, for four days. Towards the experience in medical research and a personal interest in
end of that trip, the investigator orchestrated a “meeting of the consumer technology. His personal goal was to identify a
minds” between the CEO of OneTouch and the head of the project with sustainable, wide-spread social impact that had
GMA, whereupon an initial understanding was reached. the potential to be financially self-sufficient and reproducible
Implementation was left to the two partnering agencies, which in other developing countries.
announced the program the following month and launched it C. Discussion
in January 2008. In March 2008, the investigator returned to
These differing contexts had far-reaching implications for
conduct another round of unstructured interviews with over 20
how the problem was framed. First, coming from a more
individuals, again from diverse communities, in order to
technological background, GCN investigators were in a better
assess the program’s progress.
position to attempt technological innovations and, in particular
In addition to ML’s own follow-up interviews, material for
given their knowledge of projects such as [9] and [22], on
this paper also comes from 10 interviews conducted by the
creative ways of addressing poor connectivity. The ML
GCN investigators with doctors at one of the teaching
project’s lack of technological expertise limited the possibility
hospitals, as well as their interviews with the ML investigator
of developing new technology. Second, ML, given its ties to a
directly.
major U.S. hospital, had a natural predisposition to tackle
problems internal to a specialist center, whereas GCN had no
III. THE ROLE OF PROJECT CONTEXT IN FRAMING PROBLEMS
expertise in hospital administration and hence was more
ICTD work frequently involves the interdisciplinary inclined to address the simpler logistics of the rural clinic.
participation of various communities of research and of Third, ML’s agenda for large-scale social change predisposed
practice [5], but what are the implications of working within it towards impacting the greatest number of doctors, who are
these communities, given similar projects with similar goals of course concentrated in the urban South with better mobile
over similar timeframes? Both projects clearly did some phone coverage. GCN chose to focus design efforts on the
amount of fieldwork in both the urban South and the rural problems of the North (such as those of the 14 doctors
North, both purportedly wanted to develop a solution that servicing half a million people in Ghana’s Upper West
would work for all Ghanaian doctors, but GCN focused on region), seeing this as the best use of limited resources. These
connectivity in the rural North using innovative technology different problem framings have various implications for the
while ML focused on building communication in the urban solutions developed and implemented, as we discuss in the
South using proven technology. In this section, we examine following sections.
the institutional, cultural, and personal contexts from which
each project arose.
122
Table 2 – Healthcare institutions / partners

Acronym Name and Description
GMA Ghana Medical Association
A voluntary association of 95% of Ghana’s doctors,
representing their interests nationally
MOH Ministry of Health
The national health administration, responsible (through the
Ghana Health Service) for all public hospitals
GCPS Ghana College of Physicians and Surgeons
Responsible for specialist education and certification
GPSF Ghana Physicians and Surgeons Foundation
U.S.-based non-profit organization promoting specialist
education in Ghana
Figure 5. Map of GCN current and anticipated partners.
IV. PARTNERING STRATEGY AND DEPLOYMENT

The question of partner selection is a key one for ICTD
deployments, as partners are the usual means by which
technology makes it from the laboratory to the field. In this
section, we examine the impact of local and non-local partners
on technology design and deployment.
A. GCN
GCN chose partners based on the decision to deploy
technology in hospitals to connect their doctors with the
Figure 6. Map of ML current and anticipated partners.
Ghanaian diaspora (Figure 5). The Ghana MOH was chosen
as a local institutional partner because it had the central channels, doctors took personal initiative to read the GMA
authority to allow the servers to be installed in public emails or check GMA bulletin boards; this lead to widespread
hospitals. With the initial approval of the MOH, GCN was awareness of ML. OneTouch was chosen as a technological
able to work quickly with the public hospitals, conducting 121 service provider for political and pragmatic reasons. As the
interviews with relatively little trouble alongside iterative, investigator stated:
incremental deployment of technology. Of course, the
drawback of the centralized approach is that it requires OneTouch was the national company. I wanted this to be
convincing a risk-averse bureaucracy in advance, thereby Ghanaian – by Ghanaians, for Ghanaians. That says a lot
running the risk of ‘over-selling’ the solution. In addition to more than ‘done by Ghanaians in concert with South
the added effort, another unexpected difficulty of working Africans.’ [2]
through the MOH was overloaded communication channels. OneTouch had the means and the expertise to very quickly
Advertising GCN to the doctors was a challenge because make the ML program a reality and the financial resources to
promotional material needed to work through the same support the program independently on an ongoing basis. Thus,
mechanisms through which the MOH communicated, for while GCN’s deployment was incremental, ML’s was all-or-
example, minor procedural changes and optional seminars nothing, a strategy consistent with ML’s social agenda for
from pharmaceutical companies. So there were multiple rapid, wide-scale impact. Of course, the drawback of working
instances when administration would send out notifications of entirely through OneTouch was that the project was no longer
GCN training sessions but only a small fraction of doctors under the control of the investigator but was now subject to
would show up. GCN partnered with the Ghana Physicians the organizational vagaries of a for-profit corporation. That is,
and Surgeons Foundation (GPSF), a Ghanaian medical changes in OneTouch’s business priorities could result in the
diaspora organization, to recruit medical consultants in the program being dropped as quickly as it was initiated. This
U.S. KNet, a small Ghanaian Internet service provider (ISP), question is immediately salient in light of the recent
provided local Web hosting. acquisition of OneTouch by a multinational carrier based
outside of Ghana (in fact, in South Africa).
B. ML
ML worked with a smaller number of partners, all non- C. Discussion
governmental (Figure 6). ML chose the GMA as a local In considering the non-local partners, GCN and ML bear
institutional partner because it represented the interests of striking similarities in that both were initiated in partnership
physicians and its leaders would immediately see the benefit with U.S.-based organizations hoping to foster communication
of the project to the physicians as individuals, if not to the with the Ghanaian medical community: GCN with GPSF and
healthcare system as a whole. The program is framed as a ML with a U.S. teaching hospital. However, GCN emphasized
GMA value-added member service (similar to other member the role of GPSF throughout the course of its lifetime whereas
benefits such as professional development seminars and ML stopped working with U.S. doctors in order to focus
quarterly publications). In contrast to the MOH information explicitly on the Ghanaian context. The investigator noted:
123
[B]y removing this component of having this international “network effects” to fulfill its potential. Balancing the costs
cross dialogue, and realizing that we needed to have intra- and benefits of the approaches remains an open question.
Ghanaian communication instead of inter-Ghanaian
communication, I quickly came to the conclusion that I V. ASSESSING TECHNOLOGICAL INFRASTRUCTURE
needed to do something to improve communication within
One might think that doing a baseline assessment of
Ghana. [2]
available technology infrastructure would be one of the most
The relevance of overseas medical consultants was ultimately objective elements of an ICTD project, one that is a basic part
one of the core reasons that ML could focus on mobile phones of requirements analysis. In this section, we observe that
while GCN retained a focus on the Internet. With respect to GCN and ML drew on two very different sets of infrastructure
deployment, this meant that GCN had the additional task of assumptions and show that such assessments are highly
advertising at GPSF conferences in the U.S. and to other influenced by personal and institutional context.
organizations in the West, diluting its efforts in Ghana.
GCN’s partnership with the MOH, GPSF and KNet and A. GCN
ML’s partnership with the GMA and OneTouch were factors The GCN researchers designed the system around the
in the projects’ very different rates of adoption. Arriving in pessimistic engineering assumption that the system must
the country with equipment for the initial pilot deployment, continue to function under the worst-case connectivity
GCN was deployed in four hospitals over the course of five conditions in the rural North – i.e., that all options available
weeks, garnering an initial user base of 73 doctors. Over a are unreliable and often low-throughput – and optimistic
similar stretch of five weeks, MedicareLine went from being a assumptions about PC usage. The exploratory needs
conversation between two CEOs to a national program, and by assessment fieldwork found that the quality of mobile phone
four months later over 1600 doctors had used ML to make connectivity was unacceptable in the North.
over a million phone calls across the country. It is generally so oversubscribed that if you are calling a
On the topic of sustainability, the juxtaposition of these two mobile phone from another service provider, you need to
projects raises some interesting questions surrounding the dial 10-15 times in order to successfully connect. [1]
rhetoric of “organic adoption” and its impact on long-term
GCN made no quantitative measurement of the quality and
sustainability of ICTD initiatives. That is, they reflect two
reliability of network connectivity, instead relying on
different views of what ‘organic’ or ‘bottom-up’ adoption
qualitative descriptions of regular power and network outages
means.
in formulating design requirements.
ML presented a more ‘organic’ adoption model in the sense
With respect to the technology baseline of users, the GCN
that the program was announced and doctors could sign up
needs assessment found that:
according to their individual needs and interest. On the other
hand, because one institution provides all of the technical and About a third of the physicians interviewed access email
financial resources, there is less local ownership and control regularly… whereas another third claimed that they had at
over the maintenance of each project. Essentially, there is no some point accessed email regularly. [1]
guarantee that OneTouch (or the GMA) will not unilaterally Thus, it was expected that only a minority of doctors would
end the project. This creates a situation analogous to the many actually need to acquire computer literacy.
development projects which rely on inconsistent or limited-
B. ML
term donor funding.
GCN’s model involves a more laborious adoption process The ML researcher made optimistic assumptions about
but benefits from complete in-hospital ownership. This is a mobile phone connectivity among its user base and pessimistic
different type of ‘organic’. This ‘organic’ involved more than assumptions about PC usage. As with GCN, connectivity was
just the end user; it involves the whole system of people and assessed qualitatively, drawing on experiences during the
machines that need to be in place for this network to grow. four-day trip to the North wherein calls from a OneTouch
Unfortunately, this also surfaces the issue of the lack of access phone were compared with calls from another provider. While
to ICT expertise experienced by all but the largest urban there were occasions of calls dropped or text messages
hospitals; the benefits of decentralized ownership are delayed, ML concluded that the quality was sufficient for most
compromised by the geographic concentration of ICT skills. basic uses.
GCN can be conceived as a centralized project (sponsored With regards to computer literacy, ML concluded that:
by MOH and adopted by hospitals) with decentralized The daily routine of a physician in Ghana does not revolve
deployment (core resources provided by hospitals) while ML around computers. I can’t go a day as a physician [in the
is a decentralized project (adopted independently by individual U.S.]... without using a computer – taking orders to
doctors) with centralized deployment (core resources provided checking lab values – but in Ghana it’s all people ordered,
by OneTouch). Technology innovators (such as corporations) from medical records to orders, operative notes…
often exhibit a bias towards decentralized solutions distributed Everything is done by paper! [2]
through markets. Yet there are many examples, of which GCN These observations are actually consistent with those of GCN,
would be one, of a potentially valuable tool which does not but what gives these findings an added dimension is the ability
become useful until it has sufficient infrastructure and of the ML researcher to make a direct comparison of this
124
Figure 7. Urban Hospital (DSL), Wed-Tue Figure 8. Teaching Hospital (VSAT), Thu-Wed Figure 9. Rural Hospital (VSAT), Mon-Thu
Packet Loss Rates (%) and Round-Trip Average (ms.) for three hospital servers in Ghana. Left y axis shows median, 5th and 95th percentile of RTA.
environment with his experience in a leading hospital in the GCN anticipated power outages and network
U.S. Thus, while GCN framed its findings as “most doctors do disconnections occurring several times a day, but the
indeed have some experience with computers” and would use regularity with which (nominally acceptable) bandwidth was
them more often given better access, ML concluded that unusably low was a surprise. For example, certain pieces of
“computers are not part of the daily routine of a Ghanaian systems software on the local servers needed to be configured
physician” and so are not a useful option. with estimates of the worst-case time needed to transfer an
8KB file chunk; the initial estimate of 20 seconds (~3.3Kb/s)
C. Discussion
was eventually increased to 5 minutes (~0.2Kb/s).
Part of the difficulty in producing consistent assessments of Connectivity data obtained after the deployment of GCN show
conditions on the ground is that while metrics such as the that the difference between best- and worst-case performance
frequency of power outages and the availability of DSL can be can be extreme. Figures 7, 8 and 9 illustrate the volatility of
measured and mapped in great detail, it is up to each network performance at three sample hospitals. Figure 7
investigator to determine exactly what metrics need to be shows a DSL network connection with moderately variable
measured, how rigorous the measurement needs to be, and packet loss rates and round trip times. However, Figure 8
what quality of service constitutes something usable by the illustrates the case of a satellite connection so overloaded that,
target community for the specific application. GCN had no on weekdays between 10:00 AM and 4:00 PM, packet round-
resources or expertise with which to improve the mobile trip times to the public GCN server in Ghana consistently
phone infrastructure, so it focused more on network exceeded 10 seconds and loading google.com took 15
infrastructure. ML had no expertise with network minutes. While the GCN software was successfully able to
infrastructure, but instead saw an opportunity to address a very transmit doctors’ messages when congestion decreased in the
significant, non-technical barrier in mobile phone usage. evening, a doctor who requested a consultation in the morning
A closer look at the details of connectivity suggests that would wait until the following day to receive a response (even
both projects were somewhat optimistic in downplaying the if the consultant responded immediately upon receipt of a
infrastructural limitations. The qualitative assessments verified message). (The irony here is that this hospital, with the largest
expected connectivity barriers, whereas additional issues number of dedicated IT staff - 6 people - thus experienced the
remained hidden. most apparent ‘outages’.) In contrast, Figure 9 shows the
characteristics of a satellite connection that worked reliably
and consistently.
As ML is not based on Internet connectivity, an analogous
examination of ML’s assessment of mobile phone coverage
would require assessing the quality of service of OneTouch
voice calls and text messages. While we do not have this
information, we can look instead at geographical coverage
(Figure 10). While OneTouch has more coverage than any
other provider in Ghana, the coverage map shows that vast
regions of the country remain out of coverage area, leaving
doctors working in those regions at a significant disadvantage.
Clearly, a trip to a single urban center in the North does not
systematically gauge the limitations of coverage throughout
the country, let alone quality of service within current
coverage areas.
To be clear, the issues illustrated above have not been a
cause of project failure in either case. Further, exhaustive
Figure 10. OneTouch GSM Coverage Areas. (Source: GSM Association)
125
quantitative assessment of all such issues in advance of was surprising in light of how ‘medicalized’ the investigators
deployment is not always cost-effective or even possible (for perceived the interface. Yet perhaps because their means of
example, one cannot measure DSL links that are not installed). communicating with remote doctors are so limited, it appears
What we suggest here is that care must be taken to prevent the barriers between profession and person are much more
“blind spots” in assessing infrastructure and that such blind fluid than anticipated. Second, many of the doctors discussed
spots can arise from the investigators’ backgrounds. and requested functionality in the system for sharing literature
and PowerPoint presentations, to the extent that these features
VI. SOLUTION USAGE AND APPROPRIATION were included in the upgrade from the pilot to the ongoing
In this section, we discuss the usage of the technological deployment. This is consistent with the findings mentioned
solutions that arose from each project as well as the ways in earlier that a larger screen size was important in dealing with
which the solutions were appropriated for uses other than more information-intensive tasks, and also provides insight
those they were intended to enable. We then discuss the role into the kind of tasks matched to a computer’s affordances.
of what we call technological specificity in these processes. B. ML
A. GCN As previously noted, ML experienced an incredible rate of
As described previously, GCN adoption has been relatively adoption, with 1700 of 2000 members of the GMA signing up
slow; what is notable is that GCN experienced a better rate of within the first four months. While there are no statistics
adoption in the smaller hospitals than in the large ones. At the available concerning the fraction of usage that is related to
smallest deployment hospital, with only 2 doctors on staff, one consultation, there is little doubt (judging from the interviews
of the doctors continues to log into the system and post cases and from multiple instances of observing doctors as they
on a roughly bi-weekly basis. In the mid-sized hospitals, the received calls) that ML is used frequently for consultation.
response was mixed, with a handful of doctors using the There are multiple reasons for this popularity, some of
system every week or two but the majority losing interest. At which are hard to distinguish from the affordances and
the largest deployment hospital, however, after the initial popularity of mobile phones themselves. (One doctor claimed
flurry of activity surrounding the presentations and training, his phone bill dropped from 150 USD per month to 8 USD per
none of the doctors continued to use it on a consistent basis. month after joining the program.) First, mobile phones are a
Follow-up interviews shed some light on this disparity. popular medium for medical consultation because the real-
One issue was computer access. In the smallest district time nature of voice calls is often critical to treating an
hospital, the doctors shared one computer, but since there were emergency case – the three doctors who volunteered
only two doctors, having access to the computer was never a information on the breakdown of emergency/non-emergency
problem. In contrast, access to computers was an issue at the cases reported that around 80% of cases that require further
larger hospitals. One doctor said: consultation are indeed emergency cases. One said:
Sometimes you go to the library, you see someone at the I prefer phone calls to SMS, because I prefer an immediate
computer for 15 minutes; you don’t have that time to waste answer, and also so I can make sure the phone is on. If I’m
[waiting for the computer]. dealing with a case right now, I want to know what to do
when moving ahead as soon as possible.
Another issue was the match between the system’s use case
and the needs of the pilot participants. Through its design and Second, phone skills are more widespread than computer
pool of GPSF consultants, GCN had targeted general skills. Because mobile phones are practically ubiquitous in
practitioners (GPs) in the North who wished to consult with Ghana, there is a lower learning curve as opposed to
urban and overseas medical specialists. This matched the computers, which are owned by only one in three doctors. A
needs of district hospital doctors in the South who had no few doctors claimed that texting on a mobile phone was easier
ready access to specialist consultation and saw great value in than typing on a keyboard. While many of the junior doctors
the ability to connect with specialists (either in Ghana or demonstrated great proficiency with both typing and texting –
overseas). The popularity of this system in the district the vast majority sent multiple text messages a day – a few of
hospitals is a very promising precedent for future deployments the senior doctors demonstrated great difficulty typing during
in the more rural North. In contrast, physicians at the larger training sessions. Third, it promotes more tightly-integrated
Southern hospitals could speak and consult with other doctors workflows. One doctor said:
and specialists more easily in-person than over the computer. I use [ML] a lot and I think it is wonderful. If you want to
GCN did not address the needs of urban specialists looking to talk to anybody concerning a case… concerning anything
tap into the global community of sub-specialists. (In relating to your practice... it gives you a chance to relax
evaluation interviews, many of the urban doctors requested a and really talk. It’s so good. It’s a wonderful idea.
greater number of sub-specialists within the system.) Another said:
GCN Appropriation There has been a move to ban mobile phones in certain
Beyond the expected use for consultation, there were hospitals. It is a very very big mistake, because all they are
unanticipated uses of the system. First, 6 of the 39 going to realize is that this is actually going to decrease
‘conversations’ observed were purely social in nature. This
126
efficiency rather. Consultation is not going to be working While it is tempting to conclude that providing solutions
as well as it used to. with lower specificity is strictly more desirable – and in many
ML Appropriation cases it may be more desirable – it must also be remembered
that utility comes in many forms. GCN’s higher degree of
ML was originally framed as a system to ensure continuity specificity is due to multiple factors. First, the social
of care in long-term and referral patients, but after the system networking application is required to meet the GCN goal of
had been established for three months the ML investigator enabling isolated doctors to build social capital in an extended
noticed its emergent effect in fostering camaraderie in the geographic network of colleagues. While ML assumed that
medical community. He said: the doctors that needed to work together already had each
I was talking to doctors and they were telling me. “Yeah, other’s mobile phone numbers, GCN had determined that a
I’m reconnecting with classmates.” large portion of rural, immigrant, and junior doctors did not
Indeed, many doctors were up-front about the fact that ML have a strong network of contacts [20]. Second, GCN’s
had been a boon, not just for consultation, but for facilitating emphasis on overseas consultants implies a need for low-
social interaction within their community. bandwidth, asynchronous communication as opposed to voice
The mobile phone has also integrated itself into calls. Third, GCN operated as a development project,
administrative and management processes within large maintaining statistics of usage metrics in order to facilitate
hospitals. One doctor said: evaluation; for now, ML relies entirely on the built-in
reporting mechanisms in the OneTouch network. GCN can
You don’t have walk down somewhere or you send a
map consultations made to specific case outcomes, while
patient down... it reduces the whole bureaucratic...
pushing around of patients. OneTouch – which does receive very high praise from
enrolled doctors who offered their feedback on the system –
Personal mobile phones, both on the ML program and not, are cannot. Moreover, as previously mentioned, GCN’s goals
regularly used to set up diagnostic tests at the laboratory, to include technological innovation as well as social impact.
confirm insurance claims forms, and also for the doctor on call GCN assumes connectivity in rural areas will remain an
during the night to discuss a change in a patient’s care with the ongoing challenge, while ML relies on the assumption that,
doctor who admitted that patient. In short, where a U.S. with time, OneTouch coverage will be able to reach even the
hospital might use infrastructure such as pagers and site-wide most remote doctors. This question echoes one of the
internal communications systems, large Ghanaian hospitals fundamental tensions in ICTD research: as pragmatists, we
improvised solutions using personal mobile phones. aspire towards demonstrating the greatest social impact in real
C. Discussion communities today, while as researchers, we try to identify
In reflecting upon differences between GCN and ML in what fundamental limitations exist and how these can be
usage and appropriation, we will focus on an important tackled in years to come. It is to be expected that untested
distinction that we will call ‘specificity’. Two solutions with technology would experience more hurdles in the short run,
different ‘specificity’ can be targeted at exactly the same task while its long-term contribution is yet to be seen. In short,
and be based on an equally nuanced understanding of then, a project can easily have a number of goals that might be
workflows and use cases; the difference lies in the types and frustrated by a lower-specificity solution.
number of layers of technology which make up the solution A final note about lower-specificity solutions concerns their
and the degree to which they are specific to the solution. For potential to be too widely appropriated. In the smallest district
example, the GCN system presents not only a robust hospital (where GCN experienced greater adoption), both
asynchronous communication medium, but also an in-hospital doctors tended to switch off or mute their phones while at the
server as well as an end-user Web application. ML, in its first hospital in order to minimize distraction. Even in the larger
phase, focuses exclusively on tackling the cost barrier of urban hospitals, half of the 20 doctors who discussed their
existing phones using existing networks. mobile phone usage claimed to keep their phones off while
As we have seen, these two projects with similar goals at working because of distraction. Hospital administrators have
the start ultimately resulted in two very different usage related concerns; at one teaching hospital, a memorandum was
outcomes – particular in the area of adoption. Investing time circulated to all the doctors banning the use of mobile phones
and resources in needs assessments and design process, the in many locations
GCN project produced a highly ‘specific’ system to address …to forestall the negative impact of mobile phone
not only failures in internet infrastructure but also social frequencies on medical equipments and improve the work
network gaps. However, adoption has been slowed by the ethics of staff. It is also announced… that it is a serious
need to introduce the system incrementally into hospitals. The offence to disconnect life-supporting equipment in order to
ML project achieved broader adoption over a much shorter use their power sources (socket) to charge mobile phones.
period of time, in part by relying on the existing availability In view of concerns such as these, we suggest that researchers
and high adoption levels of mobile phones. Similarly, the keep in mind both the (immediate) benefits and (eventual)
lower specificity of the ML solution seems to contribute to a costs of rapid adoption.
greater range of user appropriation behaviors.
127
VII. RELATED WORK specific technologies such as connectivity and a post-1999

trend towards specific solutions such as software and VOIP.
ICTDs for Healthcare in the Developing World
The research on ICTDs to promote healthcare has a long VIII. CONCLUSION
and rich history [15],[34]. In the context of developing
regions, remote medical consultation has been a popular and In this paper, we have presented parallel case studies of two
relatively successful approach [32]. Computers [3], mobile ICT projects addressing the need for improved medical
phones [25] and a combination of both [27] have demonstrated consultation among doctors in Ghana. We have examined
utility in a variety of settings. how the two projects have been shaped by the institutional
context and the identity of their researchers. We looked at how
Framing the Problem the partnerships formed affect the solution outcome, delivery,
[21] provides a useful overview and categorization of how and adoption; we argued that even the task of assessing
ICTs are conceptualized from a variety of different fields. technological infrastructure is far from objective; and we
Theory on ‘framing the problem’ can be drawn from social- noted the implications of the ‘specificity’ of each solution.
constructivism, although we are not advocating here for a We conclude with an open question raised by these
change from ICTD’s traditional focus on pragmatism and discussions, summarized in Table 3. In ICTD research, much
advocacy/participatory research. Much of the work on attention has been given to the socio-cultural, political, and
qualitative methodology emphasizes the importance of economic contexts of target communities – yet ‘difference’ is
reflexivity and self-awareness in order to minimize such bias a measure between communities; it is only by critically
[8], but the fact remains there are institutional accountabilities examining our own research community that we can
which no amount of methodology can shake. We can learn understand the influence and impact of communities on each
from the example of anthropologists who have challenged other. ICTD is linked with economic, social, political, and
their own role in the colonial apparatus [11],[30]. For human development agendas [6]. Regardless of whether we
example, [17] provides an interesting comparison of how take an integrated approach [10] or focus instead on specific
different sectors tackle the management of kiosks in India. local needs [14], there are still institutional, societal, and
individual layers at play in our interventions. Researchers are
Partnerships and Deployment
embedded in contexts of existing friendships, collaborations,
ICTD researchers have an important role to play at the expertise and agendas, and we need to be conscious of what
crossroads of business administration and government policy, kind of consequences our decisions to draw on them have on
identifying how both contribute to the goal of development project outcomes. In this paper we have seen impacts on
and how their contributions interplay. Many examples in everything from research agendas to infrastructure
ICTD research and development literature study the role of assumptions, yet the literature on these avenues of choice
institutional players, although these papers are typically remain fragmented across a variety of other communities
evaluative rather than action-oriented [18],[23],[24]. The including CSCW and organizational behavior. ICTD
literature on partnership selection and cultivation in ICTD is researchers need to increase the exchange of ideas between
small but growing rapidly [26],[28],[31]. these communities. For example, in the course of establishing
Specificity GCN, the task of generating a viable ‘business model’ for the
technology providers was not addressed, but a social
In reaction to the variable success of many deployments of
entrepreneurship community would never allow such an
general ICT, such as a number of telecenter and ‘hole-in-the-
omission. ML has had great adoption success, but technical
wall’ computing initiatives, much of the focus of the HCI
infrastructure for evaluating its impact was neglected. We
ICTD research community has turned towards purpose-built
need a theoretical framework under which to unite the
technology [4],[33]. Many of those doing systems research
different fields of research for ICTD practitioners, one
focus instead on providing basic computing or connectivity
surrounding not just technology design, but also technology
functionality [13],[29]. [6] presents a useful survey of ICTD
framing, partnership, assumptions, and deployment.
literature which mentions a pre-1999 trend towards non-
Table 3 – Impacts of project context on project decisions and outcomes

Framing By focusing on connectivity in the rural North, GCN targets a minority of doctors with the greatest need.
By focusing on airtime cost, ML impacts the majority of doctors, the bulk of whom are in the urban South.
Partnerships and Working through MOH allowed GCN to iterate rapidly with many hospitals early on, but with greater overhead later. GCN`s
Deployment commitment to the U.S.-based GPSF was a key factor that tied it to a computer-based platform instead of mobile phones.
GMA advertised ML effectively while OneTouch quickly took ownership of project execution and maintenance. ML’s approach is
highly dependent on the vagaries of a single national operator while GCN’s is at the mercy of individual hospitals.
Infrastructure GCN found the networks to be more challenged than anticipated.
ML found mobile phone coverage lower than anticipated.
Solution GCN`s higher specificity allowed it to incorporate evaluation indicators and address not only physical communication gaps but also
gaps in social capital. Usage limited by access to computers in larger hospitals.
ML adoption was much higher given the real-time nature of mobile phones and the fact that they were already widely tested and used.
Lower specificity facilitated adoption and appropriation while making it difficult to evaluate the program.
128
REFERENCES [18] R. Kuriyan and I. Ray. “Public Private Partnerships and Information
Technologies for Development in India,” Proc. ICTD 2007, IEEE/ACM
[1] Field notes from the GCN investigators’ journals. (2007), 310-320.
[2] Phone conversations between the GCN investigators and the ML [19] B. Latour. Science in Action. Harvard Univ. Press, Cambridge, MA,
investigator. 1987.
[3] K. Brauchli, D. O’Mahony, L. Banach, and M. Oberholzer, “iPath – A [20] R. Luk, M. Ho and P.M. Aoki. “Asynchronous Remote Medical
Telemedicine Platform to Support Health Providers in Low Resource Consultation for Ghana,” Proc. CHI 2008, ACM (2008), 743-752.
Settings,” J. IT in Healthcare 3, 4 (2005), 227-235. [21] W. Orlikowski and C.S. Iacano. “Desperately Seeking “IT” in IT
[4] P. Braund and A. Schwittay. “The Missing Piece: Human-Driven Design Research – A Call to Theorizing the IT Artifact,” Information Systems
and Research in ICT and Development,” Proc. ICTD 2006, IEEE Research 12 (2001), 121-134.
(2006), 2-10. [22] R. Patra, S. Nedevschi, S. Surana, A. Sheth, L. Subramanian and E.
[5] E Brewer, M. Demmer, B.W. Du, M. Ho, M. Kam, S. Nedevschi, J. Pal, Brewer. “WiLDNet: Design and Implementation of High Performance
R. Patra, S. Surana, and K. Fall. “The Case for Technology for WiFi Based Long Distance Networks,” Proc. NSDI 2007, USENIX
Developing Regions,” IEEE Computer 38, 6 (2005), 25-38. (2007), 87-100.
[6] J. Bussell. “International Norms on ICTs for Development: New Data, [23] L.R. Peattie, Rethinking Ciudad Guyana. Univ. of Michigan Press, Ann
Initial Findings, and Opportunities for Analysis,” American Political Arbor, MI, 1987.
Science Association Annual Meeting, APSA (2005). [24] E. Peter. “The Challenges of the Institutional Turn: New
[7] V. Cerf, S. Burleigh. A. Hooke, L. Torgerson, R. Durst, K. Scott, K. Fall Interdisciplinary Opportunities in Development Theory,” in V. Nee and
and H. Weiss. Delay-Tolerant Networking Architecture, RFC 4838, R. Swedberg (eds.), The Economic Sociology of Capitalism. Princeton
IETF, 2007. Univ. Press, Princeton, NJ, 2005.
[8] J.W. Creswell and V.L. Plano Clark. Designing and Conducting Mixed [25] K. Phipps, G. Sangudi, and S. Woolway. What Works: HealthNet
Methods Research. SAGE, Thousand Oaks, CA, 2007. Uganda’s Evolution from NGO to Sustainable Enterprise: Portable
[9] M. Demmer. A Delay Tolerant Networking and System Architecture for Healthcare Service Delivery to Uganda’s Rural Areas. World Resources
Developing Regions. Ph.D. dissertation, Univ. of California, Berkeley, Institute, 2003.
CA, 2008. [26] D. Ramachandran, M. Kam, J. Chiu, J. Canny, and J.F. Frankel. “Social
[10] R. Duncombe and R. Heeks. “Information, ICTs and Small Enterprise: Dynamics of Early Stage Co-Design in Developing Regions,” Proc. CHI
Findings from Botswana,” In H. Katrak and R. Strange (eds.), Small- 2007, ACM (2007), 1087-1096.
Scale Enterprises in Developing and Transitional Economies. Palgrave, [27] J. Royall. “SatelLife – Linking Information and People: The Last Ten
Basingstoke, UK, 285-304. Centimeters,” Development in Practice 8, 1 (1998), 85-90(6).
[11] P. Dourish. “Implications for Design,” Proc. CHI 2006, ACM (2006), [28] Y. Schwartzman and T. Parikh. “Establishing Relationships for
541-550. Designing Rural Information Systems,” Extended Abstracts, CHI 2007,
[12] K. Fall. “A Delay-Tolerant Network Architecture for Challenged ACM (2007), 1845-1850.
Internets,” Proc. SIGCOMM 2003, ACM (2003), 27-34. [29] A. Seth, D. Kroeker, M. Zaharia, S. Guo, and S. Keshav. “Low-Cost
[13] R. Fonseca and J. Pal. “Computing Devices for All: Creating and Selling Communication for Rural Internet Kiosks Using Mechanical Backhaul,”
the Low-Cost Computer,” Proc. ICTD 2006, IEEE (2006), 11-20. Proc. Mobicom 2006, ACM (2006), 334-345.
[14] R. Gerster and S. Zimmerman. “ICTs for Poverty Reduction,” [30] L. Suchman. “Practice-Based Design of Information Systems: Notes
Information for Development Online, Jul-Aug 2003. from the Hyperdeveloped World,” The Information Society 18, 139-144.
http://www.i4donline.net/issue/july03/pdfs/icts_poverty_reduction.pdf [31] S. Surana, R. Patra, S. Nedevschi, and E. Brewer. “Deploying a Rural
[15] W.R. Hersh, D.H. Hickam, S.M. Severance, T.L. Dana, K.P. Krages, Wireless Telemedicine System: Experiences in Sustainability,” IEEE
and M. Helfand. “Diagnosis, Access and Outcomes: Update of a Computer 41, 6 (2008), 48-56.
Systematic Review of Telemedicine Services,” J. Telemed. Telecare 12, [32] D.J. Vassallo, P. Swinfen, R. Swinfen, and R. Wootton, “Experience
S2 (2006), 3-31. with a Low-Cost Telemedicine System in Three Developing Countries,”
[16] D. Kuo, D.R. Gifford, and M.D. Stein. “Curbside Consultation Practices J. Telemed. Telecare 7, S1 (2001), 56-58.
and Attitudes Among Primary Care Physicians and Medical [33] M. Warschauer. Technology and Social Inclusion. MIT Press,
Subspecialists,” JAMA 280, 10 (1998), 905-909. Cambridge, MA, 2003.
[17] R. Kuriyan and J. Bussell. “Understanding Concepts of ‘Development’ [34] P. Whitten and B.D. Sypher. “Evolution of Telemedicine from an
and Linkages to Deployment Strategies in ICT4D in India,” in A. Applied Communication Perspective in the United States,” Telemedicine
Palackal and W. Shrum (eds.), Information Society and Development: and e-Health 12, 5 (2006), 590-600.
The Kerala Experience. Rawat, New Delhi, 2007.
129
Improving Child Literacy in Africa:

Experiments with an Automated Reading Tutor
G. Ayorkor Mills-Tettey, Jack Mostow, M. Bernardine Dias, Tracy Morrison Sweet,
Sarah M. Belousov, M. Frederick Dias, Haijun Gong
Abstract—This paper describes a research endeavor aimed at As a contribution to the discourse on applying information
exploring the role that technology can play in improving child and communication technologies (ICTs) to address
literacy in developing communities. An initial pilot study and development challenges, we describe a study exploring the
subsequent four-month-long controlled field study in Ghana
investigated the viability and effectiveness of an automated
potential role of computing technology in improving English
reading tutor in helping urban children enhance their reading literacy among Ghanaian and Zambian children who attend
skills in English. In addition to quantitative data suggesting that school in English but have low reading achievement levels.
automated tutoring can be useful for some children in this Low functional literacy among individuals who have
setting, these studies and an additional preliminary pilot study in completed primary school is not an uncommon problem in
Zambia yielded useful qualitative observations regarding the developing communities. For example, UNESCO reports that
feasibility of applying technology solutions to the challenge of
enhancing child literacy in developing communities. This paper
in 2000, more than 1 in 3 adults with a fifth-grade education
presents the findings, observations and lessons learned from the in Chad and Niger reported that they could not read [21]. In
field studies. other cases, individuals may finish primary school reading
below the expected level. In a representative sample of
Index Terms—Developing Regions, Education, Educational Ghanaian public schools, reading achievement levels
Technology, Literacy measured by the government-administered Criterion
Referenced Test in 2000 indicated that fewer than 10% of the
I. INTRODUCTION children in grade six were able to read with grade level
L ITERACY is a key part of the global development agenda.

It is a complex concept with multiple definitions. An
international expert meeting at UNESCO in 2003 proposed
mastery [9].
Several factors contribute to this problem. For the average
child from a rural or low-income urban background in Africa,
the following definition: “Literacy is the ability to identify, reading is not part of daily family life, and sometimes parents
understand, interpret, create, communicate and compute, using are not themselves literate. In Ghana and Zambia specifically,
printed and written materials associated with varying contexts. most children speak one of a number of local languages at
Literacy involves a continuum of learning in enabling home but attend a school taught in English, the official
individuals to achieve their goals, to develop their knowledge language for both countries. Typically, under-resourced
and potential, and to participate fully in their community and schools with overcrowded classrooms offer few opportunities
wider society.” [22] The United Nations recognizes literacy as for individual attention while developing reading skills.
a human right, noting that basic education, of which literacy is The project described in this paper is a proof-of-concept
the key learning tool, was recognized as a human right over 50 study to investigate whether an automated computer-based
years ago in the Universal Declaration of Human Rights. reading tutor that provides guided reading practice can
significantly improve the reading proficiency of children in a
developing community, even if they have no prior familiarity
The research reported here was supported in part by the discretionary gifts
to the TechBridgeWorld research group at Carnegie Mellon University, by the with computers. We focus our study on children in Accra,
Qatar Foundation for Education, Science, and Community Development, by Ghana, and Mongu, Zambia.
the National Science Foundation under ITR/IERI Grant No. REC-0326153, by As illustrated in Figure 1, we began by employing the
the Institute of Education Sciences, U.S. Department of Education through
Grant R305A080628 to Carnegie Mellon University, and by the Heinz
Reading Tutor in a preliminary three-week-long pilot study in
Endowments. The opinions expressed are those of the authors and do not Accra, Ghana. The pilot study was used to explore technical
necessarily represent the views of any of our sponsors. and operational feasibility and to motivate partnerships and
G. A. Mills-Tettey is with the Robotics Institute at Carnegie Mellon
funding for a longer term study. We followed up with a four-
University, Pittsburgh, PA 15213 (phone: 1-412-268-8645; fax: 1-412-268-
6436; email: ayorkor@cmu.edu). J. Mostow, M. B. Dias, S. Belousov, and M. month-long controlled study in Accra, in collaboration with
F. Dias are also with the Robotics Institute at Carnegie Mellon University, the Ghana-India Kofi Annan Centre for Excellence in ICT
Pittsburgh, PA 15213 (email: mostow@cs.cmu.edu, mbdias@ri.cmu.edu, (AITI-KACE) and with input from Associates for Change, an
sarahtbw@ri.cmu.edu, mfdias@ri.cmu.edu). T. M Sweet and H. Gong are
with the Statistics Department at Carnegie Mellon University, Pittsburgh, PA educational research firm in Accra. The controlled study
15213 (email: tsweet@andrew.cmu.edu, haijung@andrew.cmu.edu)
130
aimed at quantitatively measuring the educational also assist in this process: Scientific Learning’s Reading
effectiveness of the tutor with children in Accra. Observations Assistant [1] and Project LISTEN’s Reading Tutor [11] are
and lessons from these experiences fed into another pilot examples of computer-based tools that use automated speech
study in Mongu, Zambia, with the plan of following up with recognition to provide a guided reading experience for the
an additional controlled study in Zambia in the future. user. Project LISTEN’s Reading Tutor has demonstrated
success in improving the reading ability both of children
Exploring the role of technology in improving whose first language is English [11][12][13] and of children
English Literacy in Africa
learning English as a second language (ESL) in the United
States [16] and in Canada [18].
Existing tool: New tools
Automated Reading Tutor for guided oral reading practice The project described in this paper used Project LISTEN’s
Reading Tutor, developed at Carnegie Mellon University. The
Ghana Zambia Reading Tutor displays stories on a screen and “listens” to a
June 2005 July 2008 child read aloud. By using speech recognition to analyze the
Pilot study Pilot study child’s reading, the Reading Tutor is able to give graphical
• Gauge children’s reactions to • Gauge children’s & teachers’
Reading Tutor reactions to Reading Tutor
and spoken feedback. It gives help when it detects a long
• Explore out-of-school usage • Explore in-school usage pause, a severely misread word or a skipped word, and also
model model when the reader clicks for help. The tutor may speak the
• Assess feasibility of test • Explore government literacy
instruments requirements and tests whole word out loud or decompose the word into syllables or
phonemes and speak out each part while highlighting it. It
May-Sept 2007 Future work
may also give a “rhymes with” hint, or read the sentence by
4-month controlled study Year-long controlled study
• Measure reading progress • Measure reading progress
playing a fluent human narration to model expressive reading.
• Compare progress across 3 • Compare to non- The Reading Tutor includes a wide variety of stories at
socio-economic groups technological approach to different reading levels. It takes turns with the child in
• Assess sustainability of out- guided reading practice
of-school usage model • Assess sustainability of in- selecting a story to read. It monitors the child’s reading
school usage model progress and selects stories at an appropriate level for the
child. For readers at early stages of development, the Reading
Reported work Proposed future work Tutor also includes word-building exercises to develop
Figure 1 - Project outline knowledge of spelling-to-sound correspondences. Videos of
the Reading Tutor in use may be found at Project LISTEN’s
Section II describes the skills that are involved in reading website [17].
and introduces the automated tutor we used in this project.
Sections III, IV, and V describe the pilot field study in Ghana, III. GHANA PILOT STUDY
the 4-month-long controlled study in Ghana, and the pilot
study in Zambia respectively. Section VI distills practical A. Goal
lessons learned in the implementation of these field studies The 2005 Ghana pilot study aimed to evaluate the
and Section VII concludes with a discussion of future work. practicality of a technological approach to guided reading
practice in Accra, and investigate the feasibility of conducting
II. BACKGROUND a longer term controlled study.
As a cognitive proficiency, reading involves several Specifically, we wished to answer the following questions:
component skills such as phonemic awareness, decoding, 1. Partners and logistics: How feasible is it to
fluency, vocabulary, and comprehension [14]. Phonemic engage partners and arrange the logistics for such a
awareness is the ability to perceive individual sounds or study?
phonemes in words. Building on this awareness, a child learns 2. Learning to use the Reading Tutor: How quickly
the alphabetic principle that spelling generally maps do children with no prior computer experience
systematically to pronunciation. The child also learns specific learn to operate the Reading Tutor, and what
letter-sound correspondences and the ability to correctly instruction do they need in order to do so?
pronounce written words (decoding). Fluency is the ability to 3. Speech: Does the speech recognition software
read text accurately, quickly, and expressively and is an perform acceptably with Ghanaian accents, and
important foundation for comprehension. A rich vocabulary is can the students understand the narrated speech?
essential for comprehension and effective communication. 4. Tutor content: Do the children find the reading
Children develop these skills through a variety of experiences, material in the tutor engaging?
including skilled instruction. Research has shown that regular 5. Usage sessions: What is an effective length for a
guided oral reading plays an important role in developing tutoring session?
reading skills, particularly fluency and comprehension [14]. 6. Reading measures: Which test instruments can be
Such guided oral reading may happen in small groups in a used easily and effectively to assess reading
classroom setting, or with parents at home. Technology may proficiency in the Ghanaian setting?
131
B. Participants and Methodology quickly learned to recognize when it did not “hear” them
We chose to focus on the needs of children from low-income correctly and to repeat themselves when necessary. After the
families attending public school because they have a high risk initial difficulty with the built-in tutorials, the children
of low achievement in reading. In consultation with appeared to understand the prompts given by the Reading
Associates for Change, we targeted children in grades two Tutor.
through four since this is a key period for developing reading Tutor content: Overall, the children were very enthusiastic
skills after the initial adjustment to the primary school about using the Reading Tutor. Most of them appeared to
environment. We restricted our work to an urban environment enjoy the world-building exercises and the stories in the tutor.
where computers are more readily available, but where However, we noticed that a couple of children who were older
significant literacy challenges still exist. We employed an out- than the norm (11 or 12 years old) but had a kindergarten or
of-school usage model in which practice with the Reading first grade reading level were not fully engaged by the content
Tutor was supplemental to regular school activities. of the simple stories, such as “Sam sat on a mat,” available for
The pilot study involved qualitative observations of their reading level. There was a discrepancy between these
children as they used the Reading Tutor. The study was children’s maturity and their reading ability. Furthermore, we
conducted by the first author who is a native of Ghana. She believe it would be good to incorporate more local content,
was assisted by a local volunteer. Two groups of children such as Ghanaian folk tales, into the tutor.
participated in this study. One group comprised twelve Usage sessions: We noticed that the better readers could
children in grades two through four from an under-resourced use the tutor for more than half an hour at a time without
public school. They used the Reading Tutor at an internet café getting bored, whereas the more challenged readers would tire
near their school for 20-30 minutes each day over a three- after about twenty minutes, but would still look forward to
week period. The other group comprised six children from a their next turn. We thus decided to limit the length of usage
mixed low- and middle-income neighborhood. They used the sessions in the controlled study to 20-30 minutes.
tutor on laptops in the home of the researcher, for 20-30 Reading measures: Although the pilot study did not aim to
minutes each day, three days a week over the same period. quantitatively measure reading progress, we identified and
tested reading measures that we would use in the controlled
C. Results and Observations study. The instruments we chose to use to assess reading
Partners and Logistics: The school, internet café, and the ability were an oral reading fluency test [5] and the Test of
parents of all participating children agreed without hesitation Written Spelling (TWS) [8], both of which are hand-scored
to participate in the project, and logistics were arranged very tests that have been frequently used in previous studies
quickly. The internet café donated time on four desktop involving the Reading Tutor, and are psychometrically
computers; thus, four children could use the Reading Tutor at reliable, fast and easy to administer. The fluency test is timed
a time. We engaged the remaining children in other activities reading exercise in which the child is given a grade-
while they waited their turn. This worked fairly well for the appropriate story to read. It is scored as the number of words
pilot project, but would not be a feasible approach for a longer read correctly in one minute. Fluency is essential to
study involving a larger number of children. comprehension and is a sensitive measure of growth in
Learning to use the Reading Tutor: Although one child had proficiency [19]. The TWS is a dictation exercise in which the
previously played a computer game, none of the other children child writes down words read aloud by the tester in order of
had used a computer before. The children were given an initial increasing difficulty. The test ends once the child has
10-15 minute hands-on lesson introducing them to the incorrectly spelled five words in a row, at which point it is
computer and showing them how to use the mouse, keyboard assumed that the difficulty of the remaining words exceeds the
and software. The children had trouble understanding the child’s proficiency. This allows the same test to be used for
Reading Tutor’s built-in automated tutorials on how to operate multiple grade levels. TWS is scored as the number of words
it and use the keyboard, perhaps due to the unfamiliar spelled correctly. Spelling is like decoding in that it tests print-
narration accent or the use of words such as “roster” that are to-sound mappings but it is easier to assess reliably.
common in American but not Ghanaian English. This We tried out the fluency test and the TWS with the pilot
difference in language use is an example of the need to study participants, and they did not have trouble with the
localize tools for a given setting. We explained the tutorials format of either test. To combat their nervousness about being
one-on-one to the children during their first two sessions. examined, we avoided using the term “test.” We instead
Some of the questions the children asked during the first few explained that these exercises were to help figure out how the
sessions were on keyboard use and on navigating the tutor. Reading Tutor could assist them and we reassured the children
By their second or third session, most of the children were that they would not be given a grade. We modified the
able to operate the Reading Tutor without help. passages for the fluency test to use Ghanaian names rather
Speech: The speech recognition capability appeared to than American names, to make them more recognizable to the
work adequately with the children’s accents: it accepted a students. The children were familiar with the dictation format
great majority of words that were read correctly. Also, based of the TWS, but again were unfamiliar with some words such
on the graphical feedback given by the tutor, the children as “tardy” which are not regularly used in Ghanaian English.
132
additional reading intervention, and for the second half, the

roles were switched: the Control-1st group used the Reading
Tutor while the Tutor-1st group had no additional reading
intervention. Each half of the study lasted nine weeks (two
months) during which there were daily usage sessions of
approximately half an hour per child, although attendance and
usage varied considerably. This crossover study design,
illustrated in Table II, had the advantage of equity in the sense
that all participating children had the opportunity to use the
reading tutor. It also avoided having to pair similar children
for comparison since each child essentially served as his or
her own control. The potential disadvantage, which was that
carryover treatment effects from the first half may have
affected the second half, was mitigated by having half the
children participate first in the control, while the others
participated first in the treatment group.
Figure 2 - A child in Accra reading a story using the Reading Tutor
Table I - Study participants by school, grade and gender
IV. GHANA CONTROLLED STUDY Number of Children
Tutor-1st Control-1st
School Grade Female Male Female Male
A. Goal
S1 Grade 2 3 2 2 3
Following on the successful pilot study, the goal of the (29 children ) Grade 3 5 2 1 3
controlled study in Ghana was to quantitatively measure the Grade 4 1 3 3 1
efficacy of the Reading Tutor in helping children improve S2 Grade 2 2 3 3 2
their reading skills. Specifically, we wished to determine: (30 children) Grade 3 3 2 1 4
Grade 4 3 2 3 2
1. Does regular use of the Reading Tutor improve oral
S3 Intermediate 3 4 3 5
reading fluency and spelling? (30 children) Advanced 3 5 5 2
2. Do treatment effects depend on other factors such
as school/socio-economic background, gender, or
A battery of three fluency tests and one TWS were
grade level?
administered to all the children at the beginning of the study
In addition, we wanted to learn about the operational
(pre-testing), between the two halves (mid-way testing), and at
sustainability of an out-of-school usage model, taking into
the end (post-testing). The testing was conducted at the
consideration installation and maintenance of the software,
children’s schools by AITI-KACE project staff. Three fluency
training of staff responsible for the day-to-day running of the
tests were used to better estimate reading ability. The passages
project, transportation of the children between the school and
used for these tests correspond roughly to first, second, and
the project site, and other logistics. The controlled study was
third grade reading levels respectively, since a test close to the
deployed by project staff at AITI-KACE, with remote training
child’s current reading level should give a more sensitive
and support by our team in Pittsburgh.
measure of progress. We found that the scores on the three
B. Study Design / Methodology passages were highly correlated, and so for analysis purposes
The controlled study involved eighty-nine children from these scores were combined into a single mean fluency score.
three schools, representing three socio-economic
Table II – Cross-over Study Design
backgrounds. The participating schools, recruited by Tutor-1st Group Control-1st Group
Associates for Change, were S1: a private school in a middle- Pre-testing of all children
income community, S2: a public school in a low-income First half of cross-over Reading Tutor Control: no special
community, and S3: an informal educational program for (1st nine weeks) intervention
Mid-way testing of all children
highly disadvantaged children who have never attended
Second half of cross-over Control: no special Reading tutor
formal school. The study involved children in grades two (2nd nine weeks) intervention
through four of S1 and S2, and in the “Intermediate” and Post-testing of all children
“Advanced” levels (roughly corresponding to grades two and
three respectively) of S3. The study participants used the Reading Tutor in a computer
The children were split randomly across school, grade and lab at AITI-KACE because it had equipment and technical
gender boundaries into two groups, Tutor-1st and Control-1st, capability that their schools did not. The computers used for
as shown in Table I. We used a two-treatment crossover study the project had 2.4GHz Pentium IV processors, 256MB of
design: for the first half of the study, the Tutor-1st group used RAM, and 16GB hard drives.
the Reading Tutor while the Control-1st group had no
133
C. Deviation from Study Design by the within-group pooled standard deviation [3]. The effect
In a normal crossover study, experimental conditions are size is computed for each school because the treatment effect
held constant over the course of the study. However, due to depends on the school. An effect size of 0.2 is generally
logistical challenges that delayed the start of the project, the considered small, 0.5 medium, and 0.8 large [4].
first half of this study took place while the students were
attending school, whereas the second half overlapped with the Fluency pre-test scores by school and grade
school vacation. Because of the different conditions in place

180
during the two halves of the study, we analyzed the results as 160
two different experiments rather than as a single crossover
Mean fluency score

140
120
study. The first experiment measured the effect of the Reading 100
Grade 2
Grade 3
Tutor while the children attended school (and thus attended 80
Grade 4
60
English class as usual). The second experiment measured the 40
effect of the tutor while the children did not attend school (and 20
0
so did not attend English class). This modified study design is S1 S2 S3
illustrated in Table III. School
Table III – Modified Study Design Figure 3 - Fluency pre-test scores by school and grade
Tutor-1st Group Control-1st Group
Pre-testing of all children
TWS pre-test scores by school and grade
Experiment 1 School + Control: school only
(1st nine weeks) Reading Tutor
40
Mid-way testing of all children 35
Experiment 2 Control: No school No school + 30
(2nd nine weeks) Reading tutor TWS score 25 Grade 2
Post-testing of all children 20 Grade 3
15 Grade 4
D. Results and Analysis 10
5
Figures 3 and 4 illustrate the pre-test scores of children at 0
the three schools, representing their reading proficiency going S1 S2 S3
into the study. It is clear that the S1 children had much higher School
levels of reading achievement than the S2 children who in turn

Figure 4 - TWS pre-test scores by school and grade
had higher levels than the S3 children. Pre-testing for our
study occurred about two months before the end of the school Table IV - Fluency pre-test scores compared with U.S. norms
year. To provide some context for these scores, Table IV Grade
Mean pre-test fluency (standard deviation) Year-end
compares the fluency pre-test scores with end-of-year norms U.S. Norm [8]
/ Level S1 S2 S3
(50th percentile)
from schools in the United States [6]. By this standard, the
2 / Int. 82.4 (44.8) 32.3 (25.9) 11.1 (8.5) 89
average reading proficiency of the S1 students appears to be at 3 / Adv 123.3 (31.4) 45.8 (37.4) 20.4 (20.8) 107
or above the U.S. average whereas that of the S2 and S3 4 132.7 (32.7) 59.8 (41.5) -- 123
children is lower. Norms from schools in Ghana are not
available for comparison.
In analyzing each experiment, we focus on gains in fluency Experiment 1 results by school: Table V shows the results
and spelling test scores. For experiment 1, the gain is the and t-test analysis for experiment 1, when the children were
difference between pre- and mid-test scores; for experiment 2, attending school. The reading proficiency of the S2 children
the gain is the difference between the mid- and post-test who used the Reading Tutor (the “treatment” group) improved
scores. A positive gain indicates improvement in the child’s significantly more than those who did not (the “control”
reading proficiency. In each experiment, we use a standard group), as evidenced by larger gains in both fluency and TWS.
statistical t-test to compare the mean gains of the treatment The S3 treatment group significantly out-gained the control
and the control group. This test yields a p-value indicating group in fluency but not in spelling, although there was a
how significant the difference is between the means of the two positive trend. Finally, there was no significant difference in
groups. For this analysis we consider p-values of less than gains between the S1 treatment and control groups, for either
0.016 to indicate statistical significance. This value is smaller test. Omitted from this analysis are five S1 children who were
and thus more conservative than the commonly used threshold not present for the midway testing. Due to the small numbers,
of 0.05 because multiple comparisons (due to the different we did not break down the data to examine gains for each
schools) require an adjustment of significance levels [10]. We grade level within each school.
also compute a measure of effect size, that is, the magnitude
of the treatment effect. The measure we use for effect size is
Cohen’s d: the difference between group mean gains divided
134
Table V - Comparison of treatment and control gains for experiment 1 E. Discussion

Treatment Control
t-test comparison Effectiveness of the Reading Tutor: The results provide
(Tutor-1st (Control-1st
of treatment & evidence that during the school term, the S2 students (and to a
group, group,
Test School control gains
N=41) N=42) lesser extent the S3 students) who used the tutor gained
Effect considerably more than those who did not use the tutor. Thus,
Mean Gain (SD) p-value
size the Reading Tutor was helpful for the S2 children. This is a
Fluency S1 8.6 (20.5) 21.5 (11.7) 0.0833 -0.775 positive outcome since the S2 group most closely represents
(# words read S2 61.4 (22.9) 23.4 (21.0) <0.0001 1.731 our target population of children from low-income families
correctly per min) attending public school. The proficiency of the S1 children did
S3 13.4 (9.3) 5.1 (4.4) 0.0054 1.144
TWS S1 -1.9 (4.9) -0.2 (2.6) 0.2997 -0.457 not appear to be influenced by reading practice with the tutor.
(# words spelt S2 3.9 (2.7) 0.3 (3.4) 0.0039 1.153 A possible explanation for this is that the S1 children might
correctly) S3 2.8 (4.0) 0.6 (1.7) 0.0648 0.719 not have had much room to benefit from the tutor since they
were fluent readers going into the study.
Experiment 2 results by school: Table VI shows the results
Loss in proficiency over vacation period: Of the statistically
for experiment 2, when the children were not attending
significant results highlighted in Tables VI and VII, the
school. Interestingly, there were negative fluency gains for the
negative gains of the S2 children in experiment 2 stand out. A
S2 children over the vacation, and these were more dramatic
possible explanation for this might be found in studies in the
for the control group than for the treatment group. We discuss
United States which have documented that reading
this observation in the Discussion subsection that follows. For
achievement test scores for children from low-income families
the S2 children, the difference in TWS test score gains
deteriorate significantly over the summer vacation (a
between the treatment and the control group was not
phenomenon referred to as the “summer reading setback”)
statistically significant. Finally, there was no significant
whereas those for children from middle-income families
difference between the treatment and the control group in the
remain steady or increase slightly [2]. This trend has been
other two schools for either mean fluency or TWS. Omitted
attributed to the discrepancy in the reading opportunities and
from the analysis are two S1, two S2 and ten S3 children who
materials available to these two groups over the vacation
were absent for testing.
period. In light of these studies, it is interesting to note that for
Table VI - Comparison of treatment and control gains for experiment 2 the S2 children, those who used the Reading Tutor did not
Treatment Control deteriorate in reading ability as much as those who did not use
t-test comparison it. However, the sessions with the Reading Tutor (totaling, on
(Control-1st (Tutor-1st
of treatment &
group, group, average, 11.5 hours of reading per child), without English
Test School control gains
N=33) N=37)
class at school, were not enough to prevent negative gains.
Effect
Mean Gain (SD) p-value
size
Effect of pre-test scores: A complicating factor in the data
Fluency S1 0.3 (13.4) -3.8 (22.6) 0.613 0.222
analysis is the observed unequal average pre-test scores of the
(# words read S2 -9.1 (17.1) -39.8 (16.8) <0.0001 1.816
correctly per min) S3
two groups, Tutor-1st and Control-1st, despite supposed
4.8 (6.0) 17.8 (27.0) 0.1321 -0.665
random assignment of children to the two groups. Table VII
TWS S1 2.0 (4.1) 2.9 (5.1) 0.6431 -0.197 shows that this disparity is statistically significant for the S2
(# words spelt S2 0.4 (2.0) 0.6 (3.0) 0.858 -0.068 children. Higher pre-test scores can be the cause of greater
correctly) S3 0.8 (2.4) -0.3 (2.2) 0.3198 0.474 gains [20], and so to determine whether the greater gains of
the Tutor-1st group in experiment 1 were due to the Reading
Effect of Gender: In addition to exploring how test score Tutor or the higher pre-test scores, we computed the
gains varied by school, we examined the data for evidence correlation between pre-test scores and gains. We found no
that test score gains were affected by gender. We focused this significant positive correlation between pre-test scores and
analysis on experiment 1 since that data was more expressive. gains: the correlation coefficient of mean fluency gains vs.
A t-test comparison of the mean gains of females to that of mean fluency pre-test scores was −0.301 (a small negative
males in each experiment resulted in large p-values (> 0.4), correlation), and the correlation coefficient of TWS gains to
indicating that the mean gains of females and those of males TWS pre-test scores was 0.005 (no correlation). This suggests
were not significantly different from each other. An Analysis that the greater gains of the Tutor-1st group in experiment 1
of Variance (ANOVA) also confirmed that gender did not have are indeed attributable to the Reading Tutor rather than to their
a significant effect on TWS or fluency gains in experiment 1. higher pre-test scores.
Details of these analyses are omitted for brevity.
135
Table VII - Comparison of pre-test scores for the teachers to guide students in the use of the tutor?
Tutor-1st group and the Control-1st group
Tutor-1st Control-1st
3. Children’s response: How do the children respond
(N=41) (N=42) to the Reading Tutor?
Test School
Mean pre- Mean pre-
p-value
4. Test instruments and Reading Tutor content: What
test score test score test instruments and Reading Tutor content are
S1 122.0 (45.9) 99.2 (33.9) 0.137
Fluency appropriate for the Zambian setting?
S2 65.8 (33.9) 26.2 (26.9) 0.0015
(# words read
S3 18.6 (13.2) 12.1 (19.1) 0.308 5. Feasibility of long term study: Would a remote
correctly per min)
partnership between the school and our research
TWS S1 24.7 (9.7) 24.0 (9.6) 0.850
S2 11.1 (7.0) 4.3 (3.7) 0.003
group be a feasible model for a longer-term
(# words spelt controlled study?
S3 2.1 (2.2) 1.9 (3.1) 0.840
correctly)
B. Methodology and Implementation
Tutor usage: Another significant difference between the two The selection of the school for the pilot test was done after
experiments is that tutor usage in experiment 2 was much meeting with the headmasters and teachers at both schools and
lower than in experiment 1, due to a higher level of assessing the state of the computer labs and the potential for a
absenteeism from the study during the school vacation. As successful study. Subsequently, the pilot study consisted of
Table VIII shows, this is particularly true for participation of conducting interviews with teachers at the selected school,
the S3 students whose total usage per student in the second providing training for the teachers, and making qualitative
half of the study dropped to under a third of what it was in the observations of teachers and students as they used the Reading
first half. This reduced usage may partly explain why there is Tutor. We also explored with the teachers possible details of a
no statistically significant effect of the Reading Tutor for S3 longer controlled study.
children in experiment 2. Nineteen children who were absent
C. Results and Observations
for testing either at the beginning or end of each experiment
are not included in this summary. Computing infrastructure: Both schools had a computer
room with about 20 computers, most of which were 266 Hz
Table VIII - Tutor usage in each experiment Pentium II machines with 64-128MB of RAM and 10GB hard
School Experiment 1 Experiment 2 drives. Some teachers from each school had taken a computer
Number of days of tutor S1 29.6 (12.0) 22.2 (9.4) skills training course when the computers were initially
use per student S2 36.9 (2.4) 24.0 (11.7)
S3 30.7 (2.8) 12.6 (3.9)
donated, but had not had the opportunity to reinforce their
Total time spent reading S1 12.4 (5.4) 11.9 (5.2) skills through additional training or guided practice. The
with tutor per student S2 18.6 (3.1) 11.5 (6.7) schools did not have internet access and the computers were
(hours) S3 17.3 (2.7) 6.8 (2.5) being used primarily for tasks such as typing exam questions,
Average daily time S1 22.9 (7.1) 32.6 (5.3) rather than as educational tools for students or for accessing
spent reading with tutor S2 30.1 (3.7) 26.6 (7.0)
per student (hours)
online teaching resources. Given the large class sizes—which
S3 33.9 (4.6) 31.9 (5.3)
ranged from 50 to over 100 students—and strict timetable, the
teachers also faced challenges in feasibly using the one lab to
V. ZAMBIA PILOT STUDY
teach their students from grades one to nine about computers.
The work in Zambia complements the prior studies in Ghana Some computers at both schools were not functioning due to
by investigating an in-school usage model and testing the tutor broken keyboards, mice, and power strips; some computers
in a different English-speaking African country. were not being protected from dirt during the dusty winter
ProjectEDUCATE, a non-profit organization supporting some months; power outages were a daily occurrence in Mongu;
schools in Zambia, introduced us to two under-resourced and for one school, even maintaining electricity for the
public schools in Mongu, the capital of Zambia’s Western computer lab was a challenge due to limited financial
Province. The schools had received donated computers and resources. For the pilot study, we selected the school with
were enthusiastic about the possibility of testing the Reading better maintained equipment and a higher likelihood of
Tutor. However, the Mongu District Education Board maintaining communication by telephone and email since the
Secretary’s office directed us to select only one school for the school had a telephone and the headmaster had a working
pilot test. Three of the authors conducted the study in Zambia. email address that he accessed weekly.
A. Goal Training Teachers: We trained three teachers to use the
Reading Tutor. The hour-long session covered the educational
The goals of the Zambian pilot study were to answer the
features of the tutor as well as administrative tasks such as
following questions:
managing users. After basic instruction and some time to
1. Computing infrastructure: What is the state of the
practice on their own, the teachers were able to guide students
school’s computing lab, how is it currently used,
in using the tutor. They would often give feedback to the
and can it feasibly be used for sessions with a
children as a complement to the tutor when the students had
computer-based reading tutor?
difficulty reading stories. For example, they would instruct the
2. Training Teachers: How long does it take to train
136
children to click for help when they needed it. The teachers Test instruments and Reading Tutor content: We discussed
would sometimes correct a mispronounced word that the tutor with the teachers other reading assessment options for a
did not detect. We think this involvement of teachers in the longer term study, in addition to the fluency test and Test of
early stages as the child gets used to the tutor could be an Written Spelling. We learned that through the Zambia Primary
important part of longer-term use of the Reading Tutor in this Reading Programme (PRP), the students’ literacy levels are
school setting. A similar role was played by project staff in the measured by reading standardized story books aloud to their
Ghana studies. The teachers especially appreciated the ability teachers. Each color-coded book is associated with a given
to track their students’ performance using the tutor, including reading level. Once the teacher determines that the child can
the number of new words the children had read and the successfully read at a given level, he/she moves the child up to
number of words they read per minute. a higher level reading group with a different set of books. The
Children’s response: We observed one group of eleven teachers were interested in engaging the students with the
children in grades two through four as they read one or two computers in a way that would support curricular
stories from the tutor. The students were selected by the requirements from the Ministry of Education, and this was
teachers and had varying levels of reading ability, English also emphasized to us by the Ministry of Education officials.
comprehension, and speaking fluency. As in Ghana, all but Accordingly, we discussed incorporating the standardized PRP
one student were completely new to computers. We reading material into the tutor, which will also have the
introduced them as a group to the basic components of the advantage of providing additional content choices for
computer and then gave them the opportunity to demonstrate students. We saw that many of the existing Reading Tutor
use of the mouse and keyboard to each other. We provided stories, such as those related to baseball, zoos, or recycling,
verbal instruction to students individually as they began had little relevance to children’s lives in Mongu.
reading stories from the Reading Tutor. Just as in Ghana, the Feasibility of long term study: The teachers suggested that a
speech recognition capability appeared to work adequately random subset of the children in grades 2 through 4 could be
with the Zambian students’ and teachers’ accents. Some Grade selected to participate in a controlled study. They explained
2 students seemed to have difficulty understanding English. that students in these grades have 60 minutes of reading class
This is because English is introduced as a language of per day and suggested that the intervention group of students
instruction only in Grade 2, as a part of the Zambian Ministry could spend 30 minutes of that time working with the Reading
of Education’s initiative to encourage basic literacy by Tutor, with teacher supervision, while the control group of
teaching in a familiar language in the first year of school [21]. students would remain in class. We anticipate several
The students enjoyed using the Reading Tutor; when we challenges in conducting this project as a remote field study.
Although we have email and telephone contact information
returned to the school on another day to meet with the
for the key collaborators at the school, we expect
teachers, the group of students was using the word building
communication to be difficult given their limited internet
exercises in the Reading Tutor on their own time.
access and the frequent disruptions in telephone
communications due to daily power cuts. Success would
depend on the teachers taking ownership of the project as a
result of their enthusiasm. We hope that communication with
local contacts will help address some of the expected
challenges. For example, we plan to involve the Mongu
District Education Board Secretary’s office in evaluating the
project’s progress at the school. We also engaged a local
ProjectEDUCATE technical contact who joined us at our
meetings with the schools, received training from a member of
our team, and assisted us with installation and use of the
Reading Tutor software. We hope that he will stay in touch
with us remotely, help incorporate new stories into the tutor,
Figure 3 –Teachers and a child test the Reading Tutor in Mongu check in periodically with the school, and be available to the
teachers to assist with any technical challenges.
VI. LESSONS LEARNED

From the experiences in Ghana and Zambia, we can glean
many useful lessons for our future work as well as for others
implementing similar studies. These can be broadly
categorized into lessons about building relationships and
support, lessons about running the study, and lessons about the
viability of automated tutoring.
Figure 4 – Students in Mongu using word-building exercises as a group
137
A. Building Relationships about reading in English. In both locations, the children’s

An important requirement for running these studies was the natural curiosity overcame any initial apprehension of the
process of developing partnerships at each stage in the project. unfamiliar technology.
Significant time must be devoted to building relationships and In observing the children using the Reading Tutor, we
developing a shared vision. Phone conferences can be a useful noticed that children with a basic foundation in English and
tool, but some face-to-face meetings are essential, particularly not much prior experience with computers, such as the S2
in the early stages of project planning. A staged children in Ghana, were easily engaged with the tutor. It was
implementation, as in a pilot study followed by a controlled clear that insufficient familiarity with the English language
study, helps distill key questions to ask, refine design and was a challenge for some, particularly the S3 children in
implementation decisions, and identify potential problems. Ghana and some of the Grade 2 students in Zambia. For
children who did not yet understand English well enough to
B. Logistics benefit from an automated tutor that uses only English, a tool
We learned that it was important for someone involved in that bridges between the local language and English, e.g. by
the local day-to-day running of the project to have significant giving prompts or explaining words in the local language,
decision-making power and to feel ownership of the project. might have been better. Finally, we noticed that the S1
This individual must have the ability, for example, to replace children who were already fluent readers and experienced
equipment or interface with representatives of the school with computers seemed to get bored and distracted easily
administration. Otherwise, problems can easily stall progress. when using the Reading Tutor. This might have been because
It is also essential to have access to some local technical
the Reading Tutor displays the story being read one sentence
expertise to troubleshoot and repair problems—building this
at a time, and there would sometimes be a short delay in
local support base during the pilot studies is crucial.
loading the next sentence. For a fluent reader, even this short
For the controlled study in Ghana, we found that the remote
collaboration required regular, sometimes daily, delay was noticeable and sometimes frustrating. There were
communication among the project partners; email, instant- also instances that indicated that some children got bored of
messaging and voice-over-IP were a cost effective and repeated activities and enjoyed variety. For example, many
feasible means of achieving this. children enjoyed the word-building exercises in the Reading
Not surprisingly, we learned that a process that requires a Tutor but would sometimes complain that there were too many
significant change in behavior and extra work on the part of of them.
parents, such as having the children come to school or Although the Reading Tutor is designed for use by a single
participate in a study during a vacation, is hard to sustain and user at a time, we noticed that in both pilot studies children
should be avoided if possible. However, it was also clear that would gather around a single computer and try to help each
unexpected situations are bound to arise for any study that other. This also happened to a lesser extent during the
involves cross-continent collaboration and so flexibility and controlled study in Ghana and has been observed in other
the ability to adjust the study design if necessary is essential. technology interventions in developing communities [15]. It
For example, in the controlled study in Ghana, the second half suggests that we should investigate multi-user scenarios.
of the study had to be held during the school vacation Finally, there are several usability improvements that can be
although this was not the original plan. made to the Reading Tutor to reduce the required technical
In running the Ghana studies, transportation of the children support, particularly in a developing community setting. These
was the greatest expense and so although scheduled out-of- features would apply to any PC-based intervention.
school use in an ICT center was the model chosen for the
• Installation must be easy and straightforward.
controlled study, it is not a viable long-term usage model.
However, as we learned in Zambia, there are many challenges • Customization must be easy (both for adding local
to be addressed for an in-school usage model for public under- content and to control for the installation footprint for
resourced schools, even in a school that already has a machines without much hard disk capacity).
computer lab. These issues include the limited availability of • An easy administrative interface is needed for
computers, inadequate experience on the part of the school in controlling options such as level of logging, again to
maintaining computer labs, scheduling challenges with respect deal with limited-capacity machines.
to the school timetable, and also the need to work within the
constraints of the existing literacy curriculum. VII. CONCLUSIONS AND FUTURE WORK
C. Viability of Automated Tutoring This paper presents our initial experiments in Africa with an
A key lesson was that even without prior computer automated reading tutor to improve child literacy. It
experience, the participating children were quickly able to investigates the viability and effectiveness of a computer-
acquire the skills needed to use the tutor. In general, they were based reading tutor in improving the reading ability of
excited about using the computer and about the interactive children, particularly those attending under-resourced public
features of the Reading Tutor. Assistance from the project schools. Although literacy and education for all are at the top
staff (in Ghana) or from the teachers (in Zambia) helped those of the global development agenda, not much work has been
children who were initially nervous about computer use or done regarding the role of technology in this process. This
138
work is a useful contribution both because it demonstrates that [3] J. Cohen, Statistical power analysis for the behavioral sciences (2nd
ed.), Hillsdale, NJ: Lawrence Earlbaum Associates, 1988.
there is promise for the effectiveness of the approach and [4] J. Cohen, “A power primer,” Psychological Bulletin, 112, 155-159,
because of the practical lessons regarding the implementation 1992.
of the study. We have presented an initial proof-of-concept [5] S. L. Deno, “Curriculum-based measurement: The emerging alternative,”
in Exceptional Children, 52(3):219–232, 1985.
investigation. Many additional questions would need to be
[6] J. Hasbrouck and G. A. Tindal, “Oral reading fluency norms: A valuable
explored to understand the potential for large-scale application assessment tool for reading teachers,” The Reading Teacher, 59(7), 636-
of these technologies. 644, 2006.
Next steps for our project include trying to set up a year [7] M. Kam, A. Agarwal, A. Kumar, S. Lal, A. Mathur, A. Tewari, and J.
Canny, “Designing E-Learning Games for Rural Children in India: A
long in-school study in Zambia. The purpose of the study Format for Balancing Learning with Fun,” Proceedings of ACM
would be to test the feasibility of an in-school usage model in Conference on Designing Interactive Systems (DIS '08), Cape Town,
a developing community setting. It would measure the South Africa, February 25-27, 2008.
[8] S. C. Larsen, D. D. Hammill, and L. C. Moats, Test of Written Spelling,
children’s reading gains over an entire school year of using Pro-Ed, Austin, Texas, 1999.
the Reading Tutor and would compare these gains to those [9] M. Lipson and K. Wixson, “Evaluation of the BTL and ASTEP
obtained with a non-technological approach to guided reading Programs in the Northern, Eastern, and Volta Regions of Ghana,” Report
prepared by the International Reading Association for The Education
practice. An important goal of the year-long study will be to Office, USAID/Ghana, August 2004. Available online at
learn about sustainability and the feasibility of incorporating http://www.reading.org/resources/issues/reports/ghana.html
teachers closely into this work. [10] R. G. Miller, Simultaneous statistical inference, 2nd ed., Springer Verlag,
Additionally, we will continue to develop an understanding pages 6-8, 1981.
[11] J. Mostow and G. Aist, “Evaluating tutors that listen: An overview of
of what features would be required of a literacy-enhancing Project LISTEN,” in K. D. Forbus & P.J. Feltovich (Eds), Smart
tool developed specifically for use by children in developing machines in education (pp. 169-234), Cambridge, MA: AAAI Press/The
communities. What features would be needed, taking into MIT Press, 2001.
[12] J. Mostow, G. Aist, P. Burkhead, A. Corbett, A. Cuneo, S. Eitelman, C.
consideration limited resources and equipment, as well as Huang, B. Junker, M. B. Sklar, and B. Tobin, “Evaluation of an
current instructional practices? What tutorial methods are automated Reading Tutor that listens: Comparison to human tutoring
most useful? Could a lighter weight tool be designed for and classroom instruction,” Journal of Educational Computing
Research, 29(1), 61-117, 2003
alternative platforms, such as mobile phones, as has been done [13] J. Mostow, G. Aist, C. Huang, B. Junker, R. Kennedy, H. Lan, D.
with the educational games for English-as-a-second language Latimer, R. O’Connor, R. Tassone, B. Tobin, and A. Wierman, “4-Month
instruction in the MILLEE project [7]? There is significant evaluation of a learner-controlled Reading Tutor that listens,” In V. M.
Holland & F. P. Fisher (Eds.), The Path of Speech Technologies in
scope for further research on the role of computing technology
Computer Assisted Language Learning: From Research Toward
in improving child literacy in developing communities. Practice (pp. 201-219), New York: Routledge, 2008.
[14] National Institute of Child Health and Human Development, Report of
ACKNOWLEDGMENT the National Reading Panel. Teaching children to read: An evidence-
based assessment of the scientific research literature on reading and its
We are grateful to Steve Fienberg for his advice on the implications for reading instruction (NIH Publication No. 00-4769),
statistical analysis as part of the Statistical Practice course at Washington, DC: U.S. Government Printing Office, 2000.
[15] U. S. Pawar, J. Pal, K. Toyama, “Multiple Mice for Computers in
Carnegie Mellon University. We express our appreciation to Education in Developing Countries,” International Conference on
the Ghana-India Kofi Annan Centre for Excellence in ICT Information and Communication Technologies and Development, ICTD
(AITI-KACE), Accra, Ghana, for their role in the controlled 2006, pp.64-71, May 2006.
[16] R. Poulsen, P. Hastings and D. Allbritton, “Tutoring Bilingual Students
study, particularly to Dorothy Gordon for her support, Patricia with an Automated Reading Tutor That Listens,” Journal of Educational
Nyahe for her tireless management of the study, and the Computing Research, 36(2), 191-221, 2007.
UNESCO Ghana Office for sponsoring their work. We also [17] Project LISTEN Videos, available at http://www.cs.cmu.edu/~listen
[18] K. Reeder, M. Early, M. Kendrick, J. Shapiro, and J. Wakefield, “The
thank Leslie Casely-Hayford of Associates for Change,
Role of L1 in Young Multilingual Readers’ Success With a Computer-
Ghana, for her advice and contributions to the pilot and Based Reading Tutor,” Talk at the Fifth International Symposium on
controlled studies in Accra. We acknowledge the valuable Bilingualism, Barcelona, Spain, April 2005.
contributions of Cybercity Internet Café and [19] M. R. Shinn, N. Knutson, R. H. Good, W. D. Tilly, and V. L. Collins,
“Curriculum-based measurement of oral reading fluency: A confirmatory
ProjectEDUCATE to the pilot studies in Ghana and Zambia analysis of its relation to reading,” School Psychology Review, 21:459-
respectfully. Finally, we are indebted to all the participating 479, 1992.
children and schools, and to all others who contributed to the [20] K. E. Stanovich, Progress in Understanding Reading: Scientific
Foundations and New Frontiers, New York: Guilford Press, 2000.
project in various ways. [21] UNESCO, “Education for All Global Monitoring Report 2005 –
Education for All: The Quality Imperative,” United Nations Educational,
REFERENCES Cultural and Scientific Organization (UNESCO) Publishing, 2005.
Available online at http://portal.unesco.org/education/en/ev.php-
[1] M. J. Adams, "The promise of automatic speech recognition for fostering URL_ID=35939&URL_DO=DO_TOPIC&URL_SECTION=201.html
literacy growth in children and adults,” in M.C. McKenna, L.D. Labbo, [22] UNESCO, “The Plurality of Literacy and its Implications for Policies
R. D. Kieffer, & D. Reinking (Eds.), International Handbook of Literacy and Programmes.” UNESCO Education Sector Position Paper. United
and Technology, Volume 2. Mahwah, NJ: Lawrence Erlbaum Associates, Nations Educational, Scientific and Cultural Organization, Paris, 2004.
2006. Available online at
[2] R. L. Allington and A. McGill-Franzen, “The Impact of Summer Setback http://unesdoc.unesco.org/images/0013/001362/136246e.pdf
on the Reading Achievement Gap,” in Phi Delta Kappan 85(1):68-75,
September 2003.
139
Improving Literacy in Rural India:

Cellphone Games in an After-School Program
Matthew Kam, Anuj Kumar, Shirley Jain, Akhil Mathur, and John Canny
language of all professions and higher education, but also

Abstract—Literacy is one of the great challenges in the important for mid-level service jobs: retail, clerical, teaching,
developing world. But universal education is an unattainable law enforcement, etc. that are the most common steps above
dream for those children who lack access to quality educational menial labor. The value of English is widely recognized by
resources such as well-prepared teachers and schools. Worse,
many of them do not attend school regularly due to their need to
ordinary Indians [2], and it is in fact the poorest citizens who
work for the family in the agricultural fields or households. This are lobbying most strongly to expand English teaching.
work commitment puts formal education far out of their reach. English is thus the language of power in India associated
On the other hand, educational games on cellphones hold the with the middle and upper classes [3][4]. In other developing
promise of making learning more accessible and enjoyable. In regions, it is another language such as Spanish, Mandarin, or
our project’s 4th year, we reached a stage where we could French which is not native to most of the population. We
implement a semester-long pilot on cellphone-based learning.
The pilot study took the form of an after-school program in a
believe that many of our lessons will apply to other languages
village in India. This paper reports on this summative learning although our focus is on English as a Second Language (ESL).
assessment. While we found learning benefits across the board, it But the public school systems in developing regions face
seemed that more of the gains accrued to those children who insurmountable difficulties. In India, for example, we were
were better equipped to take advantage of this opportunity. We consistently unable to converse in English with those teachers
conclude with future directions for designing educational games responsible for teaching English in poor schools, where the
that target less well-prepared children in developing regions.
overwhelming majority of children in the country struggle to
Index Terms—cellphone, English as a Second Language (ESL), learn. More important, public schooling is out of the reach of
literacy, mobile game, pilot study large numbers of children in rural areas and the urban slums
who cannot attend school regularly, due to their need to work
for the family in the agricultural fields or households [5].
I. INTRODUCTION At the same time, cellphones are increasingly adopted in the
L ITERACY is one of the great challenges in developing developing world, and an increasing fraction of these phones
regions. Despite huge improvements in recent decades, feature multimedia capabilities for gaming and photos. These
literacy levels in many poor countries remain shockingly low. devices are a promising vehicle for out-of-school learning to
Even more challenging is the tension between regional and complement formal schooling. In particular, we believe that
global “power” languages, that economic opportunities are ESL learning games on cellphones present an opportunity to
often closed to those literate only in a regional language. For dramatically expand the reach of English learning, by making
instance, India is a country with 22 regional and 2 national it possible to acquire ESL in out-of-school settings that can be
languages, i.e. Hindi and English. But English, together with more convenient than school. Games can make learning more
computer skills, are the two most requested skills in surveys of engaging while incorporating good educational principles [6].
poor parents [1]. English is a great economic enabler. It is the More important, a large-scale evaluation with urban slums
children in India has shown significant learning benefits from
games that target mathematics [7]. We believe that similar
Manuscript received September 22, 2008. This work was supported in part outcomes can be replicated with e-learning games that target
by the U.S. National Science Foundation under Grant 0326582, a Qualcomm
BREW Wireless Reach award, and sponsorship-in-kind from Sony Creative
literacy.
Software. The challenge in evaluating any language learning project,
Matthew Kam is an Assistant Professor in Carnegie Mellon University’s however, is that language acquisition is a long-term process
Human-Computer Interaction Institute (phone: +1 412-268-9805; fax: +1 412-
268-1266; email: mattkam@cs.cmu.edu). He was with the Berkeley Institute
on the learner’s part. Worse, with a novel technology solution
of Design and the Department of Electrical Engineering and Computer that has yet to be institutionalized, there were tremendous
Sciences at the University of California, Berkeley when this work was carried logistical obstacles in running a pilot study over a non-trivial
out. John Canny is the founding director of BID and holds the Paul and Stacy duration. After 3 years, in which we commenced with needs
Jacobs Distinguished Professorship of Engineering at UC Berkeley.
Shirley Jain participated in the pilot study as a curriculum developer and assessments and feasibility studies, followed by subsequent
local supervisor. rounds of field testing interleaved with numerous iterations on
Anuj Kumar and Akhil Mathur are with the Dhirubhai Ambani Institute of our technology designs, we have established the necessary
Information and Communication Technology, Gujarat, India.
relationships with local partners for such an evaluation. This
140
paper describes the results from a semester-long pilot study – conducting a learning assessment for a novel educational
the longest so far in this project – which took place during the technology over a substantial timeframe when it is not yet
project’s 4th year. The study involved 27 rural children who integrated into the everyday operations of a formal entity.
participated in an after-school program that we implemented
in their village.
III. OVERVIEW OF PILOT STUDY
The pilot study was carried out in collaboration with a non-
II. RELATED WORK government organization in North India under the terms of a
Most work on technology-assisted language learning in the Memorandum of Understanding. The study took the form of
developing world does not explore the convenience that the an after-school program, which we held during the afternoons
cellphone’s mobility offers. Banerjee et al. [7] report a large- at a private village school affiliated with this NGO. However,
scale evaluation with mathematics learning games on desktop our goal was to investigate learning impacts that ESL learning
computers, carried out over 2 years with urban slums children games on cellphones have on lower-income rural children. As
in India. Mitra et al. [8] describe a study in India with slums such, students who were already enrolled in this school were
children over 5 months, which involved a “hole-in-the-wall” ineligible to participate in the study. Instead, we invited those
public computer installed with speech-to-text software. Dias et parents who could not afford the fees for this private school –
al. describe a computer-based tutor software for improving and hence sent their children to less expensive schools in the
reading in Ghana [9]. Kothari’s karaoke-like approach [10] same area – to give consent for their children to participate.
targets native language – not second language – literacy using In the after-school program, we ran three sessions per week,
television. on average. Each session lasted two hours in the afternoon.
With the cellphone’s increasing ubiquity in Africa, Brown Children from neighboring villages attended the after-school
[11] argues that it is timely to envision a future where the sessions after finishing their regular classes in the morning. In
cellphone plays a pivotal role in education in Africa. Kam et the after-school sessions, we loaned cellphones preloaded with
al. [12] describe how a set of ESL-learning cellphone games ESL learning games to participants. The after-school program
that targets children in rural India have undergone numerous took place from late December 2007 to early April 2008, and
iterations, based on successive, short-term formative studies. spanned sessions on 38 days in total.
With the exception of Kam et al., the only education-related
projects we know of in the developing world that leverage the
cellphone are Islam et al. [13] in Bangladesh, and Librero et IV. DATA COLLECTION
al. [14] in Mongolia and the Philippines. Both projects rely on As our preparation for this pilot study, we made two trips to
Short Messaging Service and target university students, unlike India, i.e. once in the summer of 2007 to familiarize ourselves
our approach. with the pilot location and end-user community, and a second
To date, Horowitz et al. [15] is the only study we know of time in December 2007 to kick-off the actual pilot. 4 local
that examines the cellphone for promoting literacy. However, staff members were hired to run the after-school sessions on
this study took place in an industrialized country (USA), even an everyday basis. 3 of them were engineering undergraduates
though participants included households below the poverty in their last semester, while the last member had graduated a
line. In the study, Sesame Street videos that target the English few years ago. We spent two weeks training them to run the
alphabet were streamed to preschool children over cellphones after-school sessions and perform data collection, and
throughout an 8-week period. Our paper therefore contributes continued to coordinate with them regularly via conference
to the literature as the first learning assessment on cellphone- calls and emails after we left India.
based language instruction in developing regions. We interviewed participants on their demographics such as
He et al. [16] describe a 2-year randomized evaluation of a their ages and the grades they were currently enrolled in in
LeapPad-like device that supports custom software modules school. During the interviews, we also asked other questions,
for English learning. This interactive system involves a paper such as the number of cellphones that their households owned,
book attached to a stylus and supports audio output. It lacks a what they currently and/or had previously used cellphones for,
visual display, unlike a cellphone, but overlaps with our goal their television watching habits and frequency, as well as their
of making literacy learning more accessible in the developing parents’ occupations. The questions on media and technology
world using portable devices. exposure were included because these variables were expected
Among the learning technologies for developing countries, to impact participant ability to learn using cellphone games.
one of the most novel devices is the multiple-mice computer To ensure that each participant has the basic numeracy and
described in Pawar et al. [17]. It was intended for collocated ESL literacy to benefit from cellphone-based learning in the
learning by a group of children around each computer, with a program, participants were required to pass a qualifying test,
mouse input device for each child. It has since been extended i.e. obtain at least 50% of the total score. The test required
to distance learning in Moraveji et al. [18]. The evaluations in them to complete one-word blanks using English words about
both papers were short-term, and underscored the difficulty in themselves, e.g. name, age, school, grade, etc. They were also
141
asked to fill in the missing letters in the alphabetic sequence, V. PARTICIPANTS

write numbers in the Arabic notation, match words with their Owing to the strong relationships that our NGO partner had
pictures, spell the words for everyday objects, and describe a built with the local community over more than a decade, we
picture of a market scene with short sentences. The qualifying were able to generate a high level of support among parents in
test was designed such that an average child in India with no this community. In total, we obtained consent for 47 children
learning disabilities who has finished 1st grade in a reputable to participate in the study. However, we needed to turn 16 of
urban school should obtain a perfect score on it. them away; 15 children did not pass the qualifying test while
By using the qualifying test as a screener, we ensured that the 16th was attending private tuition for English. The latter
participants were numerate. This was important because we represented a confounding variable. Of the 31 children whom
had previously found it difficult to teach children to use the we started the pilot with, 4 of them left the program mid-way.
cellphone keypad’s to play e-learning games when they were Reasons for attrition include time conflicts with private tuition
not familiar with the numbers from 0 to 9 in Arabic notation. (2 children) and disinterest in attending the sessions (another 2
Similarly, by ensuring that participants were familiar with the children). From post-deployment interviews, we understand
English alphabet, we could target a more advanced syllabus the latter was due to caste tensions between those 2 children,
that went beyond the alphabet. We made this decision since who belonged to the lower castes, and some upper-caste
Horowitz et al. [15] had already investigated the efficacy of children in the program.
cellphone-based learning for the English letters in the context
of preschool children in the USA. A. Demographics
Since success in acquiring a second language is correlated The 27 children who participated in the study until it ended
with literacy in one’s native language, we administered a test were aged 7 to 14 (mean = 11½ years) and belonged to grades
which evaluated the ability of the participants to read in Hindi. 2 to 9 (mean = 6th grade). There were 11 boys and 16 girls. 5
Every child was given a short passage that described a diet for children came from the upper castes while others belonged to
promoting dental health. Each child was then asked to read the the lower castes. The gender and caste breakdown seemed to
passage aloud so that we could observe his or her fluency and mirror the demographics in the community. Every participant
accuracy. These sessions were videotaped. After that, every attended between 8 and 29 sessions (mean = 20) in the after-
participant was asked to write answers to written questions school program, broken down according to the following three
that tested his or her comprehension of the passage, in Hindi. functions:
We had designed this test such that an average child who has • Cellphone training: 0 to 5 sessions (mean = 4) where we
finished 3rd grade in a reputable urban school in India should taught participants how to use the cellphones, perform
obtain a perfect score on it. alphanumeric input and play mobile games,
Our primary method of assessment was to administer pre- • ESL learning: 4 to 17 sessions (mean = 10) in which
and post-tests which evaluated participants on their ability to participants played ESL learning games on the cellphones,
spell the common nouns that the curriculum for the pilot study and
targeted. Although the curriculum targeted other competencies • Assessment: 4 to 7 sessions (mean = 6) for administrative
such as listening comprehension and the recognition of written tasks and data collection, e.g. demographics interviews and
words, our assessment emphasized spelling, which as a recall various tests.
task was cognitively more difficult than recognition tasks.
We maintained attendance records for the participants for In India, traditionally, only the upper castes owned land. As
every session. We also videotaped each session so as to have such, the upper castes earn their livelihood on the land or run
contextual data that could potentially account for their test small businesses, while lower castes graze their goats, work as
performances. The video recordings captured the classroom daily-wage laborers or perform menial jobs in the homes of
proceedings, and individual participants’ interactions with the the upper castes. Land-owning and non-land-owning families
games. The latter recordings captured participants’ levels of told us that they earned up to 100,000 (US$2,500) and 50,000
engagement with the games as shown in their facial and body rupees (US$1,250) respectively per year.
expressions. The pilot staff member who was responsible for B. Hindi and English Baseline
videotaping the sessions tried to ensure that every participant 26 of the 27 participants were enrolled in the same school,
was videotaped playing at least one level in the curriculum per where Hindi is the medium of instruction. The last participant
day. The recordings were later transcribed and translated from was a school dropout. Assuming regular school attendance,
Hindi to English. the typical participant would have taken classes on Hindi and
Finally, for every session, we asked pilot personnel to write English for 5½ and 3½ years respectively prior to the study.
a report which summarized what happened in that session, as We devised a grading rubric to evaluate each participant on
well as how well each participant interacted with the games. the Hindi literacy test and qualifying test. On the former test,
The latter not only covered usability and learning obstacles, participants scored 7.9 out of 18 on average (σ = 4.5, n = 19).
but also included pilot staff’s observations on the attitude and 2 participants turned in blank answer sheets. We observed the
persistence that each child demonstrated towards learning. following problems in the submissions:
142
• Wrong answers due to poor comprehension of the questions participants came from families that owned cellphones with a
(5% of the test-takers) or passage (53%), or responses that color screen, as opposed to monochrome display. Two of the
simply repeated the questions (32%) above cellphones – both of which belonged to 2 of the above
• Spelling errors (16%) 3 boys – contained built-in cameras.
• Grammatical errors, i.e. using the incorrect form of the verb Nonetheless, cellphone ownership and access were separate
for the subject’s gender (21%), or the incorrect form of the issues. Among the 25 children whose families owned at least
noun for the subject’s singularity vs. plurality (5%) one phone, 6 of them – 5 girls aged 7-11 and a boy aged 12 –
• Inability to phrase responses in complete sentences (5%) were prohibited by their parents from using the phone, either
entirely or most of the time. In general, in poorer families, it
TABLE I
BREAKDOWN OF PARTICIPANT PERFORMANCE ON QUALIFYING TEST
seemed that children were allowed to receive (free) incoming
Poor Fair Good calls, but not play mobile games lest they drop the devices. On
About 13% left blanks 32% filled in blanks 55% filled in blanks the other hand, in wealthier families, children were allowed to
myself empty or filled them with at least 1 with correct
in Hindi misspelling spellings for most
play on the phones. As such, although most participants were
questions familiar with cellphones, it appeared that rural parents were
Alphabet 5% filled in less than 13% filled in ~11 82% filled in at least
7 blanks in blanks in alphabetic 12 out of 13 blanks
more willing to entrust these relatively costly devices to their
alphabetic sequence sequence correctly in alphabetic sons (vs. daughters). In total, 15 of the 27 participants (56%)
correctly sequence correctly
14% matched up to 2 14% matched 4 out 72% matched at least
reported that they had played cellphone games before.
Word
recognition out of 6 words with of 6 words with 5 out of 6 words
correct pictures correct pictures with correct pictures
Spelling 41% spelt up to 2 18% spelt about 3 41% spelt at least 4
out of 6 words out of 6 words out of 6 words VI. CURRICULUM AND GAME DESIGN
correctly correctly correctly
Picture 18% made no 50% wrote answers 32% wrote One of the major challenges with carrying out a pilot study
description attempt to answer as individual words, intelligible answers over a non-trivial timeframe was that we needed to develop
this section not full sentences in sentences
sufficient digital content that could last throughout its entire
duration. We ensured that our syllabus was aligned with local
On the qualifying test, on average, participants scored 44.0
ESL learning needs in India by recruiting a local ESL teacher
out of 50 (σ = 5.5, n = 22). Our grading rubric indicated how
as our curriculum developer. She had a decade’s experience as
participants should be classified as “poor,” “fair” and “good”
an ESL teacher at a prestigious urban school, located in the
on every section of the test. Table I gives the breakdown of
same geographic region as the after-school program.
how test-takers were distributed across categories for selected
sections, and descriptions of the categories. In summary, the A. Curriculum Design
average participant had a good knowledge of the alphabet and Given the above attendance rate, the ESL curriculum for the
a fair vocabulary of written words that she could read. On the pilot was designed to be comparable to the amount of material
other hand, she was weak in recalling and spelling everyday that a qualified teacher could reasonably cover in 18 hours
nouns, and even weaker in constructing complete sentences with rural children in a classroom. The syllabus was situated
with these words. Despite the wide range in the ages of the within the classroom theme, which participants could readily
participants, it appeared from their performance on the above relate to. Concretely, the syllabus included:
tests that the variation in their English proficiency was much • Common nouns that are found in the typical classroom, e.g.
narrower. More specifically, we estimated that the average chair, table, door.
participant was comparable to an urban child in India who had • Verbs that can be performed with the above nouns, e.g. sit,
taken between 1 and 2 years of English classes. write, open, close.
Notably, only 10 children (45%) could spell their names • Sentence structures for constructing sentences out of the
correctly in English on the qualifying test papers. We had a above nouns and verbs, e.g. “This is a __.”
chance later to interview the teacher who taught them English • Sentence structures for phrasing question-and-answer
in their school. She revealed that her pedagogical approach sequences with the above nouns and verbs, e.g. “What is
revolved around having students copy sentences from English this?”, “Where is the __?”
textbooks into their notebooks. She felt that it was not worth
The curriculum design took participant performance on the
putting in more effort to teach English since she believed she
qualifying test into account. The curriculum was also based on
was underpaid.
our attempts to converse informally with participants, during
C. Technology Baseline which we learned they did not comprehend simple questions
Among the 27 participants, 25 of them came from families about themselves, did not know the English words for objects
who owned at least one cellphone; 5 participants belonged to around them (e.g. in the classroom), and made grammatical
families that owned 2 phones each while 2 participants came errors. The curriculum therefore targeted the above syllabus in
from families which owned 3 phones each. The cellphone was terms of listening comprehension, word recognition (of the
usually used by the eldest male member in every family, and written word), sentence construction and spelling.
in fact, 3 boys aged 13-14 possessed their own cellphones. 8
143
B. Game Design
We designed a set of ESL learning games for the cellphone
platform that targeted the above curriculum, and piloted them
in the after-school program. Our designs drew on 3 resources,
namely:
1) recurring patterns in state-of-the-art commercial software
applications for language learning, which represented best
practices that we reused to avoid reinventing the wheel,
2) traditional village games, which more closely matched the
expectations and understandings that rural children have
about games, in comparison to contemporary videogames Fig. 3 and 4. Figures 3 and 4 introduce additional phrases that the nouns and
that were largely Westernized, and verbs in the syllabus can be used in. Figure 3 teaches a phrase that associates
3) lessons from several previous rounds of field-testing and the verb “sit” with the noun “chair.” Figure 4 shows how to ask questions
using the “Where” keyword. Abstract phrases and function words such as
iterations with rural children elsewhere in India [12].
“where” are difficult to convey graphically. Hence, when they are taught for
In this subsection, we walk the reader through a subset of the the first time, the software explains their meanings orally in Hindi.
screen designs.
The games tested players on their comprehension and recall
of the words and phrases. For example, the game shown in
Figure 5 says the word aloud for one of the objects displayed
on screen. The player needs to identify the correct object and
push it onto the area that is blinking blue. At the same time, he
needs to avoid the balls thrown by the computer-controlled
opponent. This game was an adaptation of Giti Phod, which
was one of the traditional games that children play in Indian
villages. In Giti Phod, players in a team have to arrange some
objects (e.g. rocks) into a given configuration (e.g. a heap),
while avoiding being hit by a ball thrown by members in the
opposing team. In our experience, we have observed that rural
Fig. 1 and 2. Figure 1 introduces the English vocabulary for common nouns
in the classroom. Word-picture association is a technique employed by many children found it more intuitive to understand videogame rules
successful commercial language learning software. As the boy moves to each when the designs of these videogames drew on the rules found
object, the software highlights the corresponding word in a different (green) in the traditional village games that they play everyday.
color and says the word aloud. Figure 2 situates these objects in a classroom
scene and builds on the earlier screen by demonstrating how to use the nouns
in complete sentences. As the boy moves to each object, the software says the
“This is a __” phase aloud for the corresponding object.
In earlier field studies, we observed that rural children did

not readily associate a game with learning. It seemed that they
viewed a game as an activity to be played purely for pleasure,
and did not pay attention to the educational content embedded
within game activities. On the other hand, when educational
content such as English words and phrases were introduced in
non-interactive screens separate from interactive game
screens, the rural children appeared to grasp more intuitively
that the software was trying to teach them those English words
and phrases. Users subsequently paid more attention to the
Fig. 5. A word-picture matching game which is an adaptation of one of the
latter. Figures 1 to 4 show some screenshots in which we traditional village games that children in rural Indian play everyday.
introduced words and phrases – both written and spoken – to
the learners. Given that television has become a pervasive media among
all economic classes in India, it only made sense to draw on
popular culture in India to make our designs more appealing
to children there. One of these sources is Sesame Street, which
is a successful television program for young children that has
local co-productions around the world – in both industrialized
and developing countries. Its producers in India have found
some of its localized characters to be popular with children in
India, and we incorporated those characters into our designs
144
for teaching (Figure 4), quizzing (Figure 6) and congratulating

(Figure 7) the user.
Fig. 9. The second level of hints in the spelling activity. For every blank, a set
of possible letters are displayed to provide the player with some assistance, if
Fig. 6 and 7. Localized characters from the Indian production of Sesame he was unable to spell the word correctly after two attempts.
Street tested the player on his ability to engage in question-and-answer style
dialogues, and performed a victory dance for the player upon successful
completion of each level in the game. The curriculum is broken up into a total of 6 levels in the
games. On every screen, the player can access a menu through
The activity that targets spelling skills is shown in Figure 8. a shortcut button. Among various options, this menu permits
The player is given an image (e.g. blackboard) and is required him to move to an earlier level in the curriculum to repeat the
to spell the word corresponding to it. Some of the letters in the material, as well as to move to higher levels in the curriculum.
word are displayed, while blanks are shown for the remaining The software was designed so as not to require airtime, which
letters. The player moves between blanks with the arrow keys. was expensive for most rural families. We implemented the
Once he has filled in all the blanks and submitted his answer, games on Adobe’s Flash Lite and Qualcomm’s BREW
the correct and wrong letters are displayed in green and red (Binary Runtime Environment for Wireless) platforms. We
respectively. This feedback constitutes the first level of hints piloted the games on Motorola’s Razr V3m cellphone model,
that we have designed to help the learner arrive at the correct which has a fairly large screen.
spelling. If the player spells the word correctly, he proceeds to
the next game. Otherwise, all blanks are cleared after a short
pause and he is required to spell the same word again. VII. PILOT SESSIONS
In those sessions where participants were taught how to use
the cellphone, they were shown how to move their sprites with
the arrow buttons. They were also taught how to perform
alphanumeric text entry, since most of them did not know this.
Sprite movement and text input were essential skills for the
games we designed. Pilot staff were therefore asked to write
some simple, short sentences on the blackboard, and ensure
that each participant demonstrated his ability to enter those
sentences via text input.
Some other sessions focused on administrative tasks, such
as the above tests and demographics interviews. We learned
that a few participants had difficulty reading a small subset of
the English alphabet despite having passed the qualifying test.
Fig. 8. The first level of hints in the spelling activity. After the player has We spent two sessions coaching them on those less-frequently
tried to spell the word by filling in the blanks with letters, the correct and encountered letters, so that they would be better prepared for
incorrect letters are shown in green and red respectively. the syllabus targeted in the pilot. Next, at least 8 children had
seen the localized Sesame Street characters on television, but
If the player is unsuccessful in spelling a word correctly
did not know their names. To help participants better relate to
after two attempts, the second level of hints (Figure 9) appears
the characters, so that our games would appeal to them even
to provide him with additional learning support. Based on the
more, we introduced the characters at the start of the semester.
blank that the cursor is currently located at, the game displays
We also screened 3 localized episodes on separate occasions.
a set of possible letters for him to narrow down the choice of
These episodes were chosen such that they were educational
candidate letters.
but did not target English learning. Each episode lasted ½
hour, and we observed that participants enjoyed the humorous
acts performed by the characters.
145
Fig. 10. In the after-school sessions, each participant was loaned a cellphone Fig. 11. Frequency histogram of participant scores on the pre- and post-tests.
preloaded with English language learning games. Participants were taught
how to start the games, and were asked to focus on learning English when
playing the games on their own.
The average post-test gains was 3.4 out of 18 (σ = 3.3, n =
24). The gains exhibited a fairly large variation, and ranged
The remaining, and majority of, sessions focused on ESL from -2 (two participants exhibited negative gains) to 9 out of
learning. A 2-hour session was typically structured as follow: 18. We present the frequency histogram for post-test gains in
after an exchange of greetings, pilot staff took attendance and Figure 12.
briefed participants on the learning objectives for that day. If
new games were deployed that day, pilot staff explained and
demonstrated how to play them to the participants in small
groups. Each participant was then handed a cellphone to play
the games on her own (Figure 10), and were told to focus on
learning the English syllabus that the games covered. Children
who were absent on previous sessions received help from pilot
personnel in learning how to play those games that they were
unfamiliar with. Pilot staff were limited to providing technical
support; and were explicitly instructed not to teach English or
communicate with participants in English. There was a short
break of 10 to 15 minutes in the middle of each session. At the Fig. 12. Frequency histogram of participant post-test gains.
end of each session, pilot staff took back the phones so that
they could charge their batteries overnight and download new B. High-Gains vs. Low-Gains Learners
games onto them. Each participants received a small packet of
We sought to understand how participants’ post-test gains
biscuits after every session.
were correlated with their demographics and performance on
other tests. We also categorized participants into two groups,
namely, “high-gains learners” and “low-gains learners,” based
VIII. QUANTITATIVE RESULTS
on their post-test gains. A learner whose post-test gains
On the pre- and post-tests, test-takers were awarded 1 point exceeded the mean of 3.4 was categorized as a “high-gains
for each common noun in the syllabus that was spelt correctly. learner,” else he was classified as a “low-gains learner.” In all,
A. Post-Test Gains 9 participants were classified as high-gains learners while 15
The mean pre-test score was 5.2 out of 18 (σ = 3.3, n = 27) participants were categorized as low-gains learners. 3 of the
while the mean post-test score was 8.4 out of 18 (σ = 5.5, n = 27 participants could not be classified since they were absent
24). Participants exhibited significant post-test gains on a one- on the day when the post-test was administered.
tailed t-test (p = 0.007). We present the frequency histograms On a normalized scale, when the 27 participants were taken
for both scores in Figure 11. They illustrate that the score as one group, the average pre-test score was 29% whereas the
distribution had shifted toward the higher end of the spectrum average post-test score was 47%. The latter score did not seem
after the deployment. high in absolute terms, i.e. on average, a participant could not
spell over half of the common nouns targeted in the syllabus
by the end of the intervention. However, once the participants
had been classified, on a normalized scale, high-gains learners
scored 41% (80%) on the pre-test (post-test) whereas low-
gains learners scored 19% (27%) on the pre-test (post-test), on
average. In other words, high-gains learners not only showed
larger post-test gains but also appeared to have a higher mean
146
pre-test score at the start of the intervention. More important, The proportions in Table III were presented for the sake of
post-test gains for both high-gains (p < 0.001) and low-gains completeness. We were unable to perform any statistical tests
learners (p = 0.076) were significant, with effect sizes of 0.54 on these proportions due to the small sample size, which for
and 2.24 respectively. That is, both categories of participants example did not satisfy the standard binomial requirement.
exhibited learning gains. (But average post-test gains for low- We thus caution the reader against drawing firm conclusions
gains learners were only marginally significant, because of the from these statistics. However, when examining individual
2 participants who obtained lower scores on the post-test learners to identify surprising cases, we took the demographic
compared to their pre-test.) variables in Table III into consideration. The analysis is
deferred to the following section.
TABLE II
HIGH-GAINS VS. LOW-GAINS LEARNERS IN TERMS OF DEMOGRAPHICS TABLE IV
Age Grade Days Days HIGH-GAINS VS. LOW-GAINS LEARNERS IN TERMS OF TEST SCORES
(Years) Enrolled Spent Spent Qualify- Qualify- Hindi Pre- Post-
in School Learning Playing ing Test ing Test, Literacy Test Test
ESL ESL (out of Spelling Test (out (out
Games Games 50) Section (out of of 18) of 18)
Low- Mean 10.7 5th 3.9 10.5 (out of 6 18)
gains σ 1.8 1.9 1.4 4.2 words)*
learners Min. 7 2nd 0 5 Low- Mean 42.9 1.2 6.3 3.5 4.8
(n=15) gains 2.9 1.1 4.2 2.3 2.4
Max. 14 9th 5 16 σ
High- Mean 12.8 8th 3.8 10.2 learners Min. 37 0 0 0 2
gains σ 1.2 0.9 0.8 5.2 (n=15) Max. 46.5 4 14 10 12
learners Min. 11 7th 2 4 High- Mean 47.1 3.4 12.0 7.4 14.4
(n=9) gains 1.8 1.6 1.7 3.1 3.6
Max. 14 9th 5 17 σ
Is difference Yes Yes No No learners Min. 43.5 2 10.5 2 6
between means (p = 0.002) (p < 0.001) (p = 0.4) (p = 0.4) (n=9) Max. 49 6 14 13 18
significant? Is difference Yes Yes Yes Yes Yes
Correlation with 0.45 0.61 0.11 0.10 significant? (p < 0.001) (p = 0.001) (p < 0.001) (p = (p <
post-test gains (r) 0.003) 0.001)
Correlation with 0.57 0.70 0.45 0.46 0.86
post-test gains (r)
In Tables II and III, we examined how high-gains learners *In this column, we present the number of words that participants spelt
may differ from low-gains learners in terms of demographics. correctly on the spelling section of the qualifying test, out of a total of 6
On the whole, the high-gains learners did not appear to differ words.
significantly from low-gains learners in terms of the number
of days that they spent on learning how to play the cellphone- Table IV compares the high-gains and low-gains learners in
based games (p = 0.4) or actually playing the games to learn terms of test scores. The former outperformed the latter on the
ESL (p = 0.4). Instead, high-gains learners belonged to higher Hindi test (p < 0.001). Next, we analyzed the qualifying test
ages (p = 0.002) and were enrolled in more advanced grades results at two levels, namely, the score for the entire test as
in school (p < 0.001). In fact, post-test gains exhibited high well as the score on the spelling section. We found that high-
correlation with grade levels that participants were enrolled in gains learners outperformed low-gains learners on the entire
school (r = 0.61) and medium correlation with age (r = 0.45). test (p < 0.001) as well as on the spelling section (p = 0.001).
High-gains learners also obtained higher scores on the pre-test
(p = 0.003) and post-test (p < 0.001), vis-à-vis low-gains
TABLE III learners. In fact, participants’ post-test gains exhibited a high
HIGH-GAINS VS. LOW-GAINS LEARNERS IN TERMS OF DEMOGRAPHICS degree of correlation with their qualifying test scores, for both
Sex Caste Media Exposure Attitude* the entire test (r = 0.57) and spelling section (r = 0.70). On the
Low- 67% 77% 73% have played 24%, 38% & 38%
gains (33%) (23%) games on cellphones were described as other hand, post-test gains had a lower correlation with Hindi
learners were belonged prior to pilot; below average, literacy levels (r = 0.45) and pre-test scores (r = 0.46).
females to lower 60% (40%) watched average and
(n=15) (males) (upper) less (equal to or more) above average
castes than 1 hour of TV per learners IX. QUALITATIVE RESULTS
day respectively
High- 44% 88% 56% have played 29% and 71% The above quantitative results suggested that current levels
gains (56%) (12%) games on cellphones were described as
were belonged prior to pilot; below average
of spelling proficiency and grades enrolled in school were the
learners
(n=9)
females to lower 40% (60%) watched and above strongest predictors of success in learning how to spell new
(males) (upper) less (equal to or more) average learners
castes than 1 hour of TV per respectively
words through the cellphone-based games which we designed.
day Higher levels of Hindi literacy and academic preparation were
*The Attitude column is based on the observations that pilot personnel also associated with higher post-test gains.
have on the seriousness and aptitude that participants exhibited as learners
On the other hand, the number of sessions that participants
throughout the pilot. These qualitative comments were subsequently coded
into the “below average”, “average” and “above average” learner categories. had with the cellphone games – both for learning how to play
the ESL learning games and learning ESL through the games
147
– were not associated with post-test achievements. Among the on their first attempt in the spelling activity), or with only the
24 children whom we have post-test gains data on, 5 of them first level of hints. It seemed that they did not require much
were classified as high-gains learners despite having played scaffolding support from the software. In fact, from the video
the games on only 4 to 7 days (mean = 6 days). Conversely, 7 recordings of 9 high-gains learners, we saw that 5 (56%) and
participants were classified as low-gains learners in spite of 1 (11%) of them depended on the first and second levels of
having played the e-learning games for 13 to 16 days (mean = hints respectively. In contrast, 12 (80%) and 8 (53%) out of
14.7 days). More important, pilot personnel described 6 of the 15 low-gains learners who were videotaped relied on the
these 7 low-gains learners as “hardworking” or “serious” first and second levels of hints respectively. It seemed that the
about learning ESL. Similarly, we were curious about how the low-gains learners, as compared to the high-gains learners,
two students who exhibited negative post-test gains, as well as were less able to rectify their errors in filling in the blanks for
the school dropout, had interacted with the cellphone-based the missing letters through the first level of hints, and required
games. the second level of hints to attain the correct spellings.
The above quantitative trends raise the following questions: Worse, the inability on the part of the low-gains learners to
How did some of the high-gains learners played the games spell correctly with help from only the first level of hints made
such that they benefited despite lower attendance? In the case some of them visibly unhappy or bored when the second level
of some low-gains learners, why did they improve little on the of hints appeared. The reason for this distress was unclear.
post-test despite spending numerous days with the games and The learner could be frustrated that he was spending too much
being perceived as diligent? In order to address such questions time to learn how to spell the word. Alternatively, on seeing
at the interaction design level, we turned to our video records the second level of hints show up, he could be demoralized
and daily reports. Our hope was to recommend improvements that he had just been relegated to the ranks of the most inferior
to the technology designs and/or after-school setting. learners and needed the second level as a “crutch” in order to
succeed.
A. Interaction Patterns with the Technology
Furthermore, some learners struggled despite both levels of
At first glance, it seemed that participants needed to attend hints. In the videos, two of them turned to their neighbors and
the after-school program for more days. Our video recordings asked for the correct letters, and/or to chat. In some cases,
showed that only 3 of the participants reached the last level in participants were embarrassed to ask their neighbors for help
the curriculum by the last session in the program. This was a again after so soon, and hence used the menu to skip to other
surprise. Given that the curriculum was designed for 18 hours levels in the games.
of instruction, we expected an average attendance rate of 10 More important, we observed that participants – especially
gameplay sessions to constitute enough time with the games. among the low-gains learners – may be able to spell the words
On examining the video recordings more closely, we saw that in the spelling activity, but were not able to spell the same
at least 8 participants were using the game menu to skip ahead words on the post-test. We offer two plausible explanations.
to other levels whenever they were unable to spell the words Firstly, some children may have learned to spell the words by
in the current level correctly after a few attempts. (We note their last session in the program, but had forgotten their
that the menu was not necessarily a negative feature. Among spellings between the last session and the post-test. Secondly,
those 8 participants, at least 2 of them used the menu to skip some participants never learned to spell the words in their
those words that they already knew how to spell.) entirety, since the spelling activity only involved filling a few
We needed to understand why learners gave up on retrying blanks and did not require the learner to spell the entire word.
the spelling activity for difficult words despite the hints in the Nowhere in the video recordings did we observe any child
spelling activity. On the whole, we observed 4 different levels struggling with usability problems.
of behavior associated with the spelling activity in the videos:
1) When students encountered a word that they could spell,
they pressed the keypad buttons quickly and with ease to X. CONCLUSION
fill in the blanks for the missing letters.
Our reactions to the results of the learning assessment were
2) When students saw a word that they did not know how to
mixed. In an underdeveloped region where rural children did
spell, some of them learned to spell it correctly with the
not have access to quality English instruction in their regular
help of the first level of hints.
school or elsewhere, we were excited to see the participants --
3) Some of those students who failed to learn how to spell a
both high-gains and low-gains learners – in the after-school
word with the first level of hint eventually learned how to
program exhibit statistically significant post-test gains that
spell it correctly with the help of the second level of hints.
could be reasonably attributed to our cellphone-based English
4) Other students never succeeded in learning how to spell
learning games. On the other hand, the learning benefits were
certain words despite both levels of hints.
uneven among participants. This could be a cause for concern.
In general, we observed that high-gains learners succeeded
To begin with, high-gains learners outperformed low-gains
in learning how to spell words after having seen their written
participants on the pre-test, qualifying test and Hindi literacy
forms displayed on earlier screens (i.e. such learners were able
test. In fact, participants’ post-test gains appeared to be highly
to spell those words correctly – without requiring any hints –
correlated with their existing levels of spelling proficiency (as
148
measured by their performance on the spelling section of the the after-school program is a model that is readily replicable,
qualifying test) and the grades in school that they are currently we encourage the reader to adopt and experiment with the
enrolled in. This observation suggests that those rural children lessons from this paper. Our results suggest that the cellphone
with a stronger academic foundation are the same children – which remains a relatively scarce resource in the developing
who are most well positioned to take advantage of the benefits world – is most effectively utilized in an after-school program
that cellphone-based learning confers. that targets more advanced children. This restriction may be a
Our results are consistent with the outcomes of a study with necessarily evil until we gain a deeper understanding of how
rural and urban low-income children in India described in He to design instructional scaffolds for less well-prepared rural
et al. [16]. This study showed that weaker students benefited children.
more from a teacher-directed pedagogical intervention, while Next, children’s tendency to seek help from their neighbors
stronger students benefited more from a self-paced, machine- can be channeled productively if the latter are taught to offer
based approach to English learning. These results should not, help appropriately (e.g. instead of only telling their neighbors
however, be interpreted to mean that we rule out technology- the correct spelling, help them to associate and remember the
augmented learning completely in the context of low-income correct spelling). Such peer coaching strategies are especially
children. Horowitz et al [15] reported a study on videos for crucial since cooperative group learning is unfamiliar to many
learning the English alphabet streamed over cellphones. In rural children, whose schools (if they attend one) are more
this study, a greater proportion of lower-income parents, vis-à- likely to implement rote learning. Alternatively, such an after-
vis their higher-income counterparts, perceived the videos to school program can hire facilitators to provide academically
have improved their children’s knowledge of the alphabet. less prepared learners with similar coaching.
In the face of the above overwhelming odds, what can we
do to promote more equitable educational opportunities in the ACKNOWLEDGMENT
developing world? One possible – and perhaps cautiously We thank Suraksha (Urvashi Sahni and Shalini Mathur) and
optimistic – interpretation of the above results is that future Sesame Workshop India for collaborating with us on this pilot
research needs to be directed at understanding how e-learning study. Lauren Bailey and Anuj Tewari provided much-needed
software can provide more scaffolding support for those rural administrative assistance, while Aman Anand and Siddhartha
children who have less academic preparation. As an example, Lal provided technical assistance. We are especially indebted
the spelling activity needs to be redesigned such that the to Mehnaaz Abidi, Jatin Chaudhary, Neelima Purwar, Gautam
learner is guided to spell the entire word eventually. With this Singh and Kavish Sinha for providing ground support. Lastly,
redesign, however, gameplay becomes prolonged and can we thank the parents and their children for their participation.
potentially increase player frustration, as we have witnessed
above. One remedy is to have e-learning games track learner REFERENCES
performance, so that the software can be adaptive in skipping [1] J. Pal, M.Lakshmanan, and K. Toyama, “’My child will be respected:’
stages that are similar to those that the player has previously Parental perspectives on computers in rural India,” in Proc. of 2nd
performed well in. Another implication for instructional IEEE/ACM International Conference on Information and
Communication Technologies and Development, Bangalore, India, Dec.
design, which calls for additional investigation, is scaffolds
2007.
such as hints that are less conspicuous, so that their appearing
[2] S. Shukla, “From pre-colonial to post-colonial: Educational transitions in
on screen does not diminish the learner’s sense of self-esteem southern Asia.” in Economic and Political Weekly, 31(22), 1996, pp.
or achievement. 1344-49.
Unfortunately, adaptive educational applications require the [3] D. Faust, and R. Nagar, “Politics of development in postcolonial India:
application state to be stored and retrieved on the same mobile English-medium education and social fracturing,” in Economic and
device. From the logistics standpoint, this requirement is more Political Weekly, India, July 28, 2001.
difficult to implement in developing regions since it is harder [4] M. P. Kishwar, “Deprivations’s real language,” in The Indian Express,
September 15, 2005. Available:
to ensure that the same learner uses the same phone – which http://www.indianexpress.com/printerFriendly/12662.html.
stores his performance from prior session(s) – across sessions. [5] Azim Premji Foundation, “The social context of elementary education in
For instance, in an after-school program where attendance rural India,” 2004. Available: http://www.azimpremjifoundation.org/
fluctuates from session to session, it would be prudent to keep downloads/TheSocialContextofElementaryEductaioninRuralIndia.pdf
a shared pool of cellphones, such that children who show up [6] J. P. Gee, What Video Games Have to Teach Us About Learning and
for the day’s session can draw from. In these circumstances in Literacy, Palgrave Macmillan, 2004.
which it is not possible to reserve a cellphone for each child, a [7] A. Banerjee, S. Cole, E. Duflo, and L. Lindon, “Remedying education:
Evidence from two randomized experiments in India,” NBER Working
wireless networked mechanism for synchronizing application
Paper No. 11904, December 2005.
state across all cellphones may be necessary. This, and other
[8] S. Mitra, J. Tooley, P. Inamdar, and P. Dixon, “Improving English
issues that we have raised above, require further investigation pronunciation: an automated instructional approach,” in Journal of
for cellphone-based literacy learning to be more effective in Information Technologies and International Development, Vol. 1, No. 1,
targeting less academically advanced rural children. Fall 2003, pp. 75-84.
As we think more widely beyond the cellphone to consider [9] M. B. Dias, G. A. Mills-Tettey, and J. Mertz, “The TechBridgeWorld
initiative: Broadening perspectives in computing technology education
it as a component in the broader learning environment, since
149
and research,” in Proc. of the international symposium on Women and

ICT: Creating Global Transformation, ACM Press, 2005.
[10] B. Kothari, A. Pandey, and A. R. Chudgar, “Reading out of the ‘idiot
box’: Same-language subtitling on television in India,” in Journal of
Information Technologies and International Development, Vol. 2, No. 1,
Fall 2004, pp. 23-44.
[11] T. H. Brown, “The role of m-learning in the future of e-learning in
Africa?,” in Proc. of 21st ICDE world conference, Jun. 2003.
[12] M. Kam, A. Agarwal, A. Kumar, S. Lal, A. Mathur, A. Tewari, and J.
Canny, “Designing e-learning games for rural children in India: A
format for balancing learning with fun,” in Proc. of ACM international
conference on Designing Interactive Systems, Cape Town, South Africa,
Feb. 2008.
[13] Y. M. Islam, Z. Rahman, S. S.Razzaq, M. A. Sayed, and S. Zaman,
“Effect of feedback during lecture style delivery both in a face-to-face
classroom and during a distance education television session in a
developing country like Bangladesh without the use of Internet,” in
Proc. of 6th IEEE international conference on Advanced Learning
Technologies, pp. 469-471, 2006.
[14] F. Librero, A. J. Ramos, A. I. Ranga, J. Triñona, and D. Lambert, “Uses
of the cell phone for education in the Philippines and Mongolia,” in
Distance Education, vol. 28, no. 2, pp. 231-244, Aug. 2007.
[15] J. E. Horowitz, L. D. Sosenko, J. L. S. Hoffman, J. Ziobrowski, A.
Tafoya, A. Haagenson, and S. Hahn, “Evaluation of the PBS Ready to
Learn cell phone study: Learning letters with Elmo,” reported prepared
by WestEd, Sep. 2006.
[16] F. He, L. L. Linden, and M. MacLeod, “How to teach English in India:
Testing the relative productivity of instruction methods within the
Pratham English language education program,” working paper, Jul. 1,
2008.
[17] U.S. Pawar, J. Pal, R. Gupta, and K. Toyama, “Multiple mice for
retention tasks in disadvantaged schools,” in Proc. of ACM conference
on Human Factors in Computing Systems, San Jose, CA, Apr. 2007.
[18] N. Moraveji, T. Kim, J. Ge, U. S. Pawar, K. Inkpen, and K. Mulcahy,
“Mischief: Supporting remote teaching in developing regions,” Systems,
pp. 753-762, Florence, Italy, Apr. 2008.
150
Kelsa+: Digital Literacy

for Low-Income Office Workers
Aishwarya Lakshmi Ratan1, Sambit Satpathy2, Lilian Zia3, Kentaro Toyama1, Sean Blagsvedt4, Udai
Singh Pawar5, Thanuja Subramaniam6
menial labor in the same geography. In India, for example, the

Abstract—Almost all formal organizations employ service staff ratio of support staff to direct employees can range from 10%
for tasks such as housekeeping, security, maintenance, and transport to as high as 60% – 30 support staff for an employee base of
at their office facility. Many of these workers earn wages in line with 100 is not unusual. Given that the corporate IT and ITES
menial-labor salaries in their respective countries. They have few on- sector alone employs more than 1.6 million professionals in
the-job opportunities to upgrade their skills or learn new ones.
urban India [1], the support staff at these facilities would
Kelsa+ is an initiative through which organizations in developing
countries can increase digital literacy and skill development among
involve a population of at least 160,000 and more likely
such low-income workers, through the provision of an Internet- 320,000 urban low-income workers. Such workers typically
connected PC for the service staff’s free, unrestricted use when off have limited or low-quality education and earn between $50
duty. and $200 per month. They have few on-the-job opportunities
We study a Kelsa+ pilot implementation in Bangalore, India, to upgrade their skills or learn new ones, and therefore remain
involving an office facility with 35 service staff. In a preliminary caught in a vicious cycle of low-income work.
exploration over 18 months, we find that at a cost that is negligible We propose Kelsa+, a program that offers Internet-
for the organization, workers’ use of the Kelsa+ PC is high and can connected PCs for free, unrestricted use to the low-income
deliver benefits both to themselves and to the office. For workers, workers in modern offices in developing countries. In our 18-
broad gains were seen in confidence, self-esteem, and basic digital month pilot, we tried several things to encourage usage and
literacy, while a few individuals experienced improvements in
improve development impact, and recorded a variety of
second-language (English) proficiency and career opportunities.
These early results point in the direction of a cost-effective ICT4D positive, if limited, results. Kelsa+ could thus be a worthwhile
initiative that could be run in the developing-country offices of the project to spread to other offices in developing countries, if
very organizations promoting development off-site. these benefits could be delivered more systematically.
In this paper, we report on our pilot experiment with Kelsa+
Index Terms—ICTD, digital literacy, service staff, low-income in Bangalore, India. We discuss our design decisions, describe
workers, urban poor usage patterns, and identify explicit development-focused
opportunities and outcomes. We also discuss some of the
potential challenges to adoption by other organizations.
I. INTRODUCTION Although PCs are likely to have been made available to
Those of us working in “information and communication service staff elsewhere, perhaps on an informal basis, to our
technologies for development” (ICTD) often run projects in knowledge, a deliberate exploration of the design and value of
remote rural areas or urban slums to work with low-income such a project has never been conducted before.
communities. Meanwhile, we often neglect a low-income
group right under our very noses: workers who clean our
offices, provide security, maintain facilities, etc. II. RELATED WORK
Most offices involve a sizeable group of service staff, who There are three threads of research that are relevant to the
take care of the housekeeping, security, transport, Kelsa+ study: the Hole-in-the-Wall experiments for children,
maintenance and so forth of the facility. In developing computer-aided learning for adults, and computer kiosk or
countries, service staff tends to be employed in large numbers, telecentre initiatives aimed at promoting development in poor
with many of them earning wages equivalent to those for communities.
The ‘Hole in the Wall’ (HitW) experiments

Manuscript received September 22, 2008. Revised February 20, 2009.
1
Aishwarya Lakshmi Ratan and Kentaro Toyama are with Microsoft
The National Institute for Information Technology (NIIT)
Research India, “Scientia”, 196/36, 2nd main road, Sadashivnagar, Bangalore – [2] ran a series of computer-based education experiments with
560080, INDIA. (Corresponding author: Aishwarya Lakshmi Ratan - phone: children from disadvantaged communities in New Delhi,
+91-80-66586000; fax: +91-80- 23614657; email: aratan@microsoft.com) India, in the late 90s. Effectively, they bore a hole in the wall
2
Sambit Satpathy is with CISCO
3 that separated NIIT from the neighboring slum settlement and
Lilian Zia is with INTEL
4
Sean Blagsvedt is with Babajob.com had an Internet-connected PC set up facing the settlement,
5
Udai Singh Pawar was with Microsoft Research India with a touchpad built into the wall for navigation. They found
6
Thanuja Subramaniam is with Microsoft, Sri Lanka that children from the slum communities in the vicinity
151
(mostly 6 to 14 years old) explored PC usage on their own opportunities, improved farming knowledge and healthier
using the device, resulting in significant learning benefits life.” [13]
without any formal instruction. HitW interventions for However, many of these projects have found it difficult to
children have been conducted at various rural and urban sites, sustain the technology and to establish clear links between
and the pattern of improved digital literacy from this ICT access and development impact. In many instances, usage
Minimally-Invasive Education (MIE) intervention has been of the facility (and therefore revenues in for-profit kiosks)
consistently recorded [3]. falls over time, irregular connectivity in remote areas causes
Critiques of the project have come from those who see the periodic interruptions in service provision (and further decline
“digital divide” not as “gaps to be overcome by providing in usage), device maintenance accrues unanticipated costs and
equipment,” but as “social-development challenges to be where high usage persists, it is often for the usage of
addressed through the effective integration of technology into applications with limited welfare impact and by community
communities, institutions, and societies” [4]. Such critics find members who are not particularly disadvantaged (making it
the HitW intervention to be “technologically deterministic” hard to justify public expenditure) [14],[15]. Many programs
and lacking in community and parental involvement. Further, face a tension between pursuing financial sustainability and
they question whether simply having physical access to an IT maximizing social outreach [16]. Those kiosks that are able to
device and learning a set of basic IT skills would translate into sustain usage often involve an exceptional kiosk operator or a
any systematic improvement in the lives of these children. committed organization that is able to effectively coordinate
Kelsa+ was initially inspired by the ‘Hole in the Wall’ between the end-users and the desired application [17].
project. It examines unrestricted PC access not for children, In contrast, Kelsa+ operates in a modern office
but for adults, specifically low-income urban workers, though environment, where exactly the elements for sustainability are
within an institutional and social context. already present – IT maintenance and support, good physical,
electrical, and connectivity infrastructure, caring staff, etc.
Computer-Aided Learning for Adults
Computer-Aided Learning (CAL) has existed in various A few kiosk projects, such as the Akshaya project in
forms for a number of decades as an educational tool in Kerala, have explicitly pursued adult digital literacy as an
schools [5], universities [6], healthcare institutions, and is objective [16], [18]. Despite the project’s impressive scale,
becoming an important area for ICTD investigation in however, field studies suggest that these programs achieved
developing country contexts [7]. Early experiments in basic neither the reach nor the depth of digital literacy that was
computer-assisted instruction, comprehensive historical sought [28]. We hypothesize that one reason for this was that
overviews of which have been done [8], has led to more digital literacy courses were conducted over very short time
sophisticated recent work around using robotics for science periods (10 sessions of 90 minutes each), which did not permit
education [9], facilitated distance learning [10], and so on. learners enough time to familiarize themselves with the
However, there have been relatively few investigations technology.
around uptake and impact in the context of CAL for adult In Kelsa+, interaction with the PC is voluntary and
learning. One study identifies the importance of CAL in continuous in a process we call “digital habituation” [19]. This
maintaining the continuity of learning history of adults, allows the incremental build-up of digital literacy skills at a
arguing that such use of computers in learning will be able to user-determined pace, and through user-determined content
shift the focus of cognitive energy to cognitive creativity, and applications.
allowing adults more space for absorption of knowledge and
creativity [11]. III. BASELINE INVESTIGATION
Kelsa+ examines a subset of questions in the domain of Low-income urban workers largely live in the city’s less
CAL for low-income urban workers in developing countries. developed residential settlements with many working as part
of the informal economy. A large share of these workers are
Computer kiosk and telecentre ICTD initiatives young and recent urban migrants. Given low levels of
Most of the adult digital inclusion projects in the education or low-quality education, they are mostly employed
developing world have been undertaken as computer in low-paying service sector jobs or work as small
telecentre initiatives [12]. These typically, though not always, entrepreneurs. Among this group, some workers find jobs at
involve a small number of PCs set up in a rural area, offering formal office facilities, often associated with slightly higher
a variety of PC-related services and run by an entrepreneur, pay, additional perks, and more prestige. In Indian cities, these
salaried operator, or a community-based organization. The opportunities have mushroomed since the early 1990s, as both
expectation in these projects is often that once ICTs such as domestic and multinational corporations expanded operations
the PC are made physically available in communities where and facilities.
they did not exist before, there will be usage by members of Unfortunately, these workers have limited opportunities to
the community and socio-economic development will follow. upgrade skills or learn new ones, especially within their
This approach is evident in the project concept notes and workplaces. The training institutes that exist, for spoken
promotional material of several telecentre initiatives: “by English, typing, or IT skills, place heavy demands on workers’
providing information about employment, better farming time and finances, as they require attendance at external
techniques and health we hoped for new sustainable job training centres. As a result, even after decades of labor,
152
workers earn only marginally better incomes than when they “Even if you are poor, if you learn computers and try and get
started. used to it - you can improve” [21].
Our pilot focused on the service staff at a single urban Yet, though considered critical at a conceptual level,
corporate facility in India’s IT capital, Bangalore. The facility functional understanding of the PC was limited and based on
employs around 55 full-time employees and 35 support staff at workers’ observations from mass media and their
any given point in time. The latter group consists of environment. One worker’s comments on the PC as a learning
housekeeping or cleaning staff, drivers for the office cabs, device aptly captures this lack of a clear functional
security guards stationed at the facility’s entry and exit points, understanding, “Students now learn Windows - how to open it
and building maintenance staff to monitor the facility’s and use it, what’s inside it, games, etc. - and I can’t say why
electricity, connectivity and other infrastructure. exactly that is useful, but it is. I know there is something in the
Prior to the pilot, we conducted detailed structured computer that is important for students” [21].
interviews with nearly all of the facility’s support staff (a A few workers who had learned to use particular
sample of 30 respondents out of a total group of 35 workers), applications meaningfully on a PC had done so through
to understand their baseline socio-economic characteristics, instruction from their peers or seniors at a past workplace.
occupational history, current job requirements, use of They described having picked up these skills through a
technology and aspirations. Each interview lasted for 60-90 combination of observation and ‘learning by doing’. As one
minutes. This was complemented with participatory respondent described, “I learned to use Outlook from my boss
observation of work routines. Of the many insights gained here. I had learned to use Excel at my previous job at a travel
from this qualitative investigation, a few are described below. agency – a lady colleague taught me there. I learned to
The average age of the support staff is 26 years, and the browse the internet after observing how my friends did it. I
average worker had ~12 years of formal schooling (class 10). learned to do personal email on my own.”
All except two are men: many have migrated to the city alone
in search of better work opportunities, even as their families
continue to live in the village. The average worker earns IV. SOLUTION: KELSA+
~$100 in income per month [20]. Depending on their job and In response to the expressed desires in our baseline
the week in question, workers’ shifts varied: security guards qualitative study, we introduced an Internet-connected PC at
and maintenance staff rotated for a week each between a 9 the workplace for the exclusive and free use of the facility’s
pm – 7 am shift, 7 am – 2 pm shift, and a 2 pm – 9 pm shift; support staff, to be used during workers’ off-duty hours. The
housekeeping staff rotated between 9 am – 4 pm and 12 pm – project was named Kelsa+, with ‘kelsa’ being the local
9 pm shifts; and drivers worked 12 hour shifts from 10 am - language (Kannada) word for ‘work’. Kelsa+ was, therefore,
10 pm and 10 pm - 10 am. Workers had changed numerous meant to signify ‘after-work’, ‘beyond-work’ and ‘improving-
jobs, despite having entered the labor force recently. One work’.
worker for instance had started out in the village doing casual The intervention involved three overlapping phases of
wage-labor work, then worked as a private tailor, moved on to activity: (1) a ‘Hole-in-the-Wall’ phase inspired by the NIIT
work at an export garments factory, then changed jobs to work studies, when we were just observing what workers did on the
at a finance company, after which he moved to corporate PC when left on their own, (2) a Learning Modules phase,
housekeeping. which involved trying different things to improve productive
In the course of their daily work at the office, workers value of the PC for the workers, and (3) a Pre-Expansion
regularly saw PCs but did not touch or interact with them, phase, when we conducted interviews with management at
except to clean them. Their interactions were exclusively with other firms to understand what issues we’d need to address to
the specific ‘tools of their trade’, involving coffee machines, make Kelsa+ work in different office locations. We describe
vacuum cleaners, fax and photocopying machines, radios (for each of these below.
the drivers), and phones. A few workers had occasional
interactions with a PC at a cyber café, or at the office for a A) Phase One: An Office ‘Hole in the Wall’
specific application, e.g., managing the Building Management The first phase involved establishing the four basic
System or the Security Camera Tracking application. Such components of the project. First, a worker-dedicated PC was
workers’ general digital literacy skills were very low though – set up and integrated with existing company structures to have
a result of their repeated restricted interaction with a single full infrastructural and institutional support. The PC had a
niche application. basic Windows XP Operating System and the Office 2007
Likely because they worked in an office full of PCs (a suite of applications installed. It had a dedicated broadband
software development and research centre), PCs were internet connection, as well as peripherals such as speakers,
dominant in the discourse of the workers’ aspirations for headphones, a printer and a webcam. User log-in on this PC
themselves and their children. On a four-point Likert scale was disabled to minimize barriers to entry, so that a worker
(‘Not At All Important’ to ‘Very Important’), all except two could begin interacting with the PC as soon as s/he sat in front
respondents rated the computer as being either Important or of it. That the maintenance of this PC would be handled by the
Very Important for their own upward mobility, and all of them company’s regular IT staff was also established.
felt this way about the importance of computers for their Second, the Kelsa+ PC was set up in the office’s
children to get ahead in life. As one respondent claimed, basement,–a space that housed the maintenance office, the
153
workers’ changing rooms, the office cars, etc, and so was one office location, to multiple sustained deployments across
shared by all the service staff. The fact that it was placed in a various firms. For this, the project was consolidated as a
location that was primarily “owned” by the workers and in ‘Corporate Social Responsibility’ initiative, and proposed to a
which they felt comfortable was deliberate. Had the PC been number of major corporations in Bangalore. Senior
placed in the cafeteria or in the lobby of the building, workers management at these firms were interviewed in relation to
might have used the facility less, in deference to the office’s their own service staff populations. This data was used to
full-time staff. In some ways, this placement choice of using a compile a variety of solutions to the expressed concerns, so
space that the target group felt comfortable accessing was that the project could possibly fit into other institutional
probably just as important as the ‘Hole in the Wall’ project's contexts.
decision to place their kiosks near schools and outside of
buildings. As a worker later affirmed, “When the computers
are inside, it is difficult for us to feel comfortable touching
V. STUDY METHODOLOGY
them. Now, this is like a computer for us, outside [in the
basement]” [20]. We employed a mixture of research methods in studying
Third, in a set of initiation activities, the PC was Kelsa+ through its three phases of activity, which we describe
introduced to the workers as a facility that they could use for below.
any purpose whenever they were off duty. Workers were
informed that all activities on and around the PC would be Collective usage: quantitative metrics
recorded, both for research and security purposes, specifying The Kelsa+ PC had a logging application installed, which
that none of this material would ever be used to restrict usage. tracked all events initiated on the PC [22]. This included
Signage in the local language and English indicating this launches of all applications, as well as URLs visited. The
logging was also placed at the Kelsa+ PC location. For any logging tool also allowed an examination of how much time
questions or concerns, two of the authors were pointed out to users had spent on the Kelsa+ PC on any given day. These
the workers as the people to contact. three metrics (collective time spent using the machine,
Finally, the determination of who would use the PC when applications used, websites visited) were recorded over the
and how was left entirely to informal processes of negotiation 18-month period (Jan 11, 2007 – June 2 2008), with the
between workers. No schedules were drawn up. No analysis conducted on consecutive fortnights of activity.
instructions were given. That usage and learning was entirely Application and internet usage was measured using number of
voluntary and self-paced was established. No restrictions were launches as well as active time spent on each
laid on how many users could simultaneously work on the PC. application/website. There were two breaks in logging, one
The intent was to provide a wide open area, conducive to when the PC’s connectivity was interrupted in mid-November
groups of peers interacting with the PC together. 2007 and the other in late-March 2008 when the PC’s OS was
re-installed, both in response to virus attacks.
B) Phase Two: Learning Modules The categorization of the log data into meaningful groups,
After a year of allowing unrestricted usage of the PC, we including application categories and website categories, was
collected feedback from the workers in a series of focus group performed based on the researchers’ examination of keywords
discussions. Strong interest was displayed in achieving that tagged the data optimally despite the heterogeneity of
particular kinds of learning using the PC, including learning usage. These are available on request from the authors.
English, office productivity software, and accounting software
(Tally). To further this goal, we started by introducing English Collective usage: qualitative measures
as a Second Language (ESL) content, both as CDs and The activities on the Kelsa+ PC were also recorded using a
compiled weblinks, to the workers. No formal instruction was screenshot logger, with a screenshot of the PC taken every
given or classes were taken. The voluntary and self-paced minute. In addition, the activities around the PC were
learning nature of phase one continued. We only acted as recorded using a motion-detecting web camera. These
initiators of introducing new content at the Kelsa+ PC towards screenshot and video logs were processed manually, and
a specific productive goal, and not as evaluators or routine hence, selectively, to answer specific questions around
instructors. usability, group usage/sharing dynamics and number of
distinct users. For the latter, 14 days of video logs were
C) Phase Three: Management Discussions randomly picked from the 18-month study period to estimate
Kelsa+ is housed within an institutional context, and we the average number of distinct users per day.
made an effort to understand the effect that the intervention
was having on its social and institutional environment, to Individual impact: quantitative metrics of change
avoid some of the limitations that the Hole-in-the-Wall study Given the high job turnover rate among workers in this
had faced. As we continued with other explorations on segment (described in the baseline study’s results),
Kelsa+, we collected feedback from management to longitudinal tracking, i.e. measurements for the same
understand their reactions to the Kelsa+ facility for workers. individual over time, was very challenging. However, for a
limited subset of workers, we obtained before and after results
We also began a set of investigations to understand what it
for particular behavioral or proficiency tests and these cases
would take to expand the project from a small-scale pilot at
154
were used to understand possible patterns of change. These To assess the relevance of Kelsa+ in other institutional
included the following tests: contexts, over 15 corporations were contacted from a range of
sectors (biotechnology, energy, IT) with the majority being IT
• A Self-Esteem questionnaire: A local language- firms. They were first briefed about the Kelsa+ concept.
translation of the Rosenberg Self-Esteem questionnaire [23] Structured interviews were then conducted with them in which
was administered to a sub-set of workers prior to the Kelsa+ they shared details on the size and composition of service staff
intervention (17 workers), and then five weeks after it was at their facilities, their evaluation of the possible benefits
introduced (27 workers). Eleven respondents overlapped from, as well as concerns around, implementing Kelsa+ as a
between the two administrations (0.65 of pre-test group). The CSR initiative for the service staff at their facility.
differential in scores obtained for these workers was
calculated.
VI. STUDY RESULTS
• A Digital Literacy test: A brief digital literacy test
covering usage of basic applications on the PC was Given the multiple research methods that were used, in this
administered prior to the introduction of the Kelsa+ PC (30 section we present our results in correspondence to thehigh-
workers). The time taken to complete a task, as well as the level categories described in the methods section: collective
number of prompts needed to complete the task, were usage,individual impact and management feedback.
recorded as metrics of proficiency. The same test was
administered to three workers, 18 months after the VI.I Collective Usage
intervention, of whom two had taken the pre-test as well. The
differential in scores obtained for these workers was Time used
calculated. Normal staff turnover at the rate of ~2 workers per After the initial fortnights, the Kelsa+ PC saw high and
month had disallowed a pre- and post-test comparison for sustained adoption by the support staff (Figure 1), recording
more workers over this extended time period. average collective usage of 10.13 hours per day. Usage
peaked at 17.83 hours per day in fortnight 29. Further, usage
• An English proficiency test: An adapted version of was not restricted to a small minority of persistent users, but
Cambridge University’s ‘Key English Test’ for Beginners was spread across a broad base of workers. From the video
[24] was administered to a sub-set of workers first in March log analysis, we found an average of 13 workers to be primary
2008 (20 workers), prior to the introduction of the ESL digital users of the PC each day, with all staff workgroups
content. A second test in the same format was administered in represented in the user base. 40 unique workers were
August 2008 (17 workers), after over three months of the ESL observed using the Kelsa+ PC at least once (despite a staff
material having been available for workers’ usage. The size of only 35 at any given point in time – this occurred due
questions focused on testing basic English vocabulary, to staff turnover) during the sampled period. A quarter of the
grammar, and reading comprehension. Seven respondents workers were identified as high-frequency users, seen using
overlapped between the two administrations (0.35 share of the PC on half or more of the observed days [19].
pre-test group). The differential in scores obtained for these
workers was calculated. The post-test also asked a few
questions on workers’ Kelsa+ usage or non-usage over the
week preceding the test.
Individual impact: qualitative narratives of change

A subset of the staff from across workgroups (6 workers,
~0.17 share) was interviewed in detail 18 months after the
intervention was launched. These workers discussed their
background, prior exposure to a PC or not, their usage of the
Kelsa+ facility, and their present job, following a structured
interview protocol. Their perceptions of change since the
introduction of the Kelsa+ PC were also recorded. The results
from this analysis describe socio-economic mobility pathways
associated with usage of the Kelsa+ PC.
Management interviews
We conducted structured interviews with five facility Figure 1: Hours of active usage of the Kelsa+ PC per day
managers at the pilot location (across transport, security, (average over each fortnight)
housekeeping, etc.) and recorded their observations around the
Using a different measurement instrument, the responses
Kelsa+ intervention including: the usage patterns of workers,
to the questions on Kelsa+ usage or non-usage included in the
any effects on worker productivity (positive or negative), any
second English proficiency test revealed that all but two
other changes in service staff behavior or work since PC workers in the sample (0.88 share) were active users of the
access was given, etc.
155
Kelsa+ PC. They reported using it on an average of 4 days in Group usage of the Kelsa+ PC was very popular; in fact,
the week preceding the test, and for an average of 60 minutes some users actively sought out colleagues with whom to use
per sitting. the PC. Peer learning sometimes took the form of an informal
teacher-student relationship for a particular task or
application. For instance, as [19] describe, “many workers had
Usage dynamics gone through the email account creation process with a peer
The applications and content accessed on the Kelsa+ PC who was an existing email user. During the registration
changed over the study period. Figures 6 and 7 (on pp. 13) process, in the space where a second email address was
showcase the changing distribution of application and internet required (usually of the person creating the account), the
usage over 31 fortnights. experienced email user friend would enter his email address,
Initial application usage was spread evenly across the since for most workers, this was their first email account.” For
gamut of available software on the PC, indicating heavy other workers, group usage involved a symbiotic learning
exploratory usage. However, starting in the third fortnight, relationship: “I use it with a friend generally, so that we can
Internet Explorer becomes the dominant application launched, learn from one another. What he doesn’t know I tell him, and
followed closely by multimedia applications. There is a shift vice versa.”
away from basic interactive applications that are very popular
initially (Microsoft Paint and offline games, whose usage fell
from 15% to 0), to increased usage of the Internet with its
dynamic content and more sophisticated applications.
As Figure 7 shows, workers spent many months primarily
using the internet for entertainment – viewing music videos or
films on YouTube or Stage6. However, we see a rise in using
the Internet for communication, through email and social
networking sites like Orkut, eight months after the project was
started (correlated with 6 workers simultaneously creating
email accounts at this time).
There is a strong motivation to use the computers for self-
expression, as Figure 3 shows. The ability to create a personal
digital presence, both through images and later through email,
was a source of great pride among workers, and seemingly
altered the way they perceived themselves. Figure 3: Workers creating personalized desktop
backgrounds using the Kelsa+ PC’s webcam
Learning
Workers followed individual and group learning paths. Usability preferences
Several workers simply observed their peers using the PC for Given their educational and linguistic background, workers’
weeks before attempting to touch the PC themselves. As one revealed a number of adaptations in input techniques to
worker commented, “For the first one and half months, I just achieve functional use of the PC. Navigation using a mouse
watched how other people used the computer.” We asked why dominates since typing is problematic. Browser history is
he did this without trying to use the PC himself, to which he repeatedly used as an easy way to access content of choice,
responded that “what if something happened when I used it?” indicating implicit peer learning [Figure 4]. Search queries
[19] Observation and individual exploration established (with the “.com” tag) are repeatedly used to access online
‘learning by doing’ routines, which were instrumental to content within and across sites, given that it involves only
meaningful PC literacy gains. memorizing and typing in one keyword or a short phrase, as
opposed to an entire url. Email forwarding is the dominant
method of online communication, with one user forwarding
up to 15 emails per day to 10 or so people (mostly images).
Figure 2: Combination of workers observing and being

primary users themselves
156
Kelsa+ intervention and 18 months after it was initiated.

Churn in workers and the small sample size, curbed such a
pre- and post-test for a larger share of workers. Such testing
would be essential to establishing average impact across
workers. However, the results for these two respondents are
encouraging, given their low educational attainment and given
that in these cases we can clearly trace the digital literacy skill
gains to workers’ Kelsa+ PC usage alone.
Both workers had experienced gains in a core set of PC
literacy skills (see Table 1). While they were unable to
perform any of the tasks in the pre-test (save for turning on
the PC), both were now able to open a web browser window,
launch a search application, navigate through the results, and
open and close a document-processing application.
Figure 4: Browser history is repeatedly used as a way to Similar to the Hole-in-the-Wall gains for children, these
access content without typing out URLs measured improvements in digital literacy for two such
demographically and occupationally distinct workers, from
entirely voluntary and informal learning processes, are
VI.II Individual Impact promising. Neither of these workers had interacted with a PC
outside of the office during this time. Neither of them had
A) Self-esteem and confidence gains been schooled in English. One of them had left school after
class 7, the other after class 4. This only goes to show how
The 11 workers who took the self-esteem test just before despite the average educational attainment among the pilot
the Kelsa+ intervention and five weeks after its installation, location’s service staff group being higher at Class 10, even
recorded a 5% jump in their self-esteem score. This change those who had much lower levels of education in a local
clearly cannot be attributed just to the PC installation without language saw gains in basic digital literacy from Kelsa+
a control group. However, qualitative interview responses usage.
from workers appeared to reiterate the effect that the Kelsa+
PC had had on individuals’ sense of hope, confidence and Table 1: PC Literacy Test Results
self-esteem. One driver commented, “I see some changes in A R
my life – I couldn’t speak English before, now I can speak a Occupation/ Age Driver/ 55 Housekeeping/ 23
little… I haven’t yet changed my job or used English outside, Education/ Class 4/ Tamil Class 7 / Kannada
but I now have some hope to learn… I have that courage.” Medium of
Further, use of the Kelsa+ PC was seen as a first step towards instruction
further possibilities: “To do any computer course outside, I Annual Income $1600 $1000
think I should know some basics.... having picked up some (US$)
basics today I have the confidence that when I take up a Previous PC Never touched a Had touched a PC
course... I will be able to catch up ... and I am very inspired Exposure PC before, but never
by using the PC here to know more and learn.” used one
Given that the Kelsa+ PC was in many cases the first time a Reported usage 30-45 min each 30 min every day
worker had touched a PC, it elicited responses such as: “I felt frequency session; 3-4
so happy that day when we had the interview. For the first sessions a week
time I touched a computer and did so many things without a Key applications Local-language Local-language
mistake….. I don’t have an email account. So now maybe I used news portals, music and films,
can make one on this computer.” [20] The continuous access games, local- email, games
to a PC at a location that workers visited everyday lowered the language music
barrier to usage considerably, “Since, the computer is here, we and films
get awareness! Also because we can see the computer daily PC literacy scores Pre- Post-18 Pre- Post-18
....my desire to learn and use it has increased very much.” mths mths
And all this of course translated into changes in how workers Turn on the PC ? 9 9 9
viewed their workplace: “In all my service, this is one of the Play Windows X 9 X 9
best workplaces I have seen.” (offline) Games
Open Internet X 9 X 9
B) Digital literacy gains Browser and go to
a search engine
Two workers, a housekeeping worker and a driver, with Enter a search X ? X 9
minimal prior PC exposure and varied demographic query
characteristics, were administered PC literacy tests before the Open the best X 9 X 9
157
search result and high motivation levels, and their experiences are anecdotal. At
go back to results best, their outcomes indicate the extent of welfare impact that
list is possible when worker initiative and circumstance interact
Sign-in to E-mail X X X 9 with access to the Kelsa+ facility to produce welfare-
client improving outcomes.
Start MS-Word X 9 X ?
application Upward mobility within the same office
Type in MS-Word X X X X A housekeeping worker with a Class 7, local-language,
Save document X X X 9 rural government school education, and no prior exposure to a
Print document X X X 9 PC, began using the Kelsa+ PC for 60-90 minutes each day
Close Word X 9 X 9 after his work shift. He used it for games, internet video,
application music, etc. A few months later, he created an email ID with
Shut down PC X 9 X 9 the help of a driver colleague, began using the Learn-English
9 : Successfully completed the task CDs available at the Kelsa+ PC, and various Microsoft Office
? : Partially fulfilled the task/ completed the task with applications. He was given an initial typing job on MS Excel
extensive prompting by the IT staff in the office, which he completed and
X : Does not know/ did not attempt submitted satisfactorily. He was then given additional
Source: [19] inventory data entry jobs (on Excel), before being hired as a
dedicated worker for the IT staff.
C) Improvements in English proficiency English skill acquisition

A driver with no prior exposure to a PC and a Kannada-
Over a 3-month period of Kelsa+ usage, the English medium government school Class 10 education, took an
proficiency scores for the seven workers who took both pre- interest in using the PC for music, movies, as well as learning
and post-tests improved from 32% to 41%. Looking at their English, spending 2-8 hours of off-duty time per workshift on
individual trajectories, we find that of the seven, four saw the Kelsa+ PC. He specifically requested an extension to stay
increases in their English proficiency while the other three in this job for a few more months (a number of drivers left the
saw declines. Yet, the increments upward (average movement facility due to inconvenient shift changes at this time; contract
of +19.4%, p=0.04) appear to be significant while the workers are often moved from one location to another), since
movements downward are not distinguishable from he had begun using the Learn-English CDs at the Kelsa+ PC
measurement error of the same proficiency level (average and wanted to acquire a certain level of competence before
movement of -3.8%, p=0.13). This possibly indicates an having to leave.
interaction between individual worker motivation or initiative
and the availability of the PC-based learning material, to Shift to a new career path
produce differential skill gain outcomes. A key issue is, A security guard with a Class 12, local-language, rural
therefore, the need to understand motivational tools that might government school education, sat at the Kelsa+ PC for 60-120
make available skill-development tools more widely used minutes before or after a shift. He used the Kelsa+ PC to
across all workers. practice typing as he underwent a data-entry training course
outside the office. The practice afforded by access to the PC at
the workplace was key, in his opinion, to his successful job
Table 2: English Proficiency Test Results interview for a data-entry job. He has since moved from his
Overall Test 1 Test 2 security guard job to a data entry position at another firm. His
pride in his new job was reflected in his comments, “Today I
Average score (%) 45% 53% +8%
can stand up in front of my father and friends and say that I
Sample size (n) 20 17
am no more a watchman, but I am doing a computer job."
Pre- and post-test cases (7 workers)
Average score (%) 31.7% 41.2% +9.5% VI.III Management feedback
(p=0.14)
A) Managers at pilot location
The response from the management at the pilot location was
D) Socio-economic mobility pathways strongly positive. They felt that such exposure to technology
made their workers confident and knowledgeable, while also
While the measured estimates may indicate gains in being a source of recreation. Moreover, they commented that
particular skills for select individuals, the true impact of such providing a PC access facility for the service staff at the
facilities would need to manifest as improved socio-economic workplace had a special connotation for workers: “sitting in a
outcomes for workers. Three cases illustrate the variety of cyber [café] and learning and sitting in their office and
ways in which these skill gains have translated into individual learning is something different.”
workers’ ‘development’. These workers displayed particularly
158
Certain improvements in the workers' ability to location for the PCs due to safety measures in some
communicate in English were described. In addition, some companies, risk of downloading or sending malicious or
workers had demonstrated initiative in integrating PC usage offensive content from the PCs, legal concerns regarding
into their own everyday workflows. A worker in the company liability, the risk of computers being used for
housekeeping division, for instance, had begun to type up the entertainment purposes only, and the associated effort for
weekly stock order list for the office pantries on MS Excel regulation and supervision of usage being too huge for
and was emailing these to his supervisor, instead of writing companies to undertake. [25] The fact that most of these
them out by hand (Figure 5). workers are contracted and are not direct employees of these
companies also contributed to reluctance by some
corporations: “These workers are not our employees so why
should we educate them. Their employers may not like it.”
In general cost was not perceived to be an obstacle in
proceeding with Kelsa+ amongst most of the companies
interviewed. Connectivity charges were considered the biggest
cost component, but procuring PCs for the initiative was not
perceived to be an issue.
VII DISCUSSION
In proposing Kelsa+ as a potentially effective ICT4D
intervention for low-income office service staff, we return to
the related work in this field and discuss this intervention’s
comparative strengths and limitations.
(1) Extending ‘Hole-in-the-Wall’ possibilities
Similar to the HitW intervention, Kelsa+ highlights how

acquiring basic digital literacy skills has little to do with
socio-economic background or educational quality, but much
more to do with immersion, practice, and learning by doing.
Persistent and minimally-invasive access to the PC for
workers’ continuous exploration in the office environment
through Kelsa+ has allowed for the following primary
benefits:
(a) Improvements in basic digital literacy, including the

ability to turn a PC on, operate an input device
effectively, identify various applications, and
Figure 5 a and b: Housekeeping workers maintaining task navigate through the web to preferred content.
lists on MS Excel at the Kelsa+ PC (even as they multitask (b) Improvements in English proficiency, driven not
with playing games) only by the use of dedicated ESL content, but also by
the workers’ repeated interaction with English
B) Managers at potential implementation locations language content on the web, and sporadic use of
email and office productivity tools.
The feedback from the fifteen corporations contacted as (c) Improvements in hope, confidence and self-esteem.
potential implementers of Kelsa+ facilities fell into two Kelsa+’s achievement lies in its demonstration of
camps. One set of corporations saw Kelsa+ as a tool for their how easy PC usage can be, and its encouragement to
workers to learn English, which they considered a major workers to believe that they can learn more as they
factor in increasing workers’ confidence and employability. work and at no additional cost. Should this hope
The intervention would bring more equity and awareness to translate systematically into real outcomes
their service staff, and change the way they viewed manifested in test scores, or new jobs, workers will
technology and the world. As one manager commented, be further inspired to experiment and learn, thereby
Kelsa+ is "not just a little feel good project, but something establishing a virtuous feedback loop.
that can have an impact if done right.” [25]
On the flip side, the managers at a number of other In a departure from the HitW studies, Kelsa+ also explores
corporate facilities were very concerned about the potential the effects of the intervention on the institutional context
risks of this intervention. Security of both physical equipment that houses such an intervention. The pilot office facility
and content was perceived as the number one concern. Other saw greater staff morale and attachment to the workplace
perceived risks included difficulty in finding a suitable from having such a recreation and learning facility at the
159
office. Worker initiative to incorporate the Kelsa+ PC into

their work routines was welcomed. Managers responded
positively to improvements in worker confidence and (3) Low marginal cost for well-endowed providers
knowledge, and to enhanced worker capacity through
improved digital literacy and communication skills. In the PC kiosk franchise arrangement, there is often an
At the same time, it is unclear whether these dynamics of entire livelihood, that of the kiosk operator, depending on
security, productivity, and open exchange seen in the pilot the success or failure of the PC kiosk business. In contrast,
will repeat themselves in other Kelsa+ locations. For one of the major advantages of Kelsa+ is that it is an
instance, restrictions in the timings that workers can stay at intervention that costs marginally nothing to those financing
the office facility, before or after their workshifts, will and running it, i.e. large private corporations. The providers
indeed influence the extent to which they can experiment in this case already manage large IT budgets and
with the PC and gain from access. Resolving such issues maintaining an additional set of PCs has a negligible cost.
will involve political bargaining between stakeholders, a The Kelsa+ intervention involves a set of basic
process inherent to the effective design and implementation components, with the major costs being the upfront capital
of any sustainable ICTD intervention [26]. expense for the devices and software (operating system,
internet browser, office productivity suite, anti-virus
(2) Slow and sustained wins the race application). Additionally, there are monthly connectivity
charges.
What makes the Kelsa+ installation any different from other There exist several procurement and connectivity options
telecentre/ kiosk or digital literacy initiatives in developing that determine the final cost of a given Kelsa+ facility. For a
countries? There are three key aspects of the Kelsa+ system’s Kelsa+ facility with 40 workers, with capital costs amortized
design that distinguish the user experience it offers low- over a 3 year period, the monthly cost can range from $3.4
income service workers from other similar ICTD initiatives: per worker (when a new PC is purchased, Microsoft Office
is installed, and a 512 Kbps Unlimited Data Transfer
(a) Availability and support: The office environment connection is used) to $1.3 per worker (when a refurbished
provides 24x7 availability and all the necessary PC with Open Office is used, over a 256 Kbps Unlimited
infrastructural, technical, social and financial support Data Transfer connection) [25].
required to sustain the health of the PC for the workers.
Skilled maintenance is always at hand, more experienced (4) Development through ‘agency’ enhancement
users are all around to ask for advice, and the institution is
already an integral part of workers’ daily routines. As a A key aspect of the Kelsa+ intervention is the freedom that
result, the barrier for workers to begin experimenting with accompanies exploration on the PC. The opportunity cost for
the PC is considerably lowered. workers to invest in their own learning is minimized through
free access and convenient placement of the PC within their
(b) Digital habituation: Interaction with the Kelsa+ PC has workplace. More importantly, by refraining from any
no fee and is not limited in any sense. This allows workers restrictions on workers’ access to particular applications
to respond to the PC spontaneously and evolve behaviors versus others, the agency of the worker is respected and
incrementally as they build skill. Workers engage in a slow encouraged in determining particular usages of the PC that
process of familiarization with the technology as they learn meet his/her need [27].
basic navigation techniques and understand what the Kelsa+ encourages a peer-learning model, through which
technology can be used to achieve (or not) and how; a workers with heterogeneous exposures and skills are allowed
process [19] term ‘digital habituation’. Such habituation an opportunity to share their knowledge at a location where
“constitutes a critical intermediate step between providing they congregate organically, i.e., the work place. Learning is
PC access to a disadvantaged community and achieving driven by the user’s demands, and when not met by peers,
sustained development impact.” [19] workers are able to consult employees with more knowledge
or step up and draw from training courses offered externally.
(c) Learning by doing / learning through peers: An In all this, the worker is centre-stage and his/her decisions
advantage offered by the Kelsa+ PC and usually not dictate usage and impact, which differs from more
available at regular telecentres, formal IT training institutes, paternalistic interventions in the ICTD space where certain
or libraries, is the ability to learn passively and informally ‘developmental’ results are expected and deviations from
from peers, in addition to learning actively through doing. those results are treated as shortcomings.
As a security guard who had done two years of computer
courses at a government college noted: “They used to teach
us basics...but I didn’t pick up much from the class… I VIII LIMITATIONS AND ONGOING WORK
generally learnt things on my own after coming here...once I Though based on a simple premise, Kelsa+ is a powerful
see some people using [certain applications or features] idea because it takes the ICTD discussion and integrates it
....and then next time I generally go about following the with the daily workings of mainstream institutions in
same.”
developing countries. Many populations in need operate in or
160
close to existing IT infrastructures. As Kelsa+ shows, the ACKNOWLEDGMENT

innovations needed to have technology be useful to such To the workers involved in this project, we owe our most
groups are, therefore, mostly of political will, design, and sincere thanks, for allowing us to examine and learn from
process. their experience. Many thanks to the facilities management
and IT staff at the pilot location for their support. Thanks to
There are many shortcomings in the current study. The Itamar Kimchi for his help with refining the log analysis
evidence presented is based on a handful of cases, and such results.
anecdotal positive effects cannot be mistaken for large-scale
systematic impact. It may well be that the workers whose
results we have recorded are exceptionally motivated and do REFERENCES
not form the norm in this segment. The facility where the pilot [1] NASSCOM, ‘Indian IT Sector Score 10-in-10,” February 2007, URL:
was implemented was in fact a small software development http://www.nasscom.in/Nasscom/templates/NormalPage.aspx?id=50833
and research centre, which may have presented many unique [2] NIIT: www.niit.com
conducive factors for Kelsa+ to take off, including caring [3] R. Dangwal, S. Jha, S. Chatterjee, and S. Mitra, “A Model of How
Children Acquire Computing Skills from Hole-in-the-Wall Computers in
staff, an environment of trust, ubiquitous IT presence, strong Public Places,” Information Technologies and International
IT maintenance support, 24x7 access, and open policies Development, 2(4), 2005, pp. 41–60.
regarding employee PC usage. The interview responses from [4] M. Warschauer, Technology and social inclusion: Rethinking the digital
divide. Cambridge, MA: MIT Press, 2003.
the participants may also have been influenced by the peer [5] L. Linden, “Complement or Substitute? The Effect of Technology on
relationship they share with the researchers, which would Student Achievement in India,” unpublished manuscript, Columbia
prevent the sharing of negative feedback. university, June 2008.
URL:http://www.columbia.edu/~ll2240/Gyan_Shala_CAL_2008-06-
03.pdf . Last accessed September 22, 2008.
It is clear that objectively verifying the value of Kelsa+ as a [6] A. Lane, and M. Porch. “Computer Aided Learning and its impact on the
general ICTD intervention involves exploring it in a variety of performance of non-specialist accounting undergraduates,” Accounting
Education, 11 (3), 2002, pp. 217-233.
institutional contexts. This is our ongoing effort. In extending [7] The One Laptop Per Child (OLPC) initiative: www.laptop.org ;
the scope of Kelsa+ to become a systematic tool for socio- Multipoint for education:
economic mobility among urban low-income office service http://www.microsoft.com/unlimitedpotential/programs/multipoint.mspx
staff, our current efforts involve emphasizing longitudinal [8] T. Oppenheimer, The Flickering Mind: Saving education from the false
promise of technology. Random House Publishers, 2003.
tracking of workers to map welfare impact, deploying Kelsa+ [9] A. Mukherjee, “Build Robots Create Science – A Constructivist
in non-IT corporate facilities, introducing some structured Education Initiative for Indian Schools”, Proceedings of Development by
learning components to test their effect on learning outcomes Design (dyd 02), Bangalore, India, 2002.
[10] R. Wang, et al. “Distance Learning Technologies for Basic Education in
and worker welfare, introducing certification options to prove Disadvantaged Areas” Proceedings of the 8th Global Chinese
skill acquisition externally during job-search, offering Conference on Computers in Education (GCCCE 04), 2004.
supplementary income generation possibilities for simple [11] W. Schinagl, “New learning of adults in information and knowledge
society”, Journal of Universal Computer Science, 7 (7), 2001, pp. 623-
mini-tasks completed at the Kelsa+ PC, understanding the 628.
specific constraints that prevent women workers from using URL:http://www.jucs.org/jucs_7_7/new_learning_of_adults/Schinagl_
the Kelsa+ facility, and testing motivational tools that might W.pdf
[12] R. Heeks, “ICT4D 2.0: The Next Phase of Applying ICT for
influence the usage of the PC for explicit skill-building across
International Development,” IEEE Computer magazine, June 2008, pp.
domains and workers. 26-33.
[13] S. Bailur, “The Complexities of Community Participation in Rural
This study draws the ICTD community’s attention to a sub- Information Systems Projects: The Case of ‘Our Voices’”, Proceedings
of the 9th International Conference on Social Implications of Computers
set of the poor who spend the bulk of their time around in Developing Countries, São Paulo, Brazil, May 2007.
sophisticated IT infrastructures, but so far, do not gain from [14] K. Kiri, and D. Menon, "For Profit Rural Kiosks in India: Achievements
such proximity. We have described a pilot implementation of and Challenges," i4d magazine, June 2006. URL:
http://www.i4donline.net/articles/current-
the Kelsa+ project in an office facility with 35 service staff, article.asp?articleid=700&typ=Features
among whom the average worker earns $100 a month and has [15] R. Veeraraghavan, N. Yasodhar, and K. Toyama, “Warana Unwired:
studied till class 10 in the local language. Over a period of 18 Replacing PCs with Mobile Phones in a Rural Sugarcane Cooperative,”
Proceedings of the 2nd IEEE/ACM International Conference on
months, we saw broad improvements in workers’ confidence, Information and Communication Technologies and Development,
self-esteem, and basic digital literacy, while a few individuals December 15-16, 2007, Bangalore, India, pp. 89-98.
experienced increases in second-language (English) [16] R. Kuriyan, I. Ray, and K. Toyama, "Integrating Social Development
and Financial Sustainability: The Challenges of Rural Kiosks in Kerala,"
proficiency and career opportunities. Verifying these results
Proceedings of the 1st IEEE/ACM International Conference on
through larger-scale and wider-scoped Kelsa+ Information and Communication Technologies and Development, May
implementations and measurements is now necessary to build 2006, Berkeley, USA.
on the promise of this simple, cost-effective, yet powerful [17] N. Rangaswamy, “Social Entrepreneurship as Critical Agency: A study
of Rural Internet kiosks,” Proceedings of the 1st IEEE/ACM
ICTD intervention. International Conference on Information and Communication
Technologies and Development, May 2006, Berkeley, USA.
161
[18] J. Pal, “Examining e-literacy Using Telecenters as Public Spending: The [23] The Rosenberg Self-Esteem Index:
Case of Akshaya,” Proceedings of the 2nd IEEE/ACM International http://chipts.ucla.edu/assessment/Assessment_Instruments/Assessment_f
Conference on Information and Communication Technologies and iles_new/assess_rse.htm
Development, December 15-16, 2007, Bangalore, India, pp. 59-67. [24] Key English Test (KET) certification preparation, Cambridge
[19] A. L. Ratan, and S. Satpathy, “Digital Habituation as a Basis for Digital University. URL: http://www.cambridgeesol.org/exams/general-
Inclusion,” Microsoft Research Technical Report, June 2008. english/ket.html
[20] U.S. Pawar, A.L. Ratan and S. Blagsvedt, “An ‘Office Hole-in-the-Wall’ [25] L. Zia, “Internship report: Digital Literacy for Low-Income Workers”,
Exploration,” Microsoft Research Technical Report, January 2008. unpublished, Cambridge University, 2008.
[21] A.L. Ratan, “Lessons from Low-income Workers in Bangalore on the [26] R. De, and A.L. Ratan, “'Whose Gain is it Anyway?' Structurational
Value of Information Technology.” Paper presented at the Conference perspectives on deploying ICTs for development in India's microfinance
on Living the Information Society: The Impact of ICT on People, Work, sector,” Information Technology for Development, forthcoming.
and Communities in Asia, Manila, Philippines, April 23-24, 2007. [27] A. L. Ratan, and S. Bailur, “Welfare, Agency and ‘ICT for
[22] R. Veeraraghavan, G. Singh, K. Toyama, and D. Menon. “Kiosk Usage Development’,” Proceedings of the 2nd IEEE/ACM International
Measurement using a Software Logging Tool,” Proceedings of the 1st Conference on Information and Communication Technologies and
IEEE/ACM International Conference on Information and Development, December 15-16, 2007, Bangalore, India, pp. 119-130.
Communication Technologies and Development, May 2006, Berkeley,
USA, pp. 317-324.
Figure 6: Distribution of applications used at the Kelsa+ PC over 18 months.

Dominance of internet usage can be clearly seen. Decline in use of applications like offline games also seen.
162
Figure 7: Distribution of internet usage at the Kelsa+ PC per fortnight over 18 months.
Increase in Internet video usage is clear. Steady and high usage of email is seen after fortnight 15. Social networking
websites see increased usage starting in fortnight 21.
163
Mapping the Dynamics of Social Enterprises

and ICTD in Cambodia
Kelly Hutchinson, Alemayehu Molla
opportunities. For these groups, social enterprises can be an

Abstract— As Information and Communication Technology important source of jobs, income, training, business
(ICT) and social enterprises become drivers of economic growth, mentoring, and technical input [2], [3], [35].
the nexus provides opportunities for new models of business to As ICTs and social enterprises become drivers of economic
bring benefits to communities in developing countries.
Recognising the complex dynamics and range of actors in this
growth, it is important to investigate this nexus to see what
diverse and emerging sector, this study chooses to document the opportunities these new models of business can bring to
external influences, use and impact of ICT on social enterprises communities in developing countries. The study takes e-
(SEs) -enterprises that have both a business and social business as a typical instance of ICTs and argues that social
development goal. The problem investigated is the potential gap enterprises are well positioned to benefit from the application
between the rhetoric of the ‘promise of e-business’ versus the and use of ICTs and to deliver the promise of e-business
dynamics of enactment and impacts of ICT in practice in the
benefits for development.
social enterprise sector. The main contribution of the research is
to identify the real development impact of ICT use by social The paper explores how social enterprises are using e-
enterprises by assessing the role of ICTs in achieving their goals. business for their comparative advantage, and the outcomes
Its major finding is that SEs’ social and business missions are (benefits) of e-business to social enterprises and the impact
inextricably linked .and their use of ICTs and benefits from it (attributable to e-business) of social enterprises in the target
can be used as a proxy to assess ICTD. Thus SEs provide a new community. The study is situated in Cambodia where there are
paradigm in the study of ICTD.
many social enterprises active in IT services, handicraft,
hospitality and processing/production., Cambodia, therefore
Index Terms—ICTD, ICT, social enterprise, international provides a space where a vibrant social entrepreneurial spirit
development, community informatics, Cambodia, CICs. presented a unique perspective on the perception, use and
benefit of ICT in an international development context.
Through mapping organisations in the Cambodian social
I. INTRODUCTION
enterprise sector, this paper aims to fill the knowledge gap,
T he usefulness of ICTs to achieve development goals are

increasingly accepted by international development
agencies such as the United Nations (UN) and the World
i.e., empirical evidence on the impact of ITCD, by reviewing
donors’, associations’ and NGOs’ experience in order for
others to learn how to use ICT to benefit their communities.
Bank. As ICTs are relatively new tools within the Exploring how ICT is enacted in practice by social
development arena, their impact is emerging through the work enterprises, although descriptive in nature, represents an
of a number of researchers within the new discipline, ICT for original contribution to the body of knowledge and practice of
Development (ICTD). Whilst the position of most of the ICTD.
multilateral agencies is grossly optimistic about the potential
of ICT for addressing development needs of enterprises in II. LITERATURE REVIEW
developing countries, there is a need for empirical evidence as
ICT was promoted as offering businesses in developing
to the real impact of ICTs such as e-business in developing
countries a potential for creating new exchange mechanisms
countries
to enable them to compete on a more equal basis in world
This paper focuses on the intersection of social enterprises
markets [38]-[41], [37], [45]-[47]. Apart from donor and
(SEs) as dual social-business mission organisations, and the
multilateral agencies, there is a growing body of academic
application of ICTs by those organisations in an international
research such as SMEs’ uptake of ICTs in Botswana [12], the
development context. Social enterprises generate revenue to
contextual study into ICTs in international development
support their social mission, as well as providing employment
agenda in Vietnam [4] and the issues facing sustainability of
for often marginalised people [8], [9]. Social enterprises do
community ICTs [15]. Following on from this nexus between
business with an explicit social mission–helping poor and
ICTs and development is more applied research which builds
marginalized people to participate in sustainable business
on the agenda of donor agencies regarding the uptake of ICTs
by SMEs and the role of change agents in facilitating this
Manuscript submitted September 22, 2008. process [13].
164
Whilst the position of most of the multilateral agencies is communities, social enterprises generate revenue to support
grossly optimistic about the potential of ICT for addressing their social mission, as well as providing employment for
development needs of enterprises in developing countries in often marginalised people [8], [9]. A social mission is the
general and that of SMEs in particular, there is an clearly stated goal of providing benefit to some section of
acknowledgment that tailored initiatives (such as e-commerce) society. This requires identifying the target group the
are best suited for ICT to achieve development goals [20]- enterprise wishes to support. It also requires clear statements
[21]. However, E-commerce has found it challenging to of intent to make sure that it is understood that the purpose of
deliver the purported benefits espoused in the early studies of the organisation is to meet the social, as well as the business
UNCTAD [38] and the OECD [30]. For instance, the DFID mission.
Globalisation and Poverty study explored the benefits of e- The growing relevance of social entrepreneurship for
business for developing countries found that such benefits are development builds on the success of SMEs in reducing
largely missing and can only be realised if developing poverty and improving livelihoods The intersection of ICT
countries’ enterprises build their ability to develop tailored and social entrepreneurship therefore provides a way forward
marketplaces [22] , [31]. More specific research [27]-[28] into in building sustainable communities. As a framework Alter’s
the reality of e-commerce benefits among businesses in [1] social enterprise typology illustrates how social enterprises
developing countries also cautioned the over-optimistic have come up with ownership models, income and
expectations of e-commerce benefits. capitalisation strategies, and the unique management and
Overall, current evidence appears to suggest that for service systems designed to maximize social value. The
e-commerce to become more widespread in a way that international development focus has shifted in recent years to
benefits producer firms in developing countries, greater building sustainable enterprises and encouraging the emergent
attention will need to be given to how firms relate to each SME sector [10], [42]. In acknowledging this shift and
other within global value chains [25], as well as the specific increased focus on economic and social sustainability we find
types of transactions they are involved in. Core to this debate that social enterprises are playing an increasing role with the
is the call for empirical evidence as to the real impact of ICTs support of donors, governments and communities.
such as e-business in developing countries– ‘whose
development does e-commerce effect?–as raised by [30]. III. ASSESSMENT FRAMEWORKS
Finally the call for ‘bottom-up’ approaches in developing In order to assess the outcome and impact that ICT may
countries that are based on realistic assessments of ICTs and deliver to social enterprises this study considers how social
e-business opportunities and obstacles, and region and value enterprises’ use ICT to achieve their dual mission and how
chain specific solutions provide impetus for this current they in turn measure success. When considering how to
research. measure impact, the line of investigation reviewed a range of
This study argues that social enterprises are one form of areas including livelihood and capability frameworks, and
enterprise that are likely to benefit from the application and current work in the field of ICTD. Impact can be measured
use of ICT and can then deliver the promise of e-business using livelihood and capability frameworks [28]. In particular
benefits for development. Emerson & Twersky defined a DFID’s Sustainable Livelihoods Framework (SLF) is an
social enterprise as a ‘generic term for a non-profit enterprise, effective tool for uncovering and highlighting the
social-purpose business or revenue-generating venture relationships between the livelihoods of vulnerable people,
founded to support or create economic opportunities for poor businesses and the institutional context in which business
and disadvantaged populations while simultaneously transactions take place.
operating with reference to the financial bottom line’ [14]. The challenges facing the social enterprise sector are
Davis [8] sees social entrepreneurship as extending the similar to those of other non-profit organisations where the
definition of entrepreneurship by its emphasis on ethical social mission has driven programs yet there has been little
integrity and maximizing social value rather than private value thought how to measure the achievement of their social goals.
or profit. The diversity of the emerging field prompted Alter Some effort has been put into developing tools to assess the
to write the Social Enterprises Typology [1]. Alter’s definition impact of social enterprises in what is commonly known as
captures purpose, approach and practical applications: A ‘social accounting’. Payne clearly states that social accounting
social enterprise is any non-profit-owned revenue-generating is not a social impact assessment as the official title implies,
venture created for the purpose of contributing to a social ‘Social and Ethical Accounting, Auditing and Reporting’
cause while operating with the discipline, innovation and (SEAAR) [33]. Built around stakeholders, AA1000
determination of a for-profit business [1]. (http://www.accountability.org.uk) seeks to link the defining
Key to the understanding of social enterprises is their focus and embedding of an organisation's values to the development
on sustainability and their dual social and business mission. of performance targets, thus tying social and ethical issues
Sustainability is a core principle of social enterprises and into the organisation's strategic management’ [33, p14]. This
refers to a business venture’s simultaneous pursuit of research assumes that social enterprise impact is predicated on
beneficial outcomes along three dimensions: economic, social the organisation's mission, the social objectives it intends to
and environmental. As a way forward in building sustainable
165
achieve, and what impacts can be measured. Figure 1 based on research [6], [22], [24].
Alter [1] provides an overview of the research framework and A case study may capture reality in greater detail, analysing
Table 1 details some of the key concepts to be explored. a greater number of variables than is possible with other
research methods [15]. Yin [47] recommended the use of
case-study protocol using a triangulation of data sources. This
study looks at the key actors in the Cambodian social
enterprise environment in an effort to answer the research
question.
As this study addresses a new area of academic inquiry
there was little primary data regarding the topic, it requires a
range of research methods and active dialogue with
organisations involved in the sector. Triangulation for this
study was created through interviews, surveys and document
analysis.
TABLE 2
Fig. 1. Research Map
SAMPLE OF ORGANISATIONS
TABLE I
KEY CONSTRUCTS OF INTEREST
Sub-sectors Associations Social Enterprises
Construct Definitions Examples Handicraft Artisans Association of Hagar Design
Cambodia Handmade Textile
Beneficiaries Target Disabled, Women, Street Children, Craft Network Association
community who Orphans, People Living With Cambodian Silk Forum Colours of Cambodia
benefit from SEs HIV/AIDS, Youth, Unemployed,
Hospitality Shinta Mani
activities Trafficking Survivors, Villagers,
Farmers IT Services ICT:CAM CIST
Community Info Centres
Forms of Ways in which Employment, Training/ Skills
Digital Divide Data
Support social enterprises Development, As Consumer To Local
Yejj Info
provide support, Producers (Creating Market), As
Marketer Of Local Producer’s Processing Senteurs D’Angkor
Products (Intermediary), Supplier to Multi-sector WASMB
Local Producers, n, Assistance In
Business Development, Product
Design/ Service Development Participants were both industry protagonists and those that
Social Methods of An organisation conducts a social support them through associations and donor agencies.
accounting measuring and audit makes itself accountable to its Supplementary data was collected from a review of
and social reporting an
auditing7 organisational
stakeholders and commits itself to
following the audit's
documents and artefacts such as brochures, marketing material
social and ethical recommendations. and websites. Donor and media reports were sought regarding
performance. the activities of these social enterprises in order to ensure as
Social A set of measures Evaluate Performance against Social complete a picture of the social enterprise sector as possible.
Mission of determining if Indicators, Ability to Employ More Of the 50 organisations identified in the study, 18 were
People, Recommendations from
Indicators the social
Customers, Recommendations from
interviewed and 25 were surveyed.
mission is being
met Beneficiaries, Awards/ External
Recognition from Peers, Specify,
1) Defining Beneficiaries
Awards/ Recognition from The next step is to identify the target beneficiaries of social
Development Agencies/Donors enterprises’ blended mission. A useful tool from the UK is the
DETR 2000 Index of Deprivation – a comprehensive index
IV. METHODOLOGY for comparing levels of deficiency across a range of issues.
In order to bridge the gap between theory and practice a The index covers seven aspects of deprivation – income,
qualitative approach was chosen as applied research leans employment, poor health and disability, education skills and
towards involving practitioners, users and other stakeholders, training, housing, geographical access to services and child
in order to provide practical outcomes. The research took a poverty. The framework was used in this study to assist
holistic–multi-disciplinary, multi-method–approach to interviewees in identifying target beneficiaries. The resultant
triangulate data collection and is detailed as to sample and categories were then used in the survey and are presented in
selection, cross-sectional survey, semi-structured interviews, Figure 3. below.
information observation, document analysis and artefact
review as utilised in the case study. Examining a case of one
sector’s relationship to ICT and development allows an
exploration of particularity and complexity of a single case to
understand its activity within important circumstances (Stake
2000) utilising the case study method of exploratory IS
166
to the contention of this paper – that it is due to the very

nature of social enterprises that ICT does bring benefit
directly as it assists the overall mission and the ‘total value
creation’ that social enterprises bring their beneficiaries.
Initially it was thought that the benefit that social
enterprises bring communities could be divided into two sides
of the dual mission, business and social as illustrated in Fig. 1.
However, as the research process unfolded, the realisation that
the blended mission drives social enterprises had a profound
influence on the resulting framework for analysing how ICT
Fig. 3. Target Beneficiaries delivers benefit to target communities. To understand how this
is manifested in practice, Table 3 provides some examples of
The categories in descending order represent lesser degrees organisations’ statements of their mission from the survey
of importance. The findings showed a high incidence of results.
support for beneficiaries with disabilities; however women
and youth are also a major focus of social enterprises in TABLE 3
Cambodia. EXAMPLES OF SOCIAL ENTERPRISE MISSIONS
As this data is aggregated across different sub-sectors, CIST Bringing digital opportunities to underprivileged young
another picture emerges. The handicraft sector supports people, that is developing an IT training centre
beneficiaries from all categories, including street children and Colours of Provide fair trade channel for small producers and NGO
people living with HIV/AIDS, which neither IT services nor Cambodia products
social enterprises from production sub-sectors supported. Hagar Our social mission is to train and employ highly
Although the survey provides just one set of data, the Design disadvantaged Cambodian women while using our profits
interviews and document analysis also identified that some to financially support Hagar.
other social enterprises supported these disadvantaged people Khmer Great jobs for young people and students Buying raw
in Cambodia. Mith Samlanh works with street children Mekong agriculture from farmer to push them grow more fruit and
Food vegetable Provide nutrition and quality Safe food to
through their restaurants Friends and Romdeng, and Khana is consumers
an association working with NGOs in support of people living
with HIV/AIDS. Both recognize the potential that income Rajana Find Markets for small producers, give them the
employment, send them to school
Association
generation activities can have in creating sustainable
livelihoods for their target beneficiaries. Senteurs Good working conditions, share benefits with staff through
D’Angkor a bonus system. Create employment for local villagers
Now we turn to social enterprises themselves to measure
their impact on their target community and see how ICT does SotheaKhmer Social welfare for people with disability and vulnerable
Silver Crafts poor people. Work and wage for poverty
or does not play a role in achieving their blended mission. The
analysis will address the blended value proposition in
conjunction with the social enterprise typology to determine What is of interest is the intertwining of both business and
the outcomes and impact of ICT on social enterprises. social goals in many of the respondents answers. This
suggests that the mission of these organisations is a blended
V. KEY FINDINGS proposition providing evidence of this new ‘business’ segment
operating with specific commercial and community
Outcome and impact of ICT – achieving social enterprises’
development goals.
mission
The paper now turns to assess if ICT does deliver benefit to 1) Reframing to reflect the blended mission
social enterprises by considering how ICT helps them to As a result of this investigation, there has been a slight
achieve their blended mission which combines commercial departure from the originally perceived dual mission to the
and community goals. It is hoped that the evidence presented more blended version which is reflected in Figure 2. the
here will contribute much needed insight into the impact of Social Enterprise Blended Mission, which highlights the
ICTs on development more generally. potential outcomes of ICT in social enterprises. As this is a
qualitative study, following naturalistic inquiry, it is usual for
A. Blending the dual mission
the research questions to evolve over the period of
In addressing how the dual mission of social enterprises is investigation. What was the original understanding, that the
enacted in practice this paper asks in turn, how ICT interacts dual mission was two separate yet complementary missions,
in this context. The new organisational model of social was due to the perspective as an outsider. Through examining
enterprises embodies a shift in the pursuit of sustainable the data it became evident that this distinction is more blurred
development on all levels, financial, social and environmental. than originally expected.
This move to a ‘blended value proposition’ [14] reflects the
integration of the dual mission, which provides the context as
167
TABLE 6
INDICATORS OF IMPROVED LIVELIHOODS
Indicator % agreement
Earning a livable wage 73%
Provided new career opportunities 63%
Improved food security 71%
Fig. 2. Social Enterprise Blended Mission Improved family stability 71%
Improved health 57%
As Figure 2. demonstrates, it is difficult to differentiate Improved access to education 57%
between the business and social drivers within the blended Increased purchasing power 71%
value proposition. Improved communication can be both with
donors regarding monitoring of a social program such as DDD The nature of social development lends itself to monitoring
and UNDP, or to customers purchasing silk handbags for and evaluation [11], and often the indicators for these
resale in Australia, in the case of Hagar Design through Craft outcomes are clearly set during the start up of the venture. In
Network. the case of donor initiated enterprises there is often a logical
The findings indicate that there is not a clear distinction framework or ‘logframe’ that sets out indicators for
between an NGO that earns income from selling craft, such as monitoring and evaluation. In the case of social enterprises the
Tabitha Cambodia, to an IT company which has a policy of need to measure the triple bottom line means a range of
employing disadvantaged young people, such as Yejj Info. In indicators from both social and business areas are required. In
fact, Yejj Info provides one of the best examples of the evaluating the impact of the current sample, a number of
blended value proposition with their comprehensive mission indicators were presented to social enterprises in the survey.
statements which combines commercially, socially, ethically The 44% incidence of evaluating performance against
and environmentally sustainable internally structured social indicators is promising as it
2) Mapping the blended mission in action suggests that these organisations are very clear of the need to
Social enterprise impact is predicated on the organisation's meet their dual mission. This potentially could mean they will
mission, the social objectives it intends to achieve, and what be more successful than an organisation that is not as clear on
impacts can be measured as a result of the business initiatives. their strategic mission.
A defining ambition of social enterprises is that they support Another category for measuring the success of achieving
those disadvantaged in the community. Leading by example, the social mission is through recognition from stakeholders.
social enterprises employ disabled computer operators at This included acknowledgement from peers, awards, and
DDD, market landmine survivor’s art work at Colours of notes of appreciation from the most important people,
Cambodia, and encourage local communities to invest for beneficiaries and customers which accounted for the highest
their future through a community savings program at Tabitha response rate. Although the survey data had a low incidence
Cambodia. Exactly how this complex interweaving of social of recognition from development agencies/donors through
and financial sustainability is achieved is what this section awards, the other data collection discovered a number of
aims to address. This manifestation of dual sustainability awards and pride in showcasing social enterprises success in
illustrates the true outcomes of the blended mission. What this Cambodia is further explored below.
study acknowledges is that these are inextricably linked, so to C. Identified benefits of ICT
measure one in isolation of the other denies the new paradigm It is through the case of the Cambodian social enterprise
that social enterprises present. sector that this paper demonstrates the opportunity to address
B. Measuring the social impact social issues through business initiatives. Benefits of using
Social impact indicates a positive effect on the target ICT, either directly as in the case of IT services with skilled
population as a result of an intervention and can be measured. labour, or indirectly by selling via an ICT-enabled
Social enterprises, like all development programs, have direct intermediary. If one of the promised benefits of ICT is
as well as indirect impacts. Alter [1] provides some examples increased efficiency of both internal and external operations
of social enterprise impact measurements and corresponding then it is important to measure how ICT impacts on social
indicators. These were adapted to identify the outcomes enterprise operations. In measuring the extent of ICT
(benefits) of ICT to social enterprises and in particular to utilisation in the organisations surveyed, the perception on
assess the impact (attributable to ICT) of social enterprises in impact on operations was deemed an appropriate measure.
the target community. Survey respondents were asked ‘In The most common way ICT helped improve operations was
your opinion how has employment with your organisation through enabling external communications with clients and
improved the livelihoods of your staff?’ and their response is customers. This was further supported by the interview
shown as percentages of those who either strongly agreed or results, with most social enterprises seeing ICT as a core tool
agreed with the indicators listed in Table 6 below. for customer relationship management. Simple ICTs such as
phone and email were the dominant forms.
168
a) Access to Markets when direct sales are for goods produced by the target
Does ICT enable social enterprises to link into the global beneficiaries, customers like to know their purchase actually
trade of goods and services? Through the application of ICT goes to help those who created them. This is where ICT can
are social enterprises able to access new markets that they assist with clear communication and monitoring. An example
previously would have been excluded from? The fair trade of how this is enacted in practice can be seen on the Khmer
movement is a response by civil society to the potentially Silk Processing Association (KSPA) website (See
negative impact of globalisation [33]. It is important to http://www.phaly-craft.org) where they state clearly the
acknowledge that the networked economy can potentially promise that the income goes to those who need it. All income
further marginalise producers in developing countries. Those from the sale of these products are partly used to support
wishing to redress this imbalance see the benefits that ICT can Future Light Orphanage (FLO) activities and children whose
bring as a tool for trade [5]. The majority of enterprises numbers increase everyday.
surveyed felt that ICT provided benefit in key areas of
communication with customers (80%), and sales and c) Employment Opportunities
marketing (64%), suggesting that social enterprises are using Choosing an appropriate target group requires prior
ICT to connect to markets and facilitate trade in ways knowledge and theoretical understanding of the social
previously not imagined. environment of the specific country. DDD identified their
target group as people who fit one of these criteria: disabled,
b) Income Generation orphans (due to war or HIV/AIDS), rural poor and trafficking
One of the defining characteristics of a social enterprise is survivors. They also focus on youth providing entry-level
its income generation activities. This is in fact what positions and training in order to develop their career
differentiates them from others in the non-profit sector as the opportunities.
goal is clearly to earn income to support the social and DDD found that it is important when dealing with disabled
business development of the organisation. What is of interest people to give them the opportunity to prove their skills. On
to this study is the role that ICTs have played in achieving this first appearance a young woman with no hands below her
goal. Respondents to the survey were asked to respond to a wrist may seem to have limited potential as a data entry
number of statements which variously reflected their operator; however once given the chance she has become a
perception of the impact of ICTs on achieving increased sales valued employee who upholds the required productivity levels
or income. These statements were developed from the pilot without having any fingers. This is a prime example of how
assessment of DDD and some preliminary interviews with DDD has taken on the challenge of providing employment
other social enterprises to identify the various ways impact is opportunities for disabled people and met both individual staff
measured. The level of agreement is a combination of agreed goals and general business needs.
and strongly agreed and suggests how these social enterprises
see the relationship between income generation, employment d) Capacity Building
and ICTs. Some social enterprises’ social mission encompasses
TABLE 7 capacity building, where target beneficiaries are provided
ICT ROLE IN ACHIEVING GOALS
training and work experience in using ICTs directly. In this
How do these statements apply to your organisation? % agreement case the social mission drives the adoption of ICTs. In the
Email has increased our sales 50% hospitality industry students are trained on hotel booking
ICT has increased our profits 33% systems such as at Shinta Mani and are able to be prepared for
We have employed more staff as ICT makes it easier for
20% real life experience not just the theory that class based
us to do more work for our clients learning alone provides. By embracing the experiential
Due to increased sales from using ICTs (website/ email)
we have employed more staff
40% learning approach, ICTs can be used to simulate real life
Due to increased income from ICT related we have been experiences and provide an advantage for skill development.
40%
able to offer more social programs to staff One of the major concerns of the digital divide is not only
physical connectivity the ‘second term level digital divide’
As the blended value proposition of the social enterprise is [18]. Skills to effectively use ICTs highlight that connectivity
an interweaving of social and business indicators, it is not alone is not enough to bridge the gap. It is with this in mind
surprising that ICT helps to achieve both these that some social enterprises, particularly in the IT services
simultaneously. Hence separating the business versus social is sector, are providing direct benefit to their target communities.
not that useful, for it is this unique blended characteristic that DDD, Yejj Info and CIST all have training and skills
perhaps makes the impact of ICT more successful than if it development as a core program of assistance. Another social
was separated into a silo approach for development. enterprise in the IT sub-sector, Yejj Technology, operates a
One concern with income generation is that the proceeds go Cisco Network Academy which provides formal training
to support the beneficiaries and are not inappropriately used programs with industry recognised qualifications so that their
for administration or unrelated costs. Of course this is entirely students from disadvantaged backgrounds can engage in the
dependent on the structure of the social enterprise; however IT services sector.
169
e) Strengthening Sponsor Relations competitors. They have actively sought media coverage and
One unique benefit of sponsorship in the highly competitive see it as a way to promote the ideals of their social enterprise
non-profit sector is that a personal relationship is built to their potential market. With the backlash about outsourcing
between a sponsor and a community. It is this connection that jobs overseas a public awareness campaign in the US is very
keeps them returning or undertaking fundraising activities in important for an enterprise such as DDD. On their website
their home country continues sponsorship of either an they have a News section which lists articles and stories which
individual student as in the case of Shinta Mani, or new highlight the experience of the organisation from an external
facilities for the Future Light Orphanage, which is run from perspective. As the article on their website illustrates, DDD
the proceeds from the KSPA. Communication via email is a knows the benefit their business brings young Cambodians
key tool for this, enabling all phases of the relationship, prior like Nut Pove, the opportunity they would otherwise have
to visit and post visit, to continue seamlessly. missed. But it is sadly not enough to provide jobs alone, this is
KSPA has also branched out to provide some guest where ICT can help spread the word of the work being done
accommodation at the orphanage on the outskirts of Phnom in Cambodia and it is examples such as DDD which provide
Penh. Enabling those who wish to make their contribution to evidence of how this is being done.
the NGO through not only donations, or purchasing goods or Although social outsourcing is more commonly connected
services, guests can volunteer to work with the children with the IT services sector, the model is also used in the
teaching English or computers providing a true link to the manufacturing sector. In the case of handicraft sector in
community. The opportunity to personally interact with the Cambodia, many social enterprises work with designers from
children is actively marketed on the KSPA website and also international markets to produce their products. Small
has an online booking form to ease the process. boutique labels make connection with suppliers in Cambodia
The form is then submitted and emailed to KSPA and then through agents such as AAC and Craft Network to arrange
they communicate with the customer via email and finalise the production. For example the designer in the US or Australia
booking and program for their visit. It is the seamless ability send design, colour specifications via email and the
to engage and interact with KSPA even via a simple online production is made to order. As Cambodian enterprises can
form that shows how ICT facilitates this relationship. handle small quantities and offer individual attention to design
Cambodia unfortunately attracts paedophiles who visit houses they are increasingly popular. Bronwyn Blue a Small
Cambodia to take advantage of poor and disadvantaged Business Development Consultant, specialising in the craft
children and women. The management of FLO take the care sector, works with social enterprises to improve their quality
of their children very seriously and screen visitors to ensure control and designs to meet the needs of international market:
they are not placed at risk but benefit from interacting with The core finding presented in this paper is how the blended
only genuinely interested volunteers who match KSPA's mission is enacted in practice and in turn, how ICT interacts in
concern for the welfare of the children as a priority. this context. This study argues that it is the integration of the
dual mission, which is the very nature of social enterprises,
f) Social Outsourcing that ICT brings benefit directly as it assists the overall blended
Focusing on the philanthropic benefit of outsourcing to mission and the ‘total value creation’ that social enterprises
disadvantaged communities has been a unique strategy to bring their beneficiaries.
counteract the backlash that US firms face as they outsource
jobs overseas. Maximising this market niche, DDD shows that
business can benefit communities in developing countries in a
viable business model that does not exploit workers. DDD in
fact targets beneficiaries from marginalised backgrounds and
provides employment where others reject them. Through their
social mission they provide educational scholarships that
support their workers career development to leave DDD and
take up new job opportunities. This then provides an opening
for other disadvantaged Cambodians to join the workforce.
Focusing on the staff’s professional development and
achieving the business goal provides much more of an impact Fig. 4. Map Total Value Creation
for local communities than charity. Businesses who value
social contribution are happy to partner with social enterprises VI. DISCUSSION
such as DDD as they can satisfy their needs to reduce costs
This study found that the major characteristics of social
through accessing an ethical supply chain. All of which is
enterprises are that their development objectives are to
enabled by ICT as digitising content is the mainstay of DDD’s
achieve a blended mission meeting both social and economic
services.
goals. From this basis the assumption would be that ICTs
DDD also understand the benefit of social marketing and
should be utilised to realize these goals [19], [7]. It is with this
the role it can play in bringing a comparative advantage to
170
in mind that this study aimed to find a realistic view of the establishing the social enterprise and of those surveyed three
role ICTs play in supporting social enterprises to meet their were still registered as an NGO whilst eight others were set up
dual mission. What this study acknowledges is that these are as social enterprises from the outset. The summary of results
inextricably linked, so to measure one in isolation of the other in Figure 8.e shows a number of social enterprises as
denies the new paradigm that social enterprises present. classified according to the SE Typology. The level of ICT
Social enterprise impact is predicated on the organisation's uptake is not directly linked to the level of integration
mission, the social objectives it intends to achieve, and what however it could be claimed that there is some correlation
impacts can be measured. Social enterprises, like all social with full integration and management vision with more
programs, have direct as well as indirect impacts. Alter [1] successful ICT assimilation.
provided some examples of social enterprise impact This study argues that it is the integration of the dual
measurements and corresponding indicators. These were mission, which is the very nature of social enterprises, where
adapted to identify the outcomes (benefits) of ICT to social ICT brings benefit directly as it assists the overall mission and
enterprises and in particular to assess the impact of social the ‘total value creation’ that social enterprises bring their
enterprises in the target community. In evaluating the impact a beneficiaries. This is promising as it suggests that these
number of indicators were presented to social enterprises in organisations are very clear of the need to meet their dual
the survey and the results indicated that evaluation of mission. This potentially could mean social enterprises will be
performance was against internally structured social indicators more successful than an organisation that is not as clear on
such as improved livelihoods, increased income, access to their strategic mission. This bears direct significance on the
affordable health care and access to education. This is framing and uptake of ICTs.
promising as it suggests that these organisations are very clear The institutional form that the social enterprise takes itself
of the need to meet their dual mission. This potentially could is another potential influence as to the uptake of ICT.
mean social enterprises will be more successful than an Management awareness, capacity of staff and commitment to
organisation that is not as clear on their strategic mission. use ICT to achieve the blended mission requires both sides to
The goal for development of disadvantaged communities be successful where meeting economic needs is equally as
can take many forms and in the case of social enterprises there important and social goals.
appears to be a number of models as outlined in Alters Social
DDD Yejj Info KSPA

Hagar Design Shinta Mani Tabitha
Enterprises Typology [1]. If we consider the social enterprises
sector as a norm due to its accepted social constructs of
philanthropy and social good, then this in itself Fig.5. SE Typology - Business/Program Integration [1]
could act as an influence on the enterprises uptake of ICT. So
it could be claimed that the type of social enterprises may also The final aim of the research was to measure the role of
impact on the uptake of ICT. Perhaps it is the ways the social social enterprises’ use of ICTs in achieving their goals. This
enterprise enact their social mission that predicates how they study’s major finding was that the blended mission requires
will or will not utilise ICTs as the following examples both sides to be successful, thus meeting economic and social
illustrate. goals as equally as important. In identifying at what point ICT
The form that a social enterprise takes is an interesting intervention made an impact, the areas that stand out include
entry point for analysis as represented in Figure 5. Using income generation, access to markets, improved
Alter’s social enterprise typology to analyse the survey data, communication with customers, and strengthening of sponsor
regarding how social enterprise perceive themselves, can relations. Whilst the results indicate a positive acceptance of
provide some insight into the level of maturity and ICT, most social enterprises are still at the early stages of
development of the organisation. adoption. This means that there is a continued role for external
One of the approaches to assessing social enterprises is to institutions to support further development, with a suggested
classify them based on the level of integration between social focus on meeting the dual mission in a sustainable manner,
programs and business activities and this framework helped and therefore meeting the requisite standards for capacity and
develop the research design for this current study. This is use.
where the social enterprise typology can assist in analysis of Most importantly this study acknowledges that the social
the experience of this sector in Cambodia. Interview results and business missions are inextricably linked, so to measure
indicated that two social enterprises that were established by one in isolation of the other denies the new paradigm that
NGOs as income generating initiatives have made the social enterprises present. Perhaps the most active sub-sector
transition to being an independent social enterprise. The that best reflects this unique value proposition is the emerging
question this raises is to what extent does the relationship with IT sector. Social outsourcing represents a new model and the
the founding NGO continue once the transition has occurred? potential of the IT services sub-sector came to light during this
Interview data suggests that NGOs are instrumental in research. In summary the case of the social enterprise sector in
Cambodia provided a rich tapestry of examples with a number
171
of organisations receiving external recognition of Cambodia’s ICT can play a role in facilitating a more streamlined process
social entrepreneurs. The outlook for the other sub-sectors for all actors.
handicraft and processing have yet to fully engage and use Finally, an area that poses potential is to review the long
ICT to their maximum benefit. As such there is a need for term impact of ICT in social enterprises in developing
further support to best harness the potential that ICT can bring countries in terms of its contribution to poverty reduction and
developing countries. improved livelihoods. This study raises the issue of measuring
the impact of ICTD-SE and it is recommended that further
VII. RECOMMENDATIONS research be conducted to refine the framework. This would
If the aim of social enterprises is to have a positive outcome enable better assessment of the inherent challenges and
and impact for their target beneficiaries, then they should be potential impact that the intersection of ICT, development and
open to develop standards that ensure no-one is exploited. social enterprises presents producers and communities in
Responding to the call to address issues in the regulation of developing countries.
global supply chains, and the potential that ICT can play in the In conclusion this study contends that the foundations are
self-regulation required in ethical trade provides another there for social enterprises to progress along the continuum
perspective to this study [22]. Through adhering to the from simple ICT use to integrated ICT and e-business. This
standards of fair or ethical trade, Cambodia can build on its combined with the increased level of awareness of the benefits
reputation that workers are safe and purchases make a real and the supportive donor environment augurs well for
difference. This is an issue for workers in the home-based Cambodian social enterprises to engage in broader use and
handicraft business model and increasingly will also become therefore bring benefits of ICT in their development and
an issue for the IT outsourcing sector. positive impact in the lives of the target beneficiaries.
As such one recommendation is to develop a fair trade
charter for the social enterprise sector in order to ensure that REFERENCES
those who are meant to be benefiting from the enterprise are [1] Alter, S. K. 2004) Social Enterprise Typology [online]. Virtue Ventures
LLC. 2004.
in fact receiving a real benefit. The development of fair trade [2] Badiali, N. (2005) Tall Poppies - Enterprising the Social Economy. The
monitoring along the lines of the ILO Better Factories (See Scottish Council of Voluntary Organisations (SCVO) and Kilmarnock
http://www.betterfactories.org) program is a model that could College.
[3] Bornstein, D. (2004) How to Change the World: social entrepreneurs
address this. This is a unique initiative that uses ICT to assist
and the power of new ideas, Oxford University Press.
the process of transparent reporting in the garment sector in [4] Boyle, G. (2002) Putting Context into ICTs in International
Cambodia. Using an interactive website, this project enables Development: An Institutional Networking Project in Vietnam, Journal
buyers to log in from overseas to check the status of a of International Development, Vol.14, No.1, 101-112.
[5] Caserta, A. (2001) Creating a Fair Trade Partnership through a Fair
particular factory to see whether it has met Cambodian labour Trade Global Communication System.
standards. One issue is that this model is also seeking to make [6] Cash, J. I. & Lawrence, P. (Eds.) (1989) The Information Systems
the transition from a donor initiated program to a self- Research Challenge: Qualitative Research Methods, Boston, Harvard
Business School Research Colloquium.
sustaining model. In fact the most active representative body [7] Curtain, R. (2004) Information and Communications Technologies and
in the social enterprise sector is the Artisans Association of Development: Help or Hindrance?, Curtain Consulting for AusAID.
Cambodia, which already has a working relationship with [8] Davis, S. (2002) Social Entrepreneurship: Towards an Entrepreneurial
Culture for Social and Economic Development. Youth Employment
Garment Manufacturing Association of Cambodia via the Summit.
Government-Private Sector Forum, so it is feasible that some [9] Dees, J. G. (2003) New Definitions of Social Entrepreneurship: Free
mutually beneficial arrangements can be made. As there are Eye Exams and Wheelchair Drivers. Knowledge @ Wharton, Wharton
some correlations between the manufacturing sectors and the School, University of Pensylvania.
[10] DFID (2005) Working with the Private Sector to Eliminate Poverty.
sub-sectors of the social enterprise sector, there might be a DFID & the Private Sector. London.
possibility for a new holistic model to monitor social [11] DiMaggio, P. (2001) Measuring the Impact of the Non-profit Sector on
outsourcing, using ICT as a key tool to promote transparency. Society is Probably Impossible but Possibly Useful: A Sociological
Perspective. Measuring the Impact of the Non-profit Sector. P. Flynn
Another issue that arose from this research is the lack of and V. A. Hodgkinson. New York, Plenum: pp.249-272.
regulation of the social enterprise sector in Cambodia, which [12] Duncombe, R. and R. Heeks (2001) Enterprise Development and
the above recommendation would address. In the case of the Information and Communication Technologies in Developing Countries:
Supporting “ICT-Flyers”. Manchester, UK, IDPM, University of
NGOs making handicrafts, they are officially registered as an Manchester.
NGO with the Ministry of Interior, which has no reference for [13] Duncombe, R. and A. Molla (2005) An Evidence Based Framework for
income generation. As such, NGOs do not pay tax, and the Assessing the Role of Change Agents in the Promotion of E-Commerce
for SMEs in Developing Countries. Connecting people and places:
revenues from sales overseas are not being reinvested back
challenges and opportunities for development, Milton Keynes, UK, UK
into the broader economic development of Cambodia. Development Studies Association.
Registration with Ministry of Commerce is required for any [14] Emerson, J. (2003) Blended Value Proposition: Integrating Social and
entity selling goods or services that may require a certificate Financial Returns. California Management Review, 45.
[15] Emerson, J. and F. Twersky. (1996) New Social Entrepreneurs: The
of origin for export. These are areas that donors and other Success, Challenge and Lessons of Non-profit Enterprise Creation. San
institutions working with social enterprises can assist and that Francisco, Roberts Foundation.
172
[16] Galliers, R. (Ed.) (1992) Information systems research : Issues, methods, [44] WorldBank (1999) Knowledge for Development. World Development
and practical guidelines, Oxford, Blackwell Scientific Publications. Report. Oxford University Press, World Bank.
[17] Gurstein, M. (2005) "Sustainability of Community ICTs and its [45] WorldBank (2001) Attacking Poverty. World Development Report.
Future." The Journal of Community Informatics Vol.1 (2). Oxford University Press, World Bank.
[18] Hargittai, E. (2002)"Second-level digital divide: differences in people's [46] WorldBank (2002a) Information and Communication Technologies: A
online skills." First Monday Vol.7 (4). World Bank Group Strategy. Washington, World Bank.
[19] Harris, R. (2004) ICT for Poverty Alleviation. e-Primer. Kuala [47] WorldBank (2002b) Information Communication Technologies. A
Lumpur, APDIP. UNDP World Bank Group Strategy. Washington DC, World Bank.
[20] Harris, R. and D. Vogel (n.d.) E-Commerce for Community Based [48] Yin, R. K. (1993) Case Study Research, Design and Methods.
Tourism (Draft) Hong Kong, Harris and Associates. California, Sage Publications.
[21] Heeks, R. (2002) Failure, Success and Improvisation of Information
Systems Projects in Developing Countries. Development Informatics
Group, IDPM, University of Manchester. Working Paper 11.
[22] Heeks, R. & Duncombe, R. (2003) Ethical Trade: Issues in the
Regulation of Global Supply Chains. Centre on Regulation and
Competition. University of Manchester.
[23] Humphrey, J., M. Robin, et al. (2003) The Reality of E-commerce with
Developing Countries. Globalisation and Poverty, DFID Media @LSE
and IDS.
[24] Kaplan, B., Truex, D. P., Wastell, D., Wood-Harper, A. T. & Degross, J.
(Eds.) (2004) Information Systems Research: Relevant Theory and
Informed Practice, Norwell, MA, Kluwer Academic Publishers.
[25] Kaplinksy, R. and J. Readman (2001) Integrating SMEs in Global Value
Chains: Towards Partnership for Development, UNIDO.
[26] Klein, H. M. & Myers, M. D. (1999) A Set of Principles for Conducting
and Evaluating Interpretive Field Studies in Information Systems. MIS
Qtly.Viewed: 10 November 2006.
[27] Molla, A. (2004) The Impact of eReadiness on eCommerce Success in
Developing Countries: Firm-Level Evidence. Development Informatics
Working Paper. Viewed: 20 November 2006.
[28] Molla, A. (2005) Institutions and ICTs in the Makerere university:
Bridging digital divide or spreading donor dependence,. IFIP
Conference. Abuja, Nigeria.
[29] Murray, C. (2001) Livelihoods research: some conceptual and
methodological issues. Development Studies Association Annual
Conference-Panel on Livelihoods Frameworks and Poverty Analysis,
University of Manchester.
[30] Odedra-Straub, M. (2003) "E-commerce and development: Whose
development?" Electronic Journal of Information Systems in Developing
Countries Vol.11 (2): pp.1-5.
[31] OECD (2004) Accelerating Pro-Poor Growth Through Support for
Private Sector Development, OECD.
[32] Paré, D. J. (2001) Does This Site Deliver? B2B E-commerce Services
for Developing Countries. B2B E-commerce and Developing Countries.
Media@LSE. London, The London School of Economics & Political
Science. DFID funded project on e-commerce for developing
countries.UK
[33] Payne, J. E. (2002) E-Commerce Readiness for SMEs in Developing
Countries: A Guide for Development Professionals. Washington,
LearnLink.
[34] Redfern, A. & Snedker, P. (2002) Creating Market Opportunities for
Small Enterprises: Experiences of the Fair Trade Movement. SEED
Working Paper 30. Geneva.
[35] Spinali, L. and H. Mortimer (2001) A Scan of the Not-For-Profit
Entrepreneurship: Status of the Field and Recommendations for Action,
Kauffman Center of Entrepreneurial Leadership.
[36] Stake, R. (2000)Case Studies. Handbook of Qualitative Research. N. K.
Denzin and Y. S. Lincoln. Thousand Oaks, CA, Sage Publications.
[37] UN (2003) Tools for Development Using Information and
Communications Technology to Achieve the Millennium Development
Goals. Working Paper, United Nations ICT Task Force.
[38] UNCTAD (2001) The Digital Economy: Integrating the LDCs into the
Digital Economy. E-Commerce and Development Report 2001.
Brussels, UNCTAD, Electronic Commerce Branch.
[39] UNCTAD (2002a) World Telecommunication Development Report:
Reinventing Telecoms, UNCTAD.
[40] UNCTAD (2002b) E-Commerce and Development Report 2002.
Geneva, UNCTAD.
[41] UNCTAD (2004a) Partnership for Development: Information and
Knowledge for Development, São Paulo. UNCTAD
[42] UNCTAD (2004b) E-Commerce and Development Report 2004.
[43] UNDP (2006) Global partnership for development. Annual Report.
New York, UNDP.
173
Political Incentives and Policy Outcomes:

Who Benefits from Technology-Enabled
Service Centers?
Jennifer Bussell
delivery of government services such as identity cards, birth

Abstract—This study investigates the causes of variation in and death certificates, and licenses; the supply of welfare and
government policies to use information and communication redistributive goods; and general government-citizen
technologies to improve service delivery to citizens. I ask why
communication (see, inter alia, [4]-[6]).
state governments in India vary in the number and type of
services they offer to citizens through technology-enabled citizen Developing country governments responded to this
service centers. I argue that politicians estimate the expected opportunity and began to develop new systems for
electoral benefits from providing improved services to citizens “eGovernance” and digital technology-based service delivery
and weigh these benefits against the costs of increased (see, inter alia, [7]-[12]). Yet, after more than a decade, the
government transparency and associated reductions in corrupt question remains whether governments have utilized new
income. Politicians then design service center policies to
technologies to enable better service delivery to citizens.
maximize their chances of retaining power. Because levels of
corruption and the characteristics of electoral competition vary In India, the main subject of this study, state governments in
across the Indian states, we see related variations in technology nearly all of the twenty major states had implemented policies
policies. These variations in policy, and in particular the services to provide technology-enabled services to citizens by 2006.1
made available to citizens, have important effects on who benefits These “technology-enabled service center” policies deliver
from citizen service centers. I use evidence from sixteen Indian public and private services2 to individuals through the use of
states to test these arguments, and show that the character of the
information technologies, in particular computers and the
ruling government and the level of state corruption are robust
predictors of variation in state-level technology policies. Internet, at dedicated local centers. These centers provide a
“one-stop shop” environment for multiple government
Index Terms—Information and communication technology, departments, thus streamlining the process by which citizens
development, India, corruption, politics. access an array of services.3
Yet the outcomes of efforts in India, and thus the benefits
new technologies provide to citizens, differ tremendously
I. INTRODUCTION across the states, in terms of both the extent and character of
services provided. In the low-income states of Chhattisgarh
W hy do some governments invest in new technologies

and Orissa, the Chhattisgarh government provides nearly forty
services, while Orissa offers fewer than ten. The types of
while others do not? Why do certain governments implement services also vary, with states such as Andhra Pradesh
full-scale reforms of public service delivery, while others do providing a wide range of high-demand services, while others,
so in a superficial manner? The emergence of new information including Uttarakhand, provide only a few low-demand
and communication technologies in the 1990s raised these services.
questions in a stark way. Low-cost, digital technologies I argue that the observed variation in technology adoption,
offered prospects for increasing the effectiveness of and the related benefits derived by citizens, result from the
government, particularly through improved service delivery to strategic calculations of political elites. Drawing on extensive
citizens. International observers expected developing country sub-national analysis of technology policies in the Indian
governments, often criticized for failing to deliver services in
an effective and transparent manner [1], to be the prime 1
The constitutional allocation of responsibilities in the Indian federalist
beneficiaries of new digital technologies [2], [3]. Developing system makes sub-national state governments responsible for the provision of
country citizens were expected to benefit from improved a majority of government services to citizens, rather than the national
government. As a result, states are institutionally the most appropriate level
for implementing service centers.
2
Manuscript received September 17, 2008, accepted January 20, 2009. This Service centers often provide a mix of government services and services
work was supported in part by the National Science Foundation under Grant from the private sector, such as Internet access, digital photography, or
No. 0326582. telemedicine services.
3
Jennifer Bussell is a Ph.D. Candidate in the Department of Political While there may be related policies regarding service delivery via mobile
Science at the University of California, Berkeley, Berkeley, CA 94720-1950 centers or mobile phones, the policies studied here all involve service delivery
(email: jbussell@berkeley.edu). at an immovable center through the use of information and communication
technologies (ICTs).
174
states, I posit that politicians attempt to utilize technology funds noted first in the literature by Wade [14], citizens pay
initiatives to maximize their chances of retaining power, but bribes to bureaucrats, who then channel some portion of these
that these policies have both potential benefits and costs. funds to politicians, either in hopes of a job transfer or simply
Politicians evaluate the tradeoffs between expected effects on to satisfy a political boss’ demands (see also [15]). Politicians
their political interests, especially their potential for reelection, then use a portion of these funds to finance future elections.
of implementing new policies. When the balance of As a result, any threat to this flow of funds is a threat to
politicians’ calculations differs across states, we observe politicians’ ability to secure reelection.7
variation in the policies these states adopt, and in the I contend that the size of these costs depends primarily on
associated benefits for citizens. variation in two key characteristics of the Indian states: the
This paper looks specifically at the ways in which electoral level of corruption and the cohesion of the ruling government.
politics affect the services offered in Indian state service The level of corruption affects incumbent politicians’
center initiatives. I argue that electoral incentives, as shaped dependence on bribes as a source of campaign finance—the
by the combined effects of preexisting institutions of greater the share of bribes in overall campaign finance and
corruption and party competition, affect the number and type thus the more dependent politicians are on bribes, the less
of services made available to citizens. I then test this argument supportive they will be of policies to increase transparency in
using data on services in the sixteen Indian states 4 that service delivery. There is substantial variation in the level of
implemented technology-enabled citizen service center corruption across the Indian states, with fewer than 20% of
policies during the period from 1999 to 2006.5 The analysis is citizens in states such as Kerala and Himachal Pradesh
based on a new dataset of state service center policies that I encountering demands for bribes when interacting with
developed during fieldwork in sixteen states, in addition to government officials, while more than 60% of citizens Bihar
supplementary data collection. I utilize both quantitative and or Karnataka have direct experience with bribing in multiple
qualitative analytic techniques to evaluate the factors government departments [16].
associated with variations in technology-enabled service The cohesion of the ruling government, by which I mean
provision across the states. After considering the details of my whether a single party or a coalition of parties rules in the
argument, I address trends in the number and type of services6 state, can also affect costs. Coalition versus single-party rule is
made available. relevant because of the role that supporting minority parties
play in the stability of coalition governments. When a large
II. ELECTORAL INCENTIVES AND TECHNOLOGY POLICY party requires the support of smaller parties to form a
I argue that politicians will implement technology-enabled government, the lead party often allocates ministerial posts to
service centers, and specific services within those centers, its coalition partners in order to reinforce their support. In
when the expected electoral benefits from doing so outweigh most cases, the incentive for holding a ministerial post is not
the electoral costs. The benefits to politicians are not likely to control over policy, but access to the associated departmental
differ dramatically across states. Politicians and parties can rents. This is because ministers can demand a portion of the
amplify the electoral benefits of technology policies by bribes collected by bureaucrats in their department during
targeting services to their most important constituencies, yet service delivery. Such rents provide a concentrated economic
because every incumbent should be able to benefit from such a and political benefit to ministers, who can use these funds to
strategy, the overall size of the electoral benefit of services finance re-election campaigns. When politicians consider the
reform should be similar for political incumbents across states. potential costs of new technology policies, ministers who
The costs to politicians from service centers come control departments delivering a high volume of services to
primarily in the form of foregone corrupt income. In the citizens—for example, Transportation—may then expect
preexisting process of service delivery, politicians use the major costs from more efficient service provision.
machinery of the state to extract rents [14]. Yet in the Indian From the perspective of a coalition member, a threat to this
context they do so not simply for personal pecuniary gain but income is also a threat to the expected benefits of participating
rather to enhance their hold on political power. In a cycle of in the coalition. Moreover, minority partners in a coalition
government typically reap only a limited, often
4
disproportionately small, electoral benefit, as a party, from
I consider only those states in which the state government implemented a
service center projects; credit instead tends to accrue to the
service center policy, thereby excluding the Northeast states and Jammu &
Kashmir, where the central government initiated a service center initiative, larger majority party. Yet because minority partners are
and the four remaining states in which no service center policy was crucial to the survival of coalition governments, and because
implemented (Bihar, Goa, Jharkhand, and Madhya Pradesh). decision-making power is decentralized across ministers from
5
In some states more than one service center initiative was implemented
during the period under consideration, for example one in urban areas and a all coalition parties, key supporting party ministers often have
second in rural areas. In these cases I consider the overall services offered power to resist new technologies for delivering services. As a
across all projects, omitting any overlapping services. result, supporting party ministers have both potential
6
Evaluation of services requires an analysis of the services that are stated
to be available versus the services that are actually available in the centers.
7
While this is difficult to assess without visiting thousands of individual My argument also extends discussions of India as a “patronage
centers, a reasonable measure of available services can be made from a democracy” (Chandra, 2007; 2004) by focusing on how patronage and other
combination of site visits, interviews with project representatives, and reviews funds come from corrupt, rent-generating activities by bureaucrats and
of associated websites. politicians.
175
incentives to resist technology policies and important from decreased corrupt income. I now consider the effects of
decision-making power within the coalition that they can use these political incentives on the character of technology-
to influence policy outcomes. enabled services.
In single-party ruled states, the concentrated costs for
individual ministers are often outweighed by the overall III. CORRUPTION, COALITIONS, AND TECHNOLOGY-ENABLED
electoral benefits of technology policy to the party. Single- SERVICE DELIVERY
party governments also place decision-making authority with a A key question for an analysis of service center policies is
small number of individuals, in particular, in the government’s which citizens actually benefit from introduction of these
Chief Minister. As leader of the party, the Chief Minister technologies. Because state governments are responsible for
“internalizes” a large portion of the electoral benefit to the providing services to their entire population, if they offer
party of implementing service center projects. In single-party technology-enabled services that benefit only a portion of the
governments, the electoral benefits of implementing service population, then they are failing to offer the same benefits to
center projects are concentrated, particularly for the Chief other citizens. Fig. 1 shows the variation across Indian states
Minister, while the costs are in part diffused among Ministers in the number of services made available to citizens.
and other individual Members of the Legislative Assembly
(MLAs) who have limited capacity to resist the directives of
the Chief Minister, due largely to party discipline.
In contrast to single-party governments, then, in coalition
governments the political costs of service center
implementation are concentrated in key decision-makers with
the capacity to resist—or alter the character of—project
implementation, while the electoral benefits are, from the
perspective of these key decision-makers, diffused outwards.
This logic suggests that the type of ruling government—
namely, whether single-party or coalition—should then affect
the character of new technology policies.
There is qualitative evidence for the importance of both
corruption and ruling government cohesion in the choice of
services to include in service center initiatives. In interviews
with state level bureaucrats, representatives of states with low Fig. 1 – Available Services in the Indian States (2006)
or average levels of corruption were more optimistic that the
introduction of technology would be politically feasible. A My argument offers specific predictions for the ways in
bureaucrat from Orissa, a state with just below average which politicians’ calculations over the likely costs and
corruption, when asked about resistance to service centers benefits of technology policies may affect who benefits from
within government, noted that “There has been resistance, but these policies and to what degree. In cases where the expected
we are slowly getting rid of it…People understand now that costs from technology adoption are high, such as in states with
the story has gone past where they can stop it” [17]. Whereas high levels of corruption and in states ruled by a coalition
an official from Haryana, a state with relatively high government, governments are likely to implement policies in
corruption, stated that, “many departments are not moving ways that serve fewer citizens overall.
forward…with the implementation of front-to-back
eGovernment services” [18]. He felt that this was largely due
A. The Quantity of Services
to the threat of increased transparency. For example, in “the
Transport Department people in the department do not want to I evaluate the relationship between level of corruption and
shift to a new system because they have established ways by policy outcomes using a novel measure of state-level
which they are able to skim money off the top and they don’t bureaucratic corruption, which is uniquely suited to testing my
want to lose these sources of income” [18]. theoretical claims. This measure draws on a Transparency
Analysts have also noted the importance of coalition International survey of corruption in India [16]. The survey
governments for service center outcomes in individual state asked citizens about both their experience with corruption in
cases. Kiran [19] comments that a service center project in the acquiring services from government and their perception of
state of Kerala “was opposed by the participating departments corruption in government. Transparency International
on account of the fear that they would lose their existing provides an indexed corruption score by state, based on eleven
authority and power. This was particularly evident in Kerala, departments, including the police, municipal services,
which is ruled by a coalition government, with different electricity, and the judiciary. I use this indexed score to
political parties in charge of different departments.” develop my measure of bureaucratic corruption across the
Thus the policy-making incentives for party leaders come Indian states. The survey is particularly appropriate for the
from the potential electoral benefits of implementing “good purposes of this analysis because it focuses explicitly on
governance” technology policies and the contrasting threats bureaucratic corruption in low-level service delivery, the area
176
targeted by service center policies, rather than the high-level

corruption more often engaged in by top officials. Ruling Party Cohesion
In states with high levels of corruption, the large The next relationship to consider is between the type of
anticipated costs from more transparent service delivery government and the quantity of services. Are single party
should discourage service provision overall, leading to fewer governments associated with higher numbers of services?
services than in lower corruption states. The cohesion of the Here, the evidence is not as clear-cut. On average, states ruled
ruling government should also affect service selection, as by single parties do offer more services than coalition states;
coalition governments are likely to resist the inclusion of with a mean of 18.0 services while coalition states have a
supporting-party controlled services, leading to fewer services mean of 11.4 services. The difference between these two
overall. means however is statistically significant only at the .1 level.
In this section, I discuss descriptive statistics, difference- Given the small number of cases and the fact that the services
of-means tests, and OLS models I use to evaluate these do trend in the direction predicted by my argument, however,
hypotheses. It is worth bearing in mind that the sample size is it is reasonable to believe that there is a meaningful
small: with only sixteen state cases, statistical power may be relationship in the data between the quantity of services and
low. If I am able to uncover statistically significant the type of government. In the case studies below I discuss
relationships, it is because these relationships are particularly additional evidence for a relationship between government
strong.8 cohesion and services outcomes.
Corruption Corruption and Government Cohesion

First, the level of corruption in a state is a strong predictor Perhaps the more relevant test of my argument is an
of the number of government services offered in computer- analysis of the relationship between corruption and
enabled service centers. The major findings of this analysis are government cohesion. The effect of corruption on the number
presented in Table I. Among the states with below average of services is most evident in low corruption, single-party
bureaucratic corruption levels, the mean number of services states. These states provide, on average, a larger number of
offered is 20.1. This noticeably contrasts with the mean services than any other states, and this difference is
number of services offered in states with above average levels statistically significant.
of corruption, which is 10.6. This difference is statistically The effects of ruling government cohesion and high levels
significant at the .05 level. As noted above, any significant of corruption are somewhat more difficult to parse, as the
finding with only sixteen cases implies a particularly strong average number of service provided in high-corruption, single-
relationship between the variables under consideration. party states is not statistically different from that provided in
A univariate regression model offers similar findings. The coalition-ruled states. Coalition-governed states provide fewer
relationship between quantity of services and corruption, when services on average than single-party states, regardless of the
corruption is measured on a scale from zero to one, is not level of corruption in the state. Thus, based on this data, it is
statistically significant, but is in the predicted direction. When only possible to say that either coalition states will provide
corruption is measured dichotomously, the relationship is fewer services regardless of the level of corruption, or that
significant at the .01 level (t-statistic of -3.22) (See Model 1 in high corruption states will provide fewer services than low
Table III below). corruption states regardless of ruling government cohesion. In
Table I order to provide greater clarity on the role of ruling
Indian States, Corruption, and Quantity of Services government cohesion in service selection, I evaluate the
Corruption specific services offered by coalition governments in detail
Below Average Above Average below.
Absolute score: 240-478 Absolute score: 479-695
Scaled score: 0-.523 Scaled score: .524-1
Andhra Pradesh (41) Delhi (6)
Chhattisgarh (37) Haryana (10)
Gujarat (19) Karnataka (22)
Himachal Pradesh (24) Rajasthan (11)
Kerala (11) Tamil Nadu (10)
Maharashtra (16) Uttarakhand (2)
Punjab (16) Uttar Pradesh (13)
Orissa (7)
West Bengal (10)
Mean services: 20.11 Mean services: 10.57
Difference in means: 9.54 (t-statistic = 2.07)
8
Statistical power, of course, is a function not just of the sample size but
also the effect size.
177
Table II Table III

Mean Number of Services by Corruption and Ruling Multivariate Analysis of Candidate Explanatory Variables
Government Cohesion Variable Model 1 Model 2 Model 3 Model 4
Ruling Difference Constant 24.7 25.8 24.5 24.9
Government in Means,
Cohesion Single (7.19) (6.85) (2.29) (2.30)
Single Coalition Party - Above Average -14.0** -13.7** -13.3* -12.4
Party Coalition Corruption
(t-statistic) (-3.22) (-3.12) (-2.39) (-2.17)
Ruling Party -3.5 -3.4 -4.2
Corruption Low 27.4 11.3 16.1 Cohesion
(2.87) (-0.80) (-0.75) (-0.88)
High 10.2 11.5 -1.3 Months Since .02 .1
(-0.43) Initiation
Difference in 17.2 -0.2 (0.14) (0.38)
Means, (3.04) (-0.06) State Domestic -9.0
Product per Cap
Low - High
(-0.86)
Corruption
N 16 16 16 16
(t-statistic)
r2adj .38 .37 .32 .30
Entries are unstandardized regression coefficients with t-ratios
Alternative Explanations in parentheses. *p < .05 **p <.01
The length of time elapsed since a service center initiative
was implemented might reasonably be expected to influence
the number of available services. As governments in states B. Selection of Services
that implemented centers early will have had more time to The benefits of computerized service centers to citizens
introduce additional services. The level of economic depend not just on the quantity of services, but also on the
development might also be associated with provision of character of the specific services made available. In this
services, as states with higher incomes might have more funds section I consider first whether high demand and high
available to invest in government reforms. corruption services are made available in services centers
In order to test these alternative explanations, I used across all states. I then consider the specific types of services
multivariate regression models to evaluate the relationship of made available in coalition-government led states.
level of corruption, ruling government cohesion, time elapsed The government of India, and in particular state
since implementation, and state economic development with governments in India, provides hundreds of services to
the quantity of services available in the states. As shown in citizens on a regular basis. Table IV lists the fifteen most
Table III, the level of corruption shows a clear and robust commonly provided services in computerized centers across
relationship with the quantity of services, even when taking the states. Out of the 73 government services offered by at
into consideration these alternative explanations. However the least one service center initiative, only these services are
length of time since centers were initiated and the level of offered by more than 25% of the projects (shown in Table
economic development in a state show no relationship with the IV).9 I also include one service, ration cards, that is offered in
quantity of services. Ruling government cohesion also does 24% of initiatives, but is interesting to include in the analysis
not show a statistically significant relationship with quantity because it is provided by a different department, Food and
of services, but as noted above this may in part be due to the Civil Supplies, than any of the other services.
small number of coalition states in the dataset. Are these services in high demand by citizens? Based on
Transparency International India’s survey of corruption in
public service delivery [16], only a small number of those
services most needed by citizens are provided in computerized
service centers. Of the top twenty-five services required by
citizens, thirteen are offered in at least one state. However
only six of these are among the services offered in at least
25% of states (highlighted in bold in Table IV). Thus, in many
cases there is a disconnect between the services that are
provided by states and those that are in high demand by
citizens. This disconnect is most obvious in states with above
average levels of corruption. High corruption states provide,
9
Birth and death certificates, which are offered by 62% of initiatives, is the
only service provided in more than 50% of states.
178
on average, 2.9 of these thirteen services, while below average Table V

corruption states provide 5.2, a difference of means that is Availability of High Corruption Services in Indian Computer-
statistically significant at the .05 level. When the number of Enabled Service Centers
high demand services is regressed on the level of corruption, Reason for Est. % of < Average > Average
the relationship is in the predicted direction and significant at Bribe Annual States Corruption Corruption
the .05 level (t-statistic of -2.33). Bribes Offering States States
($ ‘000)
Table IV Police First 458,000 13% 2 0
Most Common Services offered in Indian Computer-Enabled Info. Report
Service Centers School Fees 340,000 0% 0 0
Department Service (exemption)
Municipal Corp/Rural Dev Grievances Rural 313,000 0% 0 0
Birth & Death certificates Financial
Caste/tribe certificates Loan
Property tax payment Register 307,000 50% 5 3
Food & Civil Supplies Ration cards Property
Revenue Land records School 288,000 0% 0 0
Income Certificates Certificate
Residency certificates Electricity 166,000 13% 2 0
Electricity Electricity bill payment Connection
Estimated annual bribe value based on Transparency
Telephone Company Telephone bill payment International India data [16] and author’s calculations.
Water Water bill payment
Transport Driving licenses In the case of coalition-led states, I also expect to see
Vehicle registration effects of electoral considerations on the specific services that
Home/Police Arms license are chosen for inclusion in centers. Services controlled by
supporting members of a coalition government should be less
Bold indicates that this service is one of the top 25 most
demanded services by citizens, according to [16]. likely to be included than those services controlled by the
majority party.
An alternative measure of citizen demand for service Figure 2 provides an overview of service provision in
computerization is the estimated level of corruption in a coalition-led states. As shown in this graph, politicians from
particular service. Because bureaucrats are more likely to the smaller, supporting members of a coalition are more likely,
demand a bribe for some services than others, service reforms on average, to acquire ministerial posts that involve control
that have the potential to reduce corruption should provide over services commonly offered in service centers. This is in
greater benefits to citizens when introduced for services with a line with expectations that supporting coalition members will
high likelihood of corruption. However reform of these demand ministerial appointments in departments with high
services will also provide the highest costs to bureaucrats and levels of government-citizen interaction. However, when
politicians in terms of lost income, and so computerization in supporting coalition members control these services, it is less
high-corruption potential services should be least likely in likely that they will be included in service centers, as seen
high corruption states. from the right side of the graph. In the case studies below I
When we consider high corruption services, separate from consider the specific ways in which allocation of ministerial
their overall demand by citizens relative to all services, states posts is linked to these lower levels of service provision.
with higher levels of corruption are less likely to provide these
services through computerized service centers. As seen in
Table V, below average corruption states are much more
likely to provide services associated with bribe payments than
states with above average corruption levels. Out of the six
most corrupt services, only one, land records, is available in
any of the above average corruption states.
179
only one, neither of the two ministers from these parties had
the power to bring down the coalition government on its own:
even defection by all five RSP ministers would leave the
coalition with 71 seats, sufficient to maintain its majority. The
remaining services offered through FRIENDS centers were
delivered by departments overseen by CPI(M) ministers.
Among the common services not offered through these
centers, two key departments, Revenue and Food & Civil
Supplies, were overseen by the CPI, the only party with
enough assembly seats to threaten the stability of the ruling
coalition. The other relevant departments, Local
Administration and Rural Development, were both led by
representatives of the CPI(M). The lack of services offered by
departments under the CPI(M)’s most important coalition
partner aligns clearly with predicted outcomes.
Figure 2 – Average Service Provision in Coalition States –
The Akshaya project, a second computer-enabled center
Lead vs. Supporting Parties
initiative in Kerala, offers a similar story of coalition politics,
with an interesting twist. The Akshaya initiative was launched
Kerala
under the United Democratic Front, a Congress Party-led
I begin the case discussion with Kerala. This state is
coalition that came to power in 2001. In this coalition the
viewed by many as an innovator in both development in
Congress held the largest number of seats, at 62, but needed
general [20] and ICT-enabled development in particular [21],
the support of one or the other of its two main coalition
[22]. Despite below average income per capita, Kerala has the
partners, the Muslim League of Kerala or the Kerala Congress
highest literacy rate in the country and the highest overall
(M), to reach the magic number of 71 for a majority.
human development index [23], [24]. Kerala also applied its
Akshaya was launched in 2003 through a pilot project in
development strategies to the use of technology and was one
the state's Malappurum district. The stated goal of the
of the first states to use IT to reform service delivery (Kurian,
initiative was to increase access to technology for the citizens
2003).
of Kerala, and a key part of the initiative was computer
Yet the services offered in Kerala’s two main service
literacy, or “e-literacy,” for the head of every household, a
center initiatives, FRIENDS and Akshaya, pale in comparison
goal that mimicked Kerala’s well-publicized success in
to many of their counterparts in other states. Through
boosting traditional literacy rates (www.akshaya.net).10 But
FRIENDS centers, only ten government services are offered,
these centers provided an additional opportunity for the
with just three offered in Akshaya centers. Could coalition
delivery of government services, as noted by those affiliated
politics be at least partly to blame for the low quantity of
with the project [25]
services in Kerala?
Overall, however, the Akshaya project has implemented
As noted above, analysts have pointed to the importance
next to no services through its network of what in 2008 is
of coalition dynamics in shaping Kerala’s service center
more than 1,000 rural centers. Only three common
initiatives [19]. Once the first service center initiative was
government services are on offer: electricity bill payment,
approved, the FRIENDS initiative, the ruling Left Democratic
telephone bill payment, and grievance redressal. The
Front (LDF) continued to wield their influence over the shape
Electricity Department is overseen by a Congress minister,
of the centers. This coalition was led by the Communist Party
and the Department of Rural Development, which would
of India (Marxist) (CPI(M)), with its coalition partner the
oversee grievances in the rural areas where Akshaya centers
Communist Party of India (CPI) holding the next largest
are located, is headed by a representative of the Kerala
number of seats in the state assembly during the period from
Congress (M).
1996 to 2001. The coalition in total held 76 seats, five more
While it is surprising that the KC(M) would have one of
than the necessary majority of 71. Ministerial posts were
their services offered through the Akshaya centers, it is
allocated across coalition members, with the largest number of
perhaps more telling to note the number of services overseen
ministries predictably allocated to the CPI(M).
by the Kerala Congress (M) that are not offered through these
Six of the fourteen most common services were made
centers. During the UDF government, the KC(M) was in
available through the FRIENDS centers: electricity bill,
control of both the Revenue Department and Rural
telephone bill, water bill, property tax payment, driving
Development, which together account for six of the remaining
licenses, and vehicle registration. Of these services, two of the
commonly offered services, none of which have been provided
relevant departments were overseen by supporting coalition
through Akshaya. Other coalition parties controlled the
members, the Water Department and Transport, led by the
departments overseeing nearly all other common services,
Revolutionary Socialist Party (RSP) and an Independent
member of the legislature, respectively. Because the RSP held
only five assembly seats and the independent, by default, held 10
At nearly 91%, as measured in the 2001 Census, Kerala has the highest
literacy rate of any Indian state or Union Territory.
180
with the Kerala Congress (Jacob) and the Kerala Congress (B) states in the country, with the IT hub of Bangalore as its
overseeing Water and Transport, respectively. Congress capital.
ministers oversaw only two common services that were not On the face of things, Karnataka has also been at the
offered via Akshaya, arms licenses and ration cards. forefront of using ICTs in service delivery. The Bhoomi land
For a portion of these services, the lack of provision records initiative received more media attention than most
through Akshaya is even more surprising, as these services initiatives in the country, and the former Secretary for
had already been implemented at the district level in eGovernance, Rajiv Chawla, is widely recognized as a father
FRIENDS centers. Thus even though the basic infrastructure of the Indian movement to incorporate information
was in place to extend water bill payments, property tax technologies into Indian government processes. Yet, as we
payments, driving licenses, and vehicle registrations through will see, even the presence of early initiatives and an
Akshaya outlets, this was not done under the Congress important evangelist do not guarantee de-politicized service
administration. This means that citizens must either go to the delivery in the state.
district FRIENDS office or avail of these services through A major service center initiative, Bangalore One, was
traditional department-based offices, which are still be prone initiated under the coalition government of the Congress Party
to any corruption in the system. and Janata Dal (Secular) (JD(S)) in 2005. This particular
The twist in the Akshaya case is the way in which the government was formed despite the fact that the BJP held the
policy itself was used as a tool to target the core voters of a largest number of seats in the assembly, because the BJP was
key coalition member party. The Malappurum district where short of a majority and could not agree to a coalition with
the pilot project was initiated has a largely Muslim population, other parties to acquire a majority [27]. So while the Congress
and the Muslim League is the dominant party in the region has the greater number of seats in the coalition, it is highly
[26]. When the UDF coalition regained power in 2001, a dependent on the support of its JD(S) partner.11 Bangalore One
prominent member of the Muslim League, Shri. P. K. was launched in partnership with the municipal government in
Kunhalikutty, became Minister for the Information Bangalore and was intended to provide a similar one-stop
Technology Department. government services environment as offered by Andhra
According to analyst reports, the Minister subsequently Pradesh’s eSeva initiative [28].
consulted with representatives of Malappurum and determined In the case of Bangalore One, however, despite multiple
that information technology could be a “real enabler of the years of experience with technology-enabled service delivery,
local economy” [26]. In order to facilitate IT use, a plan was many of the most commonly offered services are not
developed by the local government to provide for computer available. Of the three services that fall under the domain of
literacy in the district. This plan was subsequently modified in the Revenue department—residency certificates, income
coordination with the Kerala State IT Mission, under the certificates, and land records—none are offered in the
oversight of the IT Minister, and realized as a “telecenter” Bangalore centers. The exclusion of land records is especially
initiative for the district [26]. If successful, this pilot would surprising, given that the Bhoomi land records initiative was
then be extended to other districts in the state. So while the viewed as such a ‘success’ in the state.
stated intention was to develop an initiative that could Why are these common and highly valued services not
eventually benefit the entire state, the initial benefits would offered in Karnataka? The most plausible answer again lies in
clearly be centered on the core constituency of the standing IT coalition politics. The minister in charge of Revenue was a
Minister, representing a key ruling government coalition politician from the supporting JD(S), who also held the post of
member. Deputy Chief Minister, and so was clearly an important player
Akshaya, then, provides an important example of the in maintaining the coalition. Ration cards are also unavailable
combined effects of coalition rule. First, a second computer in the centers, and it was a minister from the JD(S) who
center initiative in the state was launched for the apparent oversaw the responsible Food & Civil Supplies department. Of
purpose of rewarding key Muslim League constituents, a clear those common services that were offered, the JD(S) was
boon for the holder of the IT Ministry post. Second, the lack responsible for only one, electricity bill payment, while the
of services offered through Akshaya reflects the prevalent use Congress was responsible for the remaining services, via
of ministerial posts by the UDF government to reward Municipal Administration, the department that was also
coalition members. Even those services that should be involved in the implementation of the initiative itself. So even
relatively easy to implement because of their inclusion in the in a state with a strong emphasis on technology and history of
earlier FRIENDS initiative have, for the most part, not been government IT initiatives, the dynamics of corruption and
made available to Akshaya patrons. coalition politics seem closely linked to the selection of
specific services made available in the state’s one-stop centers.
Karnataka
The next state, Karnataka, has a much higher level of Rajasthan
corruption than Kerala, which would, on its own, imply a
stronger emphasis among coalition members on the economic 11
The coalition eventually collapsed in early 2006, when a rebel JD(S)
side-benefits of ministerial posts. But Karnataka is also a leader, H.D. Kumaraswamy, pulled out of the government with a group of
other JD(S) MLAs in order to form an alternative coalition with the BJP
“hard” case, because it is one of the most technically advanced (Rediff, 2006).
181
In Rajasthan, the LokMitra (urban) and JanMitra (rural) service delivery. The state of Chhattisgarh was formed in the
centers were opened in early 2002 under the leadership of the year 2000 when it was carved out of Madhya Pradesh. Despite
Indian National Congress party. During this period the below average levels of economic development and a
Congress party held a majority of the seats in the state persistent and violent separatist movement in part of the state,
assembly and so did not require the support of any other the Chhattisgarh government has implemented a substantial
parties in a coalition. The initiation of service centers was eGovernment program through its Choice service centers.
largely an enterprise of the Chief Minister Ashok Gehlot, who These centers, which to date are largely located in urban
sent representatives of the state to Andhra Pradesh in 2001 to centers, provide one of the largest sets of services of any state
evaluate the eSeva initiative there and determine how in India.
something similar could be used to improve service offerings Why is it that Chhattisgarh has implemented such a robust
to citizens in Rajasthan [29]. According to one of the service center initiative? Here again we see the important role
bureaucrats who participated in that mission, a decision was played by Chief Ministers in the character of public service
made to improve on the Andhra model by offering services in reforms. Choice service centers were initially implemented
both urban and rural areas, through two separate initiatives, at under the first Chief Minister of the state, Ajit Jogi of the
least at the beginning.12 Indian National Congress. However the growth of the project
The choice of services in each set of centers was then from a pilot stage has occurred under BJP Chief Minister
determined based more on urban versus rural needs, with the Raman Singh.
LokMitra centers focused on bill payments and JanMitra According to government officials associated with the
centers emphasizing grievances and non-government services Choice initiative, Chief Minister Singh encouraged an
such as agricultural prices. However, in rural areas only four inclusive decision-making process in which the bureaucrats in
government services were offered, while in urban areas just charge of major citizen-facing departments would participate
seven services were made available. in an “empowered committee” that was tasked with making all
The availability of specific services, particularly in urban key decisions about the project moving forward. The goal of
areas, seems directly linked to the interests of the Chief this model was to ensure consensus on decisions so that no
Minister. When bureaucrats in charge of implementing participants would later attempt to block implementations of
LokMitra services attempted to convince various departments the centers or any specific services [30]. In addition, the Chief
to provide their services in the centers, they faced significant Minister took over leadership a “governing council” made up
resistance. The state telephone company did not want to allow of government ministers, which is responsible for overseeing
outsourcing of its bill payments through the computerized the implementation of the project. This meant that the
centers, and the bureaucratic officers found it necessary to implementation team could go directly to the Chief Minister in
request the intervention of the Chief Minister. In this case the case of any problems with implementation. In no other state,
Chief Minister did intervene and the Telephone company was other than perhaps Andhra Pradesh, has the Chief Minister
forced to provide services for bill payment through the taken on this type of direct oversight role in the development
LokMitra centers. On other occasions, however, such as in the and implementation of service centers.
case of income tax payments, the Chief Minister was What factors might have influenced the Chief Minister to
unwilling to take similar initiative and income tax payments take such a strong position on eGovernment in the state?
were not included in the services offered by the centers [29]. Corruption does exist in Chhattisgarh, but government
The mixed response of the Chief Minister in the case of representatives argued that the expectation of the Chief
Rajasthan may be closely tied to the high level of corruption in Minister was that improvements in service delivery would
public service delivery in the state. As one bureaucrat provide a greater potential electoral boost to the ruling
involved in the initiatives noted, politicians “want to provide government that any threat from reduced access to rents. “The
the maximum services to their vote bank,…[but] they also Chief Minister sees the benefits as greater than the costs. He
want money for the next election” [29]. Thus from the has been the brainchild behind all of these frameworks. Indian
perspective of the Chief Minister, by providing some services,
politicians…have very sharp political minds. The Chief
especially those that do not typically involve high levels of
Minister in this term is focused on good governance, on
corruption, such as bill payment, there is an opportunity to
accountability, transparency, and responsiveness of the
reap some political benefits from the centers. At the same
time, those services that may more often involve the payment government” [31]. Given the difficulties that the government
of bribes, such as income tax, can be restricted from inclusion otherwise faces in areas of development and social stability, it
in the centers in order to maintain the rents from those seems that the Chief Minister has adopted service reforms
services. In this way we can see how high levels of corruption through the use of technology as an important platform for
may affect service delivery in single party states. delivering valued goods to citizens. This is possible, at least in
part, because he does not likely feel the same threat to
Chhattisgarh campaign resources from increased government transparency
The final case to consider is Chhattisgarh, a single-party-led that exists in higher corruption states.
state that has below average levels of corruption in public
IV. CONCLUSION
12
JanMitra and LokMitra were merged into a single initiative, eMitra, in The evidence presented here provides strong support for an
2004. Both Jan Mitra and Lok Mitra are Hindi variations on the term argument about the role of corruption and coalition dynamics
“people’s friend”.
182
in affecting which citizens benefit from technology-enabled [18] Government of Haryana, IT Department official, February 4, 2008.
[19] G.R. Kiran, “Front-End First: Citizen Payment at FRIENDS Centres in
service delivery. Citizens in more corrupt states are less likely Kerala,” UNPAN eGovernment for Development Success/Failure Case
to benefit from new technologies. This is because even if Studies, Number 17, 2002.
service centers are implemented in their state, these centers are Available: http://www.egov4dev.org/friends.htm
[20] P. Heller, The Labor of Development. Ithaca: Cornell University Press,
likely to provide fewer and lower demand services than those
1999.
centers in other states. [21] S. Madon, “Governance lessons from the experience of telecentres in
Citizens may also be at a disadvantage depending on the Kerala,” European Journal of Information Systems, Vol. 14, pp. 401-
characteristics of their ruling government. In coalition-led 416, 2005.
[22] S. Madon, and G.R. Kiran. Information technology for citizen-
states, state governments will chose services that ensure the government interface: a study of FRIENDS project in Kerala. World
stability of the government before considering what services Bank Global Knowledge Sharing Program (GKSP), 2002.
might benefit voters. When MLAs from supporting parties [23] Reserve Bank of India; GOI, 2001
[24] Reserve Bank of India; GOI, 2005.
serve as ministers for departments, the services of those [25] Kerala Government, IT Department official, June 22, 2004.
departments are less likely to be implemented, particularly if [26] G.R. Kiran, “Akshaya, Malappuram, Kerala” in Information and
that party holds enough seats to disrupt the majority of the Communication Technologies for Development: A Comparative Analysis
of Impacts and Costs from India,” Bangalore: International Institute of
coalition. Information Technology, 2005.
Who benefits from technology-enabled service centers is [27] Rediff.com, “JD-S, Congress Team up in Karnataka,” May 16, 2004.
thus highly dependent on the political characteristics of a state Available: http://ia.rediff.com/election/2004/may/16karna.htm
[28] Karnataka Government, eGovernance Department official, February 22,
and in particular the extent of corruption and cohesion in the 2006.
ruling government. While there may be great potential to [29] Government of Rajasthan, former IT Department official, May 7, 2007.
improve service delivery through one-stop service centers it is [30] Government of Chhattisgarh, IT agency official, January 23, 2009.
[31] Government of Chhattisgarh, IT Department official, January 23, 2009.
clear that, at least to date, the actual benefits to citizens are
politically driven and thus, in the case of the Indian states,
highly varied.
REFERENCES
[1] World Bank, World Development Report: Making Services Work for the
Poor. Washington, D.C.: The World Bank and Oxford University Press,
2004.
[2] World Bank, ICTs and MDGs: A World Bank Group Perspective.
Washington, DC: The World Bank Group, 2003.
[3] Accenture, Markle Foundation, UNDP, Creating a Development
Dynamic: Final Report of the Digital Opportunity Initiative. 2001.
[4] R. M. Davison, C. Wagner, and L. C. K. Ma, “From government to e-
government: a transition model,” Information Technology and People,
Vol. 18(3), pp. 280-299, 2005.
[5] United Nations, World Public Sector Report: E-Government at the
Crossroads. New York: United Nations Publications, 2003.
[6] D. West, Global E-Government 2005. 2005. Available:
http://www.insidepolitics.org/egovt05int.pdf
[7] J. Bussell, “Electoral Competition and Digital Development in India and
South Africa,” Proceedings of the IEEE/ACM International Conference
on Information and Communication Technologies for Development 2007
(ICTD2007), 2007.
[8] K. De Tolly, B. Maumbe, and H. Alexander. “Rethinking E-Government
for Development: Issues, Lessons and Future Prospects for the Cape
Gateway Portal in South Africa,” presented at the IST-Africa
Conference, Pretoria, South Africa, May, 2006.
[9] S. Madon, “Evaluating the Developmental Impact of E-Governance
Initiatives: An Exploratory Framework,” The Electronic Journal on
Information Systems in Developing Countries, 20(5), pp. 1-13, 2004.
[10] J. Satyanarayana, EGovernment…the science of the possible. New Delhi:
Prentice-Hall, 2004.
[11] R. Heeks, “Causes of eGovernment Success and Failure: Factor Model,”
2003. Available: http://www.egov4dev.org/causefactor.htm
[12] S. Bhatnagar, “Lessons from eGovernment in Developing Countries”
Regional Development Dialogue, No. 24, 2002.
[13] Government of West Bengal, IT Department official, January 18, 2008.
[14] R. Wade, “The Market for Public Office: Why the Indian State is not
Better at Development,” World Development, Vol. 13(4), pp. 467-497,
1985.
[15] F. De Zwart, The Bureaucratic Merry-go-round: Manipulating the
Transfer of Indian Civil Servants. Amsterdam: Amsterdam University
Press, 1994.
[16] Transparency International India, India Corruption Study 2005. New
Delhi: Transparency International India, 2005.
[17] Government of Orissa, IT department official, January 29, 2008.
183
Results from a Study of Impact of E-

government Projects in India
Subhash C. Bhatnagar and Nupur Singh
Abstract—The paper presents the results from an assessment are making ICT investments in the public sector. On the other
study of eight e-government projects from India. The assessment hand, evidence of failed projects has drawn attention to the
framework measured the total value delivered by a project to level of risk involved in implementation. A failure rate of
various stakeholders on three dimensions: (a) cost to clients for
accessing services; (b) perception of quality of service and
more than 50 percent is widely cited in this context [1].
governance; and (c) agency cost and revenue. Data was collected A study commissioned by DFID [2] researched firm-level
from randomly selected users encapsulating their experience of impact of ICT in developing countries, 2 but similar studies
using the computerized and manual systems. Impact was have not been undertaken for the public sector. A report by
estimated as the difference between the rating of the the United Nations (2003) laments the fact that documented
computerized and the manual systems. research on the social or economic impact of e-government
development is virtually non-existent [3]. Multilateral
Citizens indicated an overwhelming preference for organizations such as the World Bank fund nearly seven
computerized service delivery. The number of trips required to
be made to the concerned offices reduced significantly and
billion dollars of ICT investment as part of their lending
waiting time came down by nearly fifty percent. Overall impact programs and grant assistance to various client countries. A
showed wide variations across projects. Of the seven agencies World Bank report noted that the largest yet the least
where corruption was reported in the manual system, five monitored investments are IT components of projects in
services saw significant reduction through computerization but different sectors, highlighting the relevance of systematic
the impact was marginal in three projects. Government of India assessment of the impact of these applications [4].
has adopted the framework used in this study to assess the
impact of 40 projects implemented at the national, state and
This paper presents an early effort at evolving a systematic
local levels. framework and methodology for assessing the impact of e-
government projects based on a review of past efforts at
Paper discusses the implications of the results for further assessment of e-government projects. Sections II and III
development of e-government projects in India and identifies the discuss the review and present the key features of the
limitations of the study. framework. The paper goes on to report the findings from a
study that used the framework to ascertain the impact of a
Index Terms—E-government, governance, impact assessment, selection of eight e-government projects from India.
India
II. EFFORTS TO ASSESS PUBLIC SECTOR IT PROJECTS
I. BACKGROUND While significant amount of academic and policy research
has focused on evaluation of public sector ICT projects, a
I N recent years a large number of countries have launched
“e-government” programs, and several development
agencies and governments have identified e-government
systematic framework for evaluation has not emerged. One
group of studies provides macro-level estimations of e-
government activity using appraisal indices focusing on
implementation as a key policy priority. 1 Driven by the supply-side, quantifiable measures such as web presence of
success of a few projects in improving delivery of services to government, network coverage, institutional and regulatory
citizens and businesses, an increasing number of governments support and human capital provision. 3 Such factor-based
assessments of e-readiness do not have immediately obvious
Manuscript received February 5, 2009. and tractable policy implications. They tend to focus almost
S. C. Bhatnagar is with the Indian Institute of Management, Ahmedabad,
India (phone: 91-79-6632-4802; fax: 91-79-6632-6896; e-mail:
2
subhash@iimahd.ernet.in). The study sponsored by DFID at the London Business Schools collected
N. Singh is with the Centre for E-Governance, Indian Institute of data from firms in India and Brazil to establish a relationship between ICT
Management, Ahmedabad, India (e-mail: nupurs@iimahd.ernet.in). investments, profitability, growth and productivity at the firm level. The
1
Examples of e-government initiatives launched by international agencies degree of collateral organizational change was studied as a mediating
can be found on the following links: World Bank variable.
3
(http://www.worldbank.org/egov); ITU (http://itu.int/itudoc/itu-t/workshop/e- Examples of such frameworks are (1) UNPAN (2004) E-Government
gov); and DFID (http://www.publictechnology.net/modules.php? Readiness Report www.unpan.org and (2) Brown University (2004). Global
op=modload&name=News&file=article&sid=1533) E-Government Report, available on www.insidepolitics.org/egovt05int.pdf
184
exclusively on measurement of physical access to certain creation in e-government projects have been documented in
types of ICT without incorporating issues such as reports by governments and external consultants [6] – [8]. An
affordability, appropriateness, ICT capacity and training, and example is the study conducted jointly by the Danish Ministry
the regulatory and macroeconomic environment [5]. The of Finance and Accenture which identifies two primary
second group of studies, largely anecdotal, done in a dimensions of value: (i) value to investors in terms of tangible
piecemeal fashion, provide project-level evaluations with little financial benefits, cost savings, cost avoidance, and increased
prospect for synthesis from past approaches. Evaluative revenue; and (ii) value to users in terms of improved services,
studies have been done to serve a variety of purposes. Some reduced cost and/or time savings to citizens, and reduced
studies looked at implementation success in terms of whether administrative burden to businesses. Investment in
the systems were functioning as they were designed to, or the government employees and technological infrastructure are
degree to which the intended outcomes were achieved. Some identified as key enablers for creating value through e-
studies looked at long term sustainability and scope for government. The focus has been on using business case
replication, 4 while some measured the benefits that were methodologies from the private sector to demonstrate the
delivered to agencies. 5 A few focused on benefits to the economic value of e-government projects. The calculation of
clients. 6 There was hardly any comprehensive study that time and money spent in finding and using public information
assessed the impact on all the stakeholders and covered both are the most direct and measurable benefit of e-government
short-term and long-term direct and indirect impacts. A few applications to users, as shown in recent OECD and EU
studies had carried out a cost-benefit analysis. Often, studies [9],[10]. These reports identify improved revenue
evaluation studies had been done by agencies that were likely collection, lower costs due to efficient processing of
to have an interest in showing a positive outcome. transactions, and a reduction of administrative burdens due to
A variety of approaches have been used for evaluation. simplification or elimination of procedures as some direct
These included surveys, expert opinion, ethnographic studies impacts that can be measured. Financial savings to users in
and internal assessments produced by lending agencies. The terms of time and money spent in finding and using public
utility of such evaluations has been limited because: information are the most direct and measurable benefits of e-
government applications for clients. A significant amount of
▪ Different studies of the same project showed very
work has also been done on studying users’ perceptions of
different outcomes, thus indicating a lack of credibility of
results. 7 Part of the reason for different outcomes was the quality in terms of attributes such as accessibility, attitude of
use of very small samples and a lack of rigor in sampling staff, cost of service, provision of information, procedural
and collecting data from clients of the systems. The fairness and convenience.
results could therefore not be easily generalized over the Calculation of net economic benefit of an IT project has
entire population of clients. proven to be challenging as outcomes are multi-dimensional
and composed of both quantitative and qualitative indicators.
▪ The studies evaluated the functioning of the Further attention is required to be given to the linkages
computerized system but were not able to assess the amongst issues of quality of service, governance, wider
difference made by ICT use, as the need for
impacts on society, and ICT investment. Newer frameworks
counterfactuals (evaluation of systems as they worked
are evolving around a notion that an exclusive focus on
before computerization) was ignored. Often, the impact
financial costs and benefits for the government ignores many
of ICT use was not separated from other interventions
that were made simultaneously with the computerization important non-economic benefits [9]. One such benefit relates
effort. to improvements in various aspects of governance activity.
Another non-economic benefit relates to the addressing of
▪ Finally, since different studies did not use a standard wider policy priorities, which, in the context of developing
methodology, it was difficult to compare the outcome of countries could constitute the priorities as articulated in the
one project with that of others. UNDP Millennium Development Goals.
Work done in EU countries has focused on understanding III. MEASUREMENT FRAMEWORK AND METHODOLOGY USED
the ‘processes’ that deliver a return on investment and IN THE STUDY
generate value [6]. Common best practice factors for value
Two approaches - MAREVA developed by the ADAE
4
(Agence pour le Development de l’Administration
Government of India, Ministry of IT has commissioned quick assessment of
29 projects. Electronique - Electronic Administration Development
5
Korea’s eProcurement agency has evaluated the impact on different Agency) in France with the help of Bearing Point and the
government agencies using the system. WiBe Economic Efficiency Assessment methodology being
6
Global Knowledge Sharing Program got 4 Indian projects evaluated where
clients were surveyed.
used by the German federal administration [11] were useful in
7
For example the Bhoomi project of issuing copies of land title has been developing the framework proposed in this chapter. These
evaluated by Public Affairs council reporting a significant positive outcomes methodologies developed by two Governments in EU
including reduction in bribes. Recent studies by a team from MIT and IIIT
Bangalore found that corruption had not declined and major benefits were
countries focus broadly on the same dimensions. They offer
derived by land sharks. two levels of impact assessment: first, in terms of how the
project provides a business case justification of expenditure
185
and second, in terms of how the project meets the goals of the of existing frameworks has been used in this study. An e-
agency concerned and, in turn, how this helps in meeting service delivery project impacts three groups of stakeholders:
wider government strategies. Guiding these assessment (i) clients receiving the service; (ii) agency (including
frameworks are strategic outcomes pursued by respective implementation partners) that delivers the service; and (iii) the
governments representing broad policy priorities that drive larger society consisting of citizens, businesses, government
the direction of government. These considerations have been as a whole and civil society. The impact can be assessed in
incorporated in the assessment framework proposed for e- terms of a variety of outcomes experienced by each type of
government projects in this section. 8 stakeholder. Table I lists key dimensions of outcomes for each
Borrowing from the above approaches and the work done type of stakeholder.
in EU countries, the framework used for the study reported in TABLE I
KEY OUTCOME DIMENSIONS
the paper focused on the idea of measuring the total value
Stakeholders Key Dimension of Impact
delivered by a project to different stakeholders and takes a
Client ▪ Economic (Direct and indirect)
balanced approach between case study and quantitative ▪ Governance (Corruption, accountability,
analysis. It recognizes that some part of the value for each transparency, participation)
stakeholder can be monetized and other part needs to be ▪ Quality of service (Decency, fairness,
assessed qualitatively. convenience, etc.)
Most of the assessment models discussed in the literature Agency (Including partners ▪ Economic (Direct and indirect)
in implementation) ▪ Governance (Corruption, accountability,
have not been used in the context of a developing country and transparency, participation)
many have not yet been applied in practice even in developed ▪ Performance on key non-economic
countries. Nor do they account for a variety of delivery objectives
▪ Process improvements
models used in developing countries such as common service
Society ▪ Long term impact on Millennium
centers and franchised outlets that can retail e-services offered Government as a whole Development Goals
by the government. Practical issues of paucity of data have ▪ Image of the government
not been taken into account, particularly in case of a
developing country where baseline surveys are not done and The primary objective of the study was to measure the
monitoring and evaluation systems are weak. Recognizing impact of computerization on clients (users) of selected
that adequate data for quantitative assessment is not available service delivery projects and to test the applicability of the
for most of the projects, collection of such data using a framework across a variety of projects. For the purpose of this
standard measurement framework for sufficiently large study a sample of eight mature projects shown in Table II was
number of projects is the first task that needs to be selected. These projects covered services to rural as well as
undertaken. Since impact assessment is the key objective, urban citizens (G2C), services to businesses (G2B) and
establishing counterfactuals is an important element of the services for internal government users (G2G). Most of the
proposed measurement framework. services are offered by state-level agencies except in eSeva
A number of empirical studies suggest that ICT has had an where services from the federal government are also offered.
impact in improving the performance of private sector In eProcurement services are accessed through a portal. In the
organizations particularly in developed countries. However, remaining projects service delivery is through assisted
regarding ICT investments by the public sector in developing computerized counters set up by the agencies at various
countries, many researchers have noted that past evaluation locations.
studies have not used a common framework or methodology For each project, the measurement framework was
and that rates of success/failure have been declared based on converted into a set of data collection instruments including:
purposive samples [12]. Since ICT is introduced mostly in the i) a profile of the project identifying services, clients and other
context of governance reform to improve efficiency, stakeholders; ii) agency level data on activity levels,
effectiveness, and transparency of governments, a crucial first investments and operating costs; iii) a client survey
stage is to ascertain to what extent these intended outcomes questionnaire covering direct cost of access, quality of service
from e-government applications have been achieved. and governance, and a few measures of overall satisfaction;
A common measurement framework evolved on the basis and iv) an employee survey for perceived impact on work,
efficiency and effectiveness.
8
The indicative items are based on a review of the following documents: The survey assessed both the manual system and the
i) Performance Reference Model of the US Federal enterprise Architecture computerized system that replaced it on all the above
framework used by the office of Management and Budgets in US Federal
Government; ii) European Commission, eGovernment Economics Project dimensions. An analysis of the differences between the old
(eGEP), Measurement Framework Interim Version Deliverable (D.2.2), 2005; and the new system provided a measure of impact. Random
iii) TP Rama Rao, V Venkata Rao; SC Bhatnagar; J Satyanarayana, ‘e- samples of about 30 users were chosen from 8 different
Governance Assessment Frameworks’, 2004, http://egov.mit.gov.in; iv)
Edwin Lau, ‘Electronic Government and the drive for growth and equity’, locations in a state. Locations were chosen to represent
OECD, 2005; and v) Lanvin Bruno, ‘METER E-strategies Monitoring and different levels of activity at the service centers, and different
Evaluation Toolkit’, in Robert Schware (ed), E-Development: From levels of development of the region. Thus nearly 250
Excitement to Effectiveness, Prepared for the World Summit on the
Information Society Tunis, 2005, World Bank.
respondents were surveyed for each project. The
questionnaires (translated into local languages) were
186
administered at the home of the sampled users by request in all applications. 9 This has resulted in increased
investigators who were trained to understand the nature of predictability in service delivery and reduced the number of
projects and to interpret individual items in the instrument. trips.
The survey was combined with secondary documentation and 4.00
primary data gathered through field visits, interviews, surveys Manual Computerised
3.37
and opinions solicited through e-mail. 3.50
TABLE II
3.00 2.81
LIST OF EIGHT PROJECTS FOR DETAILED STUDY 2.71 2.74
Project Description 2.42

2.50 2.32
2.20
Bhoomi Online issue of an RTC (among other uses, this 2.12 2.09
document is required for availing crop loans from 2.00 1.85
banks and as surety for bail) and filing of requests for 1.62
1.54
mutation for affecting changes in land record in 1.50
1.43 1.41
Karnataka at 203 Kiosks. Launched in February 1.22
1.13
2001.
1.00
Karnataka Valuation Key services delivered by 201 Sub Registrar’s
and E-Registration Offices in Karnataka are: online registration of 0.50
(KAVERI) property sale/purchase deeds; issue of non-
encumbrance certificate; and issue of copies of
0.00
registered deeds. Launched in December 2003. Bhoomi KAVERI Khajane DDO Khajane CARD eProcurement eSeva AMC
Khajane Networking and computerization of all treasuries Payee
across Karnataka. Launched in November 2002.
Computer Aided Online registration of property sale/purchase deeds; Saving 0.47 1.20 1.08 0.90 1.38 0.86 0.28 0.65
Administration of issue of non-encumbrance certificate and issue of S.E. 0.15 0.12 0.09 0.10 0.11 0.26 0.09 0.14
Registration copies of previously registered deeds at 387 Sub
Department (CARD) Registrar’s Offices in Andhra Pradesh. Launched in Fig. 1. Number of trips. A paired t-test was applied to assess whether the
November 1998. differences between the number of trips for the computerized and manual
eProcurement Online tendering for goods and services by systems were significantly different.
government departments and agencies in Andhra
Pradesh. Launched in January 2003. A reduction in the number of trips implies a reduction in
eSeva One stop service centers delivering 135 services from travel costs. The only exception is the Bhoomi project where
central, state, local governments and public utilities.
Used monthly by 3.1 million citizens at 275 locations
travel costs have increased primarily because the location of
across 190 towns. Launched in August 2003. office that issues the RTC has been shifted from the village
Ahmedabad 16 Civic Centers of AMC primarily deliver three level to a taluka 10 level government office. The opportunity
Municipal services: annual collection of property tax; issue of cost for users, which is represented by wage loss due to travel,
Corporation Civic birth and death certificates; and issue of shop
is therefore, larger. However, 800 rural kiosks are being put
Centers (AMC) licences. Launched in September 2002.
up by a private operator in Karnataka to issues RTCs. These
Computerized Inter- 10 computerized check posts use electronic weigh
state Check Posts bridges to levy fines for overloading and over- kiosks will access Bhoomi’s central server (where all records
dimensioning of commercial vehicles passing on the day’s transactions in various taluka level kiosks are
through them, inspect vehicles to check for damaged updated) to issue RTCs, which would reduce the customers’
headlights and non-standard license plates, and verify
essential documents. Launched in March 2000. cost of access.
Waiting time at the service centers has reduced in all the
projects (Fig. 2). Among those projects in which the reduction
IV. RESULTS OF CLIENT IMPACT ASSESSMENT in waiting time were statistically significant, the savings
This section presents an analysis of the eight projects from ranged from 16.16 minutes (AMC) to 96.24 minutes (CARD),
the perspective of the clients. Impacts for the key dimensions amounting to nearly 50 percent reduction from the waiting
of cost, quality, governance and overall satisfaction are time in the manual system. Reduction in waiting time has
analyzed. direct impact on user costs through reduction in foregone
wages as indicated in Table III.
A. Cost to Users
In all the eight projects, e-government has reduced the
number of trips users have to make to complete all
transactions for a service (Fig. 1). Greater formalization of
processes after process reform, quicker retrieval of data from
computerized databases, and automation of tasks such as
writing and copying documents, as compared with manual 9
In the case of AMC, the time elapsed has reduced from 9.8 days to 5.3
methods, has reduced the total processing time of a service days for certain types of transactions while in CARD, the total time required
for registrations has come down from 4.69 to 1.82 days.
10
Taluka (or sub-district) is an administrative division consisting of a city
or town that serves as its headquarters, possibly additional towns, and a
number of villages.
187
200.00 reducing additional work for the staff.

184.79
Manual Computerised 5.00
180.00
162.49 Manual Computerised
4.51 4.50 4.51
160.00 4.50 4.27
4.18 4.19 4.22
4.11
4.03
140.00 3.92
4.00 3.79 3.81
114.95 3.63 3.65
120.00 3.55
102.03 100.68 3.50
100.00 87.78 3.08 3.06
80.00 3.00
68.06 2.75
63.16 60.06
60.00 2.50
40.00 32.96 29.32 29.66

21.76 24.67 20.73 2.00
14.47 14.63
20.00
0.00 1.50
0.00
Bhoomi KAVERI Khajane DDO Khajane CARD eProcurement eSeva AMC Checkpost
1.00
Payee
Bhoomi Kaveri Khajane DDO Khajane CARD eProcurement eSeva AMC Checkpost
Payee
Saving 41.21 62.91 41.40 35.40 96.24 114.95 18.50 16.16 8.87
S.E. 9.15 7.00 6.52 4.44 7.95 7.58 1.64 1.58 1.82 Diff. 0.95 0.32 0.40 0.55 0.48 0.27 0.95 0.70 0.57
S.E. 0.08 0.04 0.07 0.07 0.05 0.05 0.04 0.05 0.04
Fig. 2. Waiting time (in minutes). A paired t-test was applied to assess
whether the differences between the waiting time in case of computerized Fig. 3. Overall service quality score (5-point scale). A paired t-test was
and manual systems were significantly different. applied to assess whether the differences between the overall service quality
scores for the computerized and manual systems were significantly different.
TABLE III
WAGE LOSS (RS) FOR MANUAL AND COMPUTERIZED SYSTEMS
16
Bhoomi Khajane
KAVERI CARD eSeva AMC 14.94
RTC Mutation - Payee 14
12.79
Wage loss 61.70 55.44 382.93 66.67 190.21 80.63 125.18 12 11.81
(Rs) Manual
N 110 91 123 9 68 27 34 10
S.E. 8.45 4.85 67.68 8.82 32.90 15.35 23.94 8 8.26 8.02
Wage loss 98.25 94.66 262.38 36.25 161.75 65.00 88.33
6
(Rs)
Computerized 4.74
4 3.88
3.31 3.49 3.56
N 81 59 122 8 63 22 24
2 1.98 1.67
S.E. 10.69 13.50 24.38 6.73 32.79 22.55 20.81 1.52 1.14 1.26
0.74
N: Sample size 0
Bhoomi-RTC Bhoomi- KAVERI Khajane- Khajane- CARD eSeva AMC
While the difference between manual and computerized Mutation DDO Payee
systems on all the cost indicators are statistically significant Manual Computerised
(in most cases at a confidence level of 99 percent), the
accuracy of the estimate of the difference in direct costs can Fig. 4. Error Rate (in percentage)
be improved by using larger samples.
C. User Perception of Governance and Corruption
B. User Perception of Service Quality
The proportion of users paying bribes has generally
For all eight projects, users rated the computerized services
declined. The decline is higher in some projects than in others
higher in overall quality (Fig. 3). The differences were
(Fig. 5). For example, in the Bhoomi project, while about 30
statistically significant. eSeva has shown a significant
percent of users were paying bribes in the manual system, less
improvement in service quality whereas eProcurement has
than one percent is continuing with the practice in the
shown a marginal improvement. Bhoomi also shows very
computerized system. The proportion of transactions in which
significant improvement in service quality, amongst the
a bribe was paid in KAVERI continues to be high at 21.61
highest in the projects assessed in the study.
The rate of error in documents (as reported by the users), percent in the computerized system. Out of the seven agencies
which is an important measure of quality, was also lower for (and eight projects) where significant corruption was reported
in the manual system, four were able to eliminate or
all the projects except the issue of RTCs (see Fig. 4). The
significantly reduce corruption through computerization. The
number of RTCs issued has gone up many-fold after
impact was marginal in the remaining three agencies as
computerization and even minor mistakes in the name are
corruption continued at a significant level even in the
now being noticed and reported. Reduction in errors saves
computerized systems. Although, in reducing corruption the
additional trips for users and improves productivity by
188
outcome is mixed, e-government seems to have the potential D. Comparison of Projects on Overall Client Impact
for significant reduction in corruption in service delivery, as Respondents were asked to rate each project on 18 common
shown by one project. attributes encompassing the three dimensions discussed
40.00
earlier. They were also asked to select the three attributes that
Manual Computeris ed they considered most important. A weighted composite score
35.00
33.82 34.32 that factors in the importance and rating of the attributes was
calculated for each project. Table IV compares the composite
30.00 29.71
28.02 scores across projects on a 5-point scale. It indicates that
25.00 Bhoomi shows the maximum improvement as its manual
23.71
21.61 version had the lowest score. eSeva also shows a very
20.42
20.00
significant improvement as the computerized counters are
15.00 14.48
rated close to very good in the composite score. KAVERI and
14.17
CARD indicate only marginal improvement over the manual
10.00 system. Ranking of projects using the composite rating, and
5.00 5.68 particularly its improvement, can represent the degree of
2.71 3.35 success of the project from the point of view of clients.
0.00 0.84 0.74 0.00 0.00 0.40
0.00 0.84
Bhoomi and eSeva can be rated as very successful projects,
Bhoomi-RTC Bhoomi- Kaveri Khajane DDO Khajane CARD eProcurement eSeva AMC Checkpost AMC as moderately successful and KAVERI and CARD as
Mutation Payee
projects where there is considerable scope for improvement.
Fig. 5. Proportion paying bribes (percentage). TABLE IV
DESCENDING ORDER OF IMPROVEMENT IN COMPOSITE SCORES
An interesting phenomenon is that the amount of bribe Manual S.E. Computerized S.E. Difference S.E.
users pay is generally higher in computerized systems in all Bhoomi 2.86 0.07 4.46 0.04 1.60 0.09
the projects. This can perhaps partly be attributed to general eSeva 3.39 0.04 4.66 0.03 1.27 0.05
inflation in the price levels. It may also suggest that the Khajane DDO 3.24 0.08 4.43 0.05 1.19 0.10
complexity of the types of favor requested through bribes has Khajane
3.08 0.07 4.19 0.05 1.10 0.10
increased. It must be noted that the estimates of bribe amounts Payee
are less accurate than other results as the sample of eProcurement 3.22 0.04 4.26 0.04 1.04 0.05
respondents paying bribes was small resulting in high Checkpost 3.48 0.05 4.32 0.04 0.84 0.05
standard error of the estimates. AMC 3.37 0.04 4.12 0.06 0.75 0.06
The overall governance rating is higher for computerized KAVERI 3.35 0.06 3.89 0.05 0.55 0.05
systems than manual systems (Fig. 6). In the case of Bhoomi, CARD 3.78 0.03 3.93 0.03 0.15 0.03
there was a marked improvement in transparency and fairness
whereas KAVERI showed hardly any improvement in the As shown in Table V, three kinds of attributes have been
quality of governance. selected more often than others. They include attributes
5.00 pertaining to: i) transactional efficiency; ii) improved
Manual Computerised
4.48
governance; and iii) quality as measured by error rate and
4.50
4.20 4.19
4.31
4.21 convenience. An important conclusion from this analysis is
4.03
3.94 3.99 that important attributes are different for different projects. If
4.00
3.58 3.61
3.74 such an exercise was to be done before designing an
3.50 3.51
3.50 3.25
3.34 3.33 3.33 application it would provide useful insights into benefits that
3.15
need to be targeted and the kind of process reform required. It
3.00 is rarely done in practice either before or after the project
implementation.
2.50
2.00
1.50
1.00
Bhoomi Kaveri Khajane DDO Khajane CARD eProcurement eSeva AMC Checkpost
Payee
Diff. 0.76 0.19 0.70 0.61 0.61 0.38 0.79 0.75 0.88
S.E. 0.07 0.04 0.06 0.06 0.04 0.04 0.04 0.04 0.06
Fig. 6. Overall governance score (5-point scale). A paired t-test was applied
to assess whether the differences between the overall governance scores for
the computerized and manual systems were significantly different.
189
TABLE V with manpower, as was seen in the case of KAVERI. The

TOP FOUR DESIRED ATTRIBUTIONS OF SERVICES
human resource and other costs incurred by the government
Project Desired Attribution of Services
have remained within a narrow range for the manual and
Bhoomi Error free No delay in Less waiting Fewer visits
transaction
computerized systems. In terms of the impact on agencies, the
transaction time
KAVERI Less Greater Error free
ability to cope with growth in transactions was enhanced in all
Less waiting
corruption transparency transaction time cases. In some cases, computerization helped in the growth of
Khajane DDO Simplicity of Convenient Friendly Error free transactions.
procedures time schedule attitude of transaction Data from employees on perceptions about impact on work
officers load and managerial processes was analyzed. The staff in
Khajane Payee No delay in Convenient Good Error free these agencies did not perceive that cost had been reduced.
transaction time schedule location transaction
However, they felt that there was a significant positive impact
CARD Less time Less waiting Less Fair
and effort time corruption treatment on the quality of governance. Computerization was also seen
required to be partially responsible for improved tax collections by
eProcurement No Easy access Equal No need to some agencies.
corruption opportunity visit
to all Government
office VI. KEY FINDINGS AND IMPLICATIONS
eSeva Less time Less waiting Convenient Fair Respondents (citizens and businesses) who have used both
and effort time time schedule treatment the manual and computerized systems indicated an
required
overwhelming preference for the computerized service
AMC Less time Less Greater Good
and effort corruption transparency complaint delivery in most projects (see Table VI). However, in the case
required handling of Bhoomi, 79 percent of respondents preferred the
system computerized system whereas in most other projects the
Checkpost No delay in Error free Error free Proper figure is 95 to 99 percent.
transaction receipt transaction queue system TABLE VI
Legend: Underline - Improved Governance; Bold - Transactional PREFERENCE FOR COMPUTERIZATION (PERCENTAGE)
Efficiency; Italics - Quality Project Preference for Computerization (%)
Bhoomi 79.34
eProcurement 83.71
V. RESULTS OF IMPACT ON AGENCIES
Checkpost 91.25
Data on transaction volumes, operating costs, investments, eSeva 96.84
tax collection (if applicable), and revenues from transaction CARD 96.98
fees was to be collected for three years prior to the AMC 97.49
introduction of computerized system and for the entire period KAVERI 98.31
since computerization. However, it was very difficult to
collect accurate data for pre-computerization as often agencies A profiling 11 of those who preferred the manual system
did not have an MIS that was reporting such data. Post- revealed that higher-income farmers with large holdings in
computerization data on transactions and revenues was easier urban areas preferred it to the computerized system. Perhaps
to collect as it had at least been recorded in individual such farmers were able to manipulate the flexibility in the
agencies. system to their advantage - something that needs to be further
Often, the prevalent costing systems do not permit investigated.
identification of operating costs for the computerization Preference for the computerized system is backed with
project alone. Many costs are joint costs being incurred to specific areas where concrete benefits have accrued to them.
support multiple activities. For different agencies, In most cases the cost of accessing service was reduced
computerized delivery of services touches different portions because the number of trips that needed to be made to the
of the overall activity portfolio. For example, AMC is an concerned offices saw a significant reduction and the waiting
organization which delivers large scale physical services such time came down by nearly fifty percent. Corruption was
as cleaning of roads and public areas, public lighting and significantly reduced or eliminated in five projects. Quality of
maintenance of urban infrastructure. Computerized delivery service delivery and quality of governance were also
of services impacts less than one percent of all employees. perceived to have improved significantly with
For agencies such as offices of the sub-registrar the impact is computerization in most cases.
more widespread as the core service is computerized. There is considerable variability in the composite scores
By and large, the major cost component was manpower. across the eight projects. The scores range from a 0.15 point
The study revealed that the staff was not reduced in any
project after computerization. In some projects work load was 11
An interaction tree was constructed by considering preference for the
reduced and therefore, additional time could be devoted to computerized/manual system as a dependent variable. The independent
other tasks. There could be cost implications in the future as variables comprised those that are indicative of the user’s profile, viz.
the transactions are growing at a much faster rate as compared occupation, income and urban/rural status.
190
difference, which represents virtually no improvement, to 1.6 of disaggregated data at the level of each service location or
(on a 5-point scale), which is quite significant. Other studies for different types of clients.
have shown a similar variation for the same project The conclusions on impact on agency are not as robust as
implemented across different ststes [13]. This suggests that those for impact on citizens because time series data on
projects are not being designed using learning from other different types of costs and revenue streams related to the
states. service being investigated could not be collected for sufficient
The distinct preference for computerized systems by the time periods for most of the projects.
clients of poorly rated projects seems to suggest that even The overall positive assessment of projects should be seen
small gains for the users can trigger major positive change in in the context of the eight projects that were selected for
perception about service delivery systems. assessment. These were mature projects, with wide reach and
Incremental operating costs for agencies (including scope of activity. These projects have been operational for at
amortized investment) per transaction could be compared with least three years, implying successful implementation. Nearly
the reduction in direct cost of accessing the service reported all the projects serve urban clients. Projects that serve rural
by the clients. This would indicate the level of user fee that clients could have a very different cost structure and demand
can be charged (should be less than the monetized gain to the pattern. Rural projects also face a greater challenge in the
user) and the degree to which the project could be made maintenance of infrastructure. For all the projects
economically viable. The fact that many projects can become improvements were measured in comparison with the manual
self-sustaining through revenues from user fee indicates that systems that were replaced. When countries move from
private sector investment can be tapped. Five of the eight computerized systems in departments to a 24x7 service
projects have private partners and the expansion of the sixth delivery through portals, the degree of perceived
project in rural areas is being planned through private improvement could be very different.
partnership. For understanding the costs and benefit of ICT investments
Further analysis of data indicated significant variability of at a macro level, more projects from different contexts need to
impact across different delivery centers in a given project. be evaluated and failed projects (that could not be
This variability should be a cause for concern for delivery successfully implemented) would have to be included in the
models in which physical service centers are created. Part of analysis. It would therefore be hasty to generalize the overall
this variability is explained by poor infrastructure (power and conclusions of positive impact and economic viability of
connectivity), particularly in rural areas. Also, given the large electronic service delivery projects from this study.
variation in activity levels at different centers, it is often Further studies need to be undertaken to explain the
difficult to match the capacity to the demand at each of these variation in impact on various dimensions; differences across
centers. Portal-based delivery accessed via the Internet can be locations for a project; and differences across projects. Studies
a solution. However, unequal access to Internet can put some could also be undertaken to understand the effectiveness of
users at a disadvantage in such systems. different delivery models and implementation modalities such
A detailed study of one of the centers in KAVERI as the use of public private partnerships. More data
indicated that any type of system breakdown leads to (quantitative as well as qualitative) will need to be collected
corruption. The breakdown can be on account of an overload from the agencies to undertake such studies.
of demand in comparison to the capacity of the system to The establishment of acceptable counter-factual was a
process registrations. Systematizing queues by appointments challenge for the study team. All the projects had
helps prevent break-down. Agents play a key role in discontinued manual delivery of service and had mandated the
promoting corruption. Private operators also exhibit rent- use of the computerized systems for the citizens. For the
seeking behavior if given an opportunity. assessment of manual system, respondents needed to rely on
memory. In case of systems that have been operational for a
large number of years, such recall can be prone to error. There
VII. LIMITATIONS OF THE STUDY AND AREAS FOR FURTHER has been no benchmarking of the service delivery in a manual
RESEARCH system prior to implementing a new computerized system -
The study was exploratory in nature undertaken with something that needs to be done for new projects that are
limited resources. More countries could have been covered taken up in the future. Other forms of counterfactuals such as
and a greater variety of applications could have been assessed a user group from an adjoining state could be used to avoid
for generalizing the conclusions. The study used a sample size the problem of recall. However, use of a different context
that was considered relatively small (and barely adequate) at could result in other types of biases.
the stage of planning the field work. While the difference The study assessed direct economic impact in terms of cost
between manual and computerized systems on all the of accessing the service. It does not measure the impact on the
performance indicators are found to be statistically significant inherent value of efficient delivery of the service for the
(in most cases at a confidence level of 99 percent), the client. For example, copy of a land title is required for
accuracy of the actual estimate of the difference in direct costs obtaining a farm loan, getting a bail in a court case, insuring
could be improved by the use of larger samples. A larger crops and for purposes of checking the veracity of the record.
sample size in subsequent studies would also permit analysis A farmer would attach different values to efficient delivery of
191
land title depending on the purpose for it was required.

Subsequent studies should also attempt to estimate the impact
on indirect economic value.
REFERENCES
[1] R. Heeks. (2008, May 8). Success and failure in eGovernment projects.
Available: http://www.egov4dev.org/success/index.shtml
[2] R. Basant, S. Commander, R. Harrison, and N. Menezes-Filho, “ICT
adoption and productivity in developing countries: New firm level
evidence from Brazil and India," IZA Discussion Paper No 2294,
Germany, Sep. 2006.
[3] “World public sector report 2003: E-government at the crossroads,”
United Nations, New York, 2003.
[4] “Information and communication technologies: A World Bank Group
strategy,” World Bank Group, 2002.
[5] bridges.org (2005, May). E-ready for what? E-readiness in developing
countries: Current status and prospects toward the millennium
development goals. InfoDev [Online]. Available:
http://www.infodev.org/files/2049_file_InfoDev_E_Rdnss_Rpt_rev11
May05.pdf
[6] “Value creation in eGovernment projects: An exploratory analysis
conducted for the Danish presidency of the eGovernment work group of
the Directors General,” European Union, 2003.
[7] “eGovernment: Realizing the vision,” Third Annual eGovernment
Benchmarking Report, Accenture, 2002.
[8] “Is IT worth it - Presenting the public sector business case,” Gartner,
2002.
[9] E. Lau, “Electronic government and the drive for growth and equity,”
OECD E-Government Project, 2005.
[10] “eGovernment economics project”, DG Information Society and Media,
European Commission., 2005.
[11] “WiBe 4.0 - Recommendations on economic efficiency assessments in
the German federal administration, in particular with regard to the use
of information technology,” Federal Ministry of the Interior, Germany,
2004.
[12] R. M. Peters, M. Janssen, and T. van Engers, “Measuring e-Government
Impact: Existing Practices and Shortcomings,” in Proc. 6th International
Conference on Electronic Commerce, Delft, The Netherlands, 2004.
[13] “State level e-governance projects in India: Overall assessment of
impact on citizens,” Centre for E-Governance, Indian Institute of
Management, Ahmedabad, unpublished.
192
The contribution of user-based subsidies to the

impact and sustainability of telecenters – the
eCenter project in Kyrgyzstan 1
Michael L. Best, Dhanaraj Thakur and Beth E. Kolko
Abstract—We examine the extent to which user-based subsidies of the program can be enhanced. For example, better social
can promote the sustainability and development impact of targeting of the coupons could have brought in more women or
telecenters, where sustainability is defined in financial and social under represented groups. From a financial sustainability point of
terms. We do this by looking at a coupon scheme used by the view, explicitly targeting new users (as opposed to people already
USAID funded eCenter network in Kyrgyzstan. The network using the center) could have also been more effective. In terms of
consisted of partnerships with existing commercial computer impact, coupons could have been distributed, for example,
centers which provided fee-based ICT services to their specifically to entrepreneurs looking to start a new business. A
communities. The eCenter program temporarily provided method of incentivizing eCenter management to perform such
subsidized coupons for Internet access and computer training to targeting is probably required.
users of these centers.
Using a mixed-method approach, we found that user-based Index Terms — Telecenters, Sustainability, Subsidy,
subsidies have to a certain degree aided financial sustainability by Kyrgyzstan, Central Asia
bringing new users to the centers, some of who will, conceivably,
become long-term customers bringing ongoing revenues. The
subsidies also helped to “kick-start” operations during the initial I. INTRODUCTION
stage of the project. However we found that the distribution of
the coupons did not lead to any significant enhancements to social As a method of improving access to information and
sustainability by, for instance, encouraging users from under- communication technologies (ICTs), the telecenter is popular
represented social groups (though we did find that women were among donors and governments in many different countries [1,
marginally more likely to take advantage of the program). 2]. The concept of the telecenter emerged from a community
Moreover, we found that the distribution of both Internet and driven movement in Scandinavia in the 1980s [3]. For poor
training coupons favored more regular users of the eCenters. rural communities, telecenters can provide access to
Finally, looking beyond sustainability to impact within the communication and content. This can include access to market
community, we found that the coupon program had a limited
and crop prices, to financial information/services, and
development impact on participating communities. For example,
the population of users who reported economic benefits from communication with friends, family, and business colleagues
using the eCenter (eg. subsequent employment or starting a new [4]. Telecenters can also support the delivery of government
business), did not make more use of the coupon program than the services [5]. Today, telecenters can be found in many countries
user population at large, suggesting that the coupon program and are referred to by a plethora of terms: tele-cottages, public
itself did not account for this benefit. information access points, public internet access points or
We argue that, even in the presence of such modest positive multi-purpose communications centers. While each of these
effects, user-based subsidies still offer an intriguing model. We terms represent slight differences in the object being discussed,
believe that if the eCenters had narrowly targeted particular the common characteristic of telecenters, as used in this paper,
participants for the coupon programs it is likely that the benefits
are publicly accessible ICT resources in one or more physical
locations that are available with or without fees.
As with any other development intervention, a critical
Michael L. Best is an Assistant Professor with the Sam Nunn School of objective is to make the telecenter sustainable. Another
International Affairs and the School of Interactive Computing, Georgia objective is to address the problem of poverty and other socio-
Institute of Technology, Atlanta, GA, 30332 USA. (e-mail:
economic concerns within the community at large; thus a
mikeb@cc.gatech.edu).
Dhanaraj Thakur is a PhD student with the School of Public Policy, successful solution should also have real impact [6, 7]2.
Georgia Institute of Technology, Atlanta, GA, 30332 USA. (e-mail: However, many telecenter projects have not realized either of
dthakur@gatech.edu). these objectives for a variety of reasons such as a lack of
Beth Kolko is an Associate Professor with the Department of Human
income, inappropriate services, little or no local content [8],
Centered Design & Engineering, University of Washington, Seattle, WA,
98195 USA. (e-mail: bkolko@u.washington.edu). inadequate infrastructure (power, roads, telecoms, etc.)
1 2
This paper is based on the results of an evaluation provided to the In fact, a successful telecenter program might be one that has had
Academy for Educational Development (AED) of the USAID funded eCenter significant impact (e.g. spun-off a few small businesses) but has no long-term
project in Kyrgyzstan in June 2007. sustainability (e.g. the program transitions and closes after a few years).
193
particularly in remote areas [9], lack of political, social, or model built upon the work of Heeks and Bhatnagar to help
managerial sustainability [10] or the challenge of identifying explain why projects that succeed initially, can still fail to
local and global partners that can move the telecenter to a enjoy long-term sustainability.
larger scale [11]. These challenges are enormous and have While the problems of sustainability are fairly well
often relegated telecenter initiatives to the “forever pilot documented (and theorized) the ultimate goals of community
syndrome” [11 pg.11]. impact (and the measurement and evaluation of such impact)
Thus, sustainability and impact remain the two principal may be less conclusively observed. To date most literature has
questions related to rural telecenters. In response, various focused on formative or process evaluation as opposed to
strategies have been developed to advance the sustainability summative or impact evaluations [21, 22]. And the literature
and impact of telecenter projects. One approach is the that does focus on social or economic impact assessments of
provision of subsidies for telecenter users. These are often telecenters to their broader community have yet to demonstrate
applied for a limited period and enable the use of ICT services an absolutely conclusive link (see [23] for a review).
by community members who might not have otherwise been From the above discussion and in tandem with previous
able to do so. While not necessarily a new approach, there is research, [see for example 5, 10, 24, 25], we have selected two
little research that examines the connection between user- main dimensions from which we examine sustainability:
based subsidies and sustainability or impact. The question that 1. Financial – This includes the financial independence,
we wish to explore in this paper is to what extent can user- business performance, and solvency of the project
based subsidy programs promote the sustainability and impact over time.
of telecenters. We do this by examining a user-subsidy coupon 2. Social – This refers to the equitable distribution of
scheme used by the USAID funded eCenter network in benefits among target groups, equal access and use,
Kyrgyzstan. and locally relevant content/services.
The rest of the paper is divided into several sections. First, In addition, we study the center’s development impact as it
we discuss issues surrounding the impact and sustainability of extends into the communities themselves. We can observe this
telecenters. We then articulate the methodology used in in both economic and social forms such as new educational
addressing our research question. Finally, we describe the opportunities, community empowerment, job creation, local
eCenter project, present our results and analyses based on our economic development, etc.
definition of sustainability and impact, and posit some Various initiatives have strived to manage the inter-related
conclusions. issues of impact and sustainability. The literature consists of
many general prescriptive reports and descriptive case studies
of such initiatives [see for example 8, 26-28]. Others have
II. THE IMPACT AND SUSTAINABILITY OF TELECENTERS reported on specific or novel approaches such as incorporating
As with any other development intervention, a critical business incubators into the telecenter and sharing resulting
objective is to make the telecenter sustainable. Typically, the profits with the telecenter organization [29]. Another approach
intention is to make a telecenter financially independent and is to provide subsidies to a telecenter initiative that is typically
solvent. This can be in terms of meeting maintenance costs, the applied to initial infrastructure or staffing costs [13].
recovery of initial investments, acquiring sufficient human Alternatively, user-based subsidies can lower the costs that
resources/staff, or adequate service delivery [12, 13]. the users themselves pay for services they seek. The hope is
Sustainability can also be viewed from a social point of view. that if well conceived, user-based subsidy programs can
Thus, sustainability hinges on having local champions, enhance both financial sustainability (by creating an early flow
sufficient community acceptance/awareness and involvement of income while developing a customer base over the long
in the running of centers, a range of users that is non- term) and social sustainability (by providing subsidies that
discriminatory and balanced, and beneficiary participation in target under-served or over-looked populations, for example
project design [4, 9]. Additionally, there are political factors to women). Additionally, if the subsidies are for activities that are
be considered such as accessing local and national political likely to lead to economic, social or political growth within the
support for the project [5]. This includes having a policy and community at large then the impact of the telecenter should,
regulatory environment which is conducive to the development ultimately, be enhanced.
of telecommunications infrastructure and a commercial This approach has been used in a variety of settings
internet market [14]. including the Cotahuasi Internet Cabina project in Peru in
Investigators have proposed a number of theoretical 1997. In that case, the use of the telecenter’s services by local
frameworks to improve our understanding of sustainability community leaders was paid for by donor funds [4]. Another
including the critical success factor (CSF) and critical failure example of user-based subsidies was the PC3 project in
factor (CFF) models [15, 16]; the ‘design-actuality’ [17] or Bulgaria. Pre-paid coupons were distributed in communities
‘design-reality’ gaps [18]; scenario analysis for long-term where the PC3 centers were located. The main goal was to
sustainability problems [19]; economic and financial promote the centers while quickly developing a client base.
sustainability models [20]; and political and institutional This helped to reduce the financial risk faced by the new PC3
models that underline the lack of commitment on the part of centers [30]. The eCenter project in Kyrgyzstan employed a
political leaders and public managers. Previous works by one similar logic in the use of coupons as an incentive both for the
of us (Best) [5, 10] have presented a sustainability failure local eCenters and clients.
194
Table 1 – Targeted and Actual sample size of user surveys by

eCenter
III. METHOD Targeted Sample Actual number of
We employed a mixed-method approach to study the effect eCenter Size collected user surveys
of the coupon program on the sustainability and impact of the Naryn 70 79
eCenters. Our research consisted of site visits to all telecenters Bosteri 50 40
that were part of the eCenter program; 7 in total. These visits Karakol 75 87
were done in collaboration with local researchers and took Karasuu 67 70
place between March and June 2007. Our research instruments Nookat 66 72
included a user survey and interviews with all the center Talas 22 27
managers, available members of their staff, and local Ivanovka 18 19
businesses. These data were supplemented by a review of user Total 369 394
logs, project reports and updates, and business proposal
guidelines.
The user survey consisted of sixty-two questions developed
around the issues of sustainability and impact. Specifically it IV. BACKGROUND – THE ECENTER PROJECT IN KYRGYZSTAN
explored how the computer training and Internet coupons were Kyrgyzstan is a small Central Asian country that was part of
used, the ways users engaged with ICTs at the eCenters, and the former Soviet Union. Although it is landlocked with
the perceived economic impact of using the centers. The limited resources, it has achieved economic growth of around
majority of the questions were close-ended with a few open- 4% between 2000-2005 [31]. Nevertheless, the poverty rate
ended questions to capture opinions on issues related to the continues to be of major concern. The national estimate of
centers. people living below the poverty line was 43% in 2005 with
To determine our sample size we estimated the overall larger percentages in rural areas [32]. In addition, the
population size as the number of users at each eCenter over the unemployment rate was estimated at 8.1% in 2005, with 13%
period for which data was available: January 2006 (start of the in urban areas [32]. This implies that a significant part of the
project) to January 2007. We defined users as those who population, though employed, is still poor.
participated in the subsidy (coupon) program of the centers for In 2002, the government approved a “National Strategy for
either Internet access or computer training. Given this ICT Development in the Kyrgyz Republic” as part of its plan
approach, we approximated the total user population at 9,497 to use ICTs to address development issues. In general, this
people and thus ensuring a confidence level of 95% required a emphasis follows what Ure [33] notes is a more open approach
sample size of 369 users. to the diffusion and use of ICTs by the Kyrgyz government
To identify subjects from each of the centers, user contact when compared to its neighbors. However, the growth of the
lists were obtained from center staff. Users were then Internet has been hampered by the monopoly held by the state
randomly selected from each list. In many cases, however, telecommunications company Kyrgyztelecom, which is deeply
these lists were incomplete with either missing contact in debt to the World Bank and seeking to return to
information or incorrect contact details. To account for this profitability. One consequence of this is that Internet
difficulty, research teams substituted or augmented random subscription costs are high relative to average incomes. Recent
sampling with subjects obtained by opportunistic sampling at estimates put Internet user rates at a relatively low level of
the eCenters during site visits. In addition, the researchers 13.3% [34]. In addition, PC ownership was estimated at 10.4%
worked through the social network of users at each site in [34]. Given the larger social and economic context, Internet
order to find both previous and current users of the eCenters. access via private, home-based means is not likely to grow
Thus, user surveys were first collected using standard random substantially anytime soon and outside the capital, public
sampling techniques when possible and then with a access is limited, expensive, and usually not fast. Given the
combination of convenience and snowball sampling. problem of poverty and the acknowledged importance of
In addition to sampling the appropriate number of users we information resources for economic development, the need to
also needed to ensure that our sampled subjects were provide alternatives to private use and to enhance public ICT
representative of the population across each center. To account venues was viewed as important.
for this we stratified the sample size according to the
proportion of users from each location. The eCenter Project
Table 1 gives the targeted sample size, and actual number of The eCenter project was launched in Kyrgyzstan in July
subjects surveyed, for all seven eCenters; in almost all cases 2005. It was funded through the Last Mile Initiative of the
we were able to over-sample the population. The one center United States Agency for International Development (USAID)
that was under-sampled (Bosteri) was closed for renovations that seeks to promote greater access to information and
during the data collection period, complicating researchers’ communication technologies, particularly in rural and
attempts to contact users. underserved areas. The goal of the project was to augment and
network a group of telecenters across the country with the aim
of promoting local economic development. Each telecenter
was established within a pre-existing business. In this way, the
195
Nookat
centers sought to improve local access to ICTs, stimulate local for those communities. These figures help to sketch the overall
business creation, improve computer skills, and increase competitive landscape of the centers. By looking at the number
opportunities in non-traditional employment training and job of employees we also get a sense of the relative size of each
creation among the local population. The local project establishment. Note that while this table focuses on Internet
manager in Kyrgyzstan was the Civil Initiative on Internet provision there were also other businesses providing related
Policy (CIIP)3, a Kyrgyz non-governmental organization that business services such as printing or photocopying.
focuses on the promotion of civil society interests in the
development of national ICT policy. Table 2 – Summary characteristics of the eCenters
The project ran from July 2005 to September 2006 with a total eCenter Number of Competing Population
budget of US$390,000. During this time, a group of seven Employees Internet Estimate -
eCenters were established across the country (see Fig.1): Centers 1999
1. Karakol Karakol 7 10 64,322
2. Bosteri Bosteri 4 0 12,000
3. Naryn
Ivanovka 6 0 20,000
4. Nookat
5. Karasuu/Osh4 Talas 3 2 32,638
6. Ivanovka Nookat 4 0 30,000
7. Talas Karasuu 3 0 19,143
(Osh) (80) (220,000)
Each of the eCenters provides a variety of fee-based Naryn 4 10 40,050
services including Internet access and email, printing,
scanning, copying, faxing, multi-media services and IP- User-based subsidies
telephony. The exact suite of services offered varies from There were two main components to the eCenter program.
center to center. Each center also delivers a curriculum of First, there were subsidies granted to the eCenters for the
computer literacy courses which consist of several modules: provision of computer literacy/accounting courses and second,
Windows, Microsoft Word, Microsoft Excel and using the free or discounted Internet access was offered to local
Internet. Additionally, some centers offer accounting courses community members. The subsidies for computer literacy or
and one offers leadership training. accounting courses came in the form of reimbursements to the
Table 2 lists the estimated number of competing Internet eCenter operator for the training costs of those who passed
cafés within each of the seven communities which had each course. People were able to participate in the courses by
eCenters. It also includes the total estimated population size receiving coupons. Similarly, Internet access was subsidized
through the use of five-hour coupons that were distributed to
3
See http://www.gipi.kg/ for an overview of CIIP. users. The subsidy level on both the computer training courses
4
The Karasuu eCenter moved to Osh city in April/May 2007.
196
and Internet coupons changed over time. Subsidy amounts The managers themselves were quick to praise the eCenter
were reduced 20% each quarter; thus while the project paid for program and saw a direct connection between their success
100% of the cost of a training course in the first quarter, by the and the use of the coupons. Some of the managers estimated
fourth quarter it paid only 20% of the cost and required the that 70-90% of their current customers would not have come
user to pay 80%. to the center without having been part of the coupon program.
Interestingly enough, a significant percentage of users did not
Partnering with local businesses actually rely on coupons. Of those surveyed, only 54% and
Each eCenter was established as part of an existing local 43% reported actually receiving coupons for Internet access
business rather than through the creation of new centers. and computer training respectively. This implies that perhaps
Suitable local business partners (including pre-existing the managers had an exaggerated belief of the effect of the
telecenters) were selected through a competitive bidding coupon program on their businesses. When asked if they
process based on criteria such as related prior business would have used the eCenter services if there were no
experience, existing level of investment, relevant telecom coupons, approximately 47% of respondents said yes and
experience, and the potential for further expansion of 19.5% said they would not have used the center in that case. It
programs. Subsequent to a center’s selection, CIIP provided should also be noted that the subsidies were only provided for
technical support on the use of the subsidy program and a limited period. Thus, by the end of the project users were
delivered training to the business owners and their staff. paying for the full cost of Internet access and computer
While some of the selected eCenter sites were already training.
providing Internet services, others were engaged in computer The existing competitive environment (Table 2) appeared to
graphics and photocopying services, gaming, computer repair influence the scale of impact of the coupons. For example, we
or providing computer training courses. found that if the manager identified more competitors in the
There was one final component of the eCenter project – a immediate community, the eCenter users were more aware of
land grant program where suitable land for investment in office and more likely to have used other cyber cafés. If there were
space and technological parks could be linked to each eCenter. fewer competing centers in the catchment then users were
It was envisioned that such investment would be supported by more likely to have used the eCenter prior to the coupon
the success of the eCenters; this component of the project, program. In other words, when the level of competition was
however, did not achieve expected outcomes and was heavy the coupon program was more successful in drawing in
eventually dropped. new users.
While there are several interesting aspects to the eCenter In general, 66% of those who received Internet coupons had
project as a whole, we have focused only on the user-based used the eCenter before while 33% of those who had received
subsidies in this study. The following sections present our a coupon had not used the eCenter before. The results are very
findings and analysis of the user-based subsidies component of similar for training coupons. Thus, during the subsidy period,
the eCenter project. coupon use was associated with the introduction of new users,
which can contribute to financial sustainability.
It should be noted that the decision to partner with existing
V. THE SUSTAINABILITY OF THE ECENTER PROJECT IN and successful local businesses was also important for the
KYRGYZSTAN5 success of the project. Generally, there is little evidence to
suggest that the one-off donor grant approach can work [24]
Financial sustainability and so some have argued for a business centered approach to
All the eCenters enthusiastically implemented the coupon telecenters [13]. The rationale behind this approach is that the
program and several have since sought to expand their offering profit motive can promote financial sustainability. In the case
of ICT services by increasing the scope and size of their of the eCenter project, it was in the interest of the local
businesses. In all cases, the businesses stated that their business partners to make sure that their centers remained in
financial situation was better off following the change to an operation and were successful. The coupon program supported
eCenter. On average, according to the managers this transition the partnership with local businesses by ensuring a sufficient
increased their revenues by an estimated 57% (with a low of number of customers during the initial subsidy period. It was
15% reported at Karakol and a high of 100% reported at critical therefore in “kick starting” the operations of the centers
Talas). In addition, almost all of the center managers agreed after which several of the local business partners were able to
that their clientele increased (some very significantly) after use this momentum to expand their businesses and tailor their
becoming an eCenter. The main sources of revenue varied suite of ICT services to their community needs.
from center to center and were based on local demand. For
example, Internet access was reported to be the main source of Social Sustainability – Types of Users
revenue in only two of the seven centers. Other revenue We examine social sustainability primarily in terms of the
sources included computer training, typing services, IP equitable distribution of benefits, access and use within the
telephony and the sale of mobile phone credit. community. Ideally, a more balanced and representative user
base should represent broader community acceptance of the
center and limited social exclusion. Table 3 below summarizes
5
Statistical results mentioned in this paper are all significant to at least the the general user distribution across the various centers. We
0.05 level.
197
found that the majority of users were women except in Nookat been exploited completely. Thus while the eCenters have
where the more traditional culture in the south seems to limit financially benefited from the coupon program; the long term
female participation at the eCenters. The majority of users in social sustainability of the project may have not.
the sample were typically young, suggesting the propensity to Two factors could help explain this type of coupon
use ICTs is linked to age particularly where the general distribution among users. First, the implied goal of the project
diffusion of such ICTs is low (i.e. rural areas). This young user was to focus on persons who had the least access to the
group is consistent with a nationally young population (31% Internet and computer courses. However the way the coupon
below 15 years in 2005 - [35]) and also partly explains the low program was operationalized did not appear to be systematic
marriage rates among our sampled population. In general, or consistent across centers. Moreover, the eCenters received
older customers used more of the basic ICT services such as no particular incentives to narrowly target users and so some
photocopying or printing and requested the assistance of may have simply distributed their coupons to people already at
younger persons or staff to help them when using the Internet. their center. Thus, there was a significant difference among
Older users (above thirty years) were, however, just as likely receivers of both Internet and training coupons from center to
to attend training courses. The education rates are similar to center though there was no discernable pattern to these
those nationally. Also, the user distribution was approximately differences.
proportional to the religious and ethnic composition of each Some centers seemed to offer coupons to whoever was
community easily available. Indeed, some were more concerned with
issuing the coupons rather than targeting and as such relied on
Table 3 – Summary of user characteristics informal means of distribution. One example of this is a
eCenter Male Female Avg. Bachelors Married neighboring Internet café owner in Karakol who said he
% % age or higher % received a training coupon for the computer literacy course.
% He was already exposed to more advanced computer courses
Karakol 44 56 21 48 10 but still chose to attend the eCenter course since it was free at
Bosteri 25 75 18 10 3 that time. Alternatively, more formal means of distribution
were used at other centers. For example, in Naryn the intention
Ivanovka 53 47 18 5 0
was to target as wide an audience as possible. This was done
Talas 33 67 23 15 19 through advertising in local media and actually did result in
Nookat 67 33 24 24 39 lines literally going out of the door of the eCenter. However, it
Karasuu 41 59 21 30 14 is unclear what type of users this form of marketing attracted.
Naryn 31 69 24 47 25 Second, the eCenters were encouraged but not compelled to
Total for give one Internet coupon per person, so as to increase the
all users 43 57 21 32 19 reach of the program. This suggestion, however, was not
generally followed. For example, in Nookat users were given
one coupon per quarter since, according to the manager there,
Social Sustainability – Targeting and Distribution of Coupons
it was difficult to continually find new Internet users.
Based on the above characteristics, we found few
Alternatively, in Karasuu the strategy was one coupon per
statistically significant differences among users who received
individual. Thus both targeting and distribution seemed to be
coupons and those who did not. Among those who received
dependent on the individual eCenters and was the result of a
Internet coupons approximately 51% were female. This could
compromise required to enable a private business to undertake
have had a marginal effect on the larger proportion of female
socially oriented goals.
users at the centers overall. Other variation in user
characteristics (ethnicity, educational level, etc.) among those
users who benefited from the coupon program and those who
VI. DEVELOPMENT IMPACTS OF THE COUPON PROGRAM
did not were not statistically significant. Also, there was no
difference between those who got Internet coupons and those Internet Access
who did not in terms of their perceptions of the importance of Although this represents a preliminary and indirect impact
the Internet to their jobs or schools. In terms of the training of the program, the Internet coupons did, of course, increase
courses there were also no discernable demographic access to the net among users. This access can be viewed as a
differences between those who received these coupons and form of impact. Indeed, in most cases overall access increased
those who did not. in participating communities. For example, several of the
What was different was whether or not the respondent had centers are located in rural areas with limited access to ICTs.
used the eCenter prior to receiving the coupon. As mentioned Thus, the introduction of the eCenters helped to meet the latent
above, 66% of those who received Internet coupons had used local demand for ICT services and training. Given the
the eCenter before. Similarly, 60% of those who received relatively high subscription costs for the Internet, this also
training coupons had already visited the eCenter. This suggests meant that the eCenter could have been the only source of
that prior users were in a better position to learn about the Internet access for some users. This was the case in at least
program and participate. The concern is that while the coupon three of the communities.
component of the project is now complete, the opportunity to For many, the communication function of the Internet was
attract larger numbers of new users to ICTs might not have most important, particularly where regular post mail services
198
were infrequent. The most common online activities were Development Impact – Local economic development
sending/receiving emails and instant messaging. Other In terms of local economic development, we observed two
important uses included reading news, doing school related types of impacts related to the introduction of user-subsidies at
work/research, downloading music and participating in chat the eCenters. First, the eCenters have stimulated the local
rooms. Several users proffered experiences of how accessing market for ICT services. For instance, there are approximately
the Internet had helped them keep in touch with friends and ten other commercial computer centers in Naryn that were
family abroad, support their work and even religious established after the eCenter opened its doors. Second, the
education. opposite is also possible. In Nookat, prior to the introduction
Finally, in addition to providing increased quality or of the eCenter there, there was one other business offering
availability of Internet access, the eCenters have also been an Internet services. This closed soon after the opening of the
important place for users to gain Internet-related skills. eCenter as it could not compete with the initial subsidies being
Approximately 70% of those who received Internet coupons provided for Internet access. That other business now operates
said that they had learned to use the Internet at their eCenter. as a café.
One of the objectives of the eCenter project was to support
Development Impact – Computer training local business development. Of all users, only 5% said that
The acquisition of relevant skills is a clearer form of they were able to utilize either the facilities or courses at the
community development impact and the eCenters provided eCenter to establish a new business. Most of the businesses
computer training courses which were well received by were located in the services sector and a few specifically in the
participants. Perhaps more than anything else, participants in ICT sector. Of these users, there were no significant
training courses were quick to inform us of the benefits of differences in terms of gender or levels of education. They
receiving the computer literacy training. The courses were were not necessarily in a higher income group either; for
basic and allowed the centers to bring those with no example they were no more likely to own a car than those who
understanding or experience with computers up to at least a did not start a business. Finally, in terms of coupon use, there
foundational level of use. Completion was contingent on was again relatively little difference between those who
passing an evaluation which tested their ability to use the reported starting a new business and those who did not.
computer and some of the Microsoft applications they had Whether we refer to the efficacy of the training courses or
studied. The majority of trainees were able to pass the test; the improved Internet access, one qualifying factor to note is the
failure rate was approximately 10%. For many users there was initial limited diffusion of ICTs in the targeted communities.
a sense of pride in completing the course. This initial lack of ICTs in these mostly rural/semi-urban
For many, these courses provided the basis for the communities meant that the potential for the subsidies to have
acquisition of job-related ICT skills. Eighty-five percent of all some modicum of impact was there from the start. We cannot
respondents stated that they acquired important job skills from be sure whether the development impact would be similar in
their eCenter. Furthermore, some users reported actually other contexts. However, we can indicatively point to the
getting jobs (usually clerical) based on these skills. example of the eCenter in Osh city where there are some 80
Approximately 15% of all users reported finding a job as a other cybercafés, various documentation centers, and in
result of using the eCenter. As noted earlier, the country has a general better infrastructure for the delivery of ICTs. In an
high national poverty rate of 43% and an unemployment rate environment such as this, the eCenter modified its strategy to
of 8.1% (2005). Therefore such impacts from the program, focus on the provision of training courses as there were few
though small, will be valuable to both the unemployed and the organizations providing this service. Thus, the business focus
working poor. We did not find, however, a link between users of the eCenters coupled with the goal of providing relevant
who participated in the coupon program and those who ICT services to targeted communities can potentially create an
reported finding jobs due to training at the center. impact in a variety of contexts.
Approximately 61% of trainees who received user-subsidy
coupons for their classes said they would pay for the service if
there was no subsidy, demonstrating the value placed on the VII. SUMMARY AND CONCLUSIONS
services and experience provided by the eCenters. Many of the The eCenter program provided subsidized coupons for
eCenters are in fact planning to continue and even expand their computer courses and Internet access to users of already
course offerings past the lifespan of the subsidy program. For established commercial computer centers. Thus, the center
at least one eCenter this is also the main source of revenue. owners provided the physical infrastructure for each eCenter
Content was an important factor in promoting the demand for and the subsidies helped them to access, and ostensibly expand
the computer literacy courses. Much of the design and content the local market for training and Internet services. The project,
for the courses came from the initial demand studies done therefore, represented a merger between public goals of
prior to the start of the project. However, a few users increasing ICT access and literacy and private goals of
suggested that having completed these courses they now want increasing profit and market share.
the center to offer more advanced computer courses. We focused on the subsidy component of this project, as an
Continually modifying content to meet the needs of users will innovative approach to telecenter development. As such we
be a factor in supporting long term development impact. attempt to address a gap in current research by exploring the
relationship between user-based subsidies and the
199
sustainability and impact of telecenter initiatives. In this case to grow their businesses. Future user-subsidy programs should
we examined sustainability in terms of its financial and social examine ways to incentivize the local center managers to
dimensions. By financial sustainability we look at an eCenter’s design and implement targeting programs that enhance social
financial independence and solvency; by social sustainability and financial sustainability as well as overall community
we are primarily concerned with the distribution of access and impact. Additionally, some broad parameters and techniques to
benefits to the community. There are of course other this targeting could be developed centrally and offered to
dimensions that could have been included, but we wanted to participating center managers.
limit our analysis in such a way as to make it focused and Even with decades of worldwide experience constructing,
useful. operating and evaluating telecenters, the research reported
In terms of financial sustainability, we found that the here makes clear that the often sited preeminent goals of the
coupon program brought some new users to the centers during telecenter movement – socio-economic sustainability and
the subsidy period. Self-reporting indicates that a third of substantial community impact – remain difficult to obtain.
Internet coupon users and 40% of the training coupon users Further experimentation with models and approaches, along
were new. More importantly, 19.5% of users stated that they with close observation and continued independent assessment
would have not have used the services of the eCenter without work, is required if we are to realize these ultimate ambitions.
the coupons. The coupons also helped the local businesses
thrive with customers during the initial subsidy period; it was ACKNOWLEDGMENTS
critical, therefore, in “kick starting” the operations of the We thank Medina Aitieva and her students from the American
centers. Thus, we argue that the coupons have to some degree University of Central Asia, members of our local research
aided financial sustainability by bringing new users (and team. We also thank the three anonymous reviewers for their
therefore additional revenue and an expanded long-term helpful inputs.
customer base) to the centers and creating a stable source of
income during the initial stage of operations. Obviously,
temporary subsidy programs may actually weaken long-term
REFERENCES
financial self-sustainability as the program winds down. But in
this case, while we observed the end of the program we did not 1. Siochrú, S.Ó., Global Governance Of Information And
find evidence for an immediate financial downturn. Communication Technologies: Implications For Transnational
In terms of social sustainability, the distribution of the Civil Society Networking. 2003, Social Science Research Council:
coupons did not lead to any significant differences in the type New York.
2. Gómez, R. and P. Hunt, eds. Telecentre Evaluation - An
of users except a marginal increase in female participation. We
International Perspective : Report of an International Meeting on
find that the distribution of both Internet and training coupons Telecentre Evaluation. 1999, International Development Research
favored more regular users of the eCenters. This implies that Centre (IDRC): Montreal.
the opportunity to expose as many new users as possible to 3. Fuchs, R., Little Engines that Did: Case Histories from the Global
ICTs may have been missed and this could have limited the Telecentre Movement. 1998, IDRC: Ottawa.
4. Proenza, F.J., R. Bastidas-Buch, and G. Montero, Telecenters for
wider social impact of the project. One of the main reasons for Socioeconomic and Rural Development in Latin America and the
this was the lack of effective targeting strategies among the Caribbean. 2001, Inter-American Development Bank: Washington
centers. DC.
Finally, the coupon program had a limited development 5. Kumar, R. and M. Best, Impact and Sustainability of E-
Government Services in Developing Countries: Lessons Learned
impact on participating communities. One preliminary impact from Tamil Nadu, India. Information Society, 2006. 22(1): p. 1-
was the enabling of Internet access for users and the learning 12.
of Internet related skills. The coupons also facilitated 6. Mercer, C., Telecentres and transformations: Modernizing
computer skills training. However, among those who reported Tanzania through the internet. African Affairs, 2006. 105(419):
p. 243-264.
economic benefits (subsequent employment or starting a new 7. Dagron, A.G., Prometheus Riding a Cadillac? Telecentres as the
business) after acquiring these skills, there was little or no promised flame of knowledge. Journal of Development
difference between those who had been coupon recipients and Communication: Special Issue on Telecentres 2001. 12(2).
just regular users who had not availed of the program. 8. Colle, R., Memo to Telecenter Planners. Electronic Journal of
Information Systems in Developing Countries, 2005. 21(1): p. 1-
We suspect that as the targeting and distribution was left up 13.
to the local eCenters, there might have been less emphasis on 9. Roman, R. and R.D. Colle, Themes and Issues in Telecentre
sustainability and impact and more emphasis on short-term Sustainability, in Development Informatics Working Paper Series.
profit. We argue that both the sustainability and impact of the 2002, Institute for Development Policy and Management,
University of Manchester.: Manchester, UK.
coupon program could have been improved with a more
10. Best, M.L. and R. Kumar, Sustainability Failure of Rural
focused targeting strategy. For example, better social targeting Telecenters: The Sustainable Access in Rural India Project.
of the coupons could have included more women, particularly Information Technologies and International Development, 2008.
in the more traditional areas of the country, or other vulnerable 4(4): p. 31-45.
groups. From a financial sustainability point of view, targeting 11. Fillip, B. and D. Foote, Making the Connection - Scaling
Telecenters for Development. 2007, Washington, DC: Academy
new users could have also been more effective. In terms of for Education Development.
impact, coupons could have been distributed, for example, 12. Harris, R.W., A. Kumar, and V. Balaji, Sustainable Telecentres?
specifically to young entrepreneurs looking for skills or access Two Cases from India. 2000, Infobridge.
200
13. Proenza, F.J., Telecenter Sustainability - Myths and 33. Ure, J., ICT Sector Development in Five Central Asian
Opportunities. Journal of Development Communications, 2001. Economies: A Policy Framework for Effective Investment
12. Promotion and Facilitation, in International Conference on
14. Whyte, A., Understanding the Role of Community Telecentres in Strengthening Regional Cooperation for Managing
Development - A Proposed Approach to Evaluation, in Telecentre Globalization,. 2005: Moscow, Russian Federation.
Evaluation - An International Perspective : Report of an 34. CAICT. 2007 CAICT Survey: Topline Results 2007 [Retrieved:
International Meeting on Telecentre Evaluation, R. Gómez and P. September 5, 2008]; Available from:
Hunt, Editors. 1999, International Development Research Centre http://depts.washington.edu/caict/surveys/2007toplines/TopLines_
(IDRC): Montreal. 2007.pdf.
15. Heeks, R. and S. Bhatnagar, Understanding success and failure in 35. UNDP, Human Development Report 2007/2008: Fighting climate
information age reform, in Reinventing government in the change: Human solidarity in a divided world. 2007, Palgrave
information age, R. Heeks, Editor. 1999, Routledge: London. p. Macmillian: New York.
49-74.
16. Heeks, R., Understanding e-Governance for Development, in i-
Government Working Paper Series No. 11. 2001, Institute for
Development Policy and Management: University of Manchester, 98
UK.
17. Heeks, R., Information Systems and Developing Countries:
Failure, Success, and Local Improvisations. The Information
Society, 2002. 18(2): p. 101 - 112.
18. Heeks, R., Most eGovernment-for-Development Projects Fail:
How Can Risks be Reduced?, in IDPM i-Government Working
Paper no.14. 2003, Institute for Development Policy and
Management: University of Manchester, UK.
19. Aichholzer, G., Scenarios of e-Government in 2010 and
implications for strategy design Electronic Journal of e-
Government 2004. 2(1): p. 1-10.
20. Best, M. and C. Maclay, Community Internet Access in Rural
Areas: Solving the Economic Sustainability Puzzle, in The Global
Information Technology Report 2001-2002: Readiness for the
Networked World. 2002, Oxford University Press: Oxford.
21. Hudson, H.E., From Rural Village to Global Village:
Telecommunications for Development in the Information Age.
2006, Mahwah, New Jersey: Lawrence Erlbaum Associates.
22. University of Washington Center for Internet Studies. CIS
Literature Review Database on the Impact of Public Access to
ICT. 2008 [Retrieved: September 9, 2008]; Available from:
http://references.cisresearch.org/index.php.
23. Kuriyan, R. and K. Toyama. Review of Research on Rural PC
Kiosks. 2007 [Retrieved: September 16, 2008]; Available from:
http://research.microsoft.com/en-
us/um/india/projects/ruralkiosks/kiosks research.doc.
24. Harris, R.W., A. Kumar, and V. Balaji, Sustainable Telecentres?
Two Cases from India, in The Digital Challenge : Information
Technology in the Development Context, S. Krishna and S.
Madon, Editors. 2003, Ashgate: Aldershot, Hants, England ;
Burlington, VT, USA. p. 124-135.
25. Bailur, S., Using Stakeholder Theory to Analyze Telecenter
Projects. Information Technologies & International Development,
2007. 3(3): p. 61-80.
26. Jensen, M. and A. Esterhuysen, The Community Telecentre
Cookbook For Africa - Recipes For Self-Sustainability. 2001,
United Nations Educational Scientific and Cultural Organization:
Paris.
27. UNDP, Telecottage Handbook: How to establish and run a
successful telecentre. 2006, UNDP, Europe and the CIS:
Bratislava.
28. Badshah, A., S. Khan, and M. Garrido, eds. Connected for
Development: Information Kiosks and Sustainability. United
Nations Information and Communication Technologies Task Force
Series. Vol. 4. 2003, United Nations Information and
Communication Technologies Task Force: New York.
29. Khelladi, Y., What Works:The Infocentros Telecenter Model, in A
Digital Dividend Study. 2001, The World Resources Institute:
Washington DC.
30. Tifft, J.D., Summative Evaluation - Bulgaria Public Computer
and Communication Centers (PC3s). 2003, USAID: Washington
DC.
31. IBRD, Joint Country Support Strategy (JCSS) for the Kyrgyz
Republic (2007-2010). 2007, The World Bank: Almaty.
32. IBRD, Kyrgyz Republic Poverty Assessment. 2007, The World
Bank: Washington DC.
201
A Speech Enabled Indian Language Text to

Braille Transliteration System
Tirthankar Dasgupta and Anupam Basu
In order to provide proper information access and to

Abstract— In this paper we present a speech enabled bridge the communication gap between the visually impaired
bidirectional automatic Indian language text to Braille and the sighted community, the need to build some advance
transliteration system. The system allows bridging the technologically supported systems like automatic Braille
communication gap between a visually impaired and a sighted
person. The present system can be configured to take Indian
transliteration and screen reading systems are indispensable.
language text document as input and based on some Several works have been done on building automatic, bi-
transliteration rules, can generate the corresponding Braille directional text to Braille transliteration system and speech
output. The system is augmented by an Indian language text-to- enabled interfaces for the visually impaired community [4]-
speech (TTS) system through which a user can get instantaneous [8]. However, most of the systems cannot be directly used for
audio feedback from the input text. We further extended the the visually impaired community of India. This is due to the
system to support transliteration of Dzongkha 1 text to Braille.
Finally we present an Audio QWERTY editor which allows a
following reasons
visually impaired person to read and write Indian language texts 1) Most of the systems are based on foreign languages like
through a computer. English, French, Germany, Spanish, Portuguese, and
Swedish [9]-[12].
Index Terms— Visual Impairment, Braille Transliteration, 2) Indian language scripts are quite different from that of
Audio QWERTY Editor, Indian Language European or American languages. Hence, separate rules
are needed to transliterate Indian language texts to
Braille.
I. INTRODUCTION 3) Foreign systems like, Duxbury [9] and JAWS [13] are
The Braille encoding system is the primary means of costly, given the Indian economic reality of the visually
representing textual documents in a readable format for the impaired population, who mostly belong to the poorer
visually impaired people [1], [2]. However, due to the scarcity section.
of Braille printed reading materials, blind people in India face In order to overcome the above mentioned challenges and
a daunting task while getting their formal education and limitations of the existing systems, we present a speech
achieve a respectable employment opportunity. The situation enabled Indian language text to Braille transliteration (ILBT)
becomes worse due to the unavailability of low-cost system. The system provides a generic framework for the
technological support. transliteration of large number of popular Indian language
The National Census of India has estimated around 21.9 texts to Braille. Further, the system can also be used as an
million disabled people in the country [3]. Out of which more Indian language document reader where a user can select a
than 15 million people in India are blind. This is considered to particular text document to get the corresponding speech
be the highest among all other disabilities. Three out of every output. This allows the present system to be used by both a
five children in the age group of 0-9 years have been reported sighted person as well as a blind person. We also extended our
to be visually impaired in India. present system to provide Braille transliteration of Dzongkha
Due to their inability in accessing information from texts keeping in view the proximity of Bhutan with India.
written text documents, blind people face tremendous Currently our system handles only four Indian languages
difficulties in communicating with sighted people in common Hindi, Bengali, Assamese and English along with Dzongkha.
places like post office, bank and other official places where However, due to the unified framework of the system, any
the primary mode of communication is through writing. other Indian language can be easily be added to it.
Manuscript received September 23, 2008. This work was supported in part II. INDIAN LANGUAGE BRAILLE ENCODING SYSTEM
by Communication Empowerment Laboratory, Indian Institute of Technology,
Kharagpur and Media Lab Asia, Mumbai. The Bharati Braille system is the standard technique of
T. Dasgupta is with the Indian Institute of Technology, Kharagpur, West representing Indian language texts to Braille [1]. The system
Bengal India (corresponding author to provide phone: +91 9433852325; e- uses 6 dot cells to represent each character. The combination
mail: iamtirthankar@ gmail.com).
A. Basu is with the Computer Science and Engineering Department, Indian of these 6 dots can generate 26-1=63 different Braille
Institute of Technology, Kharagpur, India (e-mail: anupambas@gmail.com). characters.
1
Official language of Bhutan
202
It has been observed that for all Indian languages, the 2) The system does not provide any auditory feedback to the
corresponding Braille code is the same. In other words, a user neither it gives a very good GUI for visualizing the
distinct Braille cell may correspond to different Indian transliterated output.
language character. However, a single Indian language 3) Editing of the transliterated output is very difficult and
character may require one or more than one Braille cell for its the system does not provide any interface to create or edit
representation. This process is illustrated in Table 1. input text document.
4) Working of the system depends heavily upon the user’s
Table 1: Braille representation of different Indian language
characters. knowledge, understanding and intelligence. As a result,
the system provides high cognitive load to the user which
Braille Bengali Hindi may result in increasing error rate.
Apart from computer based transliteration of Braille, a Braille
k ক क writing tutor system has been presented in [14]. The prototype
tutor system uses an E-slate device to capture a student’s
. খ ख
action and tries to develop the Braille writing skills.
IV. THE ILBT SYSTEM ARCHITECTURE

" R ঋ ऋ The architecture of our proposed Indian language text to
Braille transliteration (ILBT) system is shown in Fig. 1. The
diagram shows the essential components of the present
One of the most interesting phenomena found in Indian system. There are two major goals of the system:
language scripts is the usage of composite or conjugate • Providing audio feedback to the Indian language text.
characters. Conjugates are often constructed by the sequential
• Forward or reverse transliteration of text documents.
concatenation of two or more characters with a special
character called halant. The construction of conjugate The input to the system may be a text document written in
characters follows certain rules. Some of these rules are any popular word processor, or Indian language texts entered
illustrated in Table 2. through a keyboard. Based on the user’s requirement the
Table 2: Conjugate construction rules with examples taken from
system can generate either a speech output or can transliterate
Bengali language (C=Consonant, H=Halant, V=Vowel) the text to Braille. The details of the major components of the
system are discussed in the subsequent sections.
Rule Example Braille
A. The Text-To-Speech Engine
CHC ঙ+◌্ +ক = @+K A text-to-speech (TTS) engine is at the core of the speech
output generation module. It is converts strings of text to their
CHCV ক+ ◌্ + র+ ি◌ = িk @KRI corresponding human voice equivalent. In order to develop
the proposed transliteration architecture, we used a text-to-
CHCHC স+ ◌্ + ত+ ◌্ + র=st @S@TR speech engine called Shruti [15] that uses diphone
concatenation to synthesize speech in Indian languages using.
স+ ◌্ + ত+ ◌্ + র+ ◌্ + ◌ী
CHCHCV
= stী @S@TR9 The Shruti text-to-speech engine currently supports two
Indian languages namely Hindi and Bengali.
We built a wrapper around the TTS engine so that it can be
used by a number of applications including our transliteration
From Table 2 it can be observed that the conjugate
system. The wrapper also provides a number of other vital
characters, as constructed by clustering of consonants and
features like
vowels, may have an entirely different visual representation.
1) Conversion of text in Unicode or iTrans to ISCII 2 which
However, the corresponding transliterated Braille is
is the native representation format of the TTS being used,
represented by a sequence of Braille cells for each of the
2) Pronunciations of individual and conjugate characters.
characters.
B. Input Document Type & Keyboard Input Methods
III. RELATED WORKS In order to create an Indian language document, a text editor is
The Sparsha toolset [2] is the only system encountered by required for accepting the regional language text entered
us so far that can transliterate Indian language texts to Braille. through the keyboard and perform different operations on it,
The system provides some unique features like transliteration like formatting, printing, and saving the text. Further, it is also
of mathematical symbols and tactile graphics to Braille. desirable that the editor should support different font
However, this system suffers from several limitations like, encodings (like, Unicode, ISCII, ASCII, and iTrans) which
1) The system takes significant amount of time during
transliteration of large texts. 2
Indian Standard Code for Information Exchange
203
are
User I/P Through Keyboard Keyboard
Keyboard/Mouse Input Key
Input Hook
Methods
Operating
Regional Font Rendering System Mouse
Language Definitio Engine Hook
Fonts
Glyphs
Indian Language Text Documents

(UTF-8, UTF-16, ISCII, iTrans, ASCII) TTS SPEECH
Engine
Forward Forward Reverse Reverse

Transliteration Transliteration Transliteration Transliteration
Rule Rule
Transliteration
Braille Document Text Document
TTS
Embosser
Interface Engine
BRAILLE PRINT SPEECH
Fig. 1: System Architecture of the Speech enabled Indian Language Text to Braille Transliteration System
popularly used to create Indian language text documents. representation of any regional language text. This is
There exists no such editor that can support all the above especially important for Indian languages. Though this may
mentioned list of font encodings together. For example MS- not seem important for a blind user, it is actually essential
Word 2007 supports Unicode as well as ASCII encoding for him in order to communicate with other sighted people.
and the most common editor supporting ISCII is the iLeap The OpenType font format provides a large number of
editor. A solution to this problem is creating a new text features essential for supporting regional languages,
editor that can support all the mentioned font encodings. especially complex scripts [16]. Due to this reason it was
However, the developing a new text editor will require a found to be suitable for the mentioned purpose. In
substantial time and effort. Further, it expects a user to OpenType fonts all information related to the proper
learn the new system which is very difficult especially if rendering of the font may be included within the font file.
the user is visually impaired. Hence, in our proposed GSUB (Glyph Substitution) and GPOS (Glyph Positioning)
architecture we allow a user to create a text content in any are two tables responsible for storing such information.
of the commercially available regional language text editors These tables play a critical role in displaying regional
available in India. However, In case of unavailability of language text as, a lot of character reordering, combination
such kind of editors, our system provides a very basic form of characters and glyph reshaping may be necessary. Glyph
of a regional language text editor that is integrated along substitution is the process of replacement of one or more
with the ILBT system. Currently the editor provides some glyphs by a new glyph like while writing conjugates. Glyph
basic functionalities as compared to a full-fledged text positioning refers to process of shifting a glyph to produce
editors like, creating a new document, opening of the proper visual representation (E.g. while writing vowels
documents in multiple tabs, saving documents, searching in Indian languages called matras) as shown in Fig. 4.
regional language texts etc. However, though the visual representation of the text
changes in the above cases the original string of characters
C. Indian Language Font Rendering
remains unchanged. Both the above mentioned tables find
Proper Indian language fonts and the font rendering extensive use especially in Indian languages
engines are necessary for producing correct visual
204
D. Input and Output of the system in Fig. 3. From the figure we can see that if a character is
The proposed ILBT system can accept English text in the preceded by a halant then during transliteration, position of
form of plain text files and Microsoft Word documents. the character and the halant gets swapped. This
Apart from English the Braille transliteration system, as phenomenon can be represented by the rule:
described, can take Indian language text as input. The input
text document of the system can be of any type like: Precede(X, halant) Æ swap(X, halant)
• Unicode text – generated by any standard editor
supporting Unicode [17]. Another interesting feature found in Indian language text
transliteration is the handling of matras. A vowel may
• ISCII (Indian Script Code for Information occur either as a matra or it may appear as a distinct vowel.
Interchange) [18] documents generated by If the vowel occurs as a matra then a special symbol
applications like iLeap [19]
appears before the vowel else no special symbol appears.
The output of the ILBT system can be of two types. A This is illustrated in Fig. 4.
user can select a particular section of the text document to
get the corresponding Indian language speech output, or a
user can obtain the corresponding Braille output on a large
0981 0027 0045 090F 0947 0035
variety of commercially available Braille embossers 0982 003B 0046 0036 0000 0036
[20][21]. The current system has been tested on the 0983 002C 0047 0917 0000 0037
following Braille embossers: 0985 0041 0048 0038 0000 0038
• Index Basic-S 0986 003E 0049 0907 093F 0039
• Index Basic-D
a b
• Index 4X4 PRO
• Romeo Pro 50 Fig. 2: Code Table Structure of both Forward (a) and Reverse
• Juliet Pro (b) Transliteration Engine
E. Forward Transliteration
The input to the transliteration unit is either an English
text or an Indian language text document. The text স + ◌্ + ত + ◌্ + র + ◌ী = stী
documents are either based on Unicode, ASCII or ISCII
encodings.
As discussed in the previous sections, different Indian
language characters shares the same Braille representation.
In other words mapping rules to transliterate one Indian @ +S +@+ T+ R+9
language can be simultaneously used to transliterate any
other Indian language text to Braille. Thus, in order to Fig. 3: Transliteration of Bengali conjugate characters.
implement this, we build separate code tables for each of
the languages and based on the input language the
corresponding code table is used. An example of the code
table structure is shown in Fig. 2.
The code tables are stored in a hash table due to which
mapping of input text to Braille takes a constant amount of
ক + ি◌ = িক ক + i = কi
time resulting in a very first transliteration output. The said
. I . A I
method of implementation also makes the system highly
scalable and allows the inclusion of more languages in Special Character
future if required.
Transliteration of English text to Braille is a relatively Fig. 4: Illustration of Rule that Handles Vowels and Matras
straight forward process, where there is a direct mapping There are number of other rules present which are required
between the English character and the Braille character. for the Indian language text to Braille transliteration. The
However, Indian language text to Braille transliteration is rule files are written in such a way that a new rule can be
not as straight forward as English. Transliteration rule for a added at any point of time without disturbing the original
particular Indian language character depends upon its system.
previous and the next character. This process is illustrated
205
Fig. 5: Screenshot of the Indian Language Braille Transliteration System

We construct separate rules to handle code table for each of
F. Reverse Transliteration
the supported languages. Thus, our system can be easily
The Indian language Braille transliteration system allows extended to allow reverse transliteration of any other Indian
reverse transliteration of Braille to Indian language text language texts.
document. This enables a visually impaired to communicate
comfortably with other sighted people. The Braille code to be V. WORKING OF THE SYSTEM
transliterated may be entered into the computer using standard
The different components of the ILBT system have been
editors as discussed in the latter section. After translating the
discussed at length so far. In the following paragraph the
Braille code into text, the text may be given as an input to the
working of the entire system and how each of the above
text-to-speech engine, integrated within the transliteration
mentioned components fit into the system have been
system as discussed before, to allow audio based feedback.
summarized.
The process of reverse transliteration of the Braille text is
As mentioned previously, the ILBT system can be used either
similar to that of the forward transliteration. However, there
as an Indian language text reader or it can be used to
are certain ambiguities that are needed to be resolved in order
transliterate Indian language text to Braille. The system can
to get the proper rendering of Indian language texts. For
take input either from documents created using popular text
example, as discussed above, vowels and matras have got the
editors or the system provides its own simplified Indian
same Braille code. Hence, in order to distinguish between a
language text editor which can take inputs directly from the
vowel and matra, Braille code of the previous character is
keyboard. The keyboard input methods are integrated with
considered. An example of a rule that resolves the ambiguity
operating system in order to accept text in regional languages
between a vowel U+0990, a matra U+09C* and an ASCII
from the keyboard. Thus characters (entered through the
character U+002F is given below:
keyboard) in regional languages are fed into our text editor.
If Prev (U+002F) is U+0041 Then The editor analyses the keystrokes to identify the text
Replace (U002F, U+0990) characters and based on the users command they are
Elseif Prev (U+002F) is U+002C Then immediately sent to either the TTS engine to be spoken out or
Replace (U002F,U+002F) to the transliteration engine to instantaneously get the Braille
Else Replace (U002F, U+09C8) output. The GUI of the system also allows a user to select a
206
Fig. 6: Screenshot of the Indian Language Audio QWERTY Editor

particular portion of text and get instantaneous speech or based on the users requirements, can be changed easily. Some
Braille output. A screenshot of the ILBT system is shown in example of the mapping rules are shown in table 3. Operating
Fig. 5. the system with mouse is particularly helpful to a visually
impaired person and our experiment has shown operating the
Table 3: Mouse click events and their different GUI operations
system with mouse operations are more preferred than
Mouse click GUI operations keyboard operations for certain users.
Left, single NOP
VI. BRAILLE AUDIO QWERTY EDITOR
Left, double Speak word
The primary goal of the audio QWERTY editor is to allow
Left, triple Speak line visually challenged people to create Indian language
Text/Braille documents (see Fig. 6). This requires an interface
Middle single Open dialog
for accepting the regional language text entered through the
Middle double New tab document keyboard and performing different operations on it, like
formatting, printing and saving the text. The creation of a new
Middle triple Transliterate
editor interface was not warranted as it would put additional
Right single Close document burden on the user to learn the new system. Hence, we choose
to use some already existing standard editor with the required
Right double FREE
capabilities. Our investigations proved that Microsoft Word
Right triple FREE XP or its higher versions can be configured to accept text in
regional languages including Indian languages. Apart from
Microsoft Word, the Audio QWERTY editor plug-in can be
Apart from keystroke analysis, the GUI of the editor also
integrated to any other Unicode enabled text editors like,
captures the mouse operations as performed by the user.
Notepad, and Wordpad. However, there are several reasons
Instead of searching for the shortcut keys from the keyboard, a
for choosing Microsoft Word such as:
user can send the same keyboard commands by performing
different mouse operations. The ILBT system generates a 1) Support for Unicode – This ensures that almost all natural
languages of the world can be used with the editor. It also
mouse hook process to captures different possible mouse
implies that a document can contain text in more than one
clicks. Each mouse click operation is mapped to a certain
language.
keyboard operation. The mapping rules are customizable and
207
2) Rendering of Fonts – It uses proper rendering engines for VII. TRANSLITERATION OF DZONGKHA TEXT TO BRAILLE
correct rendering of regional language fonts including The Dzongkha script also known as the Bhutanese script is
glyph shaping and repositioning [22], [23]. used to write Dzongkha which is the national language of
3) Well documented object model – the editor in question Bhutan. The Dzongkha script has 195 alphabets out of which
exposes a comprehensive set of objects for interacting there are thirty consonants and four basic vowels. The thirty
with it, hence eliminating the need for the editor source consonants can occur twice in either nominal position or in
code. This also simplified the task of programming as it orthographic subjoined position [25]-[27]. This establishes the
allows the programmer to obtain various information fact that Dzongkha scripts can be written either from top to
including editor settings, text being edited, formatting bottom or from left to right [25]. This is illustrated with an
information, etc. by just querying the editor. example in Fig. 7.
4) Ease of Use – the existing popularity of this editor Similar to the Indian language scripts discussed above, two
predicts a low learning curve for the proposed system as or more consonants can combine to form a conjugate
it allows the user to utilize past experience with this Dzongkha character. However, these consonants are not
editor. Furthermore, the editor provides a large number of separated by any special symbols (as in the case of Indian
keyboard-shortcuts for performing many common tasks languages where the consonants within the conjugate are
which makes the system easy to use visually impaired separated by a special symbol called halant). The consonant
users. It also promotes the use of mainstream applications clustering in Dzongkha takes place between the consonant at
by blind users. the nominal position and consonant at the orthographic
One of the unique features that makes the audio QWERTY subjoined position. This is illustrated in Fig. 8.
editor different from other commercially available text editors
is that, the editor is integrated with the Indian language Text-
To-Speech system (Shruti) as a result each of the keyboard
operation performed through this editor is followed by an
Indian language voice feedback. This enables it to be used by
any visually impaired person to read and write Braille texts.
In order to write regional language or Braille documents the Fig. 7: Dzongkha Characters arranged from top to bottom taken
system provides different keyboard layouts to the user. By from [25].
selecting a particular language from the operating system, a
user actually enables the keyboard layout for that particular
language. This allows the user to write text documents in the
selected language.
A keyboard layout or an IME (Input Method Editor) allows ང + ྐ = ངྐ
a person to type in text in a particular language using the
keyboard [24]. In other words, a keyboard layout or IME can
be defined as a mapping function between different key
ས +ྐ +ྱ +ེ = སྐྱེ
strokes and corresponding characters produced by them. It is Fig. 8: Construction of Dzongkha Conjugates
usually in the form of a dynamic link library and binds
strongly with the operating system. The Windows operating
system can maintain a number of such keyboard layouts
simultaneously. It also allows the user to switch dynamically
between several keyboard layouts associated with different
languages. Hence the user can type in text in different
languages within the same document by switching keyboard
layouts.
In order to type in text in a regional language like Hindi or
Bengali the corresponding keyboard layout is required.
Microsoft Windows provides keyboard layouts for a number
Fig. 9: Construction of a Dzongkha Syllable taken from [25]
of regional languages. However, IMEs are not available for a
large number of other regional languages. Such IMEs can be Unlike any other Indian language texts, there exist no inter-
developed with the help of Windows DDK (Driver word separation in Dzongkha. Each of the Dzongkha words
Development Kit). In our case such an IME for Indian are composed of single or multiple syllables. The syllables are
language Braille was developed so as to enable the user to seperated by a special symbol called tsheg. Each of the
type Braille text directly into any Windows application. syllables contains a root letter, and may additionally contain a
prefix, suffix, vowel and post-suffix. Fig. 9 illustrates this
phenomena with an example.
208
Fig. 10: Working of the Dzongkha text to Braille transliteration system

based on the above mentioned information, we have end with either tsheg or she. This issue is handled by an
constructed separate forward transliteration rules for inbuilt auto formatting module of the transliteration system.
Dzongkha text to Braille. Most of the transliteration rules are The auto formatting module first analyzes the transliterated
very similar to Indian language text to Braille transliteration. Braille output and puts 38 Braille characters per line at the
However, there are few exceptional cases like, handling of Braille preview window. If the last character of a line does not
some special conjugates in Dzongkha script, where the belong to tsheg or she then the auto formatting module starts
transliteration rules changes. For example, Braille accumulating the previous characters into an array till it gets a
representation of certain Dzongkha characters like “ra”, “la” tsheg or she. The array elements are then printed into the next
and “sa” depends upon the occurrence of the character that line of the preview window. This result into the fact that after
follows it. This process is represented by the following rule: transliteration is over all the lines in the Braille output does
CONJ { not contain 38 characters.
PREFIX {
U+0f62|U+0f63|U+0f66, U+0F62 U+0F90 U+0F35
U+0f90|U+0F92|U+0F94
}
} Æ PutBraille (53|59|57)
The rule says, if the head letter of a Dzongkha conjugate
ར + ྐ = རྐ Æ 5 k
begins with U+0f62, U+0f63 or U+0f66 and the root
latter belongs to U+0f90, U+0F92, or U+0F94 then the
Braille representation of the head characters will changed to
“53”, “59” or “57”. Fig. 11 illustrates the above rule with an
ར + ྑ = རྑ r <
Æ
example. Fig. 10 shows a screenshot of the working of the
Dzongkha text to Braille transliteration system. U+0F62 U+0F91 U+0F52
The Dzongkha Braille allows 38 characters per line and 25
lines per page. However, the system allows the user to change Fig. 11: Two different Braille representation of the Unicode
the configuration if needed. As mentioned above Dzongkha character U+0F62 in conjunction with U+0F90 and U+0F91.
script does not allow any inter word space. However, the
syllables and sentences are separated by the special symbols VIII. FIELD TESTING AND DEPLOYMENT
like tsheg and she. An interesting feature found in Dzongkha
The ILBT system has been field-tested and deployed to
Braille formatting is that, each line of a Braille document must
209
several institutes for the visually impaired both in India and

Transliteration Errors
Bhutan. Some of those institutes are:
1) Jadavpur University, Kolkata ILBT Sparsha
2) Voice of World, Kolkata
3) Rotary Sadan, Kolkata 10
transliteration Error
4) School for the blind, Dibrughar 8 8
Number of
5) Curriculum and Professional Support Division (CAPSD) 6 6 6
Ministry of Education, Bhutan
4 4 4
As a result of this deployment, the system is undergoing 3
through a constant process of enhancement. A number of 2
1
feedbacks and suggestions from the visually impaired persons 0 0
have lead to the development of different features like 400 1000 1200 1600
integration of Indian language text to speech, accessing the Num ber of characters
system through the mouse, and inclusion of Dzongkha
language for Bhutan.
Based on the user feedback we perform a three level of Fig. 12: Comparison of Transliteration Errors between ILBT
system and Sparsha Toolset
evaluation for the current Indian language Braille
transliteration system. This includes:
B. Second level of evaluation: Transliteration Speed
1. First level of evaluation: Transliteration Accuracy
2. Second level of evaluation: Transliteration Speed In the second level of evaluation our primary aim is to
3. Third level of evaluation: Task Execution Time evaluate the transliteration time taken by the system. For this
we have created 7 Indian language text documents whose
A. First level of Evaluation: Transliteration accuracy sizes ranges from 25000 characters to 75000 characters
In the first level of evaluation our primary objective is to without considering the space. We apply the same evaluation
achieve high transliteration accuracy. The accuracy of the technique as discussed in [2]. Each of the documents was
system has been measured manually by people working in tested in an Intel Core 2 Duo 2.4 GHz CPU with 2 GB RAM
different blind schools and organizations. We create four and 160 GB HDD. The result is shown in Fig. 13.
Braille document based on Bengali texts and three Dzongkha
Estim ated Transliteration Speed (in m sec.)
Braille documents. All the documents are utf-8 encoded. The
specifications of the Bengali text documents are shown in ILBT Sparsha
table 4.
250
Table 4: Braille transliteration test data specification 223
211
200 197
Time (in msec.)
186 181
File Characters Text type
150
135 135
Doc 1 400 Only Bengali text 117 123
100 103 93
62 71
Doc 2 1000 Only Hindi text 50 46
Doc 3 1200 Bengali & English text mixed 0

25000 31000 40000 50000 67000 70000 75000
Doc 4 1600 Hindi & English text mixed Num ber of Characters
These Braille documents are read by a blind person and the Fig. 13: Comparison of the computation time taken for Indian
transliteration errors are marked manually by the users. We language Braille transliteration in Sparsha and ILBT.
use the existing Sparsha toolset as a lower level baseline for
our system. The transliteration output as generated by the two From the result it can be noted that the ILBT system
systems is compared automatically by a program. The performs considerably better than the Sparsha toolset. The
comparison is based on two parameters a) transliteration reason behind such a marginal difference may be due to the
similarity and b) Output alignment format. Fig. 12 shows the usage of separate hash table for each of the Indian languages.
result of the above experiment. We can observe that the The usage of hash table reduces the searching time complexity
transliteration accuracy of ILBT system is better than the to a constant time as a result the overall transliteration time
Sparsha system. Most of the errors are occurred due to some reduces drastically.
of the special Indian language characters like U+0950 (ॐ),

C. Third level of evaluation: Task Execution Time
In the third level of evaluation, our primary goal is to compute
U+0960 (ॠ), U+095B (ज़), and U+09F3 (৳) which are not the task execution time using the present GUI design. We
compute the task execution time using the Keystroke Level
present in the Text-Braille code table.
210
Model (KLM) [28], [29]. To accomplish a given task, a user Table 6: Action Sequence for “Mouse Movement and Selection”
must perform certain keyboard and mouse operations. The Method for ILBT System
KLM identifies these operations and assigns a timestamp Action Performed Time
value to each of them. These timestamp values are then added (in sec.)
to get the final task execution time. The keyboard and mouse
operations with their estimated time, as discussed in [28], are Move the mouse pointer to the File menu (P) 1.1
discussed in table5. Click over the File menu (BB) 2*.1=.2
Table 5: Keyboard/ Mouse operations with their estimated time
Move the mouse to the Open menu item (P) 1.1
in KLM
Operations Time Move the mouse to the Bengali menu item (P) 1.1
(in sec.) Click on the Bengali menu item (BB) 2*.1=.2
Key press and release (K) 0.28 Double click to select the file from the open 4*.1=.4
Move the mouse to an object on screen (P) 1.1 dialog box (BBBB)
Button press or release (mouse), (B) 0.1 Estimated Time 4.1
Hand from keyboard to mouse or vice versa (H) 0.4 Table 7: Action sequence for “Mouse Movement and Selection”
method for Sparsha system
Mental Preparation (M) 1.2
Action Performed Time
Type string of n characters (T(n)) n*K (in sec)
Move the mouse to the file menu (P) 1.1
We define the task of a user as “Transliterate a text document
and view the print preview of the Braille output”. To Click the file menu (BB) 2*.1=.2
accomplish this task, the ILBT system offers the following 1.1
Move the mouse to the open menu (P)
thee different methods:
• Mouse Movement and Selection Click the open menu (BB) 2*.1=.2
• Mouse Click Operations Double click to select the file from the open dialog 4*.1=.4
Table 5 and Table 6 show the different action sequences and box (BBBB)
the corresponding estimated time of the Mouse Movement and
selection method for the ILBT and Sparsha system. One of the Move the mouse to the language selection dialog 1.1
key observations we made from table 5 is that the mouse box (P)
pointer movement operation is taking around 0.9 sec higher
Click on the language bar to select language (BB) 2*.1=.2
time that the mouse click operations. Hence, we tried to
enhance our present method of Mouse Movement and Click OK button to select input text language (BB) 2*.1=.2
Selection so that we can reduce the number of mouse
Click OK button of the message box (BB) 2*.1=.2
movement operations. Consequently, we came up with a
solution by which a user can perform most of the operations Move the mouse to the transliterate menu (P) 1.1
on the ILBT system with the help of a three button mouse.
The various mouse click operations and their corresponding Click the transliterate menu (BB) 2*.1=.2
GUI commands are explained in Table 3 of section V. Move the mouse to the “text-Braille” button (P) 1.1
We try to perform the same task of “Transliterating a text
document and view the print preview of the Braille output”. Click the “text-Braille” (BB) 2*.1=.2
This task has been performed using the Mouse Click Move the mouse to the OK button of the 1.1
Operations method as explained above. Table 7 presents the transliteration dialog box (P)
different actions and the time required to accomplish the given
transliteration task. Move mouse to the OK button of the message box 1.1
From Table 8 we can see that accessing the system with the (P)
help of mouse click events considerably reduces the overall Click the OK Button (BB) 2*.1=.2
execution time of a user. We summarize the third level of
evaluation between the Sparsha toolset and the ILBT system Move the mouse to the print preview menu (P) 1.1
in Table 9. Click the print preview menu to get the preview of 2*.1=.2
Braille document (BB)
Estimated Time 11
211
Table 8: Actions sequences and estimated time for the “Mouse person.
Click Operation” method
Action Sequence Time (in REFERENCE

Sec.) [1] A. Basu, S. Roy, P. Dutta and S. Banerjee. “A PC based multi-user
Braille reading system for the blind libraries”, IEEE Transactions on
Move the mouse pointer to the language bar (P) 1.1 Rehabilitation Engineering, Vol. 6, No. 1, March 1998, pp.60—68
[2] A. Lahiri, J. S. Chattopadhyay, A. Basu, “Sparsha: A comprehensive
Click on the language bar to select the language 2*.1=.2 indian language toolset for the blind”. Proceedings of the 7th
international ACM SIGACCESS conference on Computers and
(BB) accessibility. 2005.
[3] http://v1.dpi.org/lang-en/resources/details?page=246
Middle click to open file dialog box (BB) 2*.1=.2 [4] P. Blenkhorn, “A system for converting braille to print”, IEEE
Transactions on Rehabilitation Engineering, Vol. 3, No. 2, June 1995,
Double click to select the file from open dialog 4*.1=.4 pp. 215-221
box and see the Braille print preview (BBBB) [5] HAL. Dolphin Computer Access,
www.dolphinuk.co.uk/products/hal.htm
Estimated Time 1.9 [6] C.A. Pennington and K.F. McCoy, “Providing intelligent language
feedback or augmentative communication users”, Springer-Verlag, 1998.
[7] T.V. Raman (1996). “Emacspeak – a speech interface”. Proceedings of
CHI96, April 1996
Table 9: comparison between Sparsha and ILBT system for [8] http://www.braille.se/downloads/winbraille.htm
different task completion methods [9] Duxbury Braille Translator, 2000.
www.duxburysystems.com/products.asp
Estimated Execution Time [10] www.indexbraille.com
[11] NFBTRANS. National Federation of the Blind, 2004,
Methods (in Seconds) http://www.nfb.org/nfbtrans.htm
[12] MONTY, VisuAide. http://www.visuaide.com/monty.html
ILBT Sparsha [13] JAWS for Window. Freedom Scientific.
http://www.freedomscientific.com/fs_products/software_jaws.asp
Mouse Movement and [14] N. Kalra, T. Lauwers, D. Dewey, T. Stepleton and M. B. Dias, “Iterative
4.1 11
Selection design of a Braille writing tutor to combat illiteracy”, Proceedings of the
2nd IEEE/ACM International Conference on Information and
Mouse Click Operation Feature Not Communication Technologies and Development, December, 2007.
1.9 [15] Shruti, Media Lab Asia Research Laboratory, Indian Institute of
Available
Technology, Kharagpur. www.mla.iitkgp.ernet.in/projects/shruti.html
[16] Microsoft Typography, “Specifications : overview”,
www.microsoft.com / typography/ Specifications Overview.mspx
IX. CONCLUSION [17] Unicode. http://www.unicode.org
The Indian language text to Braille transliteration system is [18] Technology Development for Indian Languages, Department of
Information Technology, Ministry of Communication & Information
an attempt to bridge the communication gap between the blind Technology, Government of India. Available at
and the sighted people. Though a considerable effort has been http://tdil.mit.gov.in/standards.htm
done previously by the Sparsha system, however, all the [19] iLeap. Centre for Development of Advanced Computing.
http://www.cdacindia.com/html/gist/products/ileap.asp
previous attempts have provided some limited capability to the
[20] A. Taylor. “Choosing your Braille embosser”, Braille Monitor, October
system. One of the objectives of our present ILBT system is to 200. Available at
improve the performance of Indian language Braille www.nfb.org/bm/bm01/bm0110/bm011007.htm
transliteration and provide more features and flexibility to the [21] www.brailler.com/
[22] J. Hudson for Microsoft Typography, “Windows glyph processing : an
system. open type primer”, November 2000,
The system can perform both forward as well as reverse [23] http://www.microsoft.com/typography/glyph%20processing/intro.mspx
transliteration of text document. This will help improve the [24] R. Rolfe “What is an IME (Input Method Editor) and how do I use it?”
http://www.microsoft.com/globaldev/handson
low literacy rate and provide more information to the visually [25] http://www.tcllab.org/events/uploads/pema-bhutan.pdf
impaired community of India. The system is integrated with [26] http://www.omniglot.com/writing/tibetan.htm
an Indian language text to speech system which gives an [27] http://www.learntibetan.net/grammar/alphabet.htm
[28] D. Kieras, “Using the keystroke-level model to estimate execution
instantaneous audio feedback corresponding to a selected text. times”. Unpublished report. University of Michigan.1993.
We also discuss about the different aspects of the Dzongkha [29] S.Card, T. P. Morn, and A. Newell. “The keystroke-level model for user
script and its difference with other Indian language or English performance with interactive systems”, Communications of the ACM, 23
script. We have extended our system to include Dzongkha text (1980), 396-210
to Braille transliteration which is the first attempt of its kind.

We have evaluated our system based on three parameters and
showed that the ILBT system performs much better that the
existing Indian language text to Braille transliteration systems.
In the next phase of our work we will develop the web
version of the ILBT system and try to integrate the system
with an automatic speech recognition system (ASR) so that
the system can be used directly by any visually impaired
212
Analyzing Statistical Relationships between Global

Indicators through Visualization
Prabath Gunawardane∗ , Erin Middleton† , Suresh Lodha∗ , Ben Crow† and James Davis∗
prabath@soe.ucsc.edu, emiddlet@ucsc.edu, lodha@soe.ucsc.edu, bencrow@ucsc.edu, davis@cs.ucsc.edu
∗ Department of Computer Science † Department of Sociology
University of California Santa Cruz
Abstract—There is a wealth of information collected about having more information is definitely better, understanding and
national level socio-economic indicators across all countries each visualizing this data becomes a harder problem.
year. These indicators are important in recognizing the level of Many websites are utilizing this large collection of socio-
development in certain aspects of a particular country, and are
also essential in international policy making. However with past economic indicator data to visualize global inequality. Popular
data spanning several decades and many hundreds of indicators websites include CISEIN [25], Gapminder [9], NationMaster
evaluated, trying to get an intuitive sense of this data has in [17], UC Atlas [24], and WorldMapper [5]. These websites
a way become more difficult. This is because simple indicator- utilize a number of classic visualization techniques including
wise visualization of data such as line/bar graphs or scatter line graphs, bar graphs, scatter plots, and geographic maps to
plots does not do a very good job of analyzing the underlying
associations or behavior. Therefore most of the socio-economic allow users to view this raw data in different ways. The tem-
analysis regarding development tends to be focused on few main poral data is almost always visualized using animation. These
economic indicators. However, we believe that there are valuable visualizations take the first step to allow users to investigate
insights to be gained from understanding how the multitude of a variety of questions: How does one country compare with
social, economic, educational and health indicators relate to each other countries in the same geographic region or with similar
other.
The focus of our work is to provide an integration of statistical GDP? How are different socio-economic indicators related to
analysis with visualization to gain new socio-economic insights each other? What policies can be implemented to improve
and knowledge. We compute correlation and linear regression health nationally and globally? However, the simple indicator-
between indicators using time-series data. We cluster countries wise visualization of data falls short of providing a deeper
based on indicator trends and analyze the results of the clustering understanding of associations between various indicators and
to identify similarities and anomalies. The results are shown on
a correlation or regression grid and can be visualized on a world countries.
map using a flexible interactive visualization system. In this work, a team of computer scientists and sociolo-
This work provides a pathway to exploring deeper relation- gists have worked together to create a novel integration of
ships between socio-economic indicators and countries in the statistical tools and visualization with a view to gain new
hands of the user, and carries the potential for identifying socio-economic knowledge. Our goal is to leverage mostly
important underpinnings of policy changes.
the familiar and well-known statistical (correlation, linear re-
gression, and clustering) and visualization techniques (scatter
I. I NTRODUCTION AND M OTIVATION
plots and geographic maps) to investigate deeper relationships
Visualization for the purpose of providing intuitive and between socio-economic indicators and countries. Is the in-
deeper understanding of global inequality is an important tuitive understanding provided by raw indicator visualization
problem. Several websites supporting these visualizations us- supported by the results of correlation and linear regression
ing raw data are becoming increasingly popular. However, analysis? Are the causality claims obtained through complex
seemingly easy to understand relationships between variables multi-regression models, often used in socio-economic lit-
visualized using line graphs, bar graphs, and scatter plots, erature, validated or contrasted by correlation or regression
can sometimes provide incomplete information and may even analysis? We view our system as a first step towards building
lead to misleading or erroneous conclusions. In this work, a bridge between the simple approach of using a raw indicator
we propose using statistical tools combined with visualization visualization and the high-powered causality or other policy-
to provide a deeper and more complete understanding of based models.
relationships between the global socio-economic indicators. Our system features an easy-to-use interface where the user
There is a large amount of data collected across all countries can interactively select and visualize multiple countries and /
annually over a range of socio-economic indicators by various or indicators. We have coupled it with the Globalization-Health
agencies including World Bank [2], United Nations [26], Nexus Database [21] to analyze the relationship between
UNESCO [7], and [18] . For example the World Development various health indicators. Furthermore, we have contrasted the
Indicators Database [2] has data that covers 225 countries and observations of both raw and statistical visualizations with the
regions, spanning 40 years for more than 500 indicators. While causal relationships between these health indicators obtained
Submission213
Id : 94
using a sophisticated econometric model by Cornia et al. [4] the sorted correlation matrix (right diagram of Figure 2)
This integration allows us to get a much better and deeper evokes interest by social scientists, and the resulting mapping
understanding of the similarities, anomalies, and evolution of of one of the columns of the correlation matrix on to the
indicators and countries. geographic map (Figure 3) is of great interest to all. We
have also developed a user interface that allows easy selection
II. R ELATED W ORK of indicators and countries from a variety of databases and
There has been considerable advances in visualizing geo- visualizations to create customized visualizations (including
graphic information data using a variety of novel techniques zooming and data mining features that allow users to gain
[13]. A majority of these techniques include using a combi- access to detailed underlying raw or computed data) that may
nation of texture and color to create a palette that can be used be helpful in analyzing the data at hand.
to display multivariate data [14], [15], [16]. Due to challenges Our main focus is to investigate whether the integrated
associated with understanding animated data, spatiotemporal statistical-visualization system can provide any new socio-
geographic data has been visualized using wedges, circles, and economic knowledge or insights. We applied our system
rings [23] and mashups [27]. Distortions of geographic areas to investigate deeper questions regarding health variables.
using rectangles, cartograms and a combination of cartograms In Section V, we present three examples of the results of
with pixelmaps [19] have also been used to convey the values our investigation. Due to simple and familiar visualizations,
of socio-economic indicators. Additional efforts to visualize social and computer scientists could share and understand the
geographic data include geographically weighted scale varying results equally well to create a meaningful dialogue. Many
visualization [6], diffusion-based density equalizing maps [10], of these investigations validated the understanding obtained
and two-tone pseudo-coloring to visualize one-dimensional through simple means, but the system produced some new
data [22]. [12] presents interactive feature section for identify- and surprising results and is also helpful in quantifying the
ing subspaces together with interactive hierachical clustering intuitive understanding.
to assist visualization.
III. V ISUALIZATION
While many of these techniques appear promising and are
very impressive from a visualization standpoint, most social Global socio-economic indicators can be captured in a 3D
scientists and users are unfamiliar with these techniques and volume as illustrated in Figure 1. Although one can attempt
remain wary of depending on these techniques to gain a better to view all the data in 3D, social scientists are much more
understanding of data. accustomed to familiar 2D visualizations. In this work, we
Integrating a statistical model with visualization has been first describe the typical visualizations associated with the 1D
also explored in the literature. Carr et. al. presented a way to and 2D of this 3D volume.
integrate statistical summaries with visualization by the use
of linked micromap and conditioned choropleth maps for spa-
tially indexed data[3]. The concept of using glyphs to visualize
a correlation matrix has been explored in [8] . Andrienko
et. al. use an iterative interactive approach to classify and
identify patterns in spatial data, by using visualization and data
mining [1]. Guo et. al. have presented an approach to cluster
and sort large multivariate datasets based on self-organizing
maps [11]. While these are general visualization toolkits, our
application is more tailored towards the needs of our target
audience, social scientists, and specifically intended to study
country/indicator based patterns relative to each other.
In the integrated geographic statistical-visualization system
that we have built, we are investigating relationships between
causality, simple statistical relationships between indicators
and countries, and intuitive understanding as obtained through
simple visualization. We have chosen to use the causality Fig. 1. The 3-dimensional volume of indicator (I), country (C)and time (T)
model for global health indicators recently proposed by Cornia data, with (i) a vertical 2D slice highlighted which shows times series data
for all indicators for a specific country, (ii) a horizontal 2D slice showing
et al. [4], that we describe in further detail in Section IV-C. time series data for a single indicator over all countries, and (iii) a vertical
In order to contrast our results with those obtained by Cornia, 2D slice for all indicators for all countries at a specific time.
we have integrated our system to draw data from the GHND
database [21] that has been used by Cornia et al. in their The three 1D slices of the 3D volume of global socio-
study. We have also integrated other databases including the economic data are C-slice (one indicator, one time, all coun-
World Bank indicators. We have introduced a visualization of tries), T-slice (one indicator, once country, all time), and I-slice
correlation and regression matrix (left diagram of Figure 2), (one country, one time, all indicators). The C-slice is typically
that has been used mostly by computer scientists; However, visualized by mapping the indicator values on a geographic
2
Submission214
Id : 94
map using pseudo-color and is perhaps one of the most popular countries to support or contrast the findings based on statistical
geographic visualizations. The T-slice is commonly visualized or regression analysis.
as a time-series line graph. The I-slice is usually presented as In addition, we have also supported multiple overlaid scatter
a table. plots (see Figure 9) where users are allowed to pick individual
In addition to these 1D slice visualizations, visualization or some group of countries.
of 1 12 D data, that is 2 or more layers of these 1D slices, 3) IC Slice for a Time Period: The Indicator-Time slice
is very common. We will refer to two layers of C-slice as represents all the indicators across all the countries for a given
a 2C-slice. A 2C-slice may represent two indicators, one year (Figure 1 (iii) ). We are not aware of any effective way
time, and all countries or one indicator, two times, and all of visualizing the whole 2D slice of this data. Again, it will
countries. In the first case, the data is ideal for visualization on be useful to reduce the dimensionality of indicator space.
a geographic map using bivariate display techniques, although As we will see soon, statistical tools allow us to quantify
there is no one commonly accepted technique except perhaps the relationship between two rows or columns of the given
for side-by-side display of two geographic maps. In the second volume of data and visualize them providing us with better
case, although animation is commonly accepted, technique understanding of relationships between the indicators or the
of small multiples is often employed in practice where two countries.
static images are displayed side-by-side. Gapminder [9] has
IV. S TATISTICAL T OOLS
developed a technique of visualizing a 4C-slice of 4 indicators
using a scatter plot where 2 variables are mapped on the x-axis A. Correlation
and y-axis and two additional variables are depicted through In this work, we utilize correlation in at least two ways
glyph size and glyph colors. – to compute correlation coefficient between two indicator
A 2T-slice may represent two indicators, all times, for one trends for a given country (shown as A in Figure 1) and
country or one indicator, all times, for two countries. This 2T to compute correlation coefficient between two countries for
slice is typically visualized using line graphs or bar graphs. a given indicator trend (shown as B in Figure 1). The first
An NT-slice can also be visualized in similar ways within the approach is useful if we want to analyze how two given socio-
space constraints. economic indicators have varied over time with relation to
A 2I-slice may represent one time, two countries and all each other. We compute their correlation for each country,
indicators or two times, one country and all indicators. The which gives us a single correlation value per country which
purpose of these 2 slices are very different. In the first case, is visualized on a geographic map. This analysis can be used
the goal is to compare the two countries, while in the second to answer questions such as ’Is an increase in immunization
case, the goal is to examine all the indicator trends for the same always correlated with a decrease in infant mortality?’
country. This data is typically presented in a tabular format or In the second analysis, we compute the correlation trends
if a subset of indicators is chosen, then this subset can be (over a period of time) between countries for a single indicator.
visualized using classical visualization techniques including Since each country pair would have a correlation value, we
line or bar graphs. visualize these results using a correlation matrix (see Figure 2).
A cellrow,column in the matrix represents the correlation between
A. Slice Visualization
the indicator trend of a countrycolumn and the indicator trend
1) TC Slice for an Indicator: The Time-Country slice (Fig- of a countryrow . A single column of this matrix corresponds
ure 1 (ii)) represents the data for a single indicator spanning to correlation of the indicator trends of a specific country with
all countries over a period of time. Most websites visualize the indicator trend of all the other countries. This analysis can
this slice using an animation of a world map where countries be used to cluster countries based on indicator trends. We can
are pseudo-colored based on indicator values for that point in get a better understanding of how close those countries match
time. up by sorting and visualizing the correlation matrix by that
2) TI Slice for a Country: This slice (Figure 1 (i)) is useful particular country (right diagram of Figure 2). This analysis
in understanding the evolution of socio-economic trends within can also be used to determine whether a particular indicator
a country. The full 2D slice includes all indicators and is trend, for example, the increase in life expectancy, has been
difficult to visualize. A useful task in this case would be to uniform in all parts of the world, and allows easy identification
reduce the dimensionality of socio-economic indicator space of anomalies (see Figure 3).
by identifying a subset of key indicators for a chosen country. We use Pearsons product-movement correlation coefficient
This would mean restrict oneself to a few rows of the TI [20] as our correlation estimator. The correlation coefficient
slice, typically by choosing specific indicators. These indicator gives a measure of positive and negative linear correlation,
trends are then visualized with line graphs and bar graphs. ranging from +1 to -1.
One can also investigate relationships between a pair of
indicators for a specific country by employing a scatter plot B. Regression
by graphing one along the x-axis and the other one along the In addition to computing the correlation between two indi-
y-axis at different time periods. In this work, we have used cator trends for a country, we have also computed the linear
this type of scatter plots (Figures 6, 7, and 8) for individual regression fit for these indicator trends by taking one of the
3
Submission215
Id : 94
Fig. 2. (left)Correlation matrix for LEB (Life expectancy at birth) for years 1980-2005 between 207 countries; (right) The
same correlation matrix sorted by the column for Sweden, indicating correlation of other countries with Sweden for LEB from
1980-2005.
Fig. 3. Correlation of other countries with Sweden for LEB from 1980-2005 shown on a world
map.
indicators to be the independent and the other the response consumption, smoking rates, unbalanced diet, migration rate,
variable. DPT immunization rate, wars, disasters, etc. Impact of these
yi = β0 + β1 xi + εi , independent variables are studied on a cluster of health vari-
ables including u5MR (infant mortality under 5), IMR (infant
where yi is the dependent or response variable, xi is the mortality rate), and LEB (life expectancy at birth).
independent variable and εi is the residual. One would expect
To improve the goodness of fit, improve the robustness
highly correlated indicators to lead to a good linear regression
of the estimates, and avoid multi-collinearity problems, some
fit and the regression coefficients β1 and β0 (which is also
variables were dropped, normalized or modified. One such
referred to as the intercept and the slope in the case of linear
variable is log (physicians/1000 people) which was divided
regression) can be used to understand the relationship between
by log (GDP per capita) to obtain an index of availability of
the two indicators. Together with the correlation visualization
distribution of health personnel relative to the GDP/c norm.
for the two selected indicators, we also provide a visualization
of the regression coefficients on a geographic map. The estimation was carried out for all the countries together,
and also for four different groupings of countries – high
C. Causality income, middle income, low income, and transitional countries
Recently, Cornia et al. [4] proposed a causality model and for two different time periods, 1960-2005 and 1980-
for global health indicators investigating five different impact 2005. Obtained results were examined for their statistical
pathways for health. These pathways are material deprivation, significance better than 1%, between 1 to 5%, between 10
progress in health technology, acute psychological stress, to 15%, and not significant.
unhealthy lifestyle pathways, and socio-economic hierarchy- Results most relevant to our work include statistically sig-
disintegration. Each of these pathways are measured by a nificant dependence of u5MR on DPT immunization rate for
cluster of socio-economic indicators that include income, all the countries as well as for all the four subgroupings of the
income inequality, unemployment rate, inflation rate, illiter- countries mentioned above and the dependence of LEB on Log
acy rate, health expenditure, number of physicians, alcohol (Physicians/1000 people)/ Log(GDP/c). In this work, we chose
4
Submission216
Id : 94
Fig. 4. Correlation coefficients between U5MR (under 5 Fig. 5. Under 5 mortality rate at zero intercept, after linear
mortality rate) and DPT immunization rate for years 1960-2005. regression between U5MR and DPT immunization rate shown
This world map depicts that U5MR is negatively correlated with only for high and low income countries from 1960 - 2005. This
DPT imm. for most countries as expected. Anamolous countries, map brings out relatively high u5MR for low income countries
such as Germany, Kazakhastan, and Congo are easily detected at comparable level of DPT immunization.
in this visualization
Fig. 6. Scatter plot between U5MR vs DPT Fig. 7. Similar scatter plot for Germany. Fig. 8. Similar scatter plot for Kazakhstan.
for Congo, from 1960 to 2000. Deviation Deviation from the norm is due to variation Deviation from the norm is due to political
from the norm is due to war. in health polivy during 1980-2005. changes.
Fig. 9. Scatter plot of u5MR vs. DPT showing 9 countries, 3 from each income group. These scatter plots reaffirm the general clustering og
high, middle, and low income countries into three separate clusters, characterized by low, middle, and high u5MR at comparable DPT levels.
5
Submission217
Id : 94
to focus on these 4 variables – u5MR, DPT immunization

rate, LEB, and Log (Physicians/1000 people)/ Log (GDP/c)
together with GDP data for classification of countries.
D. Clustering
Classification and clustering of countries and indicators
based on similarity is a common and useful endeavor. Existing
solutions deal mainly with classifying the countries based
at a fixed point in time, an example would be the world
bank classification of countries in to ’high’,’middle’ and ’low’
Fig. 10. Correlation coefficients between LEB (Life expectancy income groups. In our system, we can provide results of
at birth) and log(Physicians per 1000 people)/log(GDP per clustering, using K-means algorithm, on any socio-economic
capita) for years 1960-2005. This correlation is positive for indicator such as life expectancy and immunization rates.
most countries including high income countries as expected.
Also we allow clustering over a period of time, based on
the correlation results of indicator trends. For example, if
we could cluster countries that had an life expectancy trend
similar to Sweden for the period 1980-2000. This allows
grouping of countries with similar characteristics over a period
of time, as opposed to just a single year. In most of our
examples discussed later, we have clustered countries into four
categories, although our system allows choosing the number
of clusters.
V. A NALYSIS AND V ISUALIZATION

Fig. 11. Countries with negative correlation coefficients for We now present three examples to illustrate how the inte-
LEB (Life expectancy at birth) vs log(Physicians per 1000 peo-
ple)/log(GDP per capita) for years 1960-2005. Most countries gration of visualization with statistical tools have provided us
with negative correlation are erstwhile Russian block countries with valuable socio-economic insights. All our examples draw
and a few African countries. from the highly reliable GHND database of socio-economic
indicators [21]. We have chosen to focus primarily on health
indicators so that we can contrast or validate our results against
those obtained by Cornia et al. [4], which was described
previously. For this work, we have chosen a subset of these
indicators and variables to illustrate the utility of our integrated
visualization-statistical tool.
For health indicators, we have chosen u5MR (Infant mor-
tality rate under 5) and LEB (Life expectancy at birth). For
independent variables that impact health, we have chosen
DPT imm (DPT immunization rate) and Log (Physicians/1000
people)/ Log (GDP per capita). In addition, we have also used
Fig. 12. Countries with negative regression slopes for LEB
(Life expectancy at birth) vs log(Physicians per 1000 peo- GDP per capita. Most of the data is available for 137 countries
ple)/log(GDP per capita) for years 1960-2005. These are the for 204 indicators over the time period 1960 to 2005.
same set of countries as the countries with negative correlation
coefficients. Correlation and regression analysis agree with each
other. A. Statistical Visualization: Anomalies and Similarities
In this example, we focus on validating how correlation
analysis and visualization may be helpful in analyzing re-
lationship between indicators. To this purpose, we chose to
explore the relationship between u5MR and DPT imm for all
the countries. Correlation between these two indicators are
computed for all the countries individually for a time period
of 1960-2005. This correlation coefficient is then visualized
on the world map in Figure 4. This map clearly brings
out that there is a strong negative correlation between the
two variables, as expected, for most of the countries, with
few exceptions. This figure validates the common working
Fig. 13. Causal coefficients between LEB and log(Physicians
per 1000 people)/log(GDP per capita) obtained by Cornia et al. assumption that an increase in immunization reduces u5MR.
Causal relationship yields a surprising negative relationship for
high income countries between the variables, which is counter-
intuitive. 6
Submission218
Id : 94
Anomalies in the relationship between u5MR and DPT imm (Physicians per 1000 people)/ Log (GDP per Capita) over the
is also brought out by Figure 4. These anomalies appear as period 1960-2005.
negative or close to zero correlation for some countries. These We first discuss the derivations of the causality model
countries include Congo, Germany, and Kazakhstan. Scatter regarding the relationship between these variables. Cornia et
plots of relationships between u5MR and DPT for these 3 al. [4] derive that the regression coefficient between these
countries are shown in Figures 6, 7 and 8 respectively. Reversal two variables for middle, low, and transitional (Eastern block
or decrease in DPT immunization rate in Congo from 1990 countries) are 11.2796, 14.2350, and 8.6528, being significant
to 2005 is a result of war. Reversal of decrease in DPT at 1% level for middle income countries and being significant
imm rate between 1990 to 2000 in Germany is due to a between 1% to 5% level for low income and transitional
variation in health policy that has been checked since 2000 countries. The relationship between these variables is also
resulting in continuance of the desirable trend. Finally, the significant at 1% level for all the countries together with
increase in u5MR in Kazakhastan from 1990 to 2005 is due even higher regression coefficient of 36.89. Surprisingly, the
to political changes in the country. In summary, the correlation regression coefficient between these two variables is negative, -
visualization on the world map quickly leads us to anomalies; 28.9, also significant at 1% level. These regression coefficients
supporting scatter plots quickly helps us in validating the are visualized in Figure 13, where the negative regression
anomalies and leads us to causes of these anomalies and points coefficient is mapped to the red color, while the other three
towards possible challenges or recommendations for changes coefficients are mapped to yellow, light green, and dark green
in health policy. in increasing magnitude of the regression coefficient. These
We now examine the relationship between the same vari- causality results are in contrast with the correlation coefficients
ables, u5MR and DPT, using linear regression between the two visualized for all the countries in Figure 10. The dark green
variables. After a linear fit, we compute the y-intercept, that is, colors in Figure 10 illustrate that the relationship between the
level of u5MR at a hypothetical zero DPT level. These levels two variables are positive, as expected, that is increasing the
of u5MR are then visualized only for the high and low income number of physicians (compared to GDP per capita) ’results’
countries (excluding the middle income countries) in Figure in an increase in LEB. While the causality model in Figure
5. This visualization brings out the sharp contrast between the 13 points to a hypothesis that in a multi-regression model,
two groups of countries. the overall increase in LEB attributed to other factors such as
This observation is further validated by picking 3 sample Log GDP/volatility, female education, alcohol consumption,
countries from each of the three groups – low, middle, and and smoking, etc. is in fact offset by physicians to bring the
high income – and then visualizing the relationship between model in line with the rest of the countries. This example illus-
u5MR and DPT on a scatter plot in Figure 9. This supporting trated the utility of statistical visualization in bringing deeper
visualization using raw numbers further validates the obser- understanding and checks against the more sophisticated but
vation that the low income countries are typically clustered harder to understand multi-regression causality models.
towards the high range of u5MR and also saturate at higher Correlation and regression computations and visualizations
levels of u5MR than the middle or high income countries. bring further insight into the relationship between the two
This observation leads to the conclusion that DPT can help variables. Figures 11 and 12 show the countries with negative
reduce u5MR only up to a certain point in low and middle correlation coefficients and those with negative regression
income countries and additional health measures need to be coefficients respectively. In this case, we observe that cor-
undertaken to reduce u5MR further. Although this observation relation and linear regression results agree with each other
may seem obvious after these visualizations, the causality strongly. Furthermore, the negative relationships between the
model described by Cornia et al. [4] focus mostly on the two variables are present predominantly for transitional (East-
regression slope and not making any of the observations listed ern block countries). This is, again, a surprising result, since
above since their multi-variable regression model does not the causality model by Cornia et al. [4] computed a positive
accommodate the simple intercept view of linear regression. regression coefficient for these countries with high statistical
Nevertheless, it is to be noted that most users, when browsing significance. These visualization based observations lead us
raw data using popular websites such as Gapminder and UC to believe that the relationship between these two variables
Atlas are intuitively looking for simple relationships between is more complicated than a simple causal one and requires
variables and the closest statistical analogs are typically cor- further investigation.
relation and regression analysis. In the examples discussed
C. Clustering
so far, simple visualizations including scatter plot, correlation,
and regression visualization go a long way to provide valuable We now present our third and final example, using clus-
information regarding the relationship between these variables. tering, to illustrate the utility of integrating statistical tools
with visualization. To this purpose, we classified countries
into four categories using many different indicators. Figure
B. Correlation, Regression, and Causality
14 presents the visualization of countries classified into low
We now present a second example of relationship between (red), lower middle (yellow), higher middle (light green), and
LEB (life expectancy at birth per 1000 children) and Log high income (dark green) countries based on GDP in the
7
Submission219
Id : 94
u5MR(T)
GDP(T)
as compared the trend of Sweden. Results of comparison of
LEB(T)
DPT(T)
Phy(T)
u5MR
GDP
LEB
DPT
BRIC countries are presented in a table.
Phy
Brazil 2 1 1 1 2 1 1 4 1 1 The table along with the visualization show that India is
Russia 2 3 1 2 2 4 1 4 1 1 slightly behind other BRIC countries in u5MR while Russia
India 1 1 2 1 3 1 3 3 2 1 is slightly behind in the u5MR trend. In LEB, again, India is
China 1 1 1 1 2 1 1 3 1 1
slightly behind other BRIC countries while Russia is slightly
TABLE I behind in the LEB trend. Put together, in the two health
T HE RESULTS OF USING KM EANS TO CLUSTER COUNTRIES FOR A
SPECIFIC YEAR (2000) AND ALSO BASED ON TRENDS FOR A SPAN OF
indicator trends, u5MR and LEB, India lags behind other
YEARS (1980-2005). C LUSTERS 1 THROUGH 4 REPRESENTS THE ’ BEST ’ BRIC countries, but can catch up if it maintains its trend,
TO ’ WORST ’ CLASSIFICATIONS RESPECTIVELY. while Russia is at the greatest risk of falling behind in the
health indicators.
In terms of action or independent variables that impact
health, DPT immunization rates for all BRIC countries are in
the lowest two categories, Brazil and Russia being the lowest,
and India and China next to the lowest. However, with respect
to the DPT trend, Brazil, Russia, and China are in the strongest
category (green), while India is somewhat lagging behind. In
terms of the presence of physicians trends, all four BRIC
countries are in the strongest category, while India slightly
behind in the year 2000. Put together, all BRIC countries are
likely to improve their health indicators by increasing the DPT
Fig. 14. Clustering of countries based on GDP in year immunization rate and need to maintain their strong growth
2000 trend regarding physicians.
VI. C ONCLUSIONS AND F UTURE W ORK
In this work, we proposed an integration of statistical
computing with visualization to glean deeper understanding
of global socio-economic indicators. We utilized correlation
and linear regression to quantify relationships between pairs
of variables and between pairs of countries. We utilized these
tools to investigate static data for a fixed time period as
well as dynamic trends over a large time period. Current
Fig. 15. Clustering based on GDP trends between state-of-the-art global inequality websites provide visualization
Sweden and other countries for years 1980-2005. support using raw data without the use of any statistical
tools. Using three different examples, we demonstrate that
correlation, linear regression, and causality models can bring
year 2000. We will follow the results for the emerging BRIC out similarities and anomalies and provide better understand-
(Brazil, Russia, India, and China) countries. In this case, India ing of relationships between the variables by validating our
and China are still low income countries while Brazil and intuitions based purely on raw data visualization or sometimes
Russia are lower middle income countries. However, when yields insights that are counter-intuitive or surprising. These
we view the classification of GDP trends for the four BRIC observations or conclusions carry important implications in
countries over the period 1980-2005 in Figure 15, we observe policy making both at national and global level.
that India, China, and Brazil are classified into the same This research has opened up several new exciting oppor-
(and strongest) category as most of the developed nations tunities. Which countries can be grouped together? Based
including USA, and most European countries, while Russian on which indicators? Which socio-economic indicators can
GDP growth is the next lower category. For the purposes of be clustered together? Can we reduce the dimensionality
the trend classification, we computed the correlation between of indicators so that a profile of a country is captured by
GDP trends between all the countries and Sweden. We chose some principal socio-economic indicators? What lessons can
Sweden because it was consistently in the top for most of a nation learn from a similar group of nations? Ideally, we
the indicators that are we investigating in this study, including would like to build a system so that the empowered users can
GDP, LEB, etc. explore relationships between countries and between variables
Figures 20 to 23 classify all the countries on the four using appropriate statistical tools combined with visualization.
variables – u5MR, DPT imm, LEB, and Log (Physicians per We believe that this exploration can always be used to validate
1000 people)/ Log (GDP per Capita) using the static data from or contrast the proposed policy decisions and may also lead to
the year 2000. Figures 16 to 19 classify the countries based on important underpinnings of national or global policy decisions
the trends of these four variables over the period 1980-2005 that are not immediately obvious.
8
Submission220
Id : 94
Fig. 16. Clustering based on U5MR trends between reference Fig. 17. Clustering based on DPT immunization trends between
country (Sweden) and other countries for years 1980-2005. reference country (Sweden) and other countries for years 1980-
2005.
Fig. 19. Clustering based on log(physicians per 1000 people)

Fig. 18. Clustering based on LEB trends between reference
over log(GDP per capita) trends between reference country
country (Sweden) and other countries for years 1980-2005.
(Sweden) and other countries for years 1980-2005.
Fig. 21. Clustering of countries based on DPT immunization

Fig. 20. Clustering of countries based on U5MR in year 2000
rates in year 2000
Fig. 23. Clustering of countries based on log(physicians per

Fig. 22. Clustering of countries based on LEB in year 2000
1000 people) over log(GDP per capita) for year 2000
9
Submission221
Id : 94
VII. ACKNOWLEDGEMENTS [23] P. Shanbhag, P. Rheingans, and M. desJardins. Temporal visualization

of planning polygons for efficient partitioning of geo-spatial data. IEEE
We would like to thank Brian Fulfrost for providing us with Symposium on Information Visualization, 2005.
valuable feedback at various stages of this project. [24] UCSC. UC atlas. Website, 2008. http://ucatlas.ucsc.edu/.
[25] C. University and W. Bank. Global poverty mapping project. Website,
R EFERENCES 2008. http://www.cisein.org/povmap/atlas.html.
[26] UNSCB. United nations common database. Website, 2008. http://
[1] G. L. Andrienko and N. V. Andrienko. Data mining with C4.5 and unstats.un.org/unsd/cdb/cdb help/cdb quick start.asp.
interactive cartographic visualization. user interfaces to data intensive [27] J. Wood, J. Dykes, A. Slingsby, and K. Clarke. Interactive visual
systems. G. T. Los Alamitos, CA, IEEE Computer Society, pages 162– exploration of a large spatio-temporal dataset: Reflections on a geovi-
165, 1999. sualization mashup. IEEE Transactions on Visualization and Computer
[2] W. Bank. World development indicators. Website, 2008. http://www. Graphics, 13(6):1176–1183, 2007.
worldbank.org/data/.
[3] D. B. Carr, J. Chen, B. S. Bell, L. Pickle, and Y. Zhang. Interactive
linked micromap plots and dynamically conditioned choropleth maps.
In Proceedings of the 2002 Annual National Conference on Digital
Government Research, pages 1–7. Digital Government Society of North
America, 2002.
[4] G. A. Cornia, S. Rosignoli, and L. Tiberti. Globalisation and health:
impact pathways and recent evidence. In Proceedings of Conference on
Mapping Global Inequality, 2007.
[5] D. Dorling, A. Barford, and M. Newman. Worldmapper: The world as
you’ve never seen it before. IEEE Transactions on Visualization and
Computer Graphics, 12(5):757–764, 2006.
[6] J. Dykes and C. Brunsdon. Geographically weighted visualization: Inter-
active graphics for scale varying exploratory analysis. IEEE Transactions
on Visualization and Computer Graphics, 13(6):1161–1168, 2007.
[7] U. I. for Statistics. Global statistics. Website, 2008. http://www.uis.
unesco.org.
[8] M. Friendly. Corrgrams: Exploratory displays for correlation matrices.
The American Statistician, 56:316–324, November 2002.
[9] Gapminder. Gapminder world 2006. Website, March 2008. http://tools.
google.com/gapminder.
[10] M. T. Gastner and M. E. J. Newman. Diffusion-based method for pro-
ducing density-equalizing maps. Proceedings of the National Academy
of Sciences, 101(20):7499–7504, 2004.
[11] D. Guo. Coordinating computational and visual approaches for in-
teractive feature selection and multivariate clustering. Information
Visualization, 2(4):232–246, 2003.
[12] D. Guo, J. Chen, A. M. MacEachren, and K. Liao. A visualization
system for space-time and multivariate patterns (vis-stamp). IEEE
Transactions on Visualization and Computer Graphics, 12(6):1461–
1474, 2006.
[13] D. Guo, M. Gahegan, A. M. MacEachren, and B. Zhou. Multivariate
Analysis and Geovisualization with an Integrated Geographic Knowl-
edge Discovery Approach. Cartography and Geographic Information
Science, 32(2):113–133, 2005.
[14] H. Hagh-Shenas, S. Kim, V. Interrante, and C. Healey. Weaving
versus blending: a quantitative assessment of the information carrying
capacities of two alternative methods for conveying multivariate data
with color. IEEE Transactions on Visualization and Computer Graphics,
13(6):1270–1277, 2007.
[15] C. G. Healey and J. T. Enns. Large datasets at a glance: Combining
textures and colors in scientific visualization. IEEE Transactions on
Visualization and Computer Graphics, 5(2):145–167, 1999.
[16] V. Interrante. Harnessing natural textures for multivariate visualization.
IEEE Computer Graphics and Applications, pages 6–11, November-
December 2000.
[17] Nationmaster. Nations of the world. Website, 2008. http://www.
nationmaster.com/.
[18] OECD. Organisation for economic co-operation and development.
Website, 2008. http://www.oecd.org.
[19] M.-C. Panse, M.-M. Sips, M.-D. Keim, and S. M.-S. North. Visualization
of geo-spatial point sets via global shape transformation and local pixel
placement. IEEE Transactions on Visualization and Computer Graphics,
12(5):749–756, 2006.
[20] J. L. Rodgers and A. W. Nicewander. Thirteen ways to look at the
correlation coefficient. The American Statistician, 42(1):59–66, 1988.
[21] S. Rosignoli, L. Tiberti, and G. A. Cornia. The globalization-health
nexus database (ghnd). Website, February 2007. http://www.unifi.it/
dpssec/sviluppo/database.html.
[22] T. Saito, H. N. Miyamura, M. Yamamoto, H. Saito, Y. Hoshiya, and
T. Kaseda. Two-tone pseudo coloring: Compact visualization for one-
dimensional data. Proceedings of Information Visualization, 0, 2005.
10
222
ATMosphere: A System for ATM Microdeposit

Services in Rural Contexts
Michael Paik Lakshminarayanan Subramanian
New York University New York University
mpaik@cs.nyu.edu lakshmi@cs.nyu.edu
Abstract—This paper describes strategies to lower the cost However these, while providing a convenient method to
of providing Automated Teller Machine microdeposit services send funds from the city to more rural areas, do not provide
in rural contexts. Microdeposits represent a growing market significant capacity for savings as they are intended for
in the developing world, but the cost of running a
person-to-person money transfer rather than the
conventional ATM network is prohibitive due to the capital
investment required to deploy networks and terminals. accumulation or retention of wealth.
Our novel contributions are to use the Short Message This paper discusses findings from a simulation of ATM
Service (SMS) over high-penetration GSM cellular networks transaction behavior using real-world population and
in conjunction with a system using location awareness to geographical data from Arua, Moyo, Yumbe, and
intelligently distribute available balances among machines. Adjumani Districts in Northwestern Uganda, bordering
This allows us to provide high levels of service while Southern Sudan and the Democratic Republic of the
simultaneously reducing risk to the financial institution and
Congo, both of which have been in political upheaval
lowering per-transaction cost.
Using a simulation of ATM usage patterns and and/or civil war during the past decade. This area was
distributions, our primary results under our model are: (1) selected because 1) the population represents the poorest of
transaction cost per user per year can be optimized to less the poor, being made up primarily of refugees from
than USD 0.18 given an SMS loss rate of approximately 10% conflicts in bordering nations who have fled to an
while (2) customer withdrawal success rate can be maintained otherwise sparsely populated region[11] and 2) GSM
at approximately 98% with (3) a maximum of 5% of funds on mobile coverage of the area is relatively dense[23] as seen
deposit available in cash in ATMs at any given time.
in Figure 2. As such, this milieu is a real-world example of
These results make wide deployment of rural ATM services
by financial institutions feasible and economically viable in settlement patterns among the most disenfranchised. The
the near term using existing commodity technology. results presented here focus on satisfying a high level of
consumer withdrawal demand without imposing artificially
Index Terms—Finance, Financial data processing, low withdrawal limits. We present a novel scheme for
Simulation. balance partitioning and message flow control which
makes use of location awareness of ATMs alongside other
1 INTRODUCTION optimizations to provide a solution that is performant both
R EADY ATM access at the village level has been an

open problem in the microfinance field, as some
means for making deposits and withdrawals of small
in terms of risk to the institution and service to the end
customer.
amounts of capital is necessary for the inception and 2 THE PROBLEM

growth of microfinance markets and personal accumulation
of wealth for the rural poor. However, for-profit 2.1 The Economic Challenge
institutional financial entities have found the idea of While the precise cost breakdown of deploying an ATM
providing such access to populations living outside of in a rural environment will vary from context to context, a
urban areas with both high teledensity and population significant part of the marginal cost overhead is in the
density unattractive because the amount of capital creation and use of the network to support communication
represented by rural microdepositors is only institutionally between the managing financial institution and the ATM.
significant once aggregated, and the cost of providing There is little in the way of existing network infrastructure
access to these customers has historically been far greater (including wired telephone service) and the cost involved
than any revenue gained by doing so. This is particularly in laying even a single link given high costs of materials,
true with the high capital investment required for transportation, and scarcity of skilled labor, construction
traditional brick-and mortar retail banks, which also incur equipment, and telecoms equipment is often prohibitive.
the cost of trained staff. More recently, systems such as M- The traditional wired model (used e.g. in the United States)
PESA[27], operated by Safaricom in Kenya have made is infeasible due to the capital investment required to lay
person-to-person payment using mobile phones viable. network cables to sparsely populated regions. This is also a
223
Fig. 1. Population centers in northwestern Uganda Fig. 2. GSM network coverage in northwestern Uganda
(darker areas indicate better coverage)
primary factor in the proliferation of self-contained cellular
telephone towers which uplink via satellite or fixed underlying difficulties, not least of which is cost. A naïve
wireless transmission. Other than the existing wired model, approach to the problem might be to send a series of
there have been efforts to connect ATMs to financial messages back and forth between the ATM and the
institutions using long-distance wireless links[6] which financial institution – one to request a withdrawal, one to
also provide telephony and internet services, but the case issue an identity authorization challenge, another to supply
for capital investment of this sort for areas in which cellular the response, yet another to issue an approval, one to
telephone service already exists is more difficult to make as acknowledge the approval, etc. etc. While the absolute cost
some degree of connectivity is already present. Any of these messages is not extraordinarily high, the cost does
solution which proposes to address this problem must take not scale well as the user base increases. If the institution
into account lack of network capacity as well as keeping pays for the messaging, the cost eats into any financial gain
cost of deployment and operation low enough to be the client may realize from interest on his savings if he is
practicable for national banks in the region to afford. expected to pay, and in either case this approach does not
We use the wired ATM metaphor rather than simpler scale well as loss rates may be expected to increase if
kiosks which read smartcards with encoded balances or message traffic increases substantially while network
paper passbooks because these methods are vulnerable to capacity remains fixed.
counterfeiting, theft, loss, and damage and have poor
3 DESIGN
provisions for centralized auditing.
Our initial objective is merely to provide some form of
ATM service to rural customers with small amounts of
2.2 The Scope of our Approach personal savings. Our simulated testbed contains 46
The approach presented in this paper is aimed squarely population centers of sizes varying between 2001 and
at reducing cost and making maximum use of existing 70872, totaling 637952 in total. We simplify this by
capacity in the form of the installed base of GSM network considering households as accountholders, using the
cells. As the network’s reliability, latency and bandwidth average household size across Uganda of 4.7, yielding
are not subject to any guarantees, we create a protocol 136541 households if we consider each center separately
which is designed to be resilient against extended outage and round down.
conditions and slow message delivery while simultaneously Having ruled out wired infrastructure, our next
maintaining an extremely low message budget and high observation is that long-distance[10] wireless is feasible as
quality of service. a physical network layer, but has the disadvantage of
requiring potentially extensive infrastructure deployment
2.3 The Technical Challenge depending on the geography of the region, distances
between nodes, and the network topology necessary to
The use of the installed GSM base is a major benefit to
scale with growth of the size of the network. Newer
the design of a solution to this problem, but there are
224
technologies such as WiMAX[30] may allow connectivity techniques currently used in other devices, such as credit
between adjacent villages with commodity hardware, but card swipe readers. A possible solution to this problem is to
how much effective bandwidth a rural WiMAX mesh use a similar model to the M-PESA system and locate
installation will have is still an open question, and there is ATM kiosks within vendors’ shops, allowing the vendors
still the question of deployment lead time and network to charge a small fee for their use.
maintenance costs. GSM cellular technology is widespread We assume that any losses incurred through physical
in the developing world, and reaches into some of the most attacks on the machine are covered by the financial
sparsely populated regions. While the General Packet institution, as with machines in urban locations. In the
Radio Service (GPRS) is part of GSM Release 97 and same vein we neglect considerations about powering the
newer (colloquially referred to as 2.5G and 3G), it is not a machine (solar is an option for low-powered embedded
viable option for data transport as much of the installed computers), how deposits will be marked, validated, and
base in rural areas is comprised of cheaper 2G cellular conveyed to the host institution, and the technical details of
hardware. Due to these factors, we choose to use SMS. the tokens used to identify the users of the machine other
Once connectivity is established, a natural design than that these tokens contain some cryptographically
principle for our system is that we want to reduce network secure writable medium, one such device being a smart
traffic as much as possible, as this represents a marginal card (ICC).
cost to our system, and cost is precisely what we wish to Finally, we assume that we may neglect certain
reduce. In opposition to this is a desire to maintain overall transaction metadata, such as the exact timestamp on each
responsiveness of the system as perceived by the customer. transaction, or that several transactions occurred within a
This perception is directly related to satisfaction rate of given timeframe, preferring rather to reduce the
withdrawal requests, and tied only weakly to the communication budget required to send information about
immediacy of funds availability from deposits, as existing a given transaction by reducing timestamp granularity to an
microdeposit schemes often take up to a month[13] to hour or day-level timescale based on the aggregation
register updated balances. A final critical design principle window of transactions and potentially aggregating
is that, given the intermittent network communication multiple transactions which occurred close together in time
enforced by our traffic reduction, we need to intelligently into one larger transaction (e.g. five transactions that
distribute balances in such a way as to prevent any user happened at the same ATM on the same account within 15
from being able to game the system by visiting multiple minutes of each other may be represented as one
ATMs within a given update window and withdraw more transaction which is the sum of those individual
than his account balance. transactions, at an approximate time.)
3.1 Simplifying Assumptions 3.2 SMS

In this paper we make several simplifying assumptions. The Short Message Service (SMS) was defined in 1985
Firstly, we assume that the only parties in communication as part of the draft GSM specification[22] and as part of the
for these rural ATM systems are the machines themselves GSM standard, shares GSM’s ubiquity. As GSM is the
and some server-side system controlled by the host cellular technology most used in the developing world,
institution. This neglects considerations of legal regulations SMS as a data protocol has a significant installed base that
which may enforce some neutrality or transparency upon can be leveraged with little cost. The SMS format can carry
financial institutions. We also assume that because of this 140 bytes per message, at a cost of approximately 0.05 US
simple two-party communication when considered in the Dollars per message in the region our testbed
realm of a single ATM, that communication scales linearly represents[25]. It is important to note, however, that SMS
with transactions (O(n)) for a given ATM. In addition we does not guarantee delivery, nor does it guarantee in-order
observe that there is a limit to the number of transactions delivery for those messages that are delivered, though it is
per day based on human factors. Assuming that an ATM is a store-and-forward system. As such, our protocol has to be
available for 12 hours per day from 8AM to 8PM, and a robust against sporadic dropped packets and out-of-order
transaction takes approximately one minute to execute, we delivery. Additionally, the extremely small payload of the
estimate a rough upper bound of 720 transactions per day messages requires us to address the issues of packet
per ATM. fragmentation and careful data management. GSM does in
We further consider all physical security considerations theory support SMS concatenation of up to 255 messages,
outside the scope of our problem; we assume that any but in practice most systems do not support any more than
machine is located in some safe environment where the 10, providing a theoretical aggregate payload of 1400
likelihood of tampering is low, and also that any tampering bytes. However, SMS concatenation has a per-message
of the machine will be made evident using standard payload cost of 7 bytes per packet for every message after
225
TABLE I
the first, which is a significant amount of overhead given TP-VP GRANULARITY AND RANGE
the payload size. Also, in our testing, 1337 bytes proved to
be too few to take advantage of any generalized TP-VP Implied Validity Period
compression algorithms, with compression of most 0 - 143 (TP – VP +1) * 5 minutes
payloads resulting in an increased size of between 5 and 60 144 - 167 12 hours + (TP-VP – 143) * 30 minutes
168 - 196 (TP – VP – 166) days
percent, depending on the algorithm and the structure of
197 - 255 (TP – VP – 192) weeks
the data.
In areas where GPRS is available, most providers in the direction, including any latency introduced by the internet
developing world provide options for Multimedia Message SMS gateway.
Service (MMS) messaging, an extension of SMS which These latency and throughput numbers are used only to
allows arbitrary payload sizes. MTN provides MMS garner an overall idea of the timescale these messages
messages of up to 100 kilobytes for approximately USD operate on and are not assumed to be statistically
0.125 in urban areas of Uganda, which represents close to a significant.
300-fold increase in messaging budget per unit cost.
Naturally, as this technology expands, marginal costs will 3.2.2 Payload Optimizations
decrease commensurately. Furthermore, even where GSM At a minimum, the message payloads must contain some
coverage is not available, SMS service is available via information about depositor accounts and their balances.
satellite phone coverage (e.g. Thuraya[29]), which allows Based on our design principle of reducing network traffic,
any system built using an SMS metaphor to be generalized we wish for each payload to contain information about
to practically any location on the globe where a clear line more than one transaction. Due to the small payload size
of sight to the sky is available, albeit at far greater cost. allowed in SMS messages, we made several design
decisions to decrease the size of data passed via SMS.
3.2.1 SMS Delivery and Submission Rates We observed that the smallest denomination available in
On-field tests conducted once every 30 minutes from Uganda is 50 Ugandan Shillings (Ushs), or approximately
Adjumani (labeled in the northeast quadrant of Figure 2) USD 0.03, so we chose to encode balance information as
for the 12 hour window from 8AM to 8PM to a handset in multiples of this amount using 20 bits, which gives us a
Kampala using a handset sending 50 pre-written 140-byte maximum amount of Ushs26214350 or approximately
(160 characters with septet encoding) dummy messages by USD 15,000, fully an order of magnitude more than any
hand yielded a minimum send rate of approximately 8 microdepositor is ever likely to have on deposit, while
messages per minute during peak times. Tests using a script maintaining signing. Ugandan checking account numbers
submitting the same 50 messages over the same period on a are mandated to have exactly ten digits[26], which would
subsequent day to MTN’s internet SMS gateway[24] bound require 34 bits to encode, but we can establish a mapping
for the same handset yielded a minimum receipt rate of from account numbers to a smaller unique integer; in this
approximately 12 messages per minute. We consider these case we use 20 bits which gives us a maximum of 1048575
to be sane limits when the network is operational as we discrete accounts. This gives us a total of 40 bits for a
assume that the handset equipment used has a lower- (account number, amount) pair which sets an upper bound
powered radio and significantly less processing power than of 28 of these pairs per 140-byte payload.
an ATM might, but we do not maintain that this rate is an
absolute lower bound.
Loss rates during these tests were 0%, but we do not 3.3 Protocol
treat this result as conclusive due to the small sample size Our overall strategy is to encode a withdrawal
and the fact that these tests were conducted only across one transaction by indicating a negative delta to the balance
day in each direction. Additionally, all messages were indicated on a given machine, and encode balance updates
delivered in-order, which we also do not treat as from the financial institution’s central authority (CA)
conclusive, for similar reasons. Tseng et al. found a loss reflecting verified deposits and redistributions of balances
rate of 0.66%[15] after sending 915 messages over a 38 among machines with positive or negative deltas to the
day period in 2005 in Taiwan. As Taiwan has high GSM balance.
penetration and a more modern cellular infrastructure (all The CA maintains tables containing the GSM phone
but two of Taiwan’s GSM providers had high-bandwidth numbers of every ATM (by which incoming SMS
3G networks by October 2005) we assume that we will see messages are identified), precomputed pairwise distance
greater loss rates on the whole in our testbed. metrics of each machine to every other machine, cash
The median latency for our group of messages was balance in each machine, home ATM(s) for each account,
asymmetric, at approximately 36 seconds from Adjumani and account and balance records.
to Kampala, and approximately 21 seconds in the other
226
a single version number. However, each transaction still

counts towards the transaction limit imposed by the ATM.
During a 30-minute window after the ATM ceases
public operation each day, it will attempt to resend any
Fig. 3. Record structure
messages that are NAK for the duration of the window to
3.3.1 Message Structure compensate for extended periods of loss.
As SMS does not guarantee delivery or in-order
delivery, the addition of a sequence number at some level 3.3.2.2 CA to ATM
is critical to error-free communication. While SMS has no The server at the central authority records any incoming
built-in sequence number, a sent SMS message does have a transaction records from an ATM and applies the balance
user-specified validity period indicator (TP-VP[12]) not adjustments to the appropriate account, updating the
contained within the user data area (and therefore version number in the process. If a version number
essentially free) which guarantees expiration of the received is not the one expected based on the CA’s record,
message if it cannot be delivered to the destination system it is applied, but the missing sequence number is noted.
within a given period of time. In addition, a received SMS
3.3.2.3 Error Detection and Correction
contains a timestamp indicating when the message was
received by the SMS Center (SMSC), which is in-order for In order to protect against silent failure of the CA server
a given source as a device cannot send another SMS or an ATM, we add a heartbeat/echo message containing a
message until one sent prior has been either acknowledged random string which is triggered once every four hours by
by the SMSC or timed out. We encode each (account each ATM as necessary (i.e. after that amount of time has
number, amount) pair with four additional bits which passed without a message having been received), jittered
indicate a version number from 0-15 which wraps around. by up to 20 minutes. Any echo message received by the CA
Combined with the timestamp on delivery, this allows us to is replied to immediately with the same contents. If the CA
keep updates on a given account from a given source in- server does not hear a heartbeat or transaction message
order. We also set the daily transaction count limit for a from an ATM for 360 minutes, the ATM is flagged as
given account at a given ATM to 4, meaning that an ‘down’.
account will not experience version number wraparound An ATM being flagged as ‘down’ ensures that the
for 4 days, and set the TP-VP to be 2 hours, which operation of the local ATM continues as normal for a given
guarantees against version number collision. account until that either account’s available balance at that
Of the 1120 bits available to the payload, we use 1056 ATM or the cash in the ATM is depleted, during which
for 24 (version, account, amount) tuples of 44 bits each, time connectivity may be restored. An ATM is brought
and add a 20-bit checksum to protect against bit errors. We back ‘up’ whenever an update is received from that ATM.
order these tuples by account number. Given our one-per- Thrashing is possible, but the timescales involved reduce
minute estimate of maximum transaction execution rate due the problem, and no additional resources are consumed
to human interface considerations, this gives us an upper through the CA considering a machine ‘up’. Any machine
bound of 30 SMS messages per day for our 12 hour, 8AM which is flagged as ‘down’ during two consecutive batches
to 8PM day, with the rate of messages sent by the ATM is flagged as ‘disconnected’ and becomes ineligible for any
never exceeding one per 24 minutes. balance distribution updates from the server until manual
intervention can be scheduled.
3.3.2 Message Flows At the end of the day after the ATMs have ceased public
3.3.2.1 ATM to CA
operation, the CA waits for 1 hour for any incoming
updates that may have been lost, and applies any received
Each ATM sends a message either when a set of 24 changes, then explicitly queries ATMs for any messages
transactions has been completed or some time t has elapsed still missing using a series of messages containing only the
and at least one transaction is ready to send, where t is 120 missing version number and the account number. Any
minutes with a random jitter up to 5 minutes in order to ATM which still has questionable data at the end of one
prevent synchronization among ATMs. If a message is not hour after the final update request is sent is flagged as
acknowledged by the SMSC within 30 seconds, sending is being ‘down’ and is removed from the set of ATMs
reattempted at most three times, then the versions of the eligible for balance redistribution.
accounts contained in the message are flagged as non- After the day-end synchronization batch described
acknowledged (NAK), and operation continues. In order to above, the CA sends each ATM that is ‘up’ a message
reduce the number of updates sent, multiple transactions on containing a hash of the concatenation of all versions,
the same account within the window of a single message accounts, and balances that ATM is known to be aware of,
are aggregated and treated as a single update and are given ordered by account number and version number. Each
227
ATM computes the same hash upon receipt, and an echo In our testbed, the geography is advantageous as it offers
containing the ATM’s hash value is sent back. Any ATM only a few inter-town river routes, and no tarmac-paved
which returns an inconsistent value is flagged as roads. We also use the intuition that (1) accounts with
‘disconnected’ as above. higher available balances are more attractive to
Messages which are received but fail checksum and overwithdraw, (2) people with higher balances are similarly
cannot be repaired are discarded and considered lost. more likely to have access to faster transport such as boda-
bodas, (local motorcycle taxis), matatus (intra- and inter-
3.4 Security
town minibuses), and interdistrict buses or to own or be
SMS over GSM is susceptible to a variety of attacks[7]. able to borrow a private vehicle, (3) that fraud associated
In particular, it was indicated in a personal communication with higher-balance accounts costs the host institution
to the author by a member of MTN Uganda staff that SMS more, and (4) holders of high-balance accounts are less
is transmitted in plaintext throughout its network, and this likely to need to withdraw all or most of their available
is the default behavior[7]. As this is the case, and as both balances at once.
Originating Address (OA) spoofing and realtime sniffing of We consider the maximum feasible distance covered by
SMS messages are available in the wild, it is necessary for walking over a 3 hour period (~10km) and the maximum
the system to have mechanisms to protect the secrecy of feasible distance covered by driving along unpaved roads
balance/account information and to provide some measure during that same period (~125km). Given that the average
of authentication of messages sent between the ATMs and daily unskilled rural labor wage is in the neighborhood of
the CA. $1.2 per day[13] (approximately Ushs2100), we arbitrarily
In order to maintain effective payload size while set the ‘high balance’ mark at Ushs900000 (USD ~5000).
providing this type of protection, a stream cipher is used, Accounts with balances of 450000 and under have a
which produces ciphertext of the same length as the maximum of 100% of their account balance distributed
plaintext. This system uses RC4 due to its ease of across all the ATMs in the system, decreasing linearly to
implementation, but any such encryption system will serve. 75% at 900000 and beyond.
Each ATM maintains a separate 256-bit key, and the CA Only ATMs within a specific distance of the primary
maintains a keyring of all of these keys, using them as ATM are considered for this distribution, specified by the
appropriate in communication to a given ATM. formula
A 32-bit timestamp is embedded into all messages in
order to mitigate the problem of replay attacks; a timestamp (dmax/2 – d0) * min(1,b/c) + d0
will only be accepted once and any subsequent message
from a given source bearing the same timestamp will be where dmax represents the greatest distance between any
discarded. In order to mitigate substitution attacks, this two ATMs in the network, d0 represents the minimum
timestamp is placed between arbitrary (version, account, radius to be considered (40km in our case), b represents the
amount) tuples and will be padded with 12 bits of 1’s in balance of the account in question, and c is the cutoff ‘high
order both to maintain tuple offsets and to indicate which balance’ mark at which maximum distribution spread is
block contains the timestamp. Version number 15 then reached. This encapsulates our intuition that we want to
becomes reserved as not to restrict the range of account distribute the balance for those with higher balances over a
numbers which may be used. larger area, and the linear increase in radius produces an
increase in area covered proportional to the square of this
3.5 Balance Distribution value.
The proportion of the available balance held in the
The practice of available balance distribution among
primary ATM varies according to the formula
ATMs is a hedge against users who may wish to game the
system by trying to take advantage of network outages or
bprimary = b * pmax – ((pmax – pmin) * min(1,b/c))
periods of higher-than-normal latency in order to withdraw
more than their account balance. We assume for the
where pmax and pmin are the maximum and minimum
purposes of this practice that ATM use multifactor
proportions, respectively (.8 to .4 in our simulation). These
authentication where at least one factor is a biometric, such
numbers are based on assumptions we make about the
as a fingerprint. The hash of this biometric is stored on the
behavior of clients as listed below, and the formula again
card in order to defend against card duplication.
reflects our intuition that the further the client is likely to
We begin by associating each given user with a
travel, the further afield the balances should be distributed.
‘primary’ ATM, which is the machine closest to that user’s
The balances to the ATMs within the effective radius
residence, and then consider the other ATMs which are
defined above other than the clients’ primary ATMs are
present within one 3 hour period’s travel of that location.
distributed proportionately to the inverse of the square of
228
the distance of the machines from the primary ATM. Any proportion that is allocated is drawn from the bottom up:
available balances less than Ushs250 and any residues of any available balance outside the primary ATM is
allocated balances modulo Ushs50 are removed and added consumed first, and then primary ATM balance is
to the primary ATM’s available balance. reallocated until the correct proportion is reached.
These two formulas together have the effect of
concentrating more of a user’s balance nearer to his home 4 SIMULATION
ATM as his balance decreases. In practice the constants
and proportions are adjustable to ensure that any balances 4.1 Simplifying Assumptions
under a given amount are held entirely at the user’s primary There are several factors present in the real-world region
ATM. upon which our simulated testbed is modeled which we
ignore completely. One is familial seasonal migration from
3.5.1 Migration place to place – for the one-year duration of our simulation
The initial germ idea for this project called for support we assume that people’s residences remain fixed, i.e. the
for highly migratory populations to have continuous access primary ATM of a given account never changes; individual
to funds. We solve this problem by introducing a notion of members may go on excursions but never the entire family
carrying ‘virtual’ available balance on a credential token at once. Another is that of immigration to the region or
used to access the system, such as a smart card. This other sources of population growth such as refugee influx
or organic growth of the number of households through
‘virtual’ balance would involve ‘locking’ a specific amount
marriage. We assume that all members of our simulated
of available balance on one ATM and writing a device-
environment start life with accounts, that is, we have a
signed token for the amount onto the card including a
100% adoption rate, with some proportion of accounts
unique transaction ID allowing the user to take the card to beginning with a zero balance. As a final detail, we have
some specific other ATM and withdraw the amount found that we do not have enough information on the
‘locked’ plus any available balance on the local ATM. This overall prevalence of bitwise errors within SMS delivery,
could be useful for trips where larger amounts of capital as the literature has various accounts, and have no
are required but carrying cash is dangerous, e.g. seasonal information at all on its prevalence in the rural context,
trips to buy seed for planting or taking animals or produce where distance may increase and technology may be older
to market. Using the extra balance would cause the local or cheaper but there are far fewer sources of
machine to create a new signed token with any remaining electromagnetic noise, so we ignore it, assuming that those
‘locked’ balance and inserting it into the original machine messages not correctible by the checksum are incorporated
would unlock any such remaining portion and treat the into the loss rate.
spent portion as a withdrawal. This has the advantage that
if the card is lost or stolen, funds which are locked on the
card but have not been unlocked anywhere across the ATM
network can be recovered from the institution.
3.5.2 Multihoming
We wish to model one commonplace situation, in which
a head of household or other member may work in a
remote region for most of the year and return for several
months or a season. In such cases, the person who is
working away from home typically wants to send money
back to his household – this is the problem that approaches
like M-PESA address. We model this situation by
explicitly adding the single ATM closest to the worker to
the set of ATMs considered for balance distribution, if it is
not already in the set. While the typical usage of such an
ATM will be solely to deposit funds left over after cost of
living is paid for in cash, we arbitrarily allocate 20% of Fig. 4. Proportion of downtime versus standard deviation outage burst
balance available at the primary ATM to this ATM. The length for 4 values of λ.
rationale is the idea that the breadwinner represents, on
average, 1/5 of the total size of the household based on the
average family size of 4.7 and assuming a single primary 4.2 Simulator Overview
source of income. This is trivially extended to The simulator uses minute-scale timeslices over the
accommodate arbitrary family configurations. The course of a calendar year and contains metaphors for
229
customers, ATMs, the CA server, and the SMS network specific characteristics with which we attempt to
connecting the various components of the overall system. reasonably emulate a real population:
In our experiments we change various parameters of the 1) Primary ATM, allocated randomly using a uniform
system to see what effect they have on the cost model and distribution
customer request satisfaction rate. We consider a 2) Weekly savings rate, generated by taking an arbitrary
withdrawal request successful if, having specified some minimum savings rate (Ushs200) and adding it to
amount desired which is less than his account balance, the Ushs1000 multiplied by the absolute value of a random
machine both has enough cash on hand to satisfy the element drawn from a Gaussian distribution with mean 0
request and has enough of the balance allocated to the and standard deviation
given machine to allow the machine to dispense that cash. 3) Visit frequency, a step function generated by another
random Gaussian with mean 0 and standard deviation 1, by
4.3 Network Behavior
assigning 7 days if the random value is within 1 standard
Network behavior parameters are the factors most deviation, 14 if within 2, and 28 if more than 2
directly related to the cost model. To model bursty outage 4) Withdrawal strategy (whether the client will withdraw
behavior, we use a Poisson distribution to model burst a proportion of his balance or of his weekly withdrawal
arrival and a truncated normal distribution (with mean at 0 rate, half of the clients being in each category), selected
in which we only consider positive values) to model burst using a uniform random variable
length, and change λ and σ2, respectively for each to test 5) Probability of withdrawing from an ATM other than
the robustness of the system under various outage models, his or her primary ATM, between 0% and 20%, generated
as well as how the level of customer satisfaction changes as using a Gaussian distribution with mean 10% and standard
the network and therefore the application built atop it deviation 5%. Any random value selected outside of 0-20%
becomes less reliable. We would also like to vary latency is reselected
with a truncated normal distribution and μ set to 36 6) 0-4 arbitrarily placed seasonal brief migrations where
seconds as the experimental data in 3.2.1 indicated, and the 10% of the population has one or more migrations.
σ2 likewise set to 5 to show how the effects of variable Presence of a migration is decided by a uniform random
propagation delays; however the timescale we are using variable, and the number of migrations is generated in an
renders this variability negligible. As such we set μ and σ2 analogous manner to 3)
to 1 minute, which biases slower than experimentally and 7) multihome (second primary) ATM, where 10% of
detected, which cannot artificially improve results. Finally, the population has a secondary home as described above.
we add an additional Poisson distribution-based loss model Both the presence of a second primary ATM and its
for ‘black hole’ loss where messages simply disappear after location are selected using a uniform random variable.
having been successfully sent to the SMSC. For simplicity, we register deposits proportionate to the
weekly savings rate each time a client visits an ATM for
withdrawal. Deposits are processed every two weeks, and
only deposits received at least one week prior to processing
are applied to balances.
4.5 ATM Parameters
We use a parameter n to determine how many ATMs to
use to saturate our network, i.e. one ATM per n users.
When this figure causes the number of ATMs to exceed
one per population center, we use the notion that a given
account can only use one ATM within the center and any
others are treated as nonexistent, to obviate complexity of
balance distribution among the multiple ATMs while
maintaining semantic equivalence. We also use a ratio p to
specify the proportion of the sum of the total available
balances from a given ATM that the ATM should have on
Fig. 5. CDF of customer weekly savings rate in Ushs. hand in cash, which affects cash restocking. Finally we
specify how many days the server waits between balance
4.4 Household Behavior redistributions, tb, which affects message traffic volume as
The 136541 households in our testbed are assigned to well as funds availability following a deposit.
population centers according to real-world population To limit the number of variables in the simulation, we
levels. Each account is held by a household having certain chose to fix the frequency of cash refills at once every two
weeks.
230
Fig. 6. Loss versus Cost per User in US Dollars, averaged across all runs Fig. 7. Loss versus successful transaction proportion, averaged across all
of the simulator. runs of the simulator.
Fig. 8. Cash load proportion versus successful transaction proportion, Fig. 9. Number of users per ATM (n) versus Cost per User in US Dollars
averaged across all runs of the simulator. averaged across all runs of the simulator.
5 RESULTS Figure 7 shows that satisfaction is unaffected by the

number of messages sent, which is an encouraging
In running our simulations, we are interested in seeing
indicator that our protocol recovers well from loss.
the effect of our various parameters on two primary results:
cost per user per year, which indicates the marginal cost of
adding more users to the network, and withdrawal request 5.2 Cash Exposure
satisfaction rate. Figure 8 illustrates the relationship between the cash load
5.1 Loss Rates proportion and the proportion of requests satisfied. While
the difference is absolutely minute, constituting less than
Simulation shows that the system is affected not
0.04% difference, the trend is clear that loading the ATMs
specifically by outage frequency or outage burst length, but
with more cash provides slightly better request satisfaction
by overall downtime proportion. As shown in Figure 6, as
levels. However, it is encouraging to see that even with at
loss increases from 1% to 12%, cost increases linearly as
most 5% of the cash payable across all accounts in a given
the number of messages which need to be resent increases.
ATM loaded, satisfaction is approximately 98% of all
It is an obvious conclusion that messaging cost would
requests under this model.
increase as loss rate increased, but it is encouraging that the
cost increases linearly rather than exponentially. The
jaggedness of the curve is due to our using precomputed 5.3 ATM Density
loss models across all trials to speed the simulator up, Varying the number of users served by a given ATM has
which exacerbates certain lossy periods as they occur in the a dramatic effect on the number of messages sent as each
same place in every run. ATM sends messages after a given period of time
231
Fig. 10. Days between balance redistributions versus cost Fig. 11. Days between balance redistributions versus satisfaction rate
regardless of whether the message contains the maximum

number of transactions or not, given that at least one p = 0.05
transaction has occurred during that period of time and the n = 16,000
server must send messages to each ATM during any tb= 7
balance redistributions.
As shown in Figure 9, cost per user drops off which yields over a 98% withdrawal success rate with a
dramatically between 2000 users per ATM and 4000 users messaging cost of between USD 0.10 and 0.15 per user per
per ATM and levels off thereafter. This sharp elbow is year even accounting for nearly a 10% message loss rate.
explainable by the fact that many of the population centers
used in the simulator have between 2000 and 4000 6 DISCUSSION AND FUTURE WORK
accounts, and therefore increasing the number of users The system as described has several practical
served per ATM decreases the number of ATMs per weaknesses. An important limitation is the occasional need
location from 2 to 1. In our simulation, this change in of accountholders to withdraw their entire savings. This
density decreases the number of ATMs from 92 to 61. On can, however, be ameliorated by augmenting the system
the other end of the scale, few locations have more than with an on-demand functionality whereby a user could pay
8,000 accounts, so there is little change by increasing users
for the additional SMS messages required to withdraw his
per ATM beyond that point.
or her full balance immediately.
Other weaknesses of the analysis include the explicit
5.4 Balance Distribution Frequency exclusion of physical security considerations and legal
As shown in Figure 10, cost per user falls off restrictions, both of which are often of critical importance
dramatically with less frequent updates at the cost of user in the developing world. The former in particular is
satisfaction rate. The cost, fortunately, falls off much faster important with deposits, as they need to be conveyed to
than the request satisfaction rate as seen in Figure 11. As some physical bank branch for validation.
each balance redistribution accounts for one complete set The details of transaction auditing and reconciliation are
of account balances sent to each ATM in the system, this also neglected, but it is assumed that standard practices will
sharp increase in cost as updates become more frequent is be followed.
compounded if the number of users per ATM is low A baseline study on the field performance of SMS would
enough to introduce an unnecessarily large number of allow for further improvements to this system. We would
ATMs into the system. like to better understand outage patterns for SMSCs and
cellular networks as a whole. Some national governments,
5.5 Example Optimization such as Nigeria’s, require cellular network operators to file
To attempt to select a set of parameters to produce an reports on statistics affecting quality of service such as bit
optimal solution, we first need to observe that cost and error rate, end-to-end error rate, etc., knowledge of which
satisfaction rate are opposed to each other, as are cash load would allow us to better simulate this aspect of the
and satisfaction rate. An implementing institution would environment. Also, detailed knowledge about the location
have to determine what mixture of cost, risk, and customer of cellular towers and their respective power levels would
satisfaction it would be comfortable with, but one example greatly aid in the accurate representation of the region.
set of parameters is The next step in research for this system in particular
232
will be investigation into whether decentralization can [9] C. Olston and J. Widom, “Offering a Precision-Performance
Tradeoff for Aggregation Queries over Replicated Data,” in 26th
reduce network traffic and increase robustness against International Conference on Very Large Data Bases (VLDB). Cairo,
network failures, particularly at the cellular network cell Egypt, 2000, pp. 144-145.
hosting the institution. As balance distribution is dependent [10] R. Patra, S, Nedevschi, S. Surana, A. Sheth, L. Subramanian, and E.
on geographical location, it may prove to be wise to have Brewer, “WiLDNet: Design and Implementation of High
Performance WiFi Based Long Distance Networks,” in 4th USENIX
nodes collude with their nearest neighbors rather than Symposium on Networked Systems Design & Implementation.
receiving updates only from a central authority. Cambridge, MA, 2007.
Other potential extensions of the project are increases in [11] L. Payne, “Rebuilding Communities in a Refugee Settlement: A
Casebook from Uganda.” Oxford: Oxfam, 1998.
granularity or changes in the strategy of balance [12] G. Peersman, P. Griffiths, H. Spear, S. Cvetkovic, and C. Smythe, “A
redistribution and tracking of loss and corruption rates of tutorial overview of the short message service within GSM.”
received messages in order to better allocate message Computing & Control Engineering Journal, 11 (2), pp. 78-89.
[13] S. Staschen, “Microsavings, Microcredit and Microinsurance:
budget. A redistribution strategy which is more adaptive to Financial Products of Small Farmer Co-operatives Ltd. In Nepal.”
the past patterns of the user is also likely to yield positive Cologne: AEF University of Cologne, 2001.
results in terms of customer satisfaction. G. Taylor and M. Bekabye, “Employment Creation and Labour-
Based Technology in Roadworks: Ugandan Case Study, 1999.”
As crafted, the system could be trivially extended to
[Online] Available:
allow for payouts of microcredit installment loans or http://www4.worldbank.org/afr/ssatp/Resources/HTML/rural_transp
salaries for rural workers. ort/knowledge_base/English/Module%202%5C2_4b%20Case%20St
udy%20-%20Labour%20Based%20Techs.pdf
[14] C. Tseng, J. Jiang, R. Lee, F. Lu, C. Ouyang, Y. Chen, and C.
7 CONCLUSION Chang, “Feasibility study on application of GSM-SMS technology to
This paper presents ATMosphere, a protocol designed to field data acquisition.” Computers and Electronics in Agriculture, 53
(1), pp. 45-59.
provide ATM services to the world’s rural poor. The [15] S. Wisniwski, “Microsavings Compared to Other Sources of Funds.”
system as simulated is based on currently deployed Eschborn: GTZ, 1999.
technologies and is implementable today using low-cost [16] H. Yu, and A. Vahdat, “Design and Evaluation of a Continuous
Consistency Model for Replicated Services,” in Fourth Symposium
hardware. The result is a solution which has low start-up on Operating Systems Design and Implementation. San Diego, CA,
and marginal costs, represents low relative risk to financial 2000.
institutions, and provides sorely needed access to savings [17] H. Yu, and A. Vahdat, “Efficient Numerical Error Bounding for
Replicated Network Services,” in 26th International Conference on
and capital markets for the marginalized. Very Large Data Bases. Cairo, Egypt, 2000, pp. 144-145.
[18] M. Zeller, “Models of Rural Finance Institutions,” in Paving the Way
8 ACKNOWLEDGEMENT Forward for Rural Finance: An International Conference on Best
Practices. Washington, DC, 2003.
We would like to thank various personal friends living in [19] ETSI GSM 3.40: ‘Digital cellular telecommunication system (Phase
rural Uganda for firsthand assistance with baseline testing 2); technical realisation of the short message service point-to-point’,
of MTN’s cellular network. We would also like to thank European Telecommunications Standards Institute TC SMG, Version
4.13.0, May 1996.
Jinyang Li for her comments. [20] ETSI GSM 3.41: ‘Digital cellular telecommunication system (Phase
This material is supported by the National Science 2); technical realisation of the short message service cell broadcast
Foundation under Grant No. CNS-0831934. (SMSCB)’, European Telecommunications Standards Institute TC
SMG, Version 5.2.0, May 1996.
[21] GSM Doc 28/85: ‘Services and Facilities to be provided in the GSM
REFERENCES System’ rev2, June 1985.
[1] E. Brewer, “Technology Insights for Rural Connectivity,” in [22] MTN Uganda Ltd (MTN-UGANDA) – Interactive Coverage Map,
Wireless Communication and Development: A Global Perspective. 2007. [Online] Available: http://www.gsmworld.com/cgi-
Marina Del Rey, CA, 2005. bin/ni_map.pl?cc=ug&net=mt.
[2] L. Elser, A. Hannig, and S. Wisniwski, “Comparative Analysis of [23] MTN Uganda: A Bright Y’ello World, 2007. [Online] Available:
Savings Mobilization Strategies,” Eschborn, GTZ, 1999. http://www.mtn.co.ug.
[3] W. Enck, P. Traynor, P. McDaniel, and T. La Porta, “Exploting open [24] MTN Uganda: Tariff Guide : Mobile Solutions : Business, 2008.
functionality in SMS-capable cellular networks,” on 12th ACM [Online] Available:
Conference on Computer and Communications Security. Alexandria, http://www.mtn.co.ug/business/talktime_tariffs.htm.
VA, ACM, 2005, pp. 393-404. [25] The National Cheque Standard of Uganda, 1999. Available:
[4] A. Hannig, and S. Wisniwski, “Mobilizing Microsavings: The http://www.bou.or.ug/NPSS/Cheque%20Standard.pdf.
Experience of Seven Deposit-taking Institutions,” Eschborn: GTZ, [26] Safaricom: M-PESA, 2007. [Online] Available:
1999. http://www.safaricom.co.ke/index.php?id=745
[5] G. Ivatury and M. Pickens, “Mobile phone banking and low-income [27] Quality of Service (QoS) Indicators for Mobile Services, Unknown
customers: Evidence from South Africa,” Washington, DC, 2006 Date. Available: http://www.ncc.gov.ng/TRS/QoS_Benchmark.pdf.
[6] A. Jhunjhunwala, A. Ramachandran, and A. Bandyopadhyay, “n- [28] Thuraya Satellite Communications, 2008. [Online] Available:
Logue: The Story of a Rural Service Provider,” The Journal of http://www.thuraya.com
Community Informatics,1 (1), pp. 30-38. [29] Uganda Bureau of Statistics, 2007. [Online] Available:
[7] S. Lord, “Trouble at the Telco: When GSM Goes Bad,” Network http://www.ubos.org
Security, 2003(1), pp. 10-12. [30] WiMAX Forum – Technology, 2007. [Online] Available:
[8] C. Olston, “Approximate Replication,” Doctoral Dissertation, http://www.wimaxforum.org/technology
Stanford University, Stanford, 2003.
233
Building a Transportation Information System

Using Only GPS and Basic SMS Infrastructure
Ruth E. Anderson, Anthony Poon, Caitlin Lustig, Waylon Brunette, Gaetano Borriello, Beth E. Kolko
Kyrgyzstan, a developing region with poor infrastructure and

Abstract—This work consists of two main components: limited resources; consequently, the findings in this paper are
(a) a longitudinal ethnographic study in Kyrgyzstan that extensible to other selected developing regions with limited
demonstrates the importance of transportation resources in the resources, fragile road infrastructure, resource-constrained
developing world and how to plan for an appropriate ICT
solution, and (b) the results of a proof-of-concept system
central government authority, and ad-hoc transportation
engineered to create a bottom-up, transportation information resources for both inter- and intra-city transport.
infrastructure using only GPS and SMS. Transportation is a
very important shared resource; enabling efficient and effective II. TRANSPORTATION AND THE KYRGYZ CONTEXT
use of such resources aids overall development goals.
The system, *bus, involved the development of a hardware A. Background on the Transportation Challenge
device (a *box) containing a GSM modem and a GPS unit, that Transportation has long been recognized as having an
can be installed on a vehicle and used to track its location. The
*box communicates via SMS with a server connected to a basic impact on economic development. Factors such as levels of
GSM phone. The server runs route a prediction algorithm and investment in transportation have been shown to have a
users can send SMS messages to the server to find when a bus positive impact on economic growth [1] [2]. In the
will arrive at their location. developing world, the availability of safe and reliable
The paper discusses the system and early testing, as well as the transportation can have even broader implications. Not only
development implications for a range of urban and rural does access to transport improve access to markets and
environments where transportation is scarce or inefficient, and
mobility of workers, but it is also critical to the timely and
where a central authority or institution is not in a position to
provide robust information resources for users. We describe how affordable delivery of services such as health and education,
the solution is also situated within technology usage patterns can serve to empower vulnerable groups by increasing their
common to the developing world. independence, and is key for maintaining social
networks [3]-[5].
Index Terms—ICTD, transportation, information services, Along with providing access to transportation, it is also
mobile phones, GPS, SMS, Kyrgyzstan, developing world important to provide potential users with information about its
availability. In the developed world, users can often access
information about most bus and train schedules easily via
I. INTRODUCTION
printed schedules or web pages maintained by centrally-
T RANSPORTATION of goods and people is key to economic

and human development. This paper discusses the
development of the *bus system (pronounced “Starbus”), a
funded transportation authorities. In some cases, users can
view real time updates on the current location and expected
arrival time of their bus or train via web, phone, or SMS/text
transportation information system that uses GPS, GSM, and message [6]-[8]. However, when a community lacks the
SMS technologies. The *bus system was developed in infrastructure to provide such information resources, potential
response to transportation challenges experienced in users can find themselves unable to take advantage of
whatever (limited) transportation resources are available.
Manuscript received September 22, 2008. This work was supported in B. Methods
part by U.S. National Science Foundation grants #0326101 and #0219350 and
by the J. D. Noe Professorship at UW.
The background provided in the rest of this paper is based
R. E. Anderson, Dep’t of Computer Science & Engineering, University of on several years of research in Central Asia by the Central
Washington, Seattle, WA 98195 USA (rea@cs.washington.edu). Asia Information and Communication Technologies (CAICT)
A. Poon, Dep’t of Computer Science & Engineering, University of
Washington, Seattle, WA 98195 USA (anthop@cs.washington.edu).
project at the University of Washington. Our research has
C. Lustig, Dep’t of Computer Science & Engineering, University of included multiple periods of ethnographic research including a
Washington, Seattle, WA 98195 USA. She is now with Intermec, Inc. design ethnography, interview studies, and a longitudinal
(c.lustig@gmail.com). survey in four countries of Central Asia.
W. Brunette is with the Computer Science & Engineering, University of
Washington, Seattle, WA 98195 USA (wrb@cs.washington.edu). Ethnography has been conducted since 2000 and includes a
G. Borriello, Dep’t of Computer Science & Engineering, University of team of US and Central Asia researchers. Field notes and
Washington, Seattle, WA 98195 USA (gaetano@cs.washington.edu). photography are both used to document findings. Survey
B. E. Kolko, Dep’t of Human Centered Design & Engineering, University
of Washington, Seattle, WA 98195 USA (bkolko@u.washington.edu). results are based on a nationwide survey of 1000 respondents
234
in each of four countries (Kazakhstan, Kyrgyzstan, Tajikistan, to be places where the marshrutkas stop. There is no official
Uzbekistan) of age 15 and older, administered in 2006, 2007, schedule, so riders are not always sure how long they will
and 2008. The survey sample was based on census have to wait. Sometimes buses are full and will not stop, and
information for age, gender, ethnicity, and geographic location then riders have another unknown stretch of time to wait.
as released by each country’s government. The survey Routes can shift slightly, and because stops are often
includes multiple urban and rural sampling locations, and it demarcated by landmarks rather than actual routings, users
was administered in Russian and other regional/local sometimes have to guess what route a bus will take to get from
languages. In addition to the general sampling scheme, a listed landmark A to listed landmark B. Overall, the system
three-stage process was used to select respondents: can work well if one is traveling a familiar route that does not
• probability proportional to sample size of Primary get too crowded, but if one travels at peak times, or off hours,
sampling units (PSUs); or wants to get somewhere new, the system can be frustrating
• consecutive random sampling of households in determined to navigate because there is no clear pathway to gather the
PSU; and information relevant to a rider.
• selection of a household respondent using a Kish Grid Marshrutkas operate on inter-city as well as intra-city
method. routes. In Kyrgyzstan the inter-city system is relatively
The survey is a general social survey that asks a series of transparent, if not especially efficient. In order to travel
questions about technology usage, as well as information- between towns, one goes to the bus depot where marshrutkas
gathering habits and patterns of trust and confidence. Because are lined up. In Bishkek, the bus depot is large and well
Internet usage rates in the region tend to be low, the survey signed, with destinations labeled on placards hanging over the
was not designed to target Internet users specifically, but, bay where the minibuses are parked. In smaller cities the
rather, gauge general trends among the overall population. depots can be parking lots on the edge of town.
At the Bishkek depot, drivers stand in front of their buses
C. Inter-city and Intra-city Transportation within Kyrgyzstan with microphones, calling out their destination to the crowd.
Kyrgyzstan is a small, mountainous country in Central Asia. There is no departure time listed, and drivers will not leave
Its population is about 5.5 million and concentrated in the until the bus is full. If your destination is a popular one, you
capital of Bishkek in the north and in the heavily agricultural may not have long to wait. But if you are headed to a smaller
Ferghana Valley in the south. The country has a youth bubble, town on a not particularly well-traveled route, you could face
with the median age of 24.2 years (for comparison, the US a significant wait before the bus leaves. It is these kinds of
median age is 36.7). The population is defined as 98.7% inefficiencies that make the system ripe for technology-
literate; ethnically the population is largely Kyrgyz (64.9%), assisted solutions that could disseminate information to users
followed by Uzbek (13.8%) and Russian (12.5%). and prospective users of the transportation infrastructure.
Since independence from the Soviet Union in 1991,
Kyrgyzstan has struggled economically. Unemployment is
high and about 40% of the population lives below the poverty
line [9].
Personal vehicle ownership rates are low, and the
population depends heavily on shared and public
transportation for inter-city and intra-city transportation. Cities
have taxis and an independently operated bus system known
as the marshrutka system. Marshrutnoye taksi is a Russian
phrase for ‘routed taxicab’; the marshrutka system is a shared
taxi transportation system that runs on largely pre-plotted
routes but without standard schedules; the standard bus is a
mini-van that seats about 15 and with standing passengers can
fit 25 or more. Fares are 8 som (about US$0.19) before 8pm
and 10 som (about US$0.24) after 8pm for central city routes. Figure 1 A bus depot in Bishkek with marshrutkas and car taxis.
The drivers are often independent contractors as opposed to
employees of a municipal or national transportation authority. In addition to the bus depots, there are also more informal
Bus routes are part of local knowledge, and learning a new ways of traveling between cities. In neighboring Uzbekistan,
route is challenging even for a city resident. Numbered routes for example, if you want to travel and not take the buses
are marked by placards placed in the front windshield of the (which can be overcrowded and perilous on the mountain
bus, with the rough routing listed as major stops or landmarks. roads), you can travel to an acknowledged site – in the case of
As buses whiz by, people try to glimpse the list of stops the capital Tashkent, this is a subway stop furthest in the
written on the placard to see if the bus is going in their direction in which you want to travel. Or, in Tajikistan, a
direction. Quickly identifying an appropriate marshrutka is a mountainous country with limited public transportation
challenging step when a new route is being attempted. There resources, you similarly go to an established place and look
is a lack of official bus stops, and people stand at areas known for a driver headed in your direction; at times these trips can
235
last for days. There, drivers with cars congregate, looking for As Figure 3 illustrates, approximately one third of mobile
passengers to fill up their cars; some drivers are professionals, phone users report using SMS; interestingly, this number held
some are there to help defray the costs of getting home (often rather steady between 2007 and 2008, unlike overall mobile
after delivering agricultural produce or visiting relatives). adoption numbers. This rate of SMS usage is larger than those
The characteristic these countries have in common, and that using their phones to find information, and the percentage of
they share with many developing countries, is that shared people using their phones to connect to the Internet decreased,
transportation often happens outside the realm of a national or but is so low as to be irrelevant (4% in 2007, 1% in 2008).
metropolitan transit authority. This means there is no central There are information numbers one can call in Kyrgyzstan to
body coordinating or allocating resources to ensure efficiency get information about local businesses, so it is possible for
or at least moderate attempts to meet the needs of citizens. Our mobile phone users to seek information via voice.
research considered how, under such circumstances, one 100%
might aggregate and disseminate information about 80%

transportation options to make it more efficient or effective for
60%
both riders and drivers. Another way of stating the problem is
how might one build an information infrastructure for a shared 40%
public resource when a public government (or other 20%

centralized authority) is not positioned to do the job – either
0%
because of resources, will, or because the resource itself is Work Per- Infor- Text / Play Inter-
Music Photos
calls sonal mation SMS games net
decentralized and largely part of the informal sector. 2007 - N=451 36% 92% 16% 28% 14% 4% 13% 9%
2008 - N=712 32% 97% 21% 29% 7% 1% 13% 14%
D. Personal Technology Use in the Region
The rate of personal ICT usage in Kyrgyzstan is similar to Figure 3 Activities people engage in with mobile phones
many developing regions; that is, computer and Internet usage
remains relatively low and has seen slow growth over the past The issue of Internet use is especially key to the particular
3 years, while mobile phone use is growing at a remarkable solution this team has created. Survey findings in the region
rate. As Figure 2 shows, the mobile usage rate in Kyrgyzstan illustrate that not only is mobile usage nearly five times that of
has grown from 21% to 71% from 2006 to 2008. During that Internet usage (see Figure 2), but even when people do
same time, Internet usage has grown almost imperceptibly. identify themselves as Internet users, their model of usage is
Even in neighboring Kazakhstan, with a rapidly developing very different from that of the developed world. This point is
economy rich with oil and gas money, Internet use is holding crucial, and it feeds directly into the *bus system developed as
flat, with only 19% of the population reported as Internet users part of this study.
in 2007 and 2008. And even though computer use climbed In a city like Seattle, for example, extensive information
from 29% to 43% in Kazakhstan over the same time frame, about public transportation is available on the Internet, and
Internet usage showed no increase. people access this information either through a computer or a
The rapid increase in mobile usage is not unique to mobile phone. Information about bus routes, timetables, fares,
Kyrgyzstan, and adoption of mobiles in and of itself is not re-routings, etc. is available on a comprehensive and user-
enough to establish that a mobile solution is the best avenue friendly site. However, this model of information delivery is
for a transportation solution. However, in addition to mobile predicated on the reality that most users of the system have
usage, what we also see in Kyrgyzstan is that people use their regular access to Internet resources. The percentage of Internet
phones for personal calls more than business calls, they are users in the US is estimated at 85%, and over half of those
using SMS in measurable amounts (about one-third of users users are connected on a regular basis [10]. In Kyrgyzstan,
use SMS; see Figure 3 ), and they are not using the Internet as however, even among Internet users, the majority report using
their primary mode of ICT access. the Internet once a week or less (see Figure 4). Meanwhile,
80% mobile users use their phones on a much more frequent basis,
with 89% reporting using them at least once a day.
60%
What this means is that developing and delivering a robust
40% information delivery system that relies on people having
20% Internet access in order to take advantage of the information
resource is a fundamental mismatch between a technology and
0%
Owned Computers Used the Internet
Owned and used user habits and abilities. Recognizing what actually constitutes
mobile
an “Internet user” then, in many parts of the developing world,
2006 7% 11% 21%
helps to better envision an appropriate technological solution:
2007 8% 10% 45%
2008
delivery of information via mobile phone-based SMS.
12% 15% 71%
In addition, it is worth noting that mobile phone usage rates
Figure 2 Personal Technology Use in Kyrgyzstan. are significant among rural as well as urban populations. As
Figure 5 below illustrates, mobile usage is not only an urban
phenomenon. In many ICT and development situations, there
236
is a gap between rural and urban users. In this project, we have transportation worked was in stark contrast to the public
taken care to include the usage behavior of both urban and information resources available in many developed country
rural populations in order to develop a solution that will be communities, and because transportation was clearly a
appropriate for the broadest swath of users. fundamentally shared resource in these settings.
70% As we continued to analyze the findings from the design
60%
50%
ethnography, we found the following: (a) pricing issues for
40% marshrutkas were not a concern, however the failure of the
30%
20% marshrutka system often meant people had to take taxis which
10% were seen at prohibitively expensive, (b) safety and security
0%
Mobile Use - N Internet Use - Computer Use were issues, especially for women, and this related to
= 712 N = 149 - N = 280
uncertain routings, riding at night, uncertain waiting times,
Every few weeks or less 2% 16% 20%
and buses that would be full and unable to take new
At least once a week 9% 58% 39%
passengers, (c) learning new routes was seen as daunting,
At least once a day 89% 26% 41%
(d) there was some confusion over how much information
Figure 4 Frequency of technology use in Kyrgyzstan, 2008 about routes was available either through maps or online, and
(e) inter-city travel, especially, often created a level of
60% schedule uncertainty.
50%
40%
Among these issues, the ones that stood out as particularly
30% addressable were making the shared ride system more efficient
20% so that riders would not be forced to default to the more
10% expensive taxi system, and increasing information awareness
0%
Mobile Use** Internet Use** Computer Use** to give users the mechanism by which they could increase
Urban - N= 397 56% 26% 43% their personal safety but still make use of the shared ride
Rural - N= 603 41% 5% 18% system. Although the *bus system currently focuses on intra-
city travel, the development impact is perhaps even more
Figure 5 Urban and rural use of technology in Kyrgyzstan, 2008
** = p <.001 based on Chi-square test significant for inter-city use where long-distance
transportation resources are even more scarce.
E. Design Ethnography Findings Regarding Transportation Research into transportation information systems in the
Challenges developed world turned up municipal ride share systems such
During the summer of 2006, the CAICT team conducted as 511 Rideshare in California’s Bay Area which allows riders
design ethnography work in Kyrgyzstan to better understand and drivers to find one another; robust information resources
people’s habits of technology use within their personal available via the Internet include bus location information, and
communities, as well as the pain points of everyday life, mobile-phone based systems (e.g., in Scandinavia) that
which might best indicate potential solutions that leverage provide up-to-the-minute schedule information so people can
those preexisting technology usage patterns. time their arrival at their bus, tram, or train stop.
Four groups participated in the design ethnography: two
multi-generational family groups, and two groups of young III. DESIGNING AN APPROPRIATE TECHNOLOGICAL SOLUTION
people (18-25). Interviews were conducted in an urban area, We designed a system with the goal of improving access to
and a smaller, regional town. At each site one family and one transit information for potential bus riders at a minimum of
group of young people were interviewed. Interviews covered cost to users and without requiring the assistance of a central
issues related to technology use including Internet and mobile authority. For our solution, we chose three existing
phones, challenges of everyday life, and social networks. In technologies appropriate for Bishkek (and for much of the
particular, interviews and questionnaires focused on patterns developing world as well) and augmented them with two new
of reciprocity of goods and information exchange within the components we developed. We describe these engineered
groups, as well as whether any technologies were used to components later in Section V. The existing technologies we
facilitate such exchanges. used were SMS, GPS, and GSM, all described below.
Findings from the design ethnography indicated mobile
phones were extremely important to participants, and that both A. Short Message Service (SMS)
voice and SMS played a daily role in people’s lives. Of the 12 SMS, also known as text messaging, is a protocol to
participants in the study, only one did not own a mobile exchange short messages between mobile phones. Our system
phone. uses SMS as its primary communication mechanism. We
Transportation emerged as one of the significant problem chose SMS for several reasons. As discussed in Section II.D,
spaces for our users. Although for most people, a general mobile phone ownership and text messaging is common in
sense of lacking information resources outside of family and Bishkek. Text messages are perceived as relatively cheap,
friend networks was prevalent, the transportation issue was of costing anywhere from US$0.01 to $0.06 to send a message,
particular interest because it generated animated discussion and for many plans it is free to receive messages. Cell
among our users. This was the case because the way coverage in Bishkek is also excellent, and the country overall
237
has good coverage. Good urban and rural cell coverage is and provides the querying service to riders. These are both
becoming the case in much of the developing world. described in more detail later in Section V. The goal of the
Additionally, in the ICTD community, SMS-based solutions *bus system is to track buses’ GPS coordinates and to provide
have proven robust, flexible, and valuable to multiple users an SMS messaging system that will tell them when a bus
communities [11]-[13]. Work such as Frontline SMS and should be arriving at their location.
Warana Unwired demonstrate that information delivered via The *box is a low-cost vehicle tracking solution that utilizes
SMS can have a tangible, positive impact on individuals’ GPS satellites to locate *box-equipped vehicles; the *box
lives [14] [15]. hardware includes a GPS device and a GSM modem
Although the mobile web is a compelling solution space for integrated into a single package that just requires a local SIM
some, our research in Kyrgyzstan and elsewhere points to the card. The GSM modem allows the *box to send SMS
value of solving for cheaper devices and more lightweight messages with the GPS data to the server, taking advantage of
technology solutions that dovetail with technology patterns the widespread GSM cellular phone networks.
already in place. Text messages, as opposed to richer On the back end, the central server continually collects the
mechanisms for distributing information, such as web pages, GPS location data from all *boxes and stores it in a database.
can be supported by cheap and ubiquitous phones; The server consists of a laptop computer connected to a
additionally, because of the way the mobile phone integrates mobile phone capable of sending SMS messages and that
into how people already get information in their daily lives, it serves a gateway to the SMS services. The server does not
is a better solution space than Internet-based resources [16]. need to be connected to the Internet, and the service does not
require cooperation with mobile providers. Potential bus
B. Global Positioning System (GPS)
riders can use their cell phones to send SMS queries to the
GPS satellites transmit microwave signals to GPS receivers *box server and receive transit information in response.
that use data from the signal to determine the location of the Initially, the system will just support riders' queries for transit
receiver. Our system relies on information provided by GPS arrival times, but it has been designed to be easily expanded to
satellites to determine the current location of buses handle more complex activities such as route planning or
participating in the system. GPS was chosen over other registration for notification of bus location updates.
methods of location determination, such as GSM
triangulation [17], for its high level of accuracy – within 8-10 V. ENGINEERED COMPONENTS
meters, and its appropriateness for outdoor usage, particularly
Each bus driver participating in the *bus transportation
in rural areas or urban areas with few tall buildings and mostly
network will need to have a *box device installed or placed on
wide streets. Although GPS units continue to decrease in
their bus with them at all times to allow riders to receive
price, they currently are only included as part of high end cell
updates on their expected arrival times. At least one server,
phones. Thus, our system attempts to minimize the number of
connected to a cell phone, also needs to be running at all times
GPS units required by allowing bus riders to leverage GPS
to receive location update messages from *boxes and to accept
units in buses to “tag” locations they are interested in querying
and respond to queries from users. That server can be located
about in the future (see V.B.3)). In this way, we are leveraging
anywhere in the country with cell coverage.
an expensive piece of technology that is rare in the community
(GPS enabled phones are not a significant part of the market A. *box
in most parts of the world) in order to make its capabilities 1) *box Driver Interface
serve the needs of a larger group. In the *bus system, multiple The *box was designed to require minimal interaction with
individuals can make use of the capabilities of GPS the bus driver. In order to use the *box, the driver only needs
technology despite individuals not owning their own device. to turn it on (using a button located on the side), and then enter
C. Global System for Mobile (GSM) a route number using a keypad as shown below in Figure 7.
The selected route number is displayed on a small LCD and
GSM communications is the most popular mobile phone
stored locally on the *box to be appended to each location
network standard. Most mobile phones, including those
update message sent to the server. A version of the display
commonly used in Kyrgyzstan, are on GSM networks. In
and button labels is under development in Russian and can
2005, the GSM family of technologies had more than 1.5
easily be adapted for other languages.
billion subscribers, more than the total number of Internet
users world-wide [18].
IV. PROPOSED SOLUTION

The overall architecture of the *bus project, shown below in
Figure 6, incorporates the three existing technologies
mentioned above with two new components. These
engineered components are: a GPS vehicle tracking box (*box
pronounced “Starbox”) and a back-end server that aggregates
the location of the *boxes, runs a route prediction algorithm,
238
Figure 7 Prototype of the *box. Note the golden pigtail

connectors on the top of the box to allow easy attachment of GSM
and GPS antennas.
2) *box Internals
In addition to being able to accept route numbers via a
keypad and display the provided route number to the driver,
internally the *box must also be able to determine its location
(via GPS satellites) and send a time-stamped SMS to the
server. For our prototype device, we chose the Telit GM862-
GPS module. We chose to implement the *bus system with
this very compact module because it provides a low cost
solution with the additional convenience of combining GPS
Figure 6 Overview of the *bus System and GSM capabilities on one chip (see Figure 8). Because no
additional work is necessary to interface the GPS and GSM
The notion of a route number is a good fit for the components, developing on this chip is relatively simple. The
marshrutka system in Bishkek where buses have set routes Telit module communicates with the keypad via general
(although no set timings along those routes), indicated by purpose IO to obtain the driver's currently selected route
numbers displayed in bus windows. Changing routes is done number and stores the number in memory. An ATMega16
easily by typing in the new route number. We decided to micro-controller is used to coordinate the communication
require drivers to specify their route via the keypad interface between the Telit chip, keypad and the LCD. The module
rather than forcing the server to do the more difficult job of continuously obtains its GPS location information from GPS
attempting to determine it automatically based on the path of satellites without any action by the driver and reports this
the bus, although this is also a future possibility. information along with its currently selected route number to
One could imagine a simpler device including a hard-coded the server via an SMS message.
route number, although we chose the keypad and LCD
interface for our prototype to allow for ease of testing on B. *bus Server
different routes. The *box was designed to open easily to The primary purpose of the server is to accept SMS
allow for upgrades of the internal components and changing of messages coming from *boxes and bus riders, and to process
batteries, or inspection of device internals to allay fears of them accordingly. System requirements are intentionally
riders or drivers. Drivers can turn the *box off when not in minimal: a laptop or desktop computer running a Java virtual
service and safeguard the privacy of off-duty movements. machine and a MySQL database, connected to a phone
Since we would expect deployment models to build in driver capable of receiving SMS messages. Lower system
accountability, we do not consider theft to be a major concern. requirements allow the server to run using older and less
We built three prototype devices at a cost of about US$200 expensive hardware (such a system may cost around US$200
each, although we believe mass production would bring the to $300). To minimize SMS latency, it is recommended that
cost per box closer to US$50. In the future, as GPS-enabled the phone connected to the server be on the same cell network
cell phones become less expensive, the functionality of a *box as the *boxes, although this is not required. The server does
could potentially be replaced by an application running on a not need to be connected to the Internet, facilitating its
driver’s own cell phone. deployment in areas with low Internet connectivity. Our
Currently, the holder of the *box, in this case the driver or prototype server was implemented using MySMS [19], an
private company that owns the marshrutka, would absorb the application framework built on top of SMSLib [20]. MySMS
cost of the text messages sent out from the system. Our group provides a transport layer, allows SQL queries, notifications,
has considered cost, and the barrier it presents to adoption. remote code execution and rapid application development to
Increased efficiencies may or may not be significant enough be executed over SMS.
for an individual driver to adopt the system, but there are other
entities that might sponsor such a system for users, including a
newly formed transit authority, mobile companies which are
counting on increased SMS traffic from users, or the
employing marshrutka company deciding to adopt the system.
239
many users that will have well-known names. These can

potentially be associated with GPS coordinates in the server’s
database by the server operator. However, the *bus system
was designed to work in scenarios where little or no geo-
coded street information exists. The solution to this problem
is that we allow the clients to geo-code a location, on the fly,
by using the GPS equipment already installed on the *buses.
When a user wants to be able to refer to their current
location in future *bus queries, they can associate a name of
their choosing with the GPS coordinates of their location; they
do this by becoming a secondary user of the GPS on the bus.
The user can instruct the server to associate an arbitrary name
to their current location by waiting until a bus arrives at that
Figure 8 *box Internals. location and immediately sending the following SMS message
to the server:
The server is responsible for handling the following types of STORE <bus-id> AS <my-location>
requests which arrive via SMS messages. bus-id is the *bus id of the bus that has arrived at the user’s
location. In this case, then, a user’s location that she knows
1) Location Update Messages from *boxes only by street name is now given GPS coordinates thanks to
Messages sent from the *boxes to the server have the
the GPS receiver on the bus. The server takes the bus's GPS
following format:
coordinates at the time that the message was sent and
!LOC,<GPS-obtained-time>,<GPS-obtained-
associates it with the location name specified by the user.
location>,<bus-id>,<route-#>
Such geo-coded locations can be retrieved in the future based
The bus-id field is the unique identifier assigned to each
on the location name and the user’s phone number.
individual bus that is equipped with a *box. We would expect
Associating location names with phone numbers prevents
that bus drivers would display this bus id (in addition to their
naming conflicts between users, although other schemes are
route number) in a way that it is accessible to riders. Riders
also possible whereby location names can be shared with other
will need the bus-id in order to associate a name to a location
users, thus organically growing a set of tags for places of
(described later in 3)).
common interest. User tags could also be used to create maps
The server records this information in a database for use in
in regions where few maps exist. This is especially important
responding to user queries.
in Kyrgyzstan where the street names changed after
independence from the Soviet Union and a combination of old
2) Arrival Time Queries from Users
and new names are used. The *bus system could be used to
Information gathered about the current location of buses in
map streets in a way that is comprehensible to all residents
the system could be used to answer a variety of queries. The
regardless of which naming system they use.
server continuously receives and replies to SMS queries from
users. Potential bus riders simply text to a special *bus server
VI. PROTOTYPE DEPLOYMENT RESULTS
telephone number, and the server responds by sending an SMS
back to the user’s phone number. The most basic example is We have implemented prototype versions of the *box
to ask the server to predict the arrival time of the next bus hardware and the server software. In this section, we describe
travelling on a given route number at a given location: some preliminary data collected on the technical performance
NEXT <route-#> TO <location> of the system.
The server replies to the user with a SMS predicting: We deployed a *box on several buses in the University
Next <route-#> arrives to <location> District area of Seattle. The *box sent SMS messages to the
at about <estimated-arrival-time>. server containing its GPS coordinates at an interval of one
These examples demonstrate the types of queries we support message every 30 seconds. Based on this data the server
in English, but the exact format and language can be changed. composed a model of each of the routes. We were interested
The system has been designed so that it can accept queries and in examining the accuracy of our basic route prediction
push out information in alternate character sets – in this case, algorithm as well as the amount of time required for users to
in Cyrillic to support both the Russian and Kyrgyz languages. receive a response to their queries.
While some issues related to a multi-lingual society pose Although we have not yet field tested the system in
significant challenges (for example, when a user sends a query Kyrgyzstan, we have been careful to base the engineered
in a mixture of two languages), the system can be modified to solution on technology that will work in resource constrained
support linguistic variation. environments. Specifically, fieldwork over the years has
demonstrated to the research team that cell coverage and SMS
3) Geo-coding Requests latency in Bishkek is consistent with that experienced in
We expect that there will be some locations frequented by Seattle. Additionally, the relatively open architectural space of
Bishkek means that GPS readings will be easier to obtain.
240
Traffic patterns in the two cities are, of course, different, but

the route prediction algorithm deployed (see next section) can
be modified based on local traffic conditions.
A. Route Prediction Accuracy
1) Route Model
Much research has been done on route prediction algorithms
for public transit and vehicles in general (e.g. [21]). For our (a) Plot of the acquired GPS data for the whole route
prototype, we implemented a simple algorithm based on the (approximately 3 miles) (Google Earth)
concept of building a model by merging together multiple runs
of the route. While one could imagine an implementation that
requires the server to be pre-seeded with detailed sets of GPS
coordinates for each route, our method provides the advantage
of low setup costs for the *bus server, as no special action is
required to enter routes into its database. Furthermore,
prediction models can be made more accurate by merging
additional runs of the route, and more models for varying
traffic patterns can easily be created. More sophisticated
algorithms can be used in future versions, such as ones which
automatically account for varying traffic flow and are tolerant (b) Detailed view of the plot of acquired GPS data
of changing routes. (Google Earth)
We use our route prediction model to provide responses to Figure 9 Accuracy and resolution of GPS data collected at
different message send intervals: red – 10 seconds yellow – 20
user queries about when the next bus travelling on a certain seconds, blue – 30 seconds, purple -40 seconds, green – 60
route is due to arrive at a particular location. To make this seconds.
prediction, the server takes recent coordinates from current One option is to first produce one run with finer granularity,
runs of the route and the location entered by the user and for example, a location update every 10 seconds. This can
attempts to find these points in the model. The time between also be done manually to specify the exact path of the route.
the corresponding locations in the model provides an estimate The method we employ takes points from multiple runs of the
of the amount of time until the bus arrives at the user's route and combines them to produce a model with greater
location. resolution than a single run. Thus, more runs of the same route
The accuracy of these estimated arrival times depends on contribute to overall accuracy of the prediction.
two things: first, the fidelity of the model to the path of the In order to provide accurate estimates of arrival times, a
route and secondly, the timing similarity between the route route model must also reflect the timing characteristics of
model and the buses whose status is being queried. The model buses whose status is currently being queried. The extent to
fidelity is largely determined by the number of points in the which this is true depends on multiple factors. On a bus
model. A point consists of a set of GPS coordinates and the system such as the one in Seattle, where there are set stops,
GPS satellite-provided time of day that the data point was and drivers are expected to arrive at stops at given times,
measured. Accuracy of data points should be quite good (GPS variance can be caused by factors such as weather or
accuracy is within 8-10 meters). Thus, data points collected unusually high traffic. In Bishkek, although marshrutkas
for earlier runs on a given route should provide an accurate follow set routes, there is no schedule of any sort. Thus, we
picture of where those buses were at the time provided. would expect that there might be a wider variance in timings
Number of data points collected per route is determined by for buses in Bishkek. Collecting timing data from more bus
the frequency that *boxes are configured to send location runs and using an algorithm that incorporates time of day and
update messages to the server. More frequent location update day of week would help to improve accuracy in this case.
messages incur more cost and put more load on the server, so 2) Prediction Accuracy Results
it is desirable to keep this frequency low, yet high enough to We deployed *boxes on portions of two Metro bus routes in
give accurate arrival estimates. Seattle and collected data for four runs on each of these routes
Figure 9 below shows a plot of *box location data showing for a total of eight runs. The distance traveled was 2 miles on
the level of accuracy provided by different update message one route and 2.8 miles on the second. Each run consisted of
frequencies. As the figure shows, in general, sending update an average of 11 minutes of collected data. While this running
messages more frequently is more likely to create a model that time is shorter than the length of routes we would expect to
follows the actual path of the route. Particularly around use the system with, it allowed us to collect enough data to
curves and turns, less frequent updates can make it seem as test and provide a proof of concept.
though the bus has cut a corner. To examine the accuracy of our preliminary route prediction
algorithm, we did the following. For each route, we chose one
to three of the four runs to create a model. The remaining runs,
which did not become part of the model, were treated as a bus
241
in motion and we ran our algorithm to generate predictions

about those runs. These predictions were made for five to six
simulated user locations per route, and when the bus was
anywhere from one to nine minutes away from that simulated
user. We provided the algorithm with the model, a time to
treat as the current time of the system, calculated by
subtracting the expected prediction value from the actual
arrival time, and a set of location updates prior to this time,
accounting for location message delays (see VI. B). We
compared the prediction generated to the actual timing data
gathered for that run for a total of 547 measurements of error.
The error in a prediction is the absolute value of the
difference between the actual time it took for a bus to arrive at
the destination and the time predicted by the model. The
mean error from our predictions was approximately Figure 10 Mean error in the predicted arrival time, plotted by the
actual time to arrival (both axes in seconds).
94.9 seconds, with a standard deviation of 83.2 seconds. As
seen in Figure 10, the amount of error in the prediction
We examined 381 messages sent from the *box to the server
increases with the amount of time the bus actually is away
to measure this latency. The average latency, measured from
from the user. The error can be understood as a proportion of
the time it was sent by the *box to the time it was received by
the actual time to arrival of the bus, with the mean proportion
the server, is 19.96 seconds, with a range of 10 to 40 seconds
equal to approximately 37.9% with a standard deviation of
(standard deviation of 10.3 seconds). We consider this an
32.2%. For example, if the bus will actually arrive at a
acceptable latency for our particular application. Although
destination in 10 minutes, we can expect the prediction of this
these measurements were taken in Seattle, fieldwork over the
arrival time to be, on average about 3 minutes 45 seconds off.
years has demonstrated to the research team that cell coverage
In addition to the accuracy of our model, the amount of time
and SMS latency in Bishkek is consistent with that
required for SMS messages to travel through the network,
experienced in Seattle.
both those giving location updates and those from user
interaction, also affects the accuracy and usefulness of bus
2) User Query Latency
arrival time predictions. We examine these factors below.
Queries return information to users in the form of an
B. Message Latency estimated time (e.g. 5:30pm) that a bus is expected to arrive at
1) *box to Server Latency their location. The usefulness of this information depends
The accuracy of a particular prediction depends on both the upon both its accuracy and how quickly it is returned to the
accuracy of the route model and on receiving a recent, and user.
thus accurate, location update from the bus whose status is Factors affecting query processing latency include: server
currently being queried. Ideally, the server would have load, cell phone network load, and system limitations on rate
instantaneous information about the exact location of a of SMS message sends. Based on our measurements of
moving bus at the point it receives a user query about that bus. message latency in the previous section, we would predict the
In practice, the server must base its prediction on the last roundtrip time for the user to receive a response to their query
location update message received from that bus. Both the to be the latency of two SMS messages (assuming time spent
configured *box frequency of sending updates and any delay at the server is negligible), or approximately 40 seconds.
in that message getting from the *box to the server affects Considering that the typical usage scenario would return
accuracy of information about the current bus location. predictions on buses several minutes away, we consider this
Latency of SMS messages can be affected by a variety of latency to be acceptable.
factors, such as cell phone network congestion, signal
coverage for the *box or the server, or the need for messages VII. CONCLUSIONS AND FUTURE WORK
to cross networks. We recommend that the *box and the The *bus system is a direct response to an articulated
server be on the same cell phone network, as they were in our problem in the developing world, and it takes into close
implementation, to minimize this delay. consideration technology usage patterns and technology
infrastructure available in such resource-constrained
environments.
The system recognizes that information is valuable, and that
the lack of information inhibits certain kinds of growth and
efficiencies. However, there are not always central authorities
or institutions with the resources or motivation to supply
information infrastructure for citizens, and the *bus system
provides a mechanism by which decentralized operators of a
242
transportation system can create information resources for REFERENCES

themselves. It is also a system that uses mobile technology in
a way that does not require the direct participation of the [1] G. Weisbrod, “Models to predict the economic development impact of
transportation projects: historical experience and new applications,” The
mobile provider, thus cutting out one potential barrier to
Annals of Regional Science, Volume 42, Number 3 / September, 2008.
adoption. Finally, *bus enables community use of a single [2] S. Démurger, “Infrastructure Development and Economic Growth: An
GPS system to allow for user-generated geo-coding of the Explanation for Regional Disparities in China?” Journal of Comparative
environment. Economics, Volume 29, Issue 1, March 2001, pp. 95-117.
[3] D. Parikesit and K. Czuczman, “Transport, the missing link? A catalyst
Next steps for *bus include testing in Kyrgyzstan in early for achieving the MDGs,” id21insights, #63, July 2006,
2009. Our goal is to test both intra-city transportation in http://www.id21.org/insights/insights63/insights63.pdf
Bishkek, and inter-city transportation elsewhere, perhaps in [4] “Transport's Role in Achieving the Millennium Development Goals,”
Produced by the DFID Transport Resource Centre, August 2002,
another country. Although the *bus system has been largely Available:
built out with the concept of urban transport in mind, the http://siteresources.worldbank.org/INTWDR2004/Resources/22572_Tra
development impact for rural areas is conceivably even nsportsRole.pdf
[5] A. Bradbury, “Transport, mobility and social capital in developing
greater. In areas where distances are long, roads limited, and countries,” Proc. of the Institution of Civil Engineers, Engineering
vehicles intermittent, a system such as *bus could have Sustainability 159, June 2006 Issue ES2, pp. 79-86.
dramatic implications. [6] NextBus, Last accessed: 22 Sept. 2008. http://www.nextbus.com/
[7] Metro Tracker, Last accessed: 22 Sept. 2008.
Future work on *bus includes examining possibilities such http://transit.metrokc.gov/oltools/tracker.html
as incorporating some way to indicate the capacity of buses or [8] MyBus, Last accessed: 22 Sept. 2008. http://mybus.org/
allowing communication between drivers and riders, [9] The World Factbook, CIA, Last accessed: 22 Sept. 2008.
particularly for longer inter-city routes. A richer set of https://www.cia.gov/library/publications/the-world-factbook/
[10] Pew Internet and American Life Project, “A typology of Information
queries may be developed that allows users to subscribe to a and Communication Technology Users”, May 2006. Available:
notification system, so that they only need to send one SMS http://www.pewinternet.org/PPF/r/213/report_display.asp
message to the server in order to receive several bus arrival [11] J. Donner, “Research Approaches to Mobile Use in the Developing
World: A Review of the Literature”. The Information Society 24(3),
time messages. Another potential expansion is to allow users 2008.
to share tagged location data. Further investigation into [12] “Improving Rural Mass Transit in South Africa - Team Smile,” Last
making the system more scalable by using multiple phones per accessed: 22 Sept. 2008.
http://blogs.technet.com/unlimitedpotential/archive/2008/07/03/improvi
server for high density areas and by running stress tests is ng-rural-mass-transit-information-in-south-africa-team-smile.aspx
needed. Lastly, because the current system is based on SMS [13] MSR India SMS Toolkit, Last accessed: 22 Sept. 2008.
usage which generally requires a certain literacy level, further http://www.codeplex.com/smstoolkit
[14] FrontlineSMS. Company Website, Last accessed: 22 Sept. 2008.
work could explore the possibility of voice interfaces. http://www.frontlinesms.com/
Whether moving goods to market or people to medical [15] R. Veeraraghavan, N. Yasodhar, and K. Toyama, “Warana Unwired:
clinics, transportation is a crucial resource for any community. Replacing PCs with Mobile Phones in a Rural Sugarcane Cooperative,”
Allowing individuals to find ways to maximize such shared in Information and Communication Technologies for Development,
2007.
resources is the goal of *bus. Our approach has been a method [16] B. Kolko, E. Rose, and E. Johnson, “Communication as information-
of problem-solving that is bottom-up, that conceptualizes a seeking: the case for mobile social software for developing regions,”
technology device as shared among the community, and that Proceedings of the 16th international conference on World Wide Web,
2007.
recognizes the importance of information in people’s everyday [17] P. Javid and T. Parikh, “Augmenting Rural Supply Chains with a
lives. Our goal with this paper has also been to bring to the Location-Enhanced Mobile Information System,” in Information and
fore the importance of transportation for both economic and Communication Technologies for Development 2006, May 25-26, 2006,
Berkeley, CA.
human development – an area of ICT for development that has [18] M. Chen et al., “Practical Metropolitan-Scale Positioning for GSM
not yet received much attention. Phones,” in UbiComp 2006: Ubiquitous Computing, Springer Berlin /
Heidelberg, 2006.
[19] MySMS, Last accessed: 22 Sept. 2008. http://mysms.sourceforge.net/
ACKNOWLEDGMENTS [20] SMSLib, Last accessed: 22 Sept. 2008, http://smslib.org/
We would like to thank members of the original *bus and [21] A. Karbassi and M. Barth, “Vehicle route prediction and time of arrival
estimation techniques for improved transportation system management,”
MySMS teams of the 2007-2008 CSE477 at UW: Martin in Proc. of Intelligent Vehicles Symposium, 2003. IEEE, June 2003,
Hecko, Naasir Ramji, Jordan Walke, Brad Campbell, Ruibo pp. 511-516.
Li, and David St. Hilaire. We would also like to thank CAICT
team members Cynthia Putnam, Emma Rose, Rebecca
Walton, Erica Johnson, Mark Licata, and also Professor
Medina Aitieva, our local research team, and research
participants. We thank Aidai Seidakmatova, Shairbek Juraev,
and Natalie Linnell for their contributions. A special thanks to
Bruce Hemingway and Jesse Dosher for the design and
construction of the hardware.
243
Challenges in Health Information Systems Integration:

Zanzibar Experience
Edwin NYELLA
Department of Informatics, University of Oslo, Box 1080, Blinden 0316 Oslo, Norway
Tel: + 47 22 85 24 10, Fax: +47 22 85 24 01, Email: edwinen@ifi.uio.no
ABSTRACT -- The healthcare milieu of most developing countries is often characterized by multiplicity of health programs
supported by myriad of donors geared towards reversing disease trends in these countries. However, donor policies tend to
support implementation of vertical programs which maintain their own management structures and information systems. The
emerging picture overtime is proliferation of multiple and uncoordinated health information systems (HIS), that are often in
conflict with the primary health care goals of integrated district based health information systems. As a step towards HIS
strengthening, most countries are pursuing an integration strategy of the vertical HIS. Nevertheless, the challenges presented
by the vertical reporting HIS reinforced by funds from the donors renders the integration initiatives ineffective, some ending
up as total failure or as mere pilot projects. The failure of the systems after implementation transcends technical fixes. This
paper drew on an empirical case to analyze the challenges associated with the effort to integrate the HIS in a context
characterized by multiple vertical health programs. The study revealed the tensions that exists between the ministry of health
which strived to standardize and integrate the HIS and the vertical programs which pushed the agenda to maintain their
systems alongside the national HIS. However, as implied from the study, attaining integration entails the ability to strike a
balance between the two forces, which can be achieved by strengthening communication and collaboration linkages between
the stakeholders.
Index Terms -- Health information systems, Integration, Standardization, Vertical programs
I. INTRODUCTION Due to the donor pressure associated with the vertical

Health information systems (HIS) in developing programs, the health information systems evolved in
countries have in recent years gained more and more a rather chaotic and fragmented manner with multiple
attention as more effort by governments, international and overlapping demands from both the vertical
agencies, non governmental organizations, donors programs and the national health administrative
and other development partners seek to improve departments and ministries. The vertical programs
healthcare as a way to reverse disease trends in these usually maintain their own ‘vertical’ reporting health
countries. As part of the efforts towards reversing information systems existing alongside with the
disease burden and provision of improved healthcare national health information system where the result
services, most countries are implementing primary emerging over time is uncoordinated, disintegrated
health care (PHC) approach. However, and heterogeneous collection of systems [8] [12]. As
implementation of the PHC approach in most the corollary, many studies reported the dismal state
countries is organized in decease-focused and specific of the health information systems as being
services programs (such as Malaria, HIV/AIDS, TB predominantly ineffective, unreliable, irrelevant and
&Leprosy and Family Planning programs, etc), therefore inadequate in providing the management
dubbed as ‘vertical programs’ [2] [5] [12] [13]. The with the needed information [3] [8] [12] [13] [15].
vertical programs are mostly funded by donors who
In an attempt to ensure availability and accessibility
come with specific requirements related to the
of comprehensive health information to the national
monitoring and evaluation of their funds and the
health departments, districts and the vertical
program at large. For instance in my case study, the
programs; most countries are pursuing an integration
HIV/AIDS program alone is supported by a number
strategy of the fragmented systems. Whereas, some
of donor agencies such as the Global fund, Center for
countries have managed to standardize and integrate
Disease Control (CDC), United Nations for
some of the vertical programs into the national HIS,
Development Programmed (UNDP), United Nations
ensuring continue reliance and use of the integrated
Children's Fund (UNICEF), World Bank and the
system by the vertical programs is still a big
World Health Organization (WHO), where almost
challenge [16].
each agency has a number of indicators which require
huge amounts of data to be collected. The fragmentation of the HIS after integration,
ensuing from the non reliance and compliance to the
integrated system, has not received much attention
244
either empirically or analytically by the contemporary contextual particularities related to politics,

research in information systems integration. Much of institutional conditions, high resource constraints
the research focused on the challenges in the process (infrastructure, human resources, financial resources),
of achieving integration, for instance, [1] examined high disease burdens and the particularities of the
the potential and challenges of integrating the HIS of diseases, in which all together challenge the process
Malaria, TB and HIV/AIDs programs and also of integrating the HIS [1] [17].
integration of multiple reporting channels within each
Most of the developing countries are funded by
program. Another study looked at the challenges
international donor agencies such as the World Bank,
posed by the historicity (the conservative influence of
Global fund, and the Clinton Foundation, in order to
historically accumulated and institutionalized
support provision of health services (such as Family
practices, technologies and perceptions) and
Planning, Immunization and VCT) to the population.
heterogeneity (lacking integration and increasing
However, donor policies tend to support
fragmentation across the collection of information
implementation of disease specific programs dubbed
systems) of information systems in the development
‘vertical programs’ which maintain their own
and integration of the health information systems [8].
management structures and information systems [2],
[9] looked on the problems of fragmentation and
which are often in conflict with the primary
challenges of integration of the routine health
healthcare goals of integrated district based health
information system and the prevention of mother to
information systems. For the case of Zanzibar for
child transmission program (PMTCT).
instance, the HIV/AIDS program alone which
The thesis of this paper is the emphasis of the need to maintains its own information system is supported by
comprehend the nit gritty of what goes on after a number of donor agencies such as the Global fund,
integration. By specifically understanding the way Center for Disease Control (CDC), United Nations
vertical program managers (as users of the HIS) for Development Programme (UNDP), United
receive and relate to the newly implemented system, Nations Children's Fund (UNICEF), World Bank and
can help us answer though partly as to why some the World Health Organization (WHO), where almost
health information systems integration in the context each agency has specific requirements concerning
of developing country fail or run short of data to be collected.
expectations, after implementation.
In a study on HIS of the disease specific programs in
The rest of the paper is organized as follows: The Mozambique, a low income country, [1] identified
literature covering challenges of HIS integration in major challenges related to the integration of the HIS
developing countries are presented on the subsequent to include heterogeneity of interests among donors,
section followed by the research context and managers and health reformers; multiplicity of
methodology pursued in the study. The empirical reporting systems even within an individual program
basis of this paper is an integration effort in Zanzibar, and high disease burden. [20] discussed how poor
which is presented next and followed by the analysis infrastructural conditions and lack of transport
and discussion section. Finally, the conclusion section challenge the flow of health information from the
of the paper is put forth, where implications and district to the provincial levels of the health
contributions of the study are spelt out. administration hierarchy. Low or lack of computer
skills have also been quoted as one of the factors
II. CHALLENGES IN HIS INTEGRATION
contributing to the challenges to attain an integrated
Health information systems integration in developing
district based HIS, especially in the rural context of
countries is considered as an approach towards the
most developing countries [8] [6].
rationalization and unification of the disparate
systems in these countries. The objective of Furthermore, lack of uniform infrastructure
integrated HIS is to provide easy and equal access to development and uneven distribution of resources
relevant information to all stakeholders [12]. (e.g. humans, computers) in most developing
However, ensuring an integrated HIS in these settings countries, challenge the efforts to attain
is quite a big challenge. comprehensive integrated health information system
[18] [4]. For instance, the uneven distribution of
Generally, the challenges of HIS integration emanate
human and technical infrastructure in Mozambique
from both social and technical factors [13]
was reported as being problematic to the effort of
surrounding the integration processes. It is argued
scaling up the district-based health information
that it is more so in developing countries due to
systems because some of the remote districts do not
245
have electricity. Similar problems of uneven (HMIS), meant to provide information support to all
infrastructure development was experienced in decision making processes of the entire ministry. In
Ethiopia, where according to [18], use of standardized this paper, the term ‘Health Information System
(HIS)’ is used to refer to the HMIS and the term
data formats served as gateways between the paper
‘HMIS unit’ is used to refer to the national level
based systems at the periphery levels and computer MoHSW department, responsible for the health
software at the higher levels of the health system information system.
hierarchy. Moreover, [10] alluded to some of the
The study was done as part of an ongoing action
challenges faced by most developing countries in
research on health information system restructuring
achieving standardized and integrated HIS to include: process undertaken by the Health Information System
conflicting interests between health programs which Program (HISP) in collaboration with the MoHSW
make it difficult to reach a “final” agreement; changes and other development partners in Zanzibar. HISP is
being the only constant, where new needs keep a global action research and development network on
popping up (e.g. HIV/AIDS); and multiple software health information systems by the University of Oslo
and paper tools which are difficult to coordinate and in Norway and other stakeholders, which started in
South Africa in 1994 and thereafter spread to other
change.
developing countries such as Ethiopia, Vietnam,
Whereas, some countries have managed to India, Mozambique, Botswana, Tanzania Mainland,
standardize and integrate some of the vertical Malawi and Nigeria. In all these countries, the
programs information systems into the national HIS, primary goal of HISP is to design, implement, and
ensuring continue reliance and use of the integrated sustain Health Information System through
system by the vertical programs is still a big participatory approaches to support local management
challenge [16]. As presented by [21], some of the of health care delivery and information flows [10]. In
reasons for the categorical programs refusal to 2005, Zanzibar became a node in the HISP network,
support an integrated HIS include: fear that their when the University of Oslo was contracted by the
requirements will not get the attention needed to MoHSW to engage in the restructuring of the HIS.
ensure that their needs are met, and if the programs Case study methodology was drawn upon in the field
have pride of ownership in their existing vertical during data collection and analysis. Case study is an
systems, which meet their needs. The fragmentation appropriate approach for bringing an understanding of
of the HIS after integration, ensuing from the non a complex issue, which could be a program, event, an
reliance or refusal to support the integrated system, activity or a process involving one or more
has not received much attention empirically and individuals and using a variety of data collection
analytically by the contemporary research in procedures over sustained period of time [11]. The
information systems integration. As the literature aim of the research being to develop a comprehensive
depicts, much of the research focused on the understanding of the challenges related to the way
challenges in the process of achieving integration. users received and engaged with the integrated HIS,
The argument of this paper is the need to comprehend case study proved to be a feasible approach.
the nit gritty of what goes on after integration. By
specifically understanding the way vertical program The research was carried out in three health districts;
managers (as users of the HIS) receive and relate to two in Unguja and one in Pemba Island within a
the newly implemented system, can help us answer period of five months (June to November, 2006).The
though partly as to why some health information districts visited in Unguja were Urban and West; both
systems integration in the context of developing of them located on the west region of the island. Most
country fail or run short of expectations, after of the MoHSW main offices such as HMIS unit and
implementation. vertical programs’ offices are located in these two
districts. In Pemba the research was conducted at
III. RESEARCH CONTEXT AND METHODS Chake Chake district, which is located on the south
The research study was conducted in Zanzibar. region of the Island.
Zanzibar is a semi-autonomous region within the
United Republic of Tanzania, which is made up of Semi structured interviews were used, where
two main islands, Unguja and Pemba, and several interviewees were asked open ended questions to
others islets located in the Indian Ocean. Zanzibar has elicit their viewpoints related to the use of the
an area of 2,332 square kilometres (the total area of integrated HIS. At the health unit level, 38
Tanzania is 945,000 Km), and is divided into five informants were interviewed, 19 in Unguja and 19 in
administrative regions, each with two districts, Pemba, where impressions and perceptions of the
making a total of ten districts in the entire region. health workers in relation to the previous and the
Zanzibar maintains its own health system that is newly integrated HIS were gathered. The goal was to
administrated by a semi-autonomous Ministry of learn micro level challenges emanating from the way
Health and Social Welfare (MoHSW). Side by side users responded to the integrated HIS. At the macro
with the health system is an information system level the interviews involved the following
called Health Management Information System informants: District medical officers, District health
246
officers, vertical programs district and general adapted in South Africa and subsequently adopted in
managers and central level HMIS unit officials. other developing countries such as Malawi,
During the interviews different informants’ Mozambique, India, Ethiopia, Namibia, Zambia,
viewpoints in relation to the new datasets and tools Mali, Botswana, Vietnam and Nigeria.
were gathered. Table 1 depicts the number of
informants interviewed. The HMIS restructuring process followed a
participatory and incremental approach in the revision
Name /Position Informants of the previous datasets and creation of new once.
Health unit level staff 38 The incremental approach paved way for the
District Medical Officers 3 involvement of key stakeholders in each stage of the
Vertical programs data managers 7 revision exercise. Among the new datasets included
Programs general managers 2 was the Expanded Programme on Immunization
National level HMIS unit officials 3 (EPI), Disease surveillance, Reproductive and child
District Health Officers 3 health (RCH), STI and HIV, and Maternity dataset.
HISP team members 2 Almost each dataset brought together a number of
TOTAL 58 stakeholders for instance the Disease surveillance
dataset apart from the national HMIS, had two main
Table 1: Interviews conducted at health facility and above
stakeholders, the EPI and Malaria programs. Malaria
A number of documents and software tools were program in Zanzibar is very strong unlike in other
analysed in the field. For instance the district countries due to many interventions by donors
implementation plan was analysed in an attempt to directed towards reversing the malaria trends in the
understand the use of the new system in the country. This made malaria related data more
preparation of the plan. The software tools were valuable purposely because of the need to monitor
analysed to assess compliance to the newly integrated and evaluate the interventions. The disease
system. surveillance being the main source of the malaria
data, the program was involved in the design
IV. CASE DESCRIPTION implementation stages to ensure that the dataset met
The health information system in Zanzibar was the requirements needed. For instance, the program’s
organized haphazardly and mainly shaped by the data manager was involved in the training of the new
organization of fragmented vertical programs with disease surveillance dataset. However, despite of the
their own information systems [14]. The vertical involvement the data manager enacted separate data
programs include the Family Planning (FP), Malaria, collection tool behind the scene and circulated it in
and Expanded Program on Immunization, some districts. When asked why, the answer was very
Tuberculosis and Leprosy, Nutrition, HIV/AIDS and clear,
Safe motherhood programs. Most of the programs’ “They have taken out almost all the age
services are integrated at the health unit level, but group categories and left what they feel will
maintain separate reporting systems. A situational satisfy their needs, but what about us. So we
analysis of the health information system disclosed tried to design it to show them how it should
plethora of problems which included scarcity of look like” (Manager, Vertical program,
resources, gaps in data collection tools, poor analysis August, 2006)
of data, fragmentation of the higher levels, poor Likewise, the EPI Disease surveillance tool (see
feedback and lack of motivation and limited Figure 1) was integrated in the new dataset for
information use [14]. disease surveillance.
As an attempt to improve the situation, a strategic
plan was drafted by a joint team of stakeholders;
comprising both scientific and organizational
researchers, major Ministry of Health and Social
Welfare (MoHSW) donors (DANIDA and WHO),
University of Oslo and some officials from the
MoHSW (ibid). The roadmap detailed and agreed on
the major activities to be undertaken. The University
of Oslo under its action research program called
Health Information System Program (HISP), was
contracted to undertake the task of restructuring the
system by way of integrating the highly fragmented
HIS. HISP as explained hitherto is a research network
that aims at enhancing district health information
systems in developing countries through introduction Figure 1: Monthly disease surveillance report
and local adaptation of open source software. The
software which is known as District Health Although the new dataset was running for more than
Information Software (DHIS) was developed and eight months (till the time of writing), still EPI kept
247
on collecting data using their system. The reason, as

explained by one official was that, Though the program officers participated in the
“.. Until we are sure of getting our data from design process of the dataset, the participation as
HMIS, we can not abandon our system” explained by one officer was meant to help HMIS
(Manager, Vertical program; July, 2006) unit get HIV/AIDS related data. Lack of trust by the
vertical program to the capacity of HIMIS unit to
The aftermath of that practice was duplication of maintain and sustain the information system was
work at the point of data collection. Similarly, in envisioned as one of the reasons:
some districts, family planning was running in “……. Mostly we rely on donors in almost
parallel with the new integrated system despite the everything which sometimes lead to mistrust
fact that it was integrated in the RCH dataset, and by the vertical programs of our capability to
monthly submitted to the district. The practice was maintain and sustain the information system.
commonplace until the time of writing. When some For instance EPI are performing well
health workers at the health facility level were asked because they have enough funds. Also
why, one of them answered, HIV/AIDs have many donors which imply
“.. The new forms do not have all the enough funds, unlike HMIS which has very
required data elements as the old ones. So scarce resources both physical and human
we fill in the old one to make sure that all the resources”. (HMIS Official, July, 2006)
required information is taken to the owners
(the Family Planning program) “(RCH As an attempt to resolve the dilemma according to the
coordinator, July 2006). HMIS official, the HMIS unit resorted to
participatory approaches such as meetings,
Another health officer noted, workshops and seminars to build consensus. An
“We have not been told to abandon them; we example being mobilization of concerted efforts and
still submit them monthly to the district.” funds by HMIS unit to resolve availability of data
(MCH Aides, August 2006). collection tools problem. The problem occurred when
HMIS failed to fund production of data collection
At the district level some data officers kept on tools due to financial constraints. To resolve the
demanding submission of the family planning report problem which if left unsolved would have
using the separate data tool. This was partly attributed undermined the whole system, vertical programs as
to by the inadequate knowledge about what was one of the major stakeholders were summoned in a
supposed to be done as far as the old and the new meeting to deliberate and agree on strategies to solve
reporting system was concerned. Lack of teamwork the problem. In the meeting, it was agreed that
and sharing of information at the district level vertical programs contribute on the production of data
between those who participated in the design process collection tools. Most of the programs considered the
and those who didn’t led to this malady. For, most of idea as one of the feasible solution for the problem.
the HMIS work at the district level was in most cases Those who were at first reluctant, slowly as they saw
done by those who were not involved in the process. others responding, they also followed suit. Based on
the agreement, production of both primary and
HIV/AIDS was another program which despite being secondary data collection tools to be used for a period
involved in the process of designing new dataset for of one year was done using funds from different
HIV and STI services, kept on running their previous vertical programs.
tools separately. This program maintained its own
fragmented information system, one for VCT and As an attempt to explain the reason and solution for
another one for STI services. These subsystems were the continued use of the previous systems one
integrated into one STI and HIV/AIDs dataset. The stakeholder from the donor community argued:
dataset was functional for more than six months with “We have agreed if possible not to talk about
data routinely collected and collated from almost all integration. For when people working in
Health facilities providing the two services. Although these disparate systems hear this, they think
most of the data was submitted to the districts and of being robbed of their jobs and hence their
transmitted to higher levels, the data was not fetched salaries. This problem is more pronounced
and used by the HIV/AIDs program. Instead, the in this context where there is massive
program kept on depending entirely on their previous unemployment rate. We better talk about
systems. The reason given was that the new dataset communication and collaboration between
does not fulfil data requirements for program the vertical programs and the HMIS unit”
management and so it was not designed for the HIV (Program director, Donor Community; Nov,
program but for the HMIS unit. 2006).
“The new tools are for the higher levels
only; they can not help us in any way. We
need more information compared to what is
on the HMIS form. It is not designed for us”
(Manager, Vertical program; July, 2006)
248
V. ANALYSIS AND DISCUSSION RCH tool where the family planning data elements
The integration of the health information system were minimized. Most of the users argued that they
(HIS) involved standardization of the datasets, data were using the old tools to ensure that information
collection tools, data processing tools and associated owners get all the information they needed. In
work practices; and institutionalization of the addition, by building on their past practices, some
standards to the levels of the health system. district officers kept on enquiring about the separate
Following is the analysis and discussion of the family planning reports from the health facilities.
challenges resulted from the way vertical program
managers received and related to the newly integrated Health officers in some districts drew on their limited
system. knowledge of DHIS software to enact limited use of
the integrated HIS, where preparation of monthly
Limited use of the integrated HIS reports was done by aggregating manually a number
The integration initiative sought to align different of datasets, an exercise which took two to three days
disparate fragmented information systems to form an to finish. While at the same time, all datasets were
integrated data repository at the district level which already entered in DHIS by the data clerk, in which a
gives access of data to different stakeholders. monthly report could be printed out without much
Though, this was achieved to a certain degree, some hurdle. The decision to go back to the manual
other vertical programs whose datasets were aligned systems can be envisioned as to have been made due
with the new initiative enacted limited use of the to the dilemma of not knowing how to use the new
newly integrated HIS by running some of their systems but also by the demand of the need to
previous information systems and tools in parallel compile monthly reports.
with the new system. For instance, the disease
surveillance dataset for EPI program was aligned with Other district health officers enacted limited use by
the national disease surveillance standard. Though the engaging themselves more with a dataset for a
national dataset was in operation for more than eight particular vertical program. This came to play as the
months until the time of writing, the vertical program result of the officers drawing on their past
managers decided to run their disease surveillance experiences of vertical system mindset contrary to the
data collection tool in parallel with the national new integrated HIS where all the datasets needed to
system. This vertical program however had another be afforded equal attention. This was evident from
dataset (on immunization) which was aligned with the what one health officer provided as an answer when
new initiative and which was running smoothly. The he was confronted for poor performance of some
immunization dataset is an independent dataset, with datasets, where he asserted that his responsibility was
the EPI program as the main stakeholder. This is a particular dataset whose performance was good.
different from the disease surveillance dataset which The limited engagement with the integrated HIS was
was formed by fusing together the national disease further observed at the health unit level where health
surveillance dataset and the EPI surveillance dataset. workers dealt with particular datasets more than
To regain control of their disease surveillance system, others. This limited use at the health unit or district
the program managers decided to run their dataset level came to play as users acted on their past
alongside the integrated dataset for disease practices related to vertical programs to conceptualize
surveillance. the new system which entailed equal treatment of all
datasets to ensure availability of comprehensive data
Furthermore, the limited use came to play as the at all levels.
result of the lack of trust by the vertical program
managers to the capability of HMIS unit to ensure Completely none-use of the integrated HIS
sustainability of the integrated HIS. This was evident Some other actors determined to completely shun
from one of the program manager’s assertions that, away from the newly integrated HIS in spite of the
unless we are sure of getting our data, we can not fact that, they were involved in the design process of
abandon our system. Therefore, the managers drew datasets related to their program. HIV/AIDS program
on the history of poor economic status of HMIS unit managers participated in the design of a new dataset
and on their future prospects about the new system’s called STI and HIV dataset, which aimed at collecting
sustainability to make the decision to maintain their data related to both services. However, this dataset
system alongside the new system. was operational for about five months until the time
of writing, but since then the data collected was not
Another health program whose users enacted limited fetched and used by the program. The idea of having
use of the new system was the family planning. In both systems running in parallel may sound logical,
some districts, family planning dataset was running in because the setup gives every stakeholder access to
parallel with the newly integrated system despite the the data; contrary to the previous systems where data
fact that their data elements were integrated in the was vertically submitted to the programs and donors.
RCH dataset. This resulted as users drew on their past However, the setup has enormous implication on the
practices of submitting huge amounts of data to the workload to the data collectors. Rather than
vertical programs unlike in the new standardized rationalizing the fragmented systems and minimize
249
duplication of data, the setup intensify it and HMIS unit to solve availability of data collection
ultimately jeopardize the quality of the data collected. tools problem.
Nevertheless, it must be emphasized that the tensions
The none-use of the integrated HIS, came to play as which resulted from limited and none-use of the
the program officers’ drew on a number of newly integrated HIS were not static, but dynamic in
assumptions and on their multiple needs of data. nature. The level of dynamism was different from one
Some of the assumptions in relation to the new user group to another and from one vertical program
system are that, the new system was for HMIS unit, to another. For instance, the HIV/AIDs which initially
and so their participation aimed at helping them get opted for none-use of the integrated system, at the
their data and that the data collected in the new very end of the field study in an interview with the
system did not satisfy their needs. The assumptions data manager; he sanctioned the software to be
can be envisioned to have resulted from users installed in their computers. His acceptance came due
drawing on their past experiences of collecting to the need to make comparison between the data they
multiple data and on their desire for data to meet collect and the HMIS data, supplementing their data
future needs. in case of some missing data in their systems. Though
the manager took the decision after learning that
The none-use enacted by the HIV program as was the having the new system won’t prevent him from using
case for the limited use explained above, was also their systems, but I see this as a movement from
mediated by the poor economic conditions of the none-use to limited use. Furthermore, the health units
HMIS unit, which led to mistrust of the vertical and districts officers who opted for limited
program to the capability of HMIS unit to ensure engagement by drawing on their previous vertical
sustainability of the system over a period of time system experiences, slowly started to change as they
taking into account its almost total dependency on learned through informal trainings, feedback meetings
donors. and seminars that all the datasets needed to be
afforded equal significance. The change from none-
Tensions in the HIS Integration use to limited use conforms to the argument that,
The resulting picture from the limited and none-use other embedded practices can be changed or replaced
of the newly integrated HIS, is what I dubbed as by others over time [18]. However, contextual
‘pulling effect’ (see figure 2). particularities such as the poor economic status of
National HIS Vertical Programs most healthcare settings in developing countries
sometimes reinforce their existence and hence make
Struggle to Want to maintain
them hard to change.
standardize & their own systems
integrate the HIS (fragmentation)
Advocate on Advocate multiple The need to strengthen communication and
minimum action-led data needs to satisfy collaboration linkages between stakeholders
datasets multiple and The vertical programs in the Zanzibar case have very
No enough funds to heterogeneous
sustain the system donors
strong installed base of information systems
Use of participatory Have funds to reinforced by funds from donors. Moreover, these
approaches to build develop & maintain systems have very well defined and elaborate vertical
trust their systems organizational structures with many people employed
in there, further reinforcing their strengths and the
Pulling Effect tension towards any change attempts. For instance,
Figure 2: Tensions between the national HIS and the vertical the HIV/AIDS program maintains its own
programs administrative structure and employees working on
On one side HMIS unit under the MoHSW is the information system.
struggling to standardize and integrate the fragmented
information systems and on the other side the vertical As pointed out earlier, although most of the vertical-
programs opted for limited and none-use of the newly reporting systems were harmonized and aligned in the
integrated HIS by maintaining their own systems. The integration initiatives, some of the systems were still
upshot of that is a pulling effect on either side where running alongside the integrated system. This state of
the winner is determined by the power (e.g. to argue, inertia could be explained partly by the perceived
funds, human resources, good strategies.), which one results of integration, including fear to lose positions,
of the two sides need to have in order to haul the those with vested interests with the old systems to
opponent. lose them and the mistrust on the capability of the
national HIS in managing and sustaining the
In an attempt to alleviate or eliminate the tensions, integrated system.
the HMIS unit ventured to consensus building
through participatory approaches (PA) like meetings, The argument as implied from the case is that
workshops and seminars to try to strike a balance communication and collaboration between all the
between the two forces. This is exemplified by necessary stakeholders need to be built and
mobilization of concerted efforts and funds by the strengthened as a strategy to deal with the inertia of
250
the vertical reporting systems. The national HIS [23]. This is further reinforced by [21], who argues
however, need to take a stewardship role to ensure that pooling of resources by the categorical health
that comprehensive data is obtained from the programs to the integrated HIS is a difficult
disparate systems in a cost effective way. This further undertaking. The challenges related with the pooling
suggests that some of the vertical systems to run of resources underline the critical need of the
alongside the national HIS, but with the mandate that proposed strategy of strengthening communication
the national HIS take the driver’s seat in ensuring and collaboration linkages and the need to make use
smooth collaboration and communication between the of gateways to ensure flow of information from
stakeholders. This integration perspective is inline vertical systems to the national HIS.
with the concept of accepting to live with a
reasonable level of none integration, since no one, VI. CONCLUSION
including the national health authorities, is in Integration of HIS in developing countries
‘control’ in any strict sense; and therefore a relevant characterized by multiple vertical programs is quite a
strategy cannot be based on a planning or control challenge. It involves economic and political
approach [8]. processes in articulating interests, building alliances
and struggling over outcomes [19]. The tensions
Hand in hand with the need to strengthen
between the national HIS which strive for integration
communication linkages, is the need to use gateways
to link between the national HIS and the vertical and the vertical programs which advocate their
programs information systems which seems strong information systems supported by donors represent
and hard enough to integrate with the national HIS. both challenges and opportunities to the integration
Gateways allow the continuous existence of multiple initiatives. One of the challenges as implied from the
systems, each with their internal organization and study is ensuring compliance of the integrated HIS by
logic (ibid). Consequently, gateways can ensure the vertical programs in the face of poor economic
transfer of data from strong multiple vertical systems
status of most developing countries. However, as
to the national data warehouse, and therefore make
the data available to all stakeholders in a cost implied from this study, strengthening the
effective way. Hence, the use of gateways supports collaboration and communication linkages between
the concept of accepting a certain degree of none the national HIS and the vertical programs represent
integration, while keeping every stakeholder ‘happy’. an opportunity to curb the integration challenges.
While the paper propose this perspective as a strategy
From the case, the communication and collaboration
towards HIS integration, more empirical research is
perspective is exemplified by the approach used by
the HMIS Unit in resolving the availability of data needed to find out more how it can be achieved in
collection tools problem. The problem happened practice.
when the unit failed to fund production of data
collection tools due to financial constraints. To solve The message the paper sends to public health and
the problem, which if left unsolved would have rolled other practitioners in HIS is that misconstruing or
back the entire system; the HMIS unit summoned all under-estimating the non-technical issues of
key stakeholders in a workshop to deliberate on the integration can account for a very significant portion
strategies to solve the problem. In the workshop, it of failures of information systems in healthcare.
was unanimously agreed that each stakeholder While technically, the integrated national HIS
(vertical program) contributes some funds for consisted of datasets catering for the needs of the
production of the tools. As the results, production of vertical programs covered by this study, the programs
tools to be used for a period of one year was made however kept on using their previous systems,
possible through communication and collaboration presenting challenges to the integration initiatives.
between the stakeholders but with the HMIS unit Hence, the paper underscore the need to understand
taking the stewardship role. The national HIS using the socio economic challenges related to HIS
communication and collaboration processes can play integration which goes beyond technical fixes. The
the same role to ensure availability of comprehensive study further reiterates the need to look beyond the
data. integration process, to include the way users receive
and engage with the newly integrated HIS.
The strategy used in Zanzibar of soliciting funds from
vertical programs may point to the idea of donors as VII. REFERENCES
the main supporters of the health programs to pool [1] B. Chilundo, Integrating Information Systems of Disease-
resources together as in the case of the Sector Wide Specific Health Programmes in Low Income Countries: The
Approach (SWAp) [22], to support the integrated Case Study of Mozambique, Ph.D. thesis, Faculty of
national HIS. However, despite the SWAp policy Medicine, University of Oslo, 2004.
mandate of strengthening the local capacity to [2] B. Chilundo and M. Aanestad, Vertical or Integrated
manage funds from donors, evidence shows that Programmes? The consequences for the laboratory
information systems in Mozambique, In Proceedings of the
countries such as Tanzania is facing difficulties in
IFIP TC8 & TC9/ WG8.2+9.4, 2003.
coordinating the funds using the so called basket fund
251
[3] F. Mukama, A study of health information systems at local R. Montealegre and A. Poulymenakou (Eds), Athens, Greece:
levels in Tanzania and Mozambique, Improving the use and 35 -51, 2003.
management of information in health districts, Master thesis,
Department of Informatics, University of Oslo, 2003. [21] G. Land, N. Hoffman and R.Peterson, ”Developing the
[4] J. Braa, E. Monteiro and S. Sahay, Networks of Action, MIS
Missouri Integrated Public Health Information System”, In
Quarterly 28(3): 126, 2004.
O’Carroll PW, YasnoffWA,Ward ME, Ripp LH, Martin EL,
[5] J. Braa, O. Hanseth, W. Mohammed, A. Haywod, and V.
Shaw, Developing Health Information Systems in Developing eds. Public Health Informatics and Information Systems. New
Countries: The Flexible Standards Strategy. In press, MISQ, York: Springer; 2002, 617–64.
Special Issue on IT and Development, 2005.
[6] J. Lungo, Data flows in Health Information Systems: An [22] Brown A, “Integrating vertical health programmes into
action research study of reporting routine health delivery sector-wide approaches: Experiences and lessons”. Institute
services and implementation of computer databases in health for Health Sector Development, on behalf of the Swiss Agency
information”. Master thesis, Department of Informatics, for Development and Co-operation (SDC) 2001, Pages: 30.
University of Oslo, 2003. [23] Smith, M., Madon, S. and Anifalaje, A. ”Integrate Health
[7] J.J.Rubona, Routine health information systems that operate Information Systems in Tanzania: Experience and
in Tanzania”. In the RHINO workshop on issues and
innovation in routine health information in developing Challenges”, EJISDC (2008), 33, 1, 1-21.
countries, The Bolger Center, Protomac, MD, USA 14-16
March 2001. Arlington, VA 22209, USA: MEASURE
Evaluation, JSI Research and Training Institute, 183 –193.
[8] M. Aanestad, ., E. Monteiro, H. Kimaro, E. Macombe, G.
Macueve, F. Mukama, H. Muquingue, L.J Nhampossa, and J.
Lungo, Strategies for development and integration of health
information systems: coping with historicity and
heterogeneity. Working paper at the University of Oslo,
Department of Informatics, 2005.
[9] N. H. Shidende, Challenges and approaches to the integration
of HIS: Case study from Tanzania”, M.Sc. thesis. Department
of Informatics, University of Oslo, 2005.
[10] Braa, J. and Humberto M (2007): “Building collaborative
networks in Africa on health information systems and open
source software development – Experiences from the
HISP/BEANISH network”, IST Africa, 2007.
[11] R. Yin, Case study research: Design and methods (2nd ed.).
Thousand Oaks, CA: Sage Publishing, 1994.
[12] RHINO, The second RHINO workshop on: Enhancing the
Quality and Use of Routine Health Information at the District
Level”, September 29 – October 4, 2003.
[13] World Health Organization (WHO), Health information
systems in support of the Millennium Development Goals,
2006.
[14] Y.H. Sheikh, Improving health information systems at health
districts in Zanzibar, M.Sc. thesis. Department of Informatics,
University of Oslo, 2005.
[15] R. Sauerborn, and T. Lippeveld, Introduction. In: T.
Lippeveld, R. Sauerborn, and C. Bodart, eds. Design and
Implementation of Health Information System. Geneva,
World Health Organisation, 1-14.
[16] M. Galimoto, Integration of Health Information Systems:
Case Study from Malawi, MSc. thesis, University of Oslo
2007.
[17] Sahay, S., E. Monteiro, and M. Aanestad, “Towards a
Political Perspective of Integration in IS Research: the case of
Health Information Systems in India”. Working paper,
Department of Informatics, University of Oslo, Norway,
2006.
[18] Lagebo, B. and S. Mikonnen (2005): “Challenges and
Approaches to Scaling and Standardizing Health Information
Infrastructure in Developing Countries: Case Studies from
Ethiopia”. M.Sc. thesis, University of Oslo.
[19] B. Chilundo and M. Aanestad, Vertical or Integrated
Programmes? The consequences for the laboratory
information systems in Mozambique, In Proceedings of the
IFIP TC8 & TC9/ WG8.2+9.4, 2003
[20] Mosse, E. and S. Sahay, “Counter Networks, Communication
and Health Information Systems: A Case Study from
Mozambique”. In The IFIP TC8 & TC9/ WG8.2+9.4
Working Conference on Information Systems Perspectives
and Challenges in the Context of Globalization. M. Korpela,
Cross Technology Comparison for Information 252
Services in Rural Bangladesh

Faheem Hussain, Member, IEEE; Rahul Tongia, Member, IEEE
Abstract— This paper introduces a framework to examine the services may not be effective for disaster response. This paper
relative effectiveness and cost-effectiveness of different looks into the relative effectiveness and costs of different
information and communications technologies to deliver a range technologies to deliver different social services in rural
of social services in rural Bangladesh. It focuses particularly on
Bangladesh, focusing on major development sectors such as
major sectors like agriculture, education, disaster response and
healthcare. Expert elicitation (using both local and international agriculture, education, disaster response, healthcare etc.
group of experts) survey on ICT effectiveness by service domains This analysis has several components. First, expert
shows localization as the key determining factor for any opinions on ICT effectiveness by service domain (health,
technological intervention. Community based radio broadcasting education, etc.) for various ICTs (TV, radio, print, telecenters,
(CR) appears as the dominant option among the considered etc.), with international and local (Bangladeshi) experts in
ICTs. Human intervention has been found to be crucial in both
communications and development were gathered via expert
low (20%) and medium-high (60%) literacy populations. Our
study also shows public funded Terrestrial TV is more effective elicitation surveys, taking care to manage anchoring and other
and has higher acceptability in providing information over biases (detailed subsequently). The elicitations were for low
privately owned Satellite/Cable TV channels. Literacy doesn’t (20%) and medium-high (60%) literacy populations. Second,
seem to significantly affect the relative effectiveness of we created an extensive techno-economic model for the
Information Centers with Internet connection over Print and TV various ICTs. Finally, we made estimates of potential
based options. In addition, an extensive techno-economic model
penetrations of the ICTs (reach), to estimate the cost-
for the various ICTs and stochastic evaluation of potential
penetrations of the ICTs (reach) has also found CR to be the most effectiveness per user of the technologies for the particular
cost-effective option. We conclude with a proposed set of policy development domain.
and operating recommendations to enable effective technology The focus here is largely on the delivery of social services
based information services for rural Bangladeshi development. (supply side); analysis of issues such as citizens' rights to
Index Terms— ICTD, Rural Information Service, Community information, citizen participation in governance etc. is beyond
Radio, Telecenter.
the scope of the paper.
The primary objective of this study is to develop a robust
I. INTRODUCTION
framework that can facilitate choices by policymakers,
E nsuring the availability of community level information

practitioners, development agencies and the general public and
promote better understanding regarding the applied
efficiencies and relevance of different technologies in multiple
services (IS) is an important priority for the general population
information service domains for societal development. Given
in South Asia. With the proliferation of Information and
the difficulties comparing technologies, expert elicitation with
Communication Technology (ICT) enabled communication
normalization is a useful starting point for the analysis. Formal
and broadcasting options, the numerous positive impacts that
(pre/post) type analyses for effectiveness across ICTs are
community based information services can contribute become
rarely available, given most studies consider only one
more evident [1]. Community-based information services
technology at a time for effectiveness analyses.
usually go beyond the traditional “One size fits all”
information access model used by the incumbent regulators
II. BACKGROUND OF COMMUNITY BASED INFORMATION
and policy makers [2]. These primarily focus on the SERVICES IN SOUTH ASIA
customized needs of the target communities. Of course, not all
ICTs are equivalent. Different technological solutions may Historically, governments in S. Asian countries have been
better meet the needs of different service sectors. For example, reluctant to open up the airwaves, lower import duties and
an option that works well for agriculture related information other taxes on information technology or liberalize regulatory
policies so as to create an enabling environment for micro
level information service infrastructures. On the other hand,
ICT for Development (ICTD) friendly stakeholders (i.e.
This work is supported by a grant from the MacArthur Foundation, and by
Carnegie Mellon University. business entities, NGOs, innovators, donors etc.) are often
F. Hussain is a faculty member at Carnegie Mellon University- Qatar challenged to come up with effective methods for developing
Campus in the Department of Information Systems (phone: +974-557-2179; and managing systems to deliver information services (IS) [3].
fax: +974-454-8410; e-mail: faheem@cmu.edu).
R. Tongia is a faculty member at Carnegie Mellon University, Pittsburgh, The reasons for this split might have to do with how
in the Program on Computation, Organizations, and Society (COS), in the information has been (centrally) controlled, combined with a
School of Computer Science (ISR), and the Dept. of Engineering & Public market-type approach of first focusing on those willing or able
Policy. He is also the Associate Director of the TechBridgeWorld project at
Carnegie Mellon, and a Senior Fellow at the Center for Study of Science, to pay.
Technology, and Policy (CSTEP), Bangalore, India (phone: +1-412-268- Community oriented IS have had some successes. In Nepal,
5619; fax: +1-412-268-2338; e-mail: tongia@cmu.edu). a pioneer in community media activism in S. Asia, the
253
foundations for regulatory reform in broadcasting were first rapidly (39% in 1998, 37% in 2000, 34% in 2002 and 23% in
laid as part of the liberalization of Nepal’s constitution in 1990 2004) [15]. At the same time, the penetration level of cellular
[4]. During this period, Nepal witnessed the development of phones and internet users is increasing. Presently, the
independent, community based radio stations (CR)1, teledensity of the country is 27%, with 38.93 million cellular
telecenters2 and newspapers. The potential synergy between phone users and 1.25 million PSTN (landline) users [17]. The
telecenters and CR stations has also been explored there to a Bangladesh Telecenter Network (BTN) has also undertaken
greater extent than elsewhere in South Asia (SA) [7]. Table 1 initiatives to create 40,000 telecenters nationwide by 2011 [2].
summarizes the present CR scenario in Nepal. Only recently, the Government of Bangladesh (GoB) has
Table 1 opened up the airwaves (on an experimental basis) for
Present Status of CR in Nepal [4, 8] community based low power FM radio broadcasting. Figure 1
On Air Stations 33 shows the status of literacy and access rates of Terrestrial TV,
New Licenses issued in last Telephone, Public AM and National Newspaper in several
6 month period 93 South Asian countries. Nepal lead in the access of Public AM
Districts under total CR coverage area broadcasting but has the lowest teledensity among the four S.
(out of 75 districts nationwide) 56 Asian countries considered. On the other hand, India has
Total % of population covered by considerably higher literacy rate but lower Terrestrial TV
CR broadcasting* 62.5% penetration rate than its neighbors.4
A series of regulatory liberalization and public-private
initiatives in developing localized information service III. INTERNATIONAL EXPERIENCE AND LITERATURE ON
mechanisms are also evident in India. In November 16, 2006, COMMUNITY-BASED INFORMATION SERVICE (IS)
the government finally opened up the community airwaves for Over the last few decades, amid the disparity in access to
non-profit organizations. It also formulated a comprehensive information between the global north and south, the concept of
“Community Radio Policy”, a first such step in South Asia alternative or community based media has emerged. The New
[9]. Throughout this Indian experience, especially noteworthy World Information and Communication Order (NWICO),
are the innovative ways people have met the huge demand of proposed by UNESCO in the late 70s and early 80s [18], had
localized information, even in the face of severe legal and the goal of beginning to balance this information divide [19].
regulatory constraints[10] [9, 11, 12]. Studies have shown, with the introduction of new
technologies, valuable information needed by any community
A. Bangladesh - Promises and Problems in Community
Information Service Sector becomes more accessible and affordable [20]. At the same
time, some experts have warned against over dependence on
The overall communication sector in Bangladesh presents
any particular technology and have placed greater importance
an interesting situation, where the potential of low as well as
on the very process of providing information service [21].
high end ICT-based information access points (IAP) [13] is
There are many technology options for IS. Community
immense, even in the absence of a regulatory and development
based radio broadcasting has been found to be effective for
framework. The national literacy rate is only 43.1% [14]3, and
strengthening democracy, development and people’s
only approximately 50% of those classified as literate can
participation in the affairs of developing regions. For example,
properly read and write[15]. A recent national media survey
in Africa, organizations like COMNESA in East and Southern
found that 28.5% [15] of the total Bangladeshi population are
Africa, AMARC Africa, KCOMMNET in Kenya etc. have
in the “Information Dark”, meaning, they have never listened
been established to share resources among a wider range of
to a radio program, watched a TV program, read a newspaper
population across the continent [22].
or had any similar means of getting valuable and current
A significant amount of research has also been done to
information needed for their livelihood and wellbeing [15]!
assess the effectiveness of different technological intervention
This situation is even worse in rural areas (35.9% of rural
in various information service domains, i.e., education,
people have no access to any information service) [16], where
disaster response, agriculture etc. In the many cases evaluated,
most of the people live. Because of the absence of localized
experts have found radio-aided services to be quite effective
programming and participation, the only public radio
for different education or training purposes.[23-25]. The
broadcaster, Bangladesh Betar, is losing its listener base
effectiveness and importance of visual broadcasting media has
been supported in other studies [26, 27]
1
Community Radio: Community radio (CR) is a subset of Radio While choosing an appropriate technology to enhance
broadcasting, which is an inexpensive and popular medium for disseminating
content (information, news, entertainment) that specifically focuses on the educational delivery or other societal services, issues like
active participation of its listeners and also tries to ensure access to access, costs, maintenance, interactivity, user-friendliness,
information by all. In addition to broadcast, there are mechanisms for many- availability and speed should be considered [26]. According to
to-many discussions and community feedback [5].
2
Telecenter: Telecenters are shared premises where the public can access UNESCO, users and experts also focus on the specific abilities
information and communication technologies. A center offering only of the considered technologies in terms of delivering
telephone, computer or internet services is valid under this definition [6].
3
According to a study completed by the Government of Bangladesh in
4
2003, the literacy rate was found to be 62.66%. But a CAMPE study in 2002 Data is more limited on cable TV, which has substantial penetration in
found it to be 41.4%. some regions, but content is rarely more local community oriented.
254
Figure 1
Literacy and Access Information of Different Communication
Medium in South Asian Countries
Figure 1 represents the relative status of the four South Asian countries in literacy, teledensity, newspaper readership and radio
penetration (data collected during 2004-2008 period).[16, 28-34]
appropriate messages in a given domain. Technological

A. Significance of this Research
efficiencies can vary in terms of outreach capability,
flexibility, interactivity etc.[35]. In addition to the traditional While much of the research described has focused (at least
applications of lower end ICT options like radio and TV in a qualitatively) on the effectiveness and impacts of various
disaster response scenario, recent studies also focus on the technologies in multiple information service domains, there is
importance of computer aided, Internet based and wireless a clear need for studies that examine the comparative
technologies [36]. effectiveness of a wide range of appropriate technologies
The recent trend of technology specific effectiveness studies across a set of information services in specific socio-economic
is concentrated mainly on higher end ICTs: Computers, Web- settings. The difficulties in conceptualizing the relative
based solutions, Telecenters etc. and, to some extent, mobiles. effectiveness of multiple types of interventions, taking into
As auxiliary tools for providing education, these ICT options account a very large set of application domains, is clearly
often show superior performance in comparison with the other daunting. The objective of the present study is to develop a
alternatives. But factors like high cost, lack of training, lack of robust comparative cost-effectiveness analysis framework,
motivation, maintenance issues, and limited infrastructure which will enable different concerned stakeholders (i.e.,
often hinder the highest level of efficiency attainable by these policymakers, practitioners, international and local
relatively new technological interventions [37] [38]. development agencies etc.) to have a better understanding of
Methodologically, these past studies, as well as this one, the applied efficiencies and relevance of different technologies
struggle with the issues of penetration (which are cost-related) in multiple information service domains. While the main focus
compared to effectiveness in and off itself. of the application is SA, and more specifically on Bangladesh,
E-access enabled information access points, predominantly this study can also be used as a guideline in other developing
in rural or economically deprived areas (popularly known as regions with similar needs and aspirations.
Telecenters, Rural Information Centers etc.) have developed a
strong following in developing regions around the world. IV. RESEARCH METHOD
Studies have shown the availability of local contents and The principal research question we are addressing through
affordable services has helped many telecenters to be effective this research is: What are the most effective technological
and sustainable [39, 40]. Case studies in South America and interventions for providing Information Services in the rural
Africa provide ample evidence of micro-level South Asian (Bangladeshi) setting? There is a wide diversity
democratization, as one of the positive externalities of of Information Service domains in SA and many of these are
effective telecenter based applications [41]. interrelated and in some case overlap each other, in terms of
target population, societal objectives and modes of
communication. In order to have a better understanding of key
255
social services, we limited the number of application domains Community Radio (low power FM); Development FM Radio;
considered in this research to seven (Agriculture: General Local Newspaper; National Newspaper; Printed Brochure;
Extension Service and Market Information; Education; Public AM Radio; Rural information center (RIC-1) equipped
Disaster Response; Health Care: AIDS and STD prevention with computers, Internet connectivity, and one or more of the
and Family Planning; and Financial Services). These domains following: scanners, printers, digital cameras, cellular phones;
were selected with respect to the needs and their importance in RIC equipped similarly as RIC-1 but without any web-based
the SA region [42]. A set of technologies was then chosen, facilities (RIC-2); Satellite Radio; Satellite/Cable TV;
mainly on the basis of their availability, potentials in the Terrestrial TV (privately operated for profit) and Terrestrial
"Communication for Development" field [43], success stories TV (government operated).
and positive impacts in the real world scenario. This work B. Expert Elicitation
considers broadcasting (both audio and visual), print and the In order to quantify the relative effectiveness of various
Internet/multimedia enable Information centers. Cellular information service technologies for delivering a variety of
phone based options were not considered individually, but the social services, we administered a survey to a group of
impact and effectiveness of cell phones have been partially international (i.e., Australian, Guatemalan, Fijian, Indonesian,
accounted for in the hybrid solutions of Information centers as American, Nepali, and Sri Lankan) and local (Bangladeshi)
well as in person-to-person approaches. experts. These groups have had many years of experience in
In order to develop a framework for a comparative cost- the field of communication and development. The objective of
effectiveness model, we focused on quantifying the relative this survey was to obtain judgments of the relative
"effectiveness" of each technology using the judgment of a effectiveness of different approaches for delivering
group of international as well as local experts in the field. It information. We explicitly requested the experts to ignore
took roughly six months of correspondence using regular mail, differences in cost. At the start, twenty international and thirty
Internet and phone conversation to produce an effectiveness local (Bangladeshi) experts were solicited to take part in this
dataset from five different continents. We then built a cost study. Ultimately, eight international and fourteen local
model for calculating costs and penetration information (using experts provided detailed responses. In this study, we defined
field based South Asian datasets) through extensive techno- “information” to be most useful when it is delivered in the
economic analysis. Finally, all the analyzed and calculated language of the community, addresses topics that are of
information have been combined in a common platform to interest and concern to the community and is framed in a way
compare the relative effectiveness of the considered that is clear and understandable to members of that
technologies in information service sectors of this analysis. community. We also defined "community" as a group of a few
A. Development Domains and Technologies hundred to several thousand people who live within a few tens
of kilometers of each other in a developing country and share
1) Information Service Sectors
The seven application domains we selected for the analysis the same language and culture.
and expert surveys cover different types of information, Experts were asked to assess the relative effectiveness of
generally required at the community level in Bangladesh: different information delivery technologies for different types
a. General Agricultural Extension of information services in two types of communities:
b. Agricultural Market Information . communities with relatively high (60% or higher) and low
c. Life long learning/Adult Education level (20% or lower) of adult literacy.
d. Disaster and Emergency Response. We focused on We defined adult literacy to mean a recipient’s ability to
flood warning, mitigation and rehabilitation; drought read at an intermediate level in his or her primary language.
mitigation and awareness and cyclone/ Tsunami For each application, we asked the experts to first identify the
related awareness and protection programs. communication option that they believed would be the most
e. AIDS and Sexually Transmitted Diseases (STD): effective for a community with relatively high adult literacy
awareness and prevention campaigns and assign it a score of 100. They were then asked to assign a
f. Family Planning: awareness related activities. number to each of the other technology options to indicate
g. Financial Services. This focuses on financial how relatively less effective the experts’ believe each
information services at the personal and community communication option would be for the specific information
level. delivery identified. The whole process was conducted for both
2) Selected Technological Interventions high and low literacy scenarios. We also received qualitative
In many cases the most effective means to deliver information information from the experts.
is through one-on-one communication with a respected, As most of the experts were not as familiar with some of the
trusted, patient and informed person in the recipient's language technologies, development FM, satellite radio and private
and in an appropriate cultural setting. However in most of the terrestrial TV were omitted from further analysis.
cases in the real world, this is not feasible. In addition to the C. Cost Estimations
face-to-face (F-to-F) option (sometimes called a peer-to-peer
mechanism, though that terminology has a separate technical Our objective in estimating the costs of various delivery
meaning in networking as well), technological interventions technologies is to provide comparable descriptions or
selected for the survey were (in alphabetical order): mechanisms by which to deliver services and estimate their
256
costs in a way that would allow comparisons. In general, succinct manner.

“Cost-per Person-Year” is estimated for each option, where 1) Community Radio (CR)
“cost” is the monetary amount spent by the provider to deliver For the CR cost estimation, we included both monthly
the information in a real world setting and “person” is the operational and yearly capital expenditure cost (amortized).
generic term that covers target listener, viewer, reader, user The lifetime of radio transmission equipment was assumed to
etc. (as appropriate for any particular technology/ medium) be 10 years. The listener base was calculated in terms of CR
within the general community population. Naturally, this penetration among typical South Asian communities in India
assumes an equivalence amongst all the persons. and Nepal (there is virtually no community radio in use in
For community based information services, while Bangladesh), which usually ranges from 60%-90%[30]. After
calculating Cost-per Person-Year, both operational and deriving the total yearly cost, we can calculate a set of “cost
amortized capital expenditure are considered. For the services per listener-year” values, depending upon the probable
with national coverage, we calculate a generalized cost-per population (density) of any community, the potential CR
person-year (delivery and content development cost) that cater listener base, transmission range etc. Our observation of
to numerous communities and issues, in addition to providing global CR practices shows that approximately two-thirds of
particular information services to any target community; as total broadcast time is dedicated to societal issues [7, 44]. Of
detailed later, some fixed costs were considered sunk costs for those blocks of time, we assume an equal distribution of air
such national footprints. We assume a 2, 5 and 10 year life time among the seven different information services. Thus, the
cycle for the audio/video program, RIC equipment and CR final “cost per listener-year” for any CR based service is
station equipment, respectively, and used 7.5% as the interest computed
rate for annualized capital expenditures (sensitivity analysis as
also considered 11% and 15% but we used 7.5% as a baseline,  Total Yearly Cost   Societal Program   Selected Service 
 × × 
a common rate for development/non-profit projects in South  Number of Listeners   Total Broadcast Time   Total No. of Services 
Asia. This matches the stochastic cost modeling of CR
management in [5]. We also take into account stations’ or  Total Yearly Cost   2   1 
other facilities’ monthly operational expenditure, which or,   ×   ×  
 Number of Listeners   3   7 
include staff salaries, utility bills and space rental (if
Table 4 shows the calculated cost per person-year and
applicable).
penetration levels for all the technology based information
For the technologies with a national presence (i.e., National
service in the target community.
Newspaper, Public AM, Terrestrial TV and Satellite TV), we
2) Public AM
considered the literacy rates as the proxy of other major
In South Asian nations most Public AM stations cover the
factors while determining probable access rates within any
whole country. This centralized and “one way” broadcasting
community. We assumed that higher literacy will result in
does not have the flexibility to air localized programs nor does
higher access to certain technologies, but the rate may not be
it have the means to incorporate grassroots’ participation.
linear. In order to realistically calculate the penetration of
Considering everything, our assumptions for “cost per
these technologies in the high and low literacy region, we used
listener-year” then focused only on specific program
the ratio of 60% and 20% respectively compared to the
production and transmission costs, which in turn can be
national literacy rate (43.1%) [14], multiplied by the
divided among the potential listener base nationwide.
corresponding technology’s potential national penetration
Empirical studies support the practice of airing “twice a week”
number and a technology-specific multiplier. The values of all
programs on any particular societal issue. The assumptions for
the multipliers considered were estimated based on the
any radio program’s broadcasting time and frequency in a
respective technology’s dependence on literacy for mass
Public AM station for a certain year are stated in Table 2.
proliferation. For example, in order to estimate the penetration
Table 2
of a National newspaper in a community with 20% or lower
Parameters for Public AM Programming
literacy, we multiply the national newspaper readership
Duration of a Program 30 min
number (26%) by the ratio (20%:43.1%) and a multiplier. As
the dissemination of newspapers is directly related to the Number of Program per week 2
literacy of a target community, we estimate the baseline value
Number of Programs (yearly) 96
of this technology’s multiplier to be 1. Given the wide
uncertainty in parameters (i.e. cost, penetration), we have used Duration of the total Program time 48 hours
Analytica® stochastic modeling software to estimate a In this study, we assumed that in order to distribute
distribution in the total yearly cost for each station, giving us community based information service, we need a series of
the ability to parametrically vary a wide variety of variables. audio programs and buy the airtime from Bangladesh Betar or
Below are selected details on the cost calculations for the the regulator to broadcast it, using the cost structure available
various technologies; full details on assumptions of capital and in Bangladesh.
operating expenditures per technology are available from the 3) Terrestrial TV (Public) and Satellite TV
authors. Given the wide variance in numbers per technology, Here the majority of the assumptions regarding program
it is beyond the scope of this paper to display these in a
257
frequency, “cost per person-year”, access to programming etc. We assumed the reader ceiling of 20% for the low literacy
are similar to those for Public AM. we used the cost of video region and 60% in a high literacy community.
production and the TV transmission pricing (both terrestrial 7) Printed Brochure (PB)
and satellite). For calculating access level, national terrestrial Designing printed brochures and pamphlets for any target
TV penetration rate [16] has been used alongside the literacy community audience has long been a basic strategy utilized by
ratio and a variable multiplier. local NGOs for distributing social service information.
4) Face-to-Face (F-to-F) Readership and target community estimations are similar to
Ideally, Face-to-Face service delivery should involve one- local newspapers. The development cost of educational
to-one contact between a facilitator and a person who is a materials has been estimated to be relatively higher than the
member of the target population. But in most cases, this is not rest of the information domains for PB [45].
feasible or cost effective. F-to-F mechanisms actually vary, 8) Rural Information Center-1 (RIC-1) and Rural
depending on the type of information being provided. The Information Center-2 (RIC-2 = no Internet)
development costs of content in all the services are either Rural Information Centers (RICs) house computers with
insignificant in comparison to the calculated delivery cost or various information resources and, possibly, Internet access.
external to the system. We assume, with the exception of the The cost estimation of RICs are based on data collected
Health Care sector, that the monthly salary of the field mainly from Bangladesh, Nepal and India. The RIC model
workers remains the same across different information considered in our research has the following basic facilities:
domains. The total cost has been estimated for a year long computers, Internet Connection (available in RIC-1 and absent
program. In General Extension and Market Information in RIC-2) and facilities for basic computer training and
sectors, the percentage of people nationally involved in Printer/Cellular Phone/Fax for commercial usage. The
agriculture has been used as the base percentage in estimating Pallitathya Kendra (PK) model, used by the NGO D.Net in
the range of agro-related population within the considered Bangladesh, has been emulated in designing RIC-1 and 2’s
community. We assume that a group of 50 people (varied working infrastructure. A PK type RIC usually has an
parametrically) meets with an agricultural extension worker on outreach service, where the RIC field workers will personally
a regular basis. In the Education sector, the national literacy go from house to house within any target community and
rate was considered for estimating community illiteracy. For address their information needs by connecting users through
Disaster Response and Financial Services, community based cell phone to the pre-selected expert panel or the locally
estimations were used for calculating the target population. developed database [46]. For the penetration estimation of
Each face-to-face group’s base population and other RICs, we assume that RIC usage will be greater in regions
estimations are similar to the ones used in agriculture sector. with high literacy, resulting in the need to develop multiple
Due to the sensitive nature of Health Care information RICs within those areas, all within walking distance of all
services, we assumed monthly one-on-one interaction between community members (2 or 3 km radius). For a low literacy
public health field workers and the target community region, we estimate less usage and the presence of a single
population. RIC for the whole community. Also, the proactive approach
5) National Newspaper taken by the RIC field workers certainly has some positive
We assume that a relevant entity procures ad space to impact, at least among the literate community population and
publish specific communication material, pertaining to one of beyond. We moreover assumed that the content development
the seven information service dimensions defined in this and distribution cost for any information service will be
research. The assumptions for any national newspaper’s social similar for any RIC.
ad campaign and publishing frequency are stated in Table 3.
Table 3 V. COMPARATIVE EFFECTIVENESS ANALYSIS
Assumptions for a Newspaper based Societal Ad-Campaign We need to develop a comprehensive understanding about
Space bought in the 1/16th of the full page the comparative efficacies of different information service
front page of any (Standard Newspaper Page delivery options. For the relative effectiveness data, both
national newspaper with 36” X 24“ dimension) international and local (Bangladeshi) experts’ opinions were
initially considered. According to both of the expert groups,
Duration of the 100, 300 and 350 days (vary CR turns out to be the most effective technology in both high
Ad (per year) parametrically) and low literacy scenarios, although there was much less
familiarity with CR, and thus higher variation, in the responses
Target Population Total Community Population
of the Bangladeshi experts. All the experts showed greater
X National Newspaper
confidence in Face-to-Face intervention than most of the
Readership (%) [16]
technologies in almost all the application domains. In addition,
6) Local Newspaper
some prominent dimensions (e.g., directionality, user
The frequency of ad-publishing is assumed to be similar to
cognitive requirements etc.) within the technological
that of the national newspaper with relatively cheaper ad-rates.
“effectiveness” concept have been qualitatively analyzed,
We also took into account high and low literacy scenarios for
based on anecdotal evidence, field data and research findings
calculating potential readership within the target community.
258
Table 4
Calculated Cost Estimations*
Delivery Option Information Cost per Person-Year (US$)* Penetration Level*

Service Sector
≤ 20% Literacy ≥ 60% Literacy ≤ 20% Literacy ≥ 60% Literacy
CR 0.011 0.011 68%
Public AM 0.0025 0.001 14% 36%
Terrestrial TV 0.0045 0.0015 23% 67%
Satellite TV For any of the 0.008 0.003 9% 26%
considered
National 0.004 0.0014 12% 34%
Service Sectors
Newspaper
Local 0.35 0.08 16% 70%
Newspaper
RIC-1 0.86 0.38 33% 75%
RIC-2 0.17 0.23 33% 75%
Agriculture: 0.35 0.23
General
Extension
Agriculture: 0.55 0.43
Market 16% 70%
Information
Printed Education 0.36 0.24
Brochure
Disaster
Response
Health Care: 0.35 0.23
AIDS and STD
Health Care:
Family
Planning
Financial
Services
Agriculture:
General 1.59
Extension Ideally, 100% of the target
Agriculture: population will covered from an
availability perspective; usage and
Market 1.59
uptake may be lower.5
Information
Face-to Face Education 1.67
Disaster 0.99
Response
Health Care:
AIDS and STD 2.45
Health Care:
Family 2.45
Planning
Financial 1.63
Services
* Parametrically analyzed, using SA data
5
This challenge is similar to radio or other media, where signal availability is necessarily only one measure, lower being actual listening time.
259
Table 5
Cross Technology Comparison for Multiple Effectiveness Dimensions
Technology Penetration** Directionality Cognitive Access Type

Intervention ≤ 20% ≥ 60% Level
Literacy Literacy
CR 68% Two-way, Low Single / Shared
Participatory
Public AM 15% 36% One-way Low Single / Shared
Terrestrial TV 23% 66% One-way Low Single / Shared
Satellite TV 9% 27% One-way Low Single / Shared

National 12% 34% One-way High Mostly Individual
Newspaper
Local Newspaper 16% 70% One-way High Mostly Individual
Printed Brochure 16% 70% One-way High Mostly Individual

RIC-1 33% 75% Two-way Medium* Shared
RIC-2 33% 75% Two-way Medium* Shared
* With the help of informediaries (information intermediaries)
**Technology Penetration estimated using stochastic modeling for high and low literacy region. There is always a challenge between
availability vs. uptake
([43] and [26]). CR has been found to be a highly accessible

technology that enables two-way participation based was not found to be statistically significant7. Unlike the views
communication in a single or shared environment (two-way of international experts8, local experts rank Terrestrial TV and
capability comes from word-of-mouth and combining radio Public AM broadcasting high in both types of literacy regions.
with other modes of communication, such as cell phones, as Print based options, generally were considered highly
seen in SA). ineffective, a contrast with the international findings.
As with CR, the international experts also considered Public According to the local experts, RIC-2 (the information center
AM to be an effective literacy-independent medium. The without any internet facility) is more effective than RIC-1 in
ordinal effectiveness rankings of technological interventions low literacy regions9, significantly dissimilar from the views
diverge after this point. Localized print based options (Local of international set of experts10.
Newspaper and Printed Brochure) show clear dominance over For the in-depth analysis of the relative effectiveness data
the visual broadcasting options (Terrestrial TV, Satellite TV) (both technology and application domain), we focused on the
in a higher literacy area. On the other hand, as expected, visual international experts’ judgment that captures the technology
media are considered more effective than print media in for development experiences in multiple developing regions.
communities with lower literacy. In both cases, Rural The primary reason for this decision is the Bangladeshi
Information Center with internet connection (RIC-1) has been experts’ absence of substantial field experience with the
ranked in the middle and demonstrated statistically effectiveness of a number of technology options (i.e. CR, RIC-
insignificant differences in its usefulness as a community 1, RIC-2, Public AM, Local Newspaper etc.) due to regulatory
intervention with respect to other technologies, a promising obstacles (CR); poor program management (Public AM); high
aspect for a relatively new information service option6. Internet costs, lack of human resources and poor funding
The Bangladeshi experts’ perception about the relative (RIC); and absence of press freedom and effective
effectiveness of some of the technological options is representation of local issues (local newspaper). The
significantly different from those of the international experts.
CR remains the top choice but in a high literacy area the 7
A paired T-Test between the effectiveness scores of Terrestrial TV and
difference in effectiveness between Terrestrial TV and CR CR by the local experts found the difference statistically small (with t-stat of
4.43 and p-value of 0.002 in a 20% literacy area) has been found.
8
Two-sample T-Tests were performed between the effectiveness scores of
6
Paired T-Tests were performed on the effectiveness scores of two groups of experts in high and low literacy region. Both times, we found
International experts in high (between Local Newspaper and RIC-1) and low statistically significant difference (with t-stat of -10.04 and p-value of 7.6E-07
(between Terrestrial TV and RIC-1) literacy region. Both times, the difference in 60% literacy and t-stat of -12.54 and p-value of 9.6E-08 in 20% literacy)
was statistically small (with t-stat of 0.1068 and p-value of 0.459 in 60% between the experts’ opinions.
9
literacy and t-stat of 3.277 and p-value of 0.0084 in 20% literacy) between the With a t-stat of -22.7 and p-value=2.39E-07 in a Paired T-Test.
10
experts’ perception about relative effectiveness. With t-stat of -8.36 and p-value=-7.92E05 in a Two Sample T-Test.
260
Figure 2
Effectiveness Ranking of Different Technologies
(According to International Experts)
Figure 2: Relative effectiveness ranking of the considered technologies in both high and low literacy region, according to the judgment of the
International experts. It considers average effectiveness values of different technologies across all service domains. Individually, the experts
have shown higher variance in effectiveness estimations (details in Section V)
Bangladeshi sample also has much greater variance than the (Agriculture: General Extension). In these figures, the Y and
international sample, especially for technologies not now in X axes represent the effectiveness scores and plausibly
use in Bangladesh [47]. achievable technology penetration respectively. The spread of
This scenario has been reflected in the effectiveness ranking, effectiveness scores given by individual experts for various
which fails to reflect the large potential of RIC with proper delivery systems (from 0 to 100 in the Y-axis) can be seen
web based opportunities. Table 6 summarizes the present together with the highlighted ones representing average
access information of different technologies, in a Bangladeshi values. The size of the average value data points correspond to
setting. the Cost-per Person-Year for each considered delivery
In figures 4, 5, 6 and 7, we see the graphical representations technology. According to our assumptions, the effectiveness
of comparative cost-effectiveness data, from both international score of both expert groups are independent of the potential
and local experts, for a sample application domain technology penetration or cost scenarios in Bangladesh.
Figure 3
Effectiveness Ranking of Different Technologies
(According to Local Experts)
Figure 3: Relative effectiveness ranking of the considered technologies in both high and low literacy region, according to the judgment of the
Bangladeshi experts. It considers average effectiveness values of different technologies across all service domains. Individually,
the experts have shown higher variance in effectiveness estimations (details in Section V).
261
Table 6 to be most effective technologies with which to develop an

Present Access Status of Different Communication Services in efficient disaster response infrastructure in both high and low
Bangladesh literacy communities.
Technology/ Penetration/Access Observation 5: In the healthcare sector, experts emphasize
Information Service Information [16] the higher effectiveness of proactive Face-to-Face
Option interventions by local healthcare workers. CR remains the
Radio 30.4% most effective information service option, in all literacy and
Terrestrial TV 61% service sector scenarios.
Satellite TV 12.4% Observation 6: According to the experts, CR alongside
Newspaper 26% localized print based options are more suitable in providing
Telecenters 1061 information related to financial services among the highly
(total number, literate population. For a low literacy community,
nationwide)[2] broadcasting options are considered to offer superior
information services.
We integrate all these different types of data in a single A. Technology-specific Key Observations
graphical display for each information service domain and in Examining which technology works best for which
two types of literacy based scenarios to: 1) have a information service provides the basis to further examine the
comprehensive look at the relative effectiveness of several effectiveness trends for different types of technologies, their
technologies as information service providers in the target potential impact in future proliferation of certain information
service sectors, and 2) to examine their present positioning in service, the regulatory and policy implications on certain
terms of plausible local or national penetration level. Service technology usage etc. Here technologies here been classified
Sector specific Preliminary Observations within broader groups in order to come to some general
In order to find out the relative effectiveness of different observations.
technologies to deliver a range of information services, we 1) Broadcasting based Information Service Options
first focus on the service sector specific performance in both Observation 1: Radio broadcasting dominates the
high and low literacy scenarios. technology based service delivery mechanism with its literacy
Observation 1: For Agriculture General Extension, CR independent potentials in localized and participatory
was ranked as the most effective technological intervention, communication.
independent of the literacy rate within any community. For Observation 2: CR leads in cost effectiveness in all service
high literacy regions, print based options dominate sectors through ensuring localization and community
broadcasting alternatives. RIC-1 is ranked third overall in participation.
effectiveness, followed by Local Newspaper, Public AM Discussion: CR has been found to be an effective mode of
radio, National Newspaper and others. For communities with communication, independent of service sector type and the
low literacy, Public AM turns out to be the second most state of any target community’s literacy. The apparent
effective way to deliver general extension information among universal efficacy of CR can be attributed to its focus on
the less literate population, followed by RIC-1, Terrestrial TV, localized programming, participatory communication,
RIC-2, Local Newspaper etc. appropriate content development etc[5].
Observation 2: For Agriculture Market Information, CR Observation 3: Broadcasting based information service
remains the top choice. Here in high literacy area, Public AM options are the key behind any effective Disaster Response
radio is considered to be more effective, followed by mechanism.
technological options with higher localization options (RIC-1, Discussion: The effectiveness of broadcasting based
Local Newspaper, Printed Brochure). In low literacy areas, disaster response mechanisms is unparalleled, primarily due to
broadcasting options (Public AM, Terrestrial TV and Satellite their nearly universal acceptability, point-to-multipoint
TV) dominate the effectiveness ranking, followed by RICs and (broadcast) capabilities, and speed in disseminating the
print based options. required information.
Observation 3: In Education (Lifelong Learning), apart Observation 4: Public funded Terrestrial TV has higher
from CR, in higher literacy areas, with a wide range of options effectiveness and acceptability in providing information in
available using the Internet and multimedia, RIC-1 is comparison with privately owned Satellite/Cable TV channels.
considered to be the second most effective delivery Discussion: In this research, we assume that higher literacy
technology followed by Public AM, Local Newspaper, Printed closely correlates with higher economic affluence, an indicator
Brochure, RIC-2 and the other broadcasting alternatives. For for people to opt for “better” communication technologies. In
low literacy regions, broadcasting options overwhelmingly Bangladesh, Satellite/Cable TV is less popular (and has lower
lead the relative effectiveness ranking. penetration) than Terrestrial TV[16]. It shows the positive
Observation 4: When a country like Bangladesh is under externality a nation’s broadcasting regulations can create
threat of imminent calamities [48], radio broadcasting (CR, through mandating the incumbent public broadcasting service
Public AM) and TV (Terrestrial and Satellite) are considered air social service information programs. Generally, private
262
Figure 4
Cross Technology Comparison for providing Agriculture (General Extension)
based Information Service in a high literacy ( ≥ 60%) area (International Experts)
E 100
Face-to-Face
f CR
f
e 80
c Printed
Public AM Brochure
t RIC-1
i
v 60
e Local
n Newspaper
Satellite TV RIC-2
e
40
s National
s Newspaper
Terrestrial
20 TV
0
0% 20% 40% 60% 80% 100%
Likely Achievable Technology Penetration
in Bangladesh
Figure 4 represents the relative effectiveness (judgment of the international experts), penetration and cost-per person-year status of each
considered technologies in providing Agriculture (General Extension) related information service in a high literacy ( ≥ 60%) community. The
X and Y axes correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness
scores for various Information Service interventions (from 0 to 100 in the Y-axis) can be seen (with multiple dots) alongside the highlighted
ones representing average values. The size of the average value data points correspond to the Cost-per Person-Year for each considered
technologies. CR is the most cost-effective technology based option here, with tradeoffs of penetration, cost, and effectiveness between the
next choices of face-to-face, Rural Information Center with Internet connection, and local Print media.
3) Rural Information Centers
satellite channels have fewer regulatory obligations to abide Observation 7: Literacy doesn’t affect the relative
by and little or no financial incentives in airing societal effectiveness of RIC-1 over Print and TV based options.
programs, let alone localized versions. Satellite also has Discussion: According to the experts in this research, on
greater challenges of localization than many other average in both high and low literacy scenarios, the RIC-1
technologies based on its very wide footprint. centers are (or, perhaps, can be?) more effective than some
2) Print Based Information Service Options major TV broadcasting and locally published print based
Observation 5: Local Print based options are more effective options. Especially with the help of its information
than the national ones. intermediary (“ informadiary”) mechanism, local database and
Discussion: The general population can be effectively cellular phone integrated applications, RIC-1s have a lot of
reached with locally published newspapers or printed growth potential as a relatively new technological option[49].
brochures in comparison with the big newspapers with Observation 8: RICs without Internet can still be utilized as
nationwide circulations. Community ownership, localization a successful tool for providing community education.
of the content development and easy accessibility are reported Discussion: As we have found in this research, RICs
as some of the key factors responsible for this trend. without web enabled facilities (RIC-2) are less effective than
Observation 6: Print based options can be used as effective RIC-1s as well as the majority of the other interventions in
auxiliary mechanism for societal services that require both high and low literacy communities.11 But in the education
personalized attention. domain, RIC-2 shows higher effectiveness than mainly
Discussion: In lifelong learning, financial service advising, National Newspaper and Satellite TV. Based on discussion
health care and agriculture-based information services in any with the development experts in S. Asia, we found that the
high literacy region, community based print options are more relatively lower demand of updates in lifelong learning
effective than the visual broadcasting media, according to the curriculum is considered to be one of the main reasons behind
effectiveness ranking and comments of the international RIC-2’s better performance in this specific service sector.
experts. Usually NGOs and development agencies in the field
use this type of intervention as a supporting tool in addition to 11
An RIC-1 could always operate without Internet (RIC-2) in theory, but
their broader door-to-door personalized campaigns. in practice, this is rarely operationalized.
263
Figure 5
based Information Service in a low literacy (≤20%) area (International Experts)
100
CR Face-to-Face
E
f
80
f
e Public AM
c
t Terrestrial RIC-1
60
i TV
v
e
n 40 Satellite TV Local
e Newspaper
s
s RIC-2
20
Printed
National Brochure
Newspaper
0
0% 20% 40% 60% 80% 100%
in Bangladesh
Figure 5 represents the relative effectiveness (judgment of the international experts), penetration and cost-per person-year status of each
considered technologies in providing Agriculture (General Extension) related information service in a low literacy community. The X and Y
axes correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness scores for
various Information Service interventions (from 0 to 100 in the Y-axis) can be seen (with multiple dots) alongside the highlighted ones
representing average values. The size of the average value data points correspond to the Cost-per Person-Year for each considered
technologies. CR is the most cost-effective technology based option here, with tradeoffs in cost, penetration, and effectiveness in the next
choices of face-to-face, Public AM and Rural Information Center with Internet connection.
among the general population, irrespective of the literacy

VI. RECOMMENDATIONS FOR EFFECTIVE INFORMATION and economic status. Airing a greater number of
SERVICES development oriented programs, opening up education
While this is a comparative analysis, in all technologies, channels, ensuring participation of the local population in
availability of local, relevant, right-language content is a key regional TV stations are some of the ways to develop an
challenge, one for which the costs can and do vary effective TV based information service network.
enormously. Based on the relative effectiveness analysis • Especially for a disaster prone country like Bangladesh,
through expert elicitation, stochastic cost modeling, literature decentralized emergency warning systems and response
review, personal discussions with practitioners and anecdotal mechanisms need to be established. In addition to the
evidences, we make the following recommendations for traditional satellite based initiatives, community based
effective technology centric information services deployment. micro-broadcasting systems with local participation need
• Direct human involvement should be an integral part of to be developed for a sustainable disaster response
any "Communication for Development" initiative. infrastructure beyond the initial warning. In recent times,
Experts' overwhelming preference for "Face-to-Face" the latency in getting the weather updates from a single
interventions and research findings related to the point has resulted in the loss of thousands of lives in
importance of human interaction justifies its importance Bangladesh.
at least as an auxiliary medium for any technology based • Ideally dominant in higher literacy areas, Print media
information service options. remains the backbone of education systems, one of the
• As the majority of the Bangladeshi population lacks most important service sectors considered in our analysis.
access to effective information services, participatory The success of Print based options as a part of any
radio based broadcasting options can play important roles integrated delivery mechanism [43] makes it an important
as a socially sustainable development mechanism. element for any viable communication for development
• Over-the-air TV broadcasting has wider appeal, both as a option.
source of entertainment and societal communication
264
Figure 6
based Information Service in a high literacy ( ≥ 60%) area (Local Experts)
100 Face-to-Face
E
f
80
f CR
Terrestrial
e Public AM TV
c
t 60
i RIC-2
Satellite TV
v Local
e Newspaper
n 40 RIC-1
e
s National Printed
s Newspaper Brochure
20
0
0% 20% 40% 60% 80% 100%
in Bangladesh
Figure 6 represents the relative effectiveness (judgment of the local experts), penetration and cost-per person-year status of each considered
technologies in providing Agriculture (General Extension) related information service in a high literacy community. The X and Y axes
correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness scores for
technologies. Face-to-Face is considered to be most effective. Among the technology based options, CR is the most cost-effective one,
followed by Terrestrial TV and Public AM
.
• With literacy independent access efficiency, RIC-1 can development, management etc. [11, 52] Regular in-
provide a platform for sustainable Information (this service training for personnel and resource pooling in
implicitly assumes the use of informadiaries). The PK remote places (i.e. sharing of content, technicians etc.)
inspired proactive approach reduces the "barrier to entry" also can ensure the operational sustainability of any
factor for local information service facility [46]. information service center. This is vital for long-term
Following the successful example of Nepal [50], a sustainability beyond the initial period.
balanced integration of Face-to-Face, CR, Print and
A. Additional Policy Suggestions
Internet based multimedia activities can address the
majority of the information service needs in South Asian These draw from literature, our analysis and
region. suggestions from the surveys:
• Efforts to demystify the technologies used for information • Effective government policies and regulations need to be
service centers need to be in place to ensure greater in place to create an enabling environment for
efficacy and wider participation of the common people. development-oriented information services. The policies
For example, low cost CR options such as the "suitcase should be effective in removing bottlenecks for the free
radio" can be used in the field for on- the spot flow of information by giving away incentives like
broadcasting and human resource development purposes. lowered or zero tax rates for imported equipments,
lowered license fees etc.
• Both the literature review and collected field data [11, 51]
identified "scarcity of trained human resources" as one of • Information service providers engaged in societal and
the major roadblocks towards establishing an effective development work need separation from competing with
information service network. Experiences from South commercial entities. The absence of such support
America and from Nepal showed the effectiveness of mechanisms can result in unfair competition, pressure on
decentralized support and resource centers for training service providers to overlook societal commitment for
information service providers on technical issues, content
265
Figure 7
based Information Service in a low literacy (≤20%) area (Local Experts)
Face-to-Face
100
E
f Terrestrial CR
80 TV
f
e Public AM
c
t 60
i
v
e Printed RIC-2
n 40 Satellite TV Brochure
e
s
s RIC-1
20
Local
National
Newspaper Newspaper
0
0% 20% 40% 60% 80% 100%
in Bangladesh
Figure 7 represents the relative effectiveness (judgment of the local experts), penetration and cost-per person-year status of each considered
technologies in providing Agriculture (General Extension) related information service in a low literacy community. The X and Y axes
correspond to the likely achievable technology penetration and the effectiveness scores respectively. The spread of effectiveness scores for
technologies. Face-to-Face is considered to be most effective. Among the technology based options, CR is the most cost-effective one,
followed by Terrestrial TV and Public AM. Rural Information Center without the Internet connection has been considered to be more effective
than the one with the web-enabled options.
greater commercial success and, at times, the abrupt extends to applications such as voice enabled and GSM
closure of such initiatives[4]. based help-lines (for providing healthcare, market
• A “Development Information Service Fund” can be information etc.), more hybrid solutions or innovations
created to provide monetary help for establishing such as Internet Radio.
information infrastructure in underserved communities. • Hybrid solutions between technologies, including face-to-
All the commercial broadcasting and communication face and mobiles, might offer the most promise. This
enterprises can contribute to this fund on a mandatory implies policies and regulations should not be spoiled by
basis. This type of mechanism is the backbone for parallel technology, but, rather, facilitates such convergence.
policies worldwide on Universal Service for telephony
[53]. VII. CONCLUSION
• On modeling and comparing program and labor costs of This analysis provides a comparative framework and an
community based information service providers in initial estimate of the cost-effectiveness across technologies
isolated versus more collaborative environments, we find for delivery of information services for rural users in
found that national, regional and local cooperation in Bangladesh (and elsewhere). Effectiveness analysis using
terms of program and resource sharing should be experts’ judgment and financial data modeling in this paper
encouraged officially, in addition to pooling resources for have demonstrated the strength of Radio based information
learning, training, troubleshooting etc. (given labor is a service options as useful and applied tools for development
significant fraction of operational costs [5]. activities in Bangladesh and S. Asia. Rural Information
• With the trend of transferring everything to be digitized Centers with net connections also turn out to be very
and being transferred to cellular phone media if not the promising as hybrid information access points (under the
Internet, the policy framework should support digital assumption of informediaries for extending their reach). We
information service initiatives, content sharing and mobile find awareness is a challenge in choosing technologies, given
wireless based information services in near future. This the differences between international and local (Bangladeshi
266
experts) – this implies dissemination of data and best practices [22] A. Opoku-Menash, "Radio, Conflict and Political Transition: The
Future of Community Radion in Africa," in Communication for
to be of value. Given the enormous differences in cost, Social Change, Anthology" Hostorical and Contemporary
effectiveness and penetration, often with a tradeoff (there is Readinngs, A. Gumucio-Dagron and T. Tuft, Eds.: CFSC, 2000,
rarely a dominant solution), policy-makers will need to decide pp. 856-963.
[23] A. Dock and J. Helwig, "Interactive Radio Instruction: Impact,
how to prioritize technologies in the form of enabling
Sustainability and Future Directions," World Bank, 1999.
environment. Regulatory and policy decisions can strongly [24] I. Pringle and M. David, "The Kothmale Model: Using Radio to
impact the relative costs of technologies as well as their Make the Internet visible," in The One to Watch: Radio, New ICT
penetration. Nonetheless, the broad portfolio of technologies and Interactivity, B. Girard, Ed. Rome: FAO, 2003, pp. 97-114.
[25] D. R. Holmes, D. M. Kamacharya, and J. K. Mayo, "Radio
available for Information Services for development implies Education in Nepal," in Distance Education for Teacher Training,
great potential for the region, with relatively low costs per H. Perraton, Ed. New York: Routledge, 1993, pp. 136-195.
beneficiary for all the technologies (under $1 per person-year [26] A. W. Bates, Technology, Open Learning and Distance Education.
London: Routledge, 1995.
in scale). [27] T. Hulsmann, "Costing Open and Distance Learning," 2000.
[28] "India's foremost television network - Doordarshan." vol. 2008.
[29] "CIA - The World Factbook -- India." vol. 2008, 2008.
[30] "Future Tense of Dailies ". vol. 2008, 2001.
REFERENCES [31] "CIA - The World Factbook -- Nepal," 2008.
[32] "CIA - The World Factbook -- Pakistan," 2008.
[33] A. Riaz, "Media and Judicial Independence in Pakistan," 2004.
[1] R. Rajora, Bridging the Digital Divide: Gyandoot—The Model for [34] "Teledensity growth in Pakistan almost double in FY 2006." vol.
Community Networks. New Delhi: Tata McGraw-Hill, 2002. 2008, 2006.
[2] "Bangladesh Telecentre Network." vol. 2008. [35] L. Tobing, How To Do Community Radio: A Primer for
[3] "Case Study Series on ICT-Enabled Development:I-NetworkCase Community Radio Operators. New Delhi: UNESCO.
Study Series on ICT-Enabled Development:I-Network." vol. 2008: [36] R. Iannella and K. Henricksen, "Managing Information in the
bridges.org, 2004. Disaster Coordination Centre: Lessons and Opportunities," in 4th
[4] I. Pringle and B. Subba, "Ten Years on: The State of Community International ISCRAM Conference (B. Van de Walle, P. Burghardt
Radio in Nepal," 2007. and C. Nieuwenhuis, eds.), Delft, the Netherlands, 2007.
[5] F. Hussain and R. Tongia, "Community Radio for Development in [37] M. Carnoy, "ICT in Education: Possibilities and Challenges,"
South Asia: A Sustainability Study," in IEEE/ACM International 2004.
Conference on Information and Communication Technologies and [38] L. S. Austin and K. Husted, "Cost-Effectiveness of Television,
Development Bangalore, India: IEEE, 2007. Radio, and Print Media Programs for Public Mental Health
[6] F. J. Proenza, "Telecenter Sustainability- Myths and Education ".
Opportunities," in Bridging the rural knowledge gap: Information [39] R. Kumar and M. Best, "Social Impact and Diffusion of Telecenter
systems for improved livelihoods: FAO, 2001. Use: A Study from the Sustainable Access in Rural India Project,"
[7] "Radio Sagarmatha.." vol. 2008, 2007. ITID.
[8] R. Mainali, "Volunteering and Community Participation in [40] F. J. Proenza, "Telecenter Sustainability- Myths and
Community Radio: Experiences from Nepal.," National Opportunities," in Bridging the rural knowledge gap: Information
Consultation for Practicing and Potential Community Radio systems for improved livelihoods: FAO, 2001.
Operators 2007. [41] S. Finquelievich and G. Kisilevsky, "Community Democratization
[9] "India: Time for Community Radio has come!," UNESCO, 2007. Of Telecommunications Community Cooperatives In Argentina:
[10] F. Noronha, "India still waits: Rural poor not yet ready for the The Case Of Telpin," The Journal of Community Informatics,
promise of radio." vol. 2006, 2004. 2005.
[11] C. Fraser and S. R. Estrada, Community Radio Handbook: [42] "South Asia Economic Report – Social Sectors in Transition,"
UNESCO, 2001. 2007.
[12] J. Antin, "Cultural Assessment for Sustainable Kiosk Projects," in [43] G. M. Faheem Hussain, Rahul Tongia, "Technological
ICTD 2006 UC Berkeley, California, USA: IEEE, 2006. Interventions for BRAC's In-Service Teacher Training Program: A
[13] F. Hussain, "Community Broadcasting in South Asia: Hybrid Comparative Analysis," International Journal of Education and
Solutions for localized Information Access Points (A Bangladeshi Development (submitting for publication), 2008.
Case Study)," in 26th Scientific Conference on the International [44] "Radio Lumbini 96.8 MHz." vol. 2008, 2008.
Association for Media and Communication Research: Media and [45] B. Haiplik, "The BRAC Education Program Teacher Training and
Global Divides (IAMCR 2008) University of Stockholm, Development System Quality Improvement Initiative: Report #2,
Stockholm, Sweden, 2008. Monthly Refresher Training for BRAC Teachers," 2004.
[14] "CIA- The World Factbook -- Bangladesh." vol. 2008, 2008. [46] "Pallitathya Kendra." vol. 2008, 2008.
[15] "Media in Development: Linkages Between Poverty Reduction, [47] S. Rahman, "Internet in Bangladesh," Bangla ICT 2008.
Social Development and Diversified Media in Bangladesh," [48] "The Most Deadly 100 Natural Disasters of the 20th Century." vol.
Bangladesh Enterprize Institute, Dhaka 2006. 2008.
[16] "National Media Survey, 2002," 2002. [49] T. Köhler, M. Reitmaier, and G. J. Schulz, "Telecenters as
[17] "Bangladesh Telecommunication Regulatory Commission." vol. Instrument for Bridging the Digital Divide in Rural Areas," TU
2008, 2008. Dresden.
[18] "Many Voices, One World," Essex 1980. [50] "Community Radio Madanpokhara (CRM) Palpa." vol. 2008,
[19] C. Rodriguez, "From Alternative Media to Citizen's Media," in 2008.
Communication for Social Change, Anthology" Hostorical and [51] "Community Broadcasting Association of Australia ". vol. 2008,
Contemporary Readinngs, A. Gumucio-Dagron and T. Tufte, Eds.: 2008.
CFSC, 2001, pp. 778-786. [52] "Nepal Forum of Environmental Journalists." vol. 2008, 2008.
[20] A. E. Opubar, "If Community Media is the Answer, What is the [53] "Universal Service." vol. 2008: Federal Communications
Question?," in Communication for Social Change, Anthology" Commission, 2008.
Hostorical and Contemporary Readinngs, A. Gumucio-Dagron
and T. Tufte, Eds.: CFSC, 1999, pp. 778-786.
[21] A. Gumicio-Dagron, "Take Five: A Handful of Essentials for ICTs
in Development," in Communication for Social Change,
Anthology" Hostorical and Contemporary Readinngs, A.
Gumucio-Dagron and T. Tuft, Eds.: CFSC, 2003, pp. 976-988.
267
Decentralization, Clientelism and Popular

Participation—Is there a role for ICTs to
improve local governance?
Björn-Sören Gigler, London School of Economics
 top-down policy approach to a decentralized, participatory,

Abstract—This case study investigates under which conditions locally empowering approach [5].
ICTs can play a role in fostering the empowerment of rural The LPP transferred authorities and resources from the
communities to fully participate in the decision-making processes central state to local municipalities in order to make
of local governments. The analysis using empirical evidence from government more efficient and to improve the implementation
rural communities in Bolivia focuses hereby on the following key
of its policies for poverty alleviation1. At the core of the
questions: i) to what extent can ICTs contribute to improving the
efficiency and efficacy of local government? ii) does ICTs have
decentralization program was (i) the delegation of new
the potential to make local governments more transparent and responsibility to municipal governments and Territorial
responsive to the needs of rural communities and iii) can ICTs Grassroots Organizations (TGOs) with an allocation of 20% of
support the core objectives of the Bolivian Law of Popular national tax revenue upon approval of their Annual Operating
Participation to strengthen the role of local government in Plans (AOPs) for the social sectors—mainly health and
public-policy making and the implementation of development education and (ii) the establishment of mechanisms for social
programs? The article will explore these issues using several case control, called Vigilance Committees (VCs), responsible for
studies of rural municipalities in Bolivia that have participated in overseeing the expenditure of funds [6]-[7]. Critics of the law
the USAID-funded project Enlared Municipal. It concludes that
point out that its implementation faced severe obstacles since
the most important factors influencing whether ICT programs
can support the processes of decentralization, improved social
powerful regional elites continued to dominate local municipal
accountability and thus improve the well-being of rural planning processes and captured the resources provided by the
communities depends primarily on political, social and cultural central government [8]. Furthermore, the results from the
factors, whereby economic and technical factors are secondary. bottom-up participatory process had to fit with national and
departmental plans, and respond to the spending priorities of
the national government [9]. It is however important to
I. INTRODUCTION acknowledge both the LLP‘s shortcomings and its successes—
O ver the past few decades, most developing countries have in particular that, with the LLP, the Bolivian government
embarked on decentralization programs with the finally recognized traditional indigenous forms of organization
ambitious aims of: i) increasing government‘s efficiency and and transferred important resources to local governments. It
efficacy; ii) enhancing the quality of service delivery in such was, in short, a formidable attempt to shift power relations
areas as health and education; and iii) to enhance the within Bolivia‘s mestizo-dominated society, so that indigenous
responsiveness and poverty-focused of governments [1]-[3]. peoples and other marginalized groups would have the space
Most of these programs operate under the assumption that to voice their concerns, interests, and needs.
decentralized governments achieve more efficient results and This case study investigates the role ICTs can play in
are less corrupt since they are closer to the people, better able supporting such a decentralization process and in helping to
to understand the local dynamics of economic and social improve the well-being of indigenous peoples. In particular,
development and are thus more accountable for their programs the case study addresses the following questions: (i) do ICTs
to rural communities [4]. These programs are usually contribute to improving the efficiency and efficacy of local
embedded into broader policy reforms of the public sector and governments, thus strengthening their institutional capacity to
go frequently hand in hand with major privatization programs. provide basic services to local communities?; (ii) can ICTs
In Bolivia, the 1994 Law of Popular Participation (LPP) enhance the transparency of local governments and make them
signaled a dramatic shift in national policy from a centralized more responsible and accountable to the needs of local
communities?; (iii) was Enlared Municipal, the centralized
government-led program, able to achieve its main objectives,
and which constraints and limitations did it face in the
implementation of the program?; (iv) in which dimensions
Manuscript received September, 22, 2008. Bjӧrn-Sӧren Gigler is a PhD
Student at the Development Studies Institute (DESTIN), London School of
1
Economics and Political Science, Houghton Street, London WC2A 2AE, UK Bolivian Government in cooperation with the IMF and the World Bank,
(e-mail: b.gigler@lse.ac.uk). ―Economic Policy Framework Paper for 1997-99‖, June 1997
268
(i.e., political, organizational, and social) do ICTs have the of local governments [17]-[20]. Most of these authors have an
biggest influence on local governments and to what extent do optimistic view on ICTs and their role in development and
they enhance the human and social capabilities of the thus recommend that governments proactively embrace e-
program‘s participants?; and (v) did the project lead to any government applications in order reap the benefits from the
institutional changes in local governments, thus altering the use of ICTs within the public sector.
relationship between local government officials and More critical views of e-government highlight that in praxis
indigenous communities? the majority of e-government programs in developing
The case study investigates these central issues, based on countries have failed [21]-[23]. In fact, Heeks [24] ascertains
the empirical evidence from the government-led ―Enlared that the empirical evidence shows that 35% of e-government
Municipal‖ project—the principal ICT program in support of programs are total failures (the program was not implemented
local governments in Bolivia, coordinated by the Federation of or was immediately abandoned after its development), and
Municipal Associations of Bolivia (FAM). 50% are partial failures (major goals were not attained and/or
The overall objectives of the program were (i) to improve there were undesirable outcomes). He introduces the ‘reality-
the transparency and accountability of local governments; (ii) design‘ gap framework to analyze the underlying causes for
to enhance local governance structures and to improve the the numerous cases of failures and shows that the major
relationship between local government and indigenous reason for these failures is the common gap between the
communities; and (iii) to strengthen the institutional capacity current realities within public administrations and the design
and thus the performance of local governments in particular in of the e-government project. Heeks concludes that the larger
relations to the delivery of social services. Thus, the program‘s the design-reality gap is in the following seven dimensions
main focus was the political dimension, as its principal aim (information, technology, process, objectives and values,
was to improve local democratic processes and to strengthen staffing and skills, management systems and structures and
the popular participation of indigenous peoples in local other resources) the greater the risk for-government programs
governments. Nevertheless, the program also included in its to fail [25].
overall goals institutional and social aspects of local Madon demonstrates based on an interpretive set of case
development as evidenced through its third general objective. studies in Kerala India, that e-government programs have
Based on these objectives the project focused on the frequently neglected such critical factors as adequate
following specific activities: (i) implementing an ICT resources, skill-levels, values, beliefs and motivations and thus
capacity-building program for local government officials; (ii) poor people and communities were unable to derive real
developing and promoting the use of a national Portal on benefits out of such investments [26]. The author proposes a
issues related to municipal development www.Enlared.org.bo framework for evaluation the impact of e-governance projects
among local, regional, and national government officials; (iii) based on Sen‘s capability approach in order to be able to
assisting municipalities to develop and implementation their assess the developmental impact of these programs on
own municipal websites at the local level and (iv) helping peoples‘ quality of live. Madon concludes that a critical
formulate a national ICT strategy within the framework of the barrier for the successful implementation of e-government
ongoing UNDP-sponsored ETIC-initiative2. programs is that far-reaching back-end administrative reforms
need to be implemented in parallel for these programs to have
II. THE RELATIONSHIP BETWEEN E-GOVERNMENT, GOOD any significant positive effects on local governance [27].
GOVERNANCE AND DECENTRALIZATION The literature on the e-government and decentralization has
E-government is defined in the literature “as the use of stressed that the principal value of ICTs lays in both in
ICTs to improve the activities of public sector organizations‖ enhancing the participatory (good governance) and the
[10]. More specifically, the main objectives of e-government managerial (improved performance) dimensions of local
are (i) to make government more accessible, effective and government [28]-[30]. Lawrence Pratchett emphasizes that
accountable; (ii) improve the relationship between government ICT projects have the potential to contribute to support local
and citizens; (iii) strengthen the coordination and cooperation governments in the following three dimensions of their work:
within the public administration; and (iv) enhance government (i) enhance local democracy; (ii) promote public policy-
performance in the delivery of public services [11]-12]. making; and (iii) improve the quality of their service delivery
The proponents of e-government frequently argue that the [31]. Most scholars however highlight that local governments
application of ICTs in the public sector can play a catalytic use ICTs primarily as instruments to improve the delivery of
role for greater democracy, improved government services and neglect its uses to enhance participatory
performance and socio-economic development more generally mechanism conducive for good governance [32-35]. In fact,
[13]-[16] particular, the literature frequently points to the an empirical study of the ICT use of 270 local governments in
potential of e-government programs to promote good California highlights that most municipalities do not see ICTs
governance by enhancing the performance and accountability as an instrument for promoting good governance at the local
level, but as an additional communications tool—failing to
2
The Estrategia Nacional de las Tecnologías de Información y explore their value as change agents and their potential to alter
Comunicación (ETIC) is a UNDP-financed initiative which developed the fundamental relationship between government and citizen
through a participatory process including all sectors of the Bolivian society a
National Strategy for the Information Society for Bolivia (www.etic.org.bo) [36].
269
This case study draws on previous e-government studies by

Ciborra (2005); Ciborra and Navarra (2005); Madon (1993, ―Frequently the needs of local communities
2004), Walsham (1993) and Heeks [37]-[42] in its approach to are not considered, the projects are in fact
apply Pettigrew‘s (1985; 1998) methodology of a contextual designed on the desks of the Ministries or of
approach which emphasizes the importance of the economic, the international donors. A consequence of
social, political, and cultural context into which the this approach is that many projects are not
technologies are introduced [43]-[44]. These factors thus will sustainable at the local level and finish once
be essential for assessing the effects Enlared program had on the financing dries up. This program must be
local governments and the well-being of indigenous peoples. one of those unsustainable projects—at least
for my municipality there are currently no
III. THE ENLARED MUNICIPAL PROGRAM concrete benefits visible from this project‖.3
The Enlared Municipal project was implemented in two
distinct phases. The first phase of the project, from March The findings of the fieldwork showed that the first phase of
2001 to June 2004, was implemented by the International the project did not have any lasting development impact on
City/County Management Association (ICMA) and financed indigenous communities, so the research is focused instead on
by the U.S. Agency for International Development (USAID). the second phase of the project. Based on its experience with
ICMA is a U.S.-based international professional organization the first phase of the project, USAID decided to refocus the
for managers, administrators, and assistants in cities, towns, program and strengthened its partnership with Bolivia‘s
and counties and provides technical and managerial assistance, national association of local governments—the Federación de
training, and information resources to its members and the Asociaciones Municipales de Bolivia (FAM)4. During this
local government community (ICMA website). phase—from the beginning of September 2004 to the end of
During its first phase, the project developed an online November 2005—the FAM assumed responsibility for project
central Portal for municipal government coordination. The Federation organized a competitive bidding
(www.Enlared.org.bo), attempted to introduce a national process in mid-2004, which was won by the Swiss-funded
bidding platform for Bolivia, and provided technical support program Programa de Apoyo a la Democracia Municipal
to the Mancomunidad of the Chiquitania for the development (PADEM)5. The program focuses on increasing the capacity,
of an integrated Management Information System for the accessibility, and accountability of municipal governments in
fourteen municipalities belonging to the Mancumidad of the order to strengthen the participation of community-based
Chiquitania [45]. organizations and citizens in local governance. This program
However, the program‘s impact on local governments was was selected, in part, because of its seven-year experience in
limited by its ambitious and poorly focused objectives, its working with local governments on issues related to good
concentrated, top-down organization, and the perception governance, participation, and empowerment6.
among participants that the ICMA was a foreign and distant The new Enlared project team aimed to make the program
organization unconnected with the particular realities of the more responsive to specific local needs and integrated ICTs
Bolivian municipalities. During field visits to the Chiquitania into the ongoing local and regional processes of policy-
region in the Eastern lowlands in from April 20-25 and August making and citizen participation. There was a strong emphasis
2-7, 2005, it became apparent that the project did not have any on finding new ways of achieving the overall development
sustainable impact on local governments and communities. objectives of the project.
Interviews with local indigenous leaders, several government A. Socio-economic, Cultural and Technological Diversity
officials, mayors, and Juan Burgos, the ICT project of Municipalities
coordinator of APCOB—a local NGO promoting indigenous In order to set realistic objectives, the project team focused
people‘s development in the eastern lowlands—revealed that its technical assistance on fifteen municipalities and nine
the municipalities and communities were not aware of the regional municipal associations, covering all of Bolivia‘s nine
suggested Information Management System, and that in fact states.7 A major challenge for the project team, however, was
the fourteen communities of the Mancomunidad did not have
any connectivity to the Internet. The tremendous ―design-
reality‖ gap between project descriptions, consultant reports, 3
Interview with Justo Seoane, Mayor of Concepción, and Ex Vice-Minister of
and the realities on the ground was highlighted by the Indigenous Affairs, April 22, 2005.
4
FAM is the National Association of all the Bolivian municipalities with the
evidence through the results from my fieldwork. Most people exception of the major cities of La Paz, El Alto, Cochabamba and Santa Cruz.
interviewed were not aware that the project existed, even 5
PADEM stands for the Program for the support of the democratization of
though its first phase had been completed less than a year municipalities, which is being coordinated by the Swiss Non-governmental
before the fieldwork was carried out. In the municipality of organization Ayuda Obrera Suiza (AOS), which works towards the
strengthening of local democratic processes and the empowerment of poor
Concepcion, located in the Chuiqutania, approximately five communities throughout Latin America.
hours from the city of Santa Cruz, the newly elected mayor 6
Interview with Carlos Soria, coordinator of the Enlared project. August 25th
clearly expressed his frustration with the overly centralized 2005.
7
The 15 municipalities include: Cobija (Pando), Trinidad (Beni), Montero
planning and execution of the projects: and Comarapa (Santa Cruz); Sipe Sipe and Villa Tunari (Cochabamba),
Monteagudo and Tarabuco (Chuquisaca); El Puente (Tarija); Batallas and
270
that the fifteen municipalities differed extremely between municipal elections in December 2004) and their contact
rural, poor indigenous communities, such as Batallas in La information, as well as the most recent socio-economic and
Paz, or El Puente in Tarija, rural more mestizo dominated human development data on all 314 municipalities. In addition
municipalities such as Comarapa or Montero in Santa Cruz to the provision of content, the Portal provided more
and large urban municipalities such as Trinidad or Cobija— interactive features, such as a weekly forum with national or
the capitals of Beni and Pando, respectively. In addition to the local policy makers including relevant issues of interest such
tremendous difference in the socio-economic, political, and as the Constitutional Assembly or political and social
cultural context, these municipalities differed in their access to exclusion of the extreme poor due to their lack of appropriate
electricity and telecommunications services. To address and documentation.
understand the heterogeneity and the local context of the Concerning the capacity-building program, the project
fifteen participating municipalities described above, the trained 60 municipal staff in the use of ICTs (30 from the
program team initiated its activities in October 2004 with a regional associations and 30 from the 15 municipalities). The
detailed ICT needs assessment at the local level. staff from the municipalities participated in two one-week
Its main objective was to assess the differences in national training workshops, one organized in June in La Paz
awareness, ICT uses, and proficiency of use among local and the other one organized in late August in Cochabamba.
government officials in the fifteen participating municipalities. The content of the training focused on (i) basic computer
As part of this baseline study, the team carried out community skills, such as word processing or use of spreadsheets; (ii)
meetings, interviews, and structured surveys with mayors, introduction to the use of the Internet; (iii) the design and
council members, technical staff of the municipalities and with development of websites and; (iv) an introduction to ‗digital‘
representatives of local Territorial Grassroots Organizations journalism. The main purpose of these training workshops was
(TGOs), members of the Vigilance Committees (VCs), and to create a network of technical specialists in the
other community leaders. In total, 634 people—424 men and municipalities, who then could become the municipal focal
210 women—participated in this baseline study and provided points for Enlared.
valuable information on their information and
communications needs and their current use of ICTs [46]. IV. EVALUATION OF THE PROJECTS IMPACT
The assessment found that there was little awareness and
uses of the Internet in any of the rural municipalities, and that, Based on the above description of the project‘s main
even in the urban municipalities like Tupiza or Uyuni, local achievement or outputs, this section will focus on evaluating
governments were not using the Internet for work. In fact, out the impact the project had on the well-being of rural
of a total of 127 local government officials, only a quarter communities. The evaluation of the project impact is hereby
(24%) used the Internet for work; the overwhelming two- defined as the “systematic analysis of the lasting or significant
thirds majority did not use the Internet at all, and one-tenth of changes— positive or negative, intended or not—in people’s
participants did not even use computers. The results of the lives brought about by a given action or series of action [47].
needs assessment reveal profound differences in ICT readiness As such, the analysis aims to unpack the extent to which this
among the fifteen municipalities studied. project has made a difference in peoples‘ lives and has in fact
B. Main Activities met its overall development objectives by asking the
following: (i) Has the project contributed to strengthen the
Based on the baseline study, the project team focused its
managerial and technical capacity of local governments, thus
activities on the following four strategic areas: (i) technical
allowing them to play a more pro-active role in public-policy
redesign of the Portal; (ii) content development and online
making? (ii) Has the project supported the transparency of
services; (iii) capacity-building in the use of ICTs; and (iv)
municipalities and thus promoted participatory democratic
news desk and communications. Within these different areas
processes at the local level; and finally (iii) Did the project
the project was able to achieve the following specific results.
enhance the efficiency and efficacy of local government thus
First, the Portal www.enlared.org.bo was redesigned using
improving its delivery of services, such as health and
local technology and focusing on providing technical support
education to its community. This categorization is based on
to the Federation of Municipal Associations, as such helping 9
the theoretical framework developed by Lawrence Prachett
municipal associations and individual municipalities in the
described above in the case study [48]. The article applies the
development of their own websites. At the completion of the
alternative evaluation framework based on Sen‘s capability
second phase of the project, the Portal hosted 116 sites from
approach developed by the author in an earlier article to
these organizations. The Portal placed significant effort on
evaluate the impact of the e-government program on people‘s
providing targeted and timely content for policy makers, local
well-being [49].
government officials, donor agencies and the general public
interested in issue of municipal and local development. For
instance, the Portal provided for the first time a A. The program’s Overall Impact
comprehensive list of all newly elected mayors (from the The principal finding presented in this section is that
the program had a very limited positive overall impact on its
Coroico (La Paz); Machacamarca and Pampa Aullagas (Oruro) and Uyuni and participants‘ human and social well-being. The empirical
Tupiza (Postosi).
271
results are based on two user surveys carried out by the project
Knowlede about the EnlaRed Municipal Program
team and a sub-sample from the impact survey (n=22), carried
among local government officials
out as part of my research in July 2005. The first user survey (user survey: n=190)
(n=190), administrated as part of the ICT needs assessment in
October 2004, was designed to generate a baseline about the
N/A, 22, 12%
awareness and proficiency of ICT use in the fifteen
municipalities. The second user survey (n=186), was carried Yes, 70, 40%
out by the project team in September 2005 in the same fifteen

municipalities, as part of its monitoring and evaluation
activities. The main difference between these two surveys and No , 85, 48%
the impact survey that I carried out as part of my research is
that they did not include any questions on people‘s perceptions
Yes No N/A
of the project‘s impact on their well-being and the well-being
of their communities, but instead assessed the participants‘ Fig. 1: Low Awareness of the Enlared Municipal program
proficiency in ICT use and on evaluating the ICT training
component of the program. The principal finding of this section—that the project had a
Before analyzing participants‘ perceptions of the project‘s very low overall impact—was also confirmed by the empirical
impact on their human and social capabilities, we can observe results from the impact survey I carried out during my
several basic statistical indicators about people‘s ICT use and fieldwork. Results from this survey indicate that—based on
conclude that the project had a limited impact in the fifteen the participants‘ self-perceptions—the program did not
municipalities it targeted. enhance in any significant manner the human and social
The first critical finding from the data analysis is that the capabilities of its participants. The graphic below shows that
large majority of local government officials from the fifteen the participants perceived the Internet to have had only a very
municipalities that participated in the project were entirely minor positive impact across all dimensions of their lives.
unaware of the project‘s existence. Only a small percentage of With the exception of the organizational dimension, in all
respondents to the baseline survey administrated in 2004 and other dimensions (i.e., political, social, and economic) the
the user survey from 2005 indicated that they knew of the majority of participants expressed serious doubts about the
project, and an even smaller percentage of the respondents had positive impact of the Internet on their lives, indicating that
actually used the services provided by the project. In fact, the they believed that the Internet could only ―somewhat‖ or ―not
data from the first survey indicate that out of 127 respondents, at all‖ enhance their well-being.
only 16 (13%) knew about the project and only 14 (11%) were Another remarkable finding was that the Enlared Municipal
using the first municipal Portal developed by ICMA. This project participants were very skeptical about the Internet‘s
figure is astonishing, particularly since in October, 2004 the impact in all dimensions of their lives. For instance, only 56%
project had already been operating in Bolivia at a national of the Enlared participants indicated that the Internet had a
level for three years. These data confirm that, particularly in positive impact on the social dimension of their well-being.
its first phase, the project failed to make any long-term The negative perceptions of the participants is particularly
positive impact on local governments or on the participating striking in the political and social dimension, since the core
communities. objective of the Enlared Municipal project was to improve the
The graphic below shows that awareness of the program performance of local governments in these two dimensions.
among local government officials significantly increased In fact, more than two-thirds (70%) of the program‘s
during the second phase but the project continued to suffer participants believed that the Internet could only ―somewhat‖
from relatively low awareness about its activities at the local or ―not at all‖ improve their well-being in the political sphere
level. The empirical data from 2005 show that out of a total of of their lives. In the social dimension only 31% of its
190 respondents, 70 people (about 37%) indicated that they participants indicate that the Internet could significantly
knew of the project—a major improvement from the enhance their well-being in this dimension.
extremely low figure of 13% a year earlier. The project In sum, the empirical data reveal that the participants were
continued to be under-utilized: only 31 respondents (17.5%) particularly skeptical of the Internet‘s ability to improve their
indicated that they frequently used the services provided by own well-being and the living conditions of their communities
the project. politically or socially, constituting the program‘s failure.
In spite of the significant improvements during the second On the other hand, the organizational dimension is the only
phase of the project, the continuously low levels of awareness dimension in which its program participants‘ have favorable
and usage among local government officials raise major views. In fact, 59% of the respondents believe that the Internet
questions about the program‘s overall effectiveness and are a can significantly enhance their well-being in this dimension.
good initial indicator for the limited overall impact of the With respect to the economic dimension, respondents‘
project. views are most skeptical; an overwhelming majority (70%)
272
findings clearly indicate the overall limited impact of the

Enlared: Perceptions about the Impact of the Internet
on peoples' well-being program in this dimension.
What are the underlying causes that explain this finding?
Organizational 59 18 23 Which factors explain why the program was so unsuccessful
Personal 41 24 35
in reaching its main objectives? What were the effects of the
program on local governance structures and which barriers did
Cultural 35 35 30
it encounter in trying to enhance indigenous peoples‘ political
Social 31 25 44 well-being?
Political 30 35 35 The following analysis draws on the empirical results from
Economic 24 6 70 the impact survey and presents more in-depth qualitative
evidence from the two municipalities of Batallas and
0% 20% 40% 60% 80% 100%
Comarapa. Based on Pratchett‘s above mentioned
A lot Somewhat None classification and the principal objectives of the Enlared
program, I have broken down the political dimension into the
Fig. 2: Enlared Municipal Participants‘ overall perceptions on following two areas: (i) enhanced transparency and
ICT impact accountability and (ii) improved local governance.
indicated that the Internet would not have any impact at all on B.1 The Impact on Enhanced Transparency & Accountability
either their personal or community well-being. The empirical The first critical finding is that the Enlared Municipal
evidence from the impact survey thus emphasizes the program had only very limited effects on the transparency and
participants‘ overall negative perceptions of the Internet‘s vertical accountability of local governments and thus could
effects on their well-being. In contrast with the overall results, not significantly strengthen local democratic processes. The
Enlared Municipal participants were skeptical of the Internet‘s graphic below shows the effect the program had on the major
impact on their individual well-being and on the social four areas of accountability.
capabilities of local governments and their communities. In The disaggregated results illustrate that the program helped
fact, the data analysis demonstrates that the program failed to improve people‘s access to information about their own rights
enhance people‘s individual capabilities and help local and national government policies, but failed to enhance the
governments to improve their performance, accountability, transparency of local governments. In fact, less then one
and delivery of social services to the mostly rural quarter of the respondents (23.5%) believed that the Internet
communities. could make an important contribution in this area, and only
25% were convinced of its value for enhancing people‘s
B. Impact on the Political Dimension of Well-being access to information on government programs.
This section investigates the underlying reasons that the
program had only a limited effect on the political dimension. 60
Enlared: The internet and Vertical Accountability
This issue is central to the evaluation of e-government 51.8

50 45.2
programs, and specifically the Enlared Municipal program, 41.2 40.4
40
since its main objectives were political in nature.
Percentage
35.3 34.4
28.6 31.7
The analysis of survey results and interview responses in 30
23.5 25.1
23.1
the previous section shows that the Enlared Municipal project 20
19.6
failed to significantly improve participants‘ informational 10

capabilities; this section will investigate the project‘s failure to 0
promote ICT use as a way to improve governmental inproved improved improved improved
responsiveness, transparency, and accountability. A glaring information about know ledge about transparency information about
nat. gov. policies citizen rights gov. programs
result from the impact survey was that less than one-third of
the respondents (30%) thought the Internet had any significant A lot Somew hat None/Na
effect on the political dimension. The data indicate that 35% Fig. 3: Continued lack of governmental transparency and
of the programs participants believed that the Internet had no accountability
effect at all on this dimension. The particular pessimism of
Enlared project participants becomes apparent when their The results point to a principal incongruity in the program:
responses are compared with those from the overall sample, while it had some positive impact through its Portal on
among whom 36% of respondents indicated that they were improving the access to information related to municipal
convinced that the Internet could play a significant role in this development at the national level, it did not improve
area and only 20% expressed serious doubts about its efficacy. information flows between municipalities and local
It is critical to note again that the Enlared Municipal program communities. The program had an overly abstract national
was the only ICT program that specifically prioritized the use agenda and improved the access to information for a small
of ICTs to enhance peoples‘ political well-being. These elite of national policy makers, international donors and the
273
media, however failed to enhance in any significant manner population and attempted to suppress the influence of
the dynamics between local governments and its citizens at the indigenous peoples in local politics. The president of the local
community-level. agricultural producers associations and a Quechua indigenous
This major finding was also confirmed by the semi- leader describes the exclusivity of local politics:
structured interviews with local government officials, NGO
representatives and community leaders in Batallas and ―For us it is very clear that the municipality
Comarapa. In an interview, the ―official mayor‖8 of Comarapa is quite biased against our communities. It is
clearly expressed his skepticism about the project and the role always the same issue. We can wait for
of the FAM in supporting local governments in general: hours in the municipality with our demands,
the mayor and the councilmember won‘t
―We have a very diverse population in listen to our concerns. They are too busy
Comarapa and the surrounding with their political games and to represent
communities; about half the population are the interest of the rich people of the city—
mestizos and the other half are indigenous we just don‘t have any voice in their
migrants from the highlands. Thus for us to decisions and we have no information at all
improve our communications with all the about what type of programs and activities
communities is very important, but I don‘t they are planning for this year.‘10
see how this program can help us to better
deal with this problems. The program has In short, local authorities lacked the will to improve access
developed a website—How does this help us to information for the general population and instead allowed
to improve our communications with more a small circle of mostly urban elites to dominate the local
remote indigenous communities? How do political decision-making process. The mostly rural
these communities benefit from this—the indigenous population was entirely excluded from any
majority of them do not even have access to political decision-making processes and was forced to choose
electricity, then how can they access a the path of street blockages and civil disobedience to gain a
website. I really can‘t see the value of this voice in the political process. Questions remain, however:
type of program for us.‖9 What was the role of Enlared Municipal project in such a
complicated socio-political environment? Did the program
This comment needs to be understood in the context of have any influence at all on the local political process? Why
Comarapa, a municipality that, in spite of relatively favorable did the program fail to promote in any way the popular
socio-economic indicators, was confronted with significant participation of indigenous communities in the decision
social tensions between its primarily rural indigenous making process of local government.
population and its urban mestizo population. In the statement One reason was the programs‘ inability to react flexibly to
above, the senior advisor stresses the challenge of improving the given socio-political context. Instead of working with local
his municipality‘s capacity to communicate with local government officials to make the existing municipal planning
communities in order to enhance the popular participation of a processes more transparent and to gradually enhance the
broad range of people in municipal planning processes. This is popular participation of indigenous communities in local
a critical shortcoming the implementation of the LPP has decision-making processes, the project continued to promote
confronted in most rural communities throughout Bolivia. its original concept to assist the municipality in developing a
While the law has succeeded in significantly enhancing the website which was supposed to enhance its transparency. This
participation of communities in the identification of municipal activity provided by the program clearly did not match the
development projects through broad-based consultations, it local socio-political context nor was it an appropriate
has faced in most municipalities serious limitations in terms of mechanism to address the existing democratic shortcomings in
improving the participation of local communities in municipal Comarapa. Thus, the program did not alter the prebendal
decision-making processes about critical issues of local culture 11 of local government. The same few powerful groups
development such as for instance the approval of planned continued to dominate the decision-making process within
public investment projects (Goudsmit and Blackburn, 2001). local government and the mostly indigenous population
The above quote highlights that the program failed to continued to be excluded and marginalized in political terms.
enhance the transparency of the municipalities and to improve In brief, the program failed (i) to improve the transparency and
its communications with more remote indigenous accountability of local government and (ii) to enhance
communities. In Comarapa a critical issue furthermore was peoples‘ political capabilities to better participate in the
that the local government sided frequently with the mestizo decision-making processes. The empirical evidence presented
8 10
The ―official mayor‖ is the most senior government official in local Interview with Julio Sanchez, president of the agricultural producer
government, who is in charge of the all processes of the public administration association, Comarapa, May 24, 2005.
11
(own translation). A prebendal local political culture is defined by local political practices
9 9
Interview with Jose Luis, official mayor of the municipality in that are characterized by patrimonial, clientelistic and caudillist features
Comarapa, July 15, 2005. (Blackburn, 2000)
274
above, based on participatory observations and unstructured of our own program. Our idea was that we
interviews with key policy-makers and local actors, clearly could share the connectivity costs with
demonstrates that the local stake-holders perceived the several institutions working in Comarapa
Enlared project to be distant from their own realities and and asked the mayor for his support.
vividly illustrates the program‘s inability to meaningfully Unfortunately, the mayor did not see any
improve local democratic processes. value in this program and instead promised
that the municipality will build its own a
B.2 Impact on Enhanced Local Governance
telecenter for the community. Now, two
Another critical aspect of enhanced political capabilities at years later, Comarapa still does not have the
the local government level is the improved capacity of promised telecenter and we have never
municipalities for public policy making in local communities. received an official response from the
Within the increasingly fragmented and disaggregated municipality on our proposal we submitted
structures of local communities many local governments face in writing to the mayor‘s office.‖12
considerable challenges in playing a central role in promoting
local governance structures that are conducive for economic This statement reflects the all too common lack of
and social development [50] The following analysis uses the coordination among local municipalities, local NGOs, and the
concept of ―good local governance,‖ developed by Judith central government. Despite the stated interest from the
Tendler, which refers to a healthy ―three-way dynamics municipal government and the Enlared project, after two years
among local government, civil society, and an active central Comarapa still did not have the telecenter it had been
government‖ [51]. promised. A key aspect in the failure of the Enlared Municipal
At the core of the concepts stands the notion that a good program to promote good governance at the local level was
local governance structure is essential for effective public the frequent discrepancy between the political mandate
policy making. Tendler stresses that coordination between designating local governments the executive agencies of
local governments, local civil society organizations, and development programs like the ICT program and their actual
communities, as well as coordination between local and institutional capacity and technical knowledge to actually
central government, are critical for good governance. Local implement such programs. The following comment by the
government is more of an effective facilitator of partnerships mayor of Comarpa about the planned telecenter illustrates this
between different local actors and stakeholders than a sole point:
implementer of development programs. ―For me it is very clear. We in the
A critical question this raises is what role ICTs can play in municipality are the executive branch of
such a new system of local governance? The case study will government at the local level. It is our
focus on the extent to which the Enlared project was able to mandate to plan and implement all programs
strengthen the capacity of local governments to assume the in our municipality. In the case of the
role of a ―facilitator‖ of local development. In brief, to what planned telecenter, I had several meetings
extent was the project successful in enhancing healthy local with the staff from ICO and explained to
governance in local communities? them that it is our role to coordinate this
The principal finding from both the municipalities of project. I still think that it is a good idea to
Comarapa and Batallas demonstrate that the Enlared project build a telecenter here in Comarapa,
was unable to play a central role in enhancing the capabilities particularly for your youth, however it needs
of local governments in this dimension, primarily because it to be clear that we need to oversee all
did not effectively promote a cultural change within local activities related to this project.‖13
governments to a more democratic and inclusive approach of
local public policy making. Instead, local government This statement clearly demonstrates that the local mayor
continued to manage local development programs in their considered it the prerogative of local government to
traditional authoritarian and bureaucratic styles. coordinate and oversee all development programs in his
Asked about the role the Enlared program played in local municipality. His defensive attitude illustrates the tensions
politics, Claudia Camacho from ICO cited the program‘s between his local administration and the NGO. While local
inability to facilitate a partnership between the local government officials frequently insist that the municipality
government and her NGOs in order to work together on the needs to spearhead all the development programs at the local
construction of a telecenter for Comarapa. level, the severe crisis of the state has led to a situation where
national and international NGOs have frequently filled the
―When we first heard about the Enlared institutional vacuum and are de facto implementing many
Municipal program and that Comarapa was programs, such as in health, education or infrastructure which
selected to participate in its pilot phase, we under normal circumstance would be carried out by the state.
were very excited and thought that we could
closely collaborate with the municipality on 12
Interview with Claudia Camacho, ICT program coordinator of ICO in
promoting the use of ICTs, a key objective Comarapa, May 23, 2005.
13
Interview with mayor of Comarapa, May 23, 2005.
275
The above-mentioned telecenter is a good example of a significant impact on enhancing the inter-agency coordination
situation in which the local government lacks the institutional between local government, civil society organizations, and
and technical capacity to implement the program and a non- local communities; neither did it enhance the relationship
governmental organization brings its own expertise and between local and central government.
resources to provide the population with a needed service.
Another reason for the poor local governance structure in
C. Impact on the Organizational Dimension
Comarapa is the role international donors frequently play in
creating dependencies in both the public sector and civil The overall empirical results from the impact survey seem
society organizations at the local level. The Enlared to indicate that the program had considerable positive effects
Municipal program, for instance, had defined its priorities and on the organizational dimension. The data show that more
operational methods through a highly donor-centric approach. than three-fourths of the program‘s participants (77%) rated
In fact the program did not allow local actors to come together the impact of the Internet on this dimension either as very
in order to adjust the program‘s planned activities to their significant (59%) or somewhat significant (18%). This section
particular social, political, and cultural context and to define a disaggregates these overall results in order to investigate the
way on how to collaborate on its implementation in extent to which the Enlared project did indeed lead to
Comarapa. In fact, the provision of internet services and ICT significant organizational changes in local governments and
training workshops would have lend itself to such a thus had a positive impact on their institutional capability.
collaborative approach, since many organizations within the This issue is critical for evaluating the overall impact of the
municipality had initially expressed their strong interest in program, since one of its core objectives was to enhance the
sharing the connectivity services with each other. An efficiency and efficacy of local governments by introducing
important consequence of the common donor-driven approach ICTs into local public administrations.
is that it provides important disincentives for local actors to First, the analysis uses the ―social capital index‖ in order to
work together and to coordinate their activities with each carry out a deeper analysis of the impact of the Internet on the
other, since each of the organizations is primarily accountable organizational dimension. The graphic below highlights that
to their international donor and not to local stakeholders. while a two-thirds majority of participants (69%) perceived
The example provided above—about the inability of the the Internet as critical to strengthening their horizontal
municipality to react to the proposal by ICO to share the costs networks with colleagues in other municipalities, only a
of the Internet connectivity—demonstrates that programs like minority of respondents (49%) believed that the Internet had a
the Enlared Municipal project focus solely on achieving their significant effect on strengthening the organizational
centrally defined development objectives rather than on capabilities of local government. The participants‘ skeptical
facilitating partnerships among the different local actors. In views about the program‘s impact on organizational structures
fact, the program worked exclusively together with the of local governments is also evidenced by the fact that more
municipality and entirely neglected to reach out to the other than one-third (35%) of respondents doubted that the Internet
organizations within the municipality. could play any role in this dimension. Furthermore, it is
In this sense the Enlared program is exemplary of many noteworthy that the large majority of respondents (64%), in
other development programs in Bolivia in that it was unable to spite of their skepticism about the Internet‘s ability to affect
facilitate a partnership between the different local stakeholders organizational change within government, had a very
that would have significantly enhanced the impact and favorable view of the positive effects of the Internet on local
sustainability of the program [52]. In the case of the Comarapa organizations outside of government, such community-based
municipality, the local government officials even preferred to organizations. This seems to indicate that the participants
delay the municipality‘s Internet access for many months believed that there are major barriers to introducing ICTs into
rather than develop a collaborative relationship with a local public administrations, while they consider this to be easier to
NGO. be achieved outside of government.
In sum, the Enlared program failed to enhance the local The more in-depth analysis of the data provides much richer
governance structure at the community level, since its program insight into the effects of the program on the organizational
had an overly abstract national agenda and its centralized dimension. In fact, the results demonstrate that the program‘s
approach prevented it from affecting in any significant manner organizational impact was limited to establishing a strong
institutional changes at the local level. Furthermore, the social network among its participants, instead of instituting
program did not facilitate effective partnerships between far-reaching organizational changes within local
different local actors and instead narrowly focused on trying to bureaucracies. Furthermore, the observed strengthening of
implement its pre-conceived objectives irrespective of the social capital and horizontal communication networks was the
local socio-political context. In this sense the program result of ad hoc formation of informal networks between the
represents the common shortcomings of similar donor-centric program‘s participants, rather than a sign of improved
programs in Bolivia. Finally, the program failed because it institutional linkages between the different local
lacked the necessary institutional credibility with local municipalities.
government officials based on the severe crisis of central
government institutions. In brief, the program had no
276
majority (62%) of the respondents who had an e-mail account

Enlared: Impact on Social Capital
used it only once every three days.
strengthens horizontal Moreover, the graphic below illustrates the resistance or
69 20 11
netw orks apathy of key decision-makers to these technologies. In fact,
the large majority of mayors and/or council members (82%)
strengthens local
49 16 35 did not consider the Internet important and thus did not use it
gov ernments
for their work. The data indicate that only among technical
staff did the majority (58.2%) use the Internet for their work.
strengthens local
organizations
64 18 18 Technical staff also demonstrated higher rates of e-mail use:
54% of them had personal e-mail accounts, compared to just
0 20 40
Percentage
60 80 100 18% of mayors and council members.
A lot Somew hat None/NA The very low local appropriation of ICTs in local
government can be confirmed by comparing the results from
Fig. 4: ICT—a powerful tool for horizontal networking
the baseline study from October 2004 with those from the user
The above data also have to be seen in the context of the survey from September 2005 described above.
significant barriers the program participants faced in
developing and promoting the use of information systems Internet Use for work by local government officials
within local bureaucracies, because of local government (user survey: n=190)
officials‘ limited understanding of ICTs. In fact, a second

critical indicator for the program‘s lack of organizational N/A, 41, 22%
impact on local governments is the continuously low levels of Yes, 69, 36%
Internet use in the large majority of municipalities.
The following section draws on the empirical results from
the project team‘s two user surveys14 carried out in October No , 80, 42%
2004 and September 2005 with local government officials in
all of the fifteen participating municipalities. The analysis of
Yes No N/A
this dataset adds significant value to the statistical analysis,
since these surveys were carried out with a broad range of Fig. 5: Limited Internet use within local governments
local government officials in each of the participating
municipalities15, while the impact survey focused only the IT The percentage of respondents who used the Internet at
and communication specialists who directly participated in the work remained unchanged at about 36% in both years. The use
program‘s training workshops. of e-mail accounts actually decreased from 47% in 2004 to
The first finding from the empirical evidence is that the only 40% in 2005. This empirical evidence clearly
program failed to institutionalize the use of the Internet in demonstrates that the use of ICTs and the Internet continued to
public administrative processes in local governments. As the be low priorities in the fifteen municipalities, in spite of their
graphic below illustrates, by September 2005—twelve months nominal participation in the Enlared Municipal project. One
into the second phase of the project—only about one-third reason for this low level of ICT use is the extremely high turn-
(36%) of the local government officials were using the over rate of local government officials. The municipal
Internet for work; although about two-thirds of respondents elections in December 2004, four months into the project
indicated that they had some experience in using the Internet, implementation, resulted in profound political changes in a
most continued to use the Internet for personal, rather than large majority of local governments and negatively affected
professional, communication. In fact, the survey results show the project.
that only one-third of the respondents who were Internet users
(29%) used it at their workplace; most people (49%) used it in
Internet use by position in municipality
telecenters or Internet cafés.
90 82.2
The failure of the program to promote the use of ICTs in 80 66.9
70 63.8
local governments is confirmed by the fact that only 40% of
percentage
60
41.8
government officials had a personal e-mail account (such as 50
40 32.7
―.yahoo‖) and none of the participants had an institutional e- 30 10.7
25.5
14.818.3 15.7
20.5
20
mail account (―.gov.bo‖). It is also noteworthy that the large 10 7.1
0
Mayor/Council Technical Member of Total
member specialist oversight
14 committee
Both of these user surveys focused exclusively on issues related to the
awareness, readiness and use of ICTs in local governments (Enlared project Position
A lot Somew hat None/NA
files).
15
The user survey from September 2004 was based on a sample of 190 Fig. 6: Government officials‘ skepticism about Internet use
respondents and the second survey from September 2005 on a sample of 186
participants.
277
In fact, 68% of the respondents from the 2005 survey

indicated that they had been working for less than one year in
the municipality and just eight of the training participants, out Enlared: Impact on Social Services
of a total of ninety-four, had participated in a previous training

workshop during the first phase of the project. Improv ed access to
38 25 37
education
The principal finding about the marginal role ICTs play
within local governments is also confirmed by the fact that Improv ed access to health 14 38 50
only technical staff (i.e., IT specialists and communication
officers) participated in the ICT training workshops organized Improv ed gov ernment
13 17 70
by the Enlared program, and that key decision-makers (i.e., serv ices
senior administrative staff and/or council members) did not

0 10 20 30 40 50 60 70 80 90 100
show any interest in participating. In sum, the program failed %
to reach one of its main objectives—to promote the use of A lot Somewhat None
ICTs in local public administrations in order to strengthen the

organizational capabilities of local government. Why did the Fig. 7: Internet‘s limited role in improving service delivery
project fail to promote organizational changes?
V. CONCLUSION
D. IMPACT ON THE SOCIAL DIMENSION OF WELL-BEING
This case study demonstrates that the Enlared Municipal
The empirical findings reveal that the Enlared program had program failed to significantly impact the accountability,
no substantial impact on the social dimension of people‘s transparency, and performance of local governments. The
well-being. In fact, as the graphic above illustrates, only one- reasons for this failure were both internal and external. The
third (31%) of its participants indicated that the Internet has an main internal factors were (i) the overly centralized approach
significant impact on their human capabilities in the social taken in the implementation of the project, compounded with
dimension. This finding is striking, since the social dimension the project‘s assumption that the regional municipal
was the one dimension in which participants in the overall associations would be an effective intermediary in reaching
survey had the most positive views, with 60% believing that local municipalities which turned out to be false; (ii) the
the Internet had a major impact on their lives. These data need political nature of the Federation of Municipal Associations
to be seen in the context of the third major objective of the (FAM), which forced the project to engage with local
Enlared program, which aimed to improve the performance of governments that were unwilling and unprepared to participate
local governments in the delivery of social services by in such a program; (iii) the overemphasis on the technical
strengthening their institutional capabilities. This aspect of development and maintenance of a Portal on municipal
municipal development issues is particularly important for development; and (iv) the lack of adequate resources and
indigenous peoples, since the large majority of indigenous staffing to carry out a more systematic and long-term ICT
communities lacks access to basic social services, such as capacity-building program at the local level. In addition to
water and sanitation, education and health. these internal factors, there were several key external factors:
Furthermore, one of the principal objectives of the entire (i) the overall political instability in Bolivia, beginning in
decentralization program in Bolivia was to substantially October 2003, which involved the dissolution of two
improve the delivery of social services to rural, marginalized governments in one year alone (2005) and which provoked
communities. In fact, a critical argument in favor of the LPP disruptive forms of civic unrest; (ii) the lack of a coherent and
has been that local municipalities are supposed to have a much comprehensive national strategy for developing a
closer relationship to communities due to their geographic telecommunications and ICT infrastructure in Bolivia,
proximity. The literature, however, indicates that this major particularly in rural areas, meaning that ICT projects were
objective of the LPP was not realized due to capture of local implemented in technically unprepared regions; and (iii) the
municipal planning processes through local elites (Blackburn, limited institutional and technical capacity of many local
2000). The empirical findings from the impact survey seem to governments to develop public policies and to implement
confirm this skeptical view in the sense that the participants development projects.
expressed very critical opinions about the role ICTs can play In sum, the program was too centralized, which led to a
in improving the delivery of social services to indigenous critical design-reality gap that undermined its ability to
communities. While at first sight this general finding seems enhance local governments‘ transparency, accountability, and
unforeseen, it is fully consistent with the failure of the performance. The program was never integrated into broader
program to significantly enhance the organizational issues related to local governance and did not have any
capabilities of local government. In other words, since the significant impact on its participants‘ human and social
program did not succeed in strengthening the institutional capabilities. In order to more significantly affect indigenous
capacities of municipalities it is to be expected that it also peoples‘ well-being, it would have to be much more precisely
failed to have any positive impact on the delivery of social tailored to the specific socio-economic, political, and cultural
services. realities of local governments and indigenous communities.
278
Bibliography Communications Services: A Comparison of Malaysia

and South Africa. Journal of Developing Societies, vol.
[1] J. Samoff, The bureaucracy and the bourgeoisie: 15, pp.47-60, 1999
Decentralization and class structure in Tanzania. [17] M. Gasco, New Technologies and Institutional Change
Comparative Study of Society and History 21 (1), pp.30- in Public Administration, Social Science Computer
62, 1979. Review, 21, 1, 6-14, 2003.
[2] W. Oyugi, ―Decentralization for good governance and [18] R. Heeks, 1999; Information and communication
development: The unending debate‖. Regional technologies, poverty and development, University of
Development Dialogue, 21 (1): 3-22. Manchester. Institute for Development Policy and
[3] J. P. Faguet, ―Decentralization and Local Government in Management. Working Paper Paper No. 5,1999.
Bolivia: An overview from the Bottom-up‖, Crisis State [19] W‘O Okot-Uma, R, Electronic Governance: Re-
Programme, Working Paper No. 29, 2003. inventing Good Governance, 2001, World Bank.
[4] World Bank, The state in a changing world: World [20] OECD, Engaging Citizens in Policy-Making:
Development report, 1997. New York: Oxford University Information, Consultation and Policy Participation,
Press. Puma Policy Brief No. 10, 2001.
[5] Booth at al., Empowering the poor through institutional [21] C. Ciborra, Unveiling E-Government and Development:
reform? An initial appraisal of the Bolivian experience. Governing at a distance in the new war. Department
Stockholm: Department of Social Anthropology, of Information Systems, Working Paper Series. London,
Stockholm University. 1996. London School of Economics and Political
[6] B. Kohl, ―Democratizing Decentralization in Bolivia— Science.2005;
The Law of Popular Participation Journal of Planning [22] R. Heeks and S. Bhatnagar, 'Understanding success and
Education and Research 23, pp. 153-164, 2003. failure in information age reform', in Heeks, R. (ed.)
[7] H. Blair, ―Participation and accountability at the Reinventing Government in the Information Age –
periphery: Democratic local governance in six countries‖, International Practice in IT-enabled Public Sector
World Development 28 (1), pp.21-39, 2000. Reform, London and New York: Routledge, pp. 49-
[8] J. Blackburn, Popular participation in a prebendal society 74, 1999.
: a case study of participatory municipal planning in [23] R. Heeks, Most e-Government-for-Development Projects
Sucre, Bolivia, Thesis (Ph.D.)--University of Sussex, Fail How Can Risks be Reduced? Institute for
2000 Development Policy and Management (IDPM),
[9] Godsmith and Blackburn, 2001) Participatory Municipal iGovernment Working Paper Series, Paper no. 14, 2003.
Planning in Bolivia: an ambiguous experience, in: [24] ibid.
Development in Practice, Volume 11, Number 5, 1 [25] ibid.
November 2001, pp. 587-596 (10). [26] S. Madon, (2004). ―Evaluating the Developmental
[10] R. Heeks, Most e-Government-for-Development Projects Impact of e-Governance Initiatives: An Exploratory
Fail How Can Risks be Reduced? Institute for Framework.‖ Electronic Journal of Information Systems
Development Policy and Management (IDPM), in Developing Countries 20(5): 1–13.adon, 2004:2
iGovernment Working Paper Series, Paper no. 14, 2003. [27] ibid. p.9.
[11] T. Yigitcanlar, ―Bridging the Gap between Citizens and [28] Moon, The evolution of e-government among
Local Authorities via e-government‖. in municipalities: rhetoric or reality‖, Public Administration
Symposium on E-government, 10–12 May 2003, Review, Vol. 62 No. 4, pp. 424-33. 2002;
Muscat, Oman. [29] J.C. Musso, Weare and H. Hale, ―Designing Web
[12] C. Ciborra and D. Navarra, "Good Governance, Technologies for local Governance
Development Theory and Aid Policy: Risks and Reform: Good Management or Good Democracy?‖
Challenges of E-Government in Jordan." Journal of Political Communication, vol. 17, no. 1, pp.1-19, 2000.
Information Technology for International Development [30] L. Pratchett, New Technologies and the Modernization
11(2), 2005. of Local Government: An Analysis of Biases and
[13] S. Bhatnager; ―Egovernment: Lessons from Constraints‖ in: Public Administration, Vol. 77, no.4,
Implementation in Developing Countries‖, Regional p.731-750.1999
Development Dialogue, Vol. 24, UNCRD, Autumn Issue [31] ibid.
pp. 164-174, 2002. [32] ibid
[14] S. Krishna and G. Walsham, ―Implementing Public
[33] Moon, The evolution of e-government among
Information Systems in Developing Countries: municipalities: rhetoric or reality‖, Public Administration
Learning From a Success Story‖, Information Review, Vol. 62 No. 4, pp. 424-33. 2002
Technology for Development, 2005. Vol. 11 (2), pp. 123- [35] W. J. Tettley, ICT, Local Government Capacity
140. Building, and Civic Engagement: An Evaluation of the
[15] N. Negoroponte, Being Digital. New York, Vintage, Sample Initiative in Ghana. In: Perspectives on Global
1995. Development and Technology, Vol. 1, no. 2d, pp-165-
[16] Wilson III, Development of National Information and
279
192, 2002
[36] J.C. Mussa et al, 2000.
[37] C. Ciborra (2005).
[38] C. Ciborra and D. Navarra (2005).
[39] S. Madon ―Introducing administrative reform through the
application of computer-based information systems:
a case study in India.‖ Public Administration and
Development 13: 37-48, 1993.
[40] S. Madon, (2004)
[41] G. Walsham, Interpreting Information Systems in
Organizations. Chichester: John Wiley & Sons, 1993.
[42] R. Heeks, (2003)
[43] Pettigrew, A.M., (1985) 'Contextual Research: A Natural
Way To Link Theory And Practice', in E.E. Lawler
(ed.), Doing Research that is Useful in Theory and
Practice, San Francisco: Jossey Bass,
[44] Pettigrew, A., Success and Failure in Corporate
Transformation Initiatives, In: Galliers, R.D., Baets,
W.R., Information Technology and Organizational
Transformation, Chichester, Wiley, 1998.
[45] ICMA Final Enlared project report, ICMA 2003.
[46] Enlared Municipal Consultation Report, Enlared project
files, 2004.
[47] C. Roche, Roche, Impact assessment for development
agencies: learning to value change. Oxford: Oxfam GB,
1999.
[48] L. Pratchett, New Technologies and the Modernization
of Local Government: An Analysis of Biases and
Constraints‖ in: Public Administration, Vol. 77, no.4,
p.731-750.1999
[49] Gigler, B., Enacting and Interpreting Technology - From
Usage to Well-Bring: Experience of Indigenous
Peoples with ICTs. In H. Rahman (Ed.), Empowering
Marginal Communities with Information Networking
(pp. 124-164). London: Idea, Group Publishing, 2006.
[50] L. Pratchett, 1999: p. 735.
[51] J. Tendler. Good governance in the tropics. Baltimore:
Johns Hopkins University Press; 1997:145
[52] Van Niekerk, Desarrollo rural en los Andes, Un studio
sobre los programas de desarrollo de Organizaciones no
governamentales, Leiden Development Studies, 1994.
280
Design and Deployment of a Blood Safety

Monitoring Tool
S. Thomas, A. Osuntogun, J. Pitman, B. Mulenga, and S. Vempala
Abstract—Blood is a scarce resource critical to the developed as a collaboration between CDC, the Georgia
management of a variety of life-threatening medical conditions. Institute of Technology and the participating countries. The
It can also be a medium for the transmission of infections, tool expands on the Excel-based system to facilitate the
including HIV. Monitoring the quality and quantity of available efficient and accurate collection and analysis of more than 80
blood, essential to utilizing it as effectively as possible, has been a
indicators from 14 countries.
challenge in many developing countries. This paper describes the
design and implementation of a web-based tool to monitor the
collection, screening and distribution of blood in developing
countries. This project was conducted under the auspices of the
II. RELATED WORK
US President's Emergency Plan for AIDS Relief (PEPFAR),
which funds blood safety projects in 14 countries in Africa and
Several ICT projects have focused on improving the health
the Caribbean. We report results from a usability study,
care system in developing countries. These projects have had
formulate relevant design principles and discuss prospects for several different themes including development of web-based
long-term sustainability. tools for information tracking, implementations of electronic
medical record systems and telemedicine systems to support
Index Terms—developing nations, health, ICT and education of clinicians and referral of difficult cases for
development, Internet, medical services further review.
In [2] a web based information system supported medical
I. INTRODUCTION record keeping and tracking of HIV treatment in Haiti. While
this work is similar to our work in that there is a centralized
Blood is a scarce resource worldwide. But the scarcity is database and an online interface for certified users to enter and
particularly acute in the developing world. In 2008, the World track data, the population and scale of users is quite different.
Health Organization estimated that although 80% of the In our work we focus on users in 14 different countries
world's population lives in developing countries, these worldwide with varying levels of internet access and technical
countries only collect 45% of the global blood supply [1]. support. We also present evaluation of the system by
Effectively managing this scarcity—deciding who gets blood prospective users which indicated that the system was easy to
and when—is a challenge for health care systems throughout adapt to and use.
the developing world. Blood is also a highly effective medium
for the transmission of blood borne illnesses, including HIV, Blaya et al in [3] created a web based tool, e-Chasqui, to
hepatitis and syphilis. The WHO estimates that between improve the timeliness and quality of reporting tuberculosis
95-10% of all HIV infections worldwide may be linked to laboratory data in Peru. This national system is able to provide
unsafe blood transfusions. In 2004, the US President's access to a hierarchy of users in health centers, regional
Emergency Plan for AIDS Relief (PEPFAR), identified blood laboratories and the national laboratory to test results. These
safety as a key element in its comprehensive HIV prevention results are then used to prescribe treatment regimens for
strategy. Since then, PEPFAR has committed $192 million in patients at the local level.
funding to National Blood Transfusion Services (NBTS) in 14 In analyzing the use of web-based tools, Marquez found
countries (Botswana, Cote d'Ivoire, Ethiopia, Guyana, Haiti, that such tools provide notable enhancements in quality
Kenya, Mozambique, Namibia, Nigeria, Rwanda, South improvements in health care [4]. The web-based tool
Africa, Tanzania, Uganda, and Zambia). To track, monitor and described in this paper is an example of such tools.
evaluate the impact of this funding, the Centers for Disease
Control and Prevention (CDC) worked with global blood
safety experts to develop a set of program indicators and a III. PROBLEM AND BACKGROUND
Microsoft Excel-based data collection tool. In this paper, we
describe a web-based data collection and management tool Monitoring the development of a national blood service
requires data from a diverse set of sources. The WHO, in its
Aide Memoire for National Blood Programmes [5], advises
S. Thomas, A. Osuntogun, and S. Vempala are with the Georgia Institute of
Technology, Atlanta, GA 30332, supported in part by a Raytheon Fellowship developing countries to build national blood strategies around
and the NSF. four "key areas", including: The establishment of a nationally-
J. Pitman is with the Centers for Disease Control and Prevention, Global coordinated blood transfusion service; the collection of blood
AIDS Program, Atlanta, GA 30333 only from voluntary non-remunerated blood donors; universal
B. Mulenga is with the Zambian National Blood Transfusion Service, testing of all donated blood, including screening for
Lusaka, Zambia
transfusion-transmissible infections, blood grouping and
281
compatibility testing; and the reduction of unnecessary These challenges increased the possibility of inaccurate, or
transfusions. These broad recommendations can be out-of-date, reporting. To illustrate, in several countries,
further categorized by technical area: regional or provincial blood service offices would would fill
out and email the Excel-based tool to the national manager, or
1) Policy information is needed to track progress in
save the report on a disk or flash drive and send the disk back
standardizing procedures (e.g., laboratory testing)
to the national manager by road. The national office would
and establishing a blood service's legal authority over
then compile all the results from the different provinces and
the national blood supply,
send a final report to the CDC. Each update from a regional
2) Social and demographic information is needed to identify, center required re-entry and re-aggregation of the data, leading
recruit and retain blood donors with low behavioral risk to multiple versions and increasing the chance of error.
profiles for infection with HIV and other transfusion-
To address the weaknesses of the Excel-based tool, the team
transmissible infections (TTI),
developed an Internet-hosted database accessible with a
3) Scientific data are necessary to track the prevalence of standard Web browser. Storing the data in a Web-based
infectious markers in donated blood and to monitor relational database provides several key benefits compared to
laboratory performance, ad hoc circulation of a spreadsheet file:
4) Administrative data are needed to measure human 1) There is a reliable, authoritative source of reported data.
resource and logistical capacity, and, in many cases, to (No more wondering which file among several is most
manage large grants from bilateral or multilateral donors, accurate.)
and
2) Data is continuously available throughout the reporting
5) Medical information is critical to ensure that blood is used period allowing immediate modifications and corrections.
appropriately and to monitor patient outcomes. (No files "in transit" or lost via email.)
A blood service could potentially collect data on hundreds 3) The system can support automated, real-time aggregation
of variables. of reported data from multiple sources with less risk of
transcription errors.
With PEPFAR support, CDC and its international partners
developed a limited, but comprehensive, set of programmatic 4) Full change tracking with audit logs is available.
indicators. These indicators are broken down into a subset of
Additional benefits derive from the system's use of a
variables which cover the five technical categories described
standard web browser interface:
above.
5) Access is ubiquitous and available on all modern
Collecting these data from 14 developing countries with
computing platforms.
different data systems was the principal challenge associated
with this project. To begin to address this problem, CDC 6) End users are familiar with web browsers (e.g., Internet
developed a standardized data collection tool using Microsoft Explorer, Firefox, Safari).
Excel. This format was selected for its ease of use (data
managers in all of the PEPFAR countries were familiar with 7) Updates and enhancements can be easily managed and
deployed.
Excel) and its ability to be transferred via e-mail over low
bandwidth internet connections. A version of this tool, with The specific design principles for the system follow directly
pre-programmed formulae and skip patterns, was field tested from its requirements:
in 2007. While countries reported the tool was easy to use and
met the programs' PEPFAR-related reporting needs, the pilot 1) [Network compatibility] The system must support access
phase also revealed several weaknesses in the use of a through low bandwidth, dial-up connections, as high
spreadsheet-based tool, including: bandwidth connections cannot be assumed for all users.
1) Difficulty in tracking versions of multiple files in 2) [Intuitive interface] The system must provide a user
circulation as new data were entered and re-saved by experience appropriate for blood safety staff who may be
different users, familiar with the use of standard office software (e.g.,
web browsers, spreadsheets, word processing), but who
2) Countries' inability to quickly modify, clean or correct a lack high-level IT training.
data set after a file was submitted to CDC,
3) [Security] The system must provide appropriate security
3) The potential for transcription and other errors as multiple and access control for aggregated health information.
versions of the spreadsheet were merged (e.g., from
multiple regional centers) prior to submission to CDC. 4) [Flexible and adaptable] the system must be easy to
manage, adapt, and expand. For example, it must be easy
4) PEPFAR requires quarterly reports from each country, but for administrators (many of whom will not be trained IT
blood is collected, tested and utilized continuously. This managers) to add new countries, regions or users, and
meant the Excel-based tool could sit idle for weeks or ensure that data entered by these new sources are properly
months at a time. aggregated. (Based on previous experience, the CDC did
not feel the need to support languages other than English
282
in the initial deployment. The site was designed, however, users are able to view as much or as little of the information as
to accommodate localization to alternate languages should they wish, all on one page, as Figure 1 highlights.
that be desirable in the future.)
The tool was designed via collaboration between CDC,
Georgia Tech, and NBTS data managers in the 14 PEPFAR
countries. The Zambian National Blood Transfusion Service
(ZNBTS) agreed to host the pilot phase described in this
paper. This partnership was made possible through a program
at Georgia Tech which seeks to leverage student and faculty
expertise to address social and development problems in low-
income settings. Funding for logistics during the pilot
project was provided, in part, by PEPFAR and the ZNBTS.
The Georgia Tech team's time and expertise were donated
gratis as part of a course.
After multiple rounds of testing, user feedback and usability
studies (described later), it was evident that the web tool was
highly welcomed by NBTS data managers in Zambia and
elsewhere. Their feedback led to changes both in the user Fig. 1. Selective disclosure allows the user to show only those sections of the
interface and in the architecture of the system. Further, data page that are relevant for the task at hand.
managers expressed a strong desire to use the system not just
for reporting but also as a decision informing tool. We Although it is theoretically possible to implement selective
redesigned the system to allow individual countries and disclosure strictly through PHP scripts executing on the web
individual centers within countries to choose the frequency server, most users would find the resulting lag between their
of data entry (to match the frequency at which data was action and the interface's response to be intolerable. For a
actually being aggregated physically), and to choose the type suitably responsive interface, selective disclosure is
of reports and views of historic data that they desired. This implemented solely on the web browser client, using
was all done while maintaining a uniform quarterly PEPFAR javascript to show or hide the appropriate interface elements.
report for all participating countries. However, not all web browsers can fully support javascript,
and some users may choose to disable javascript even when
their browser supports it. To accommodate these users,
IV. INITIAL DESIGN unobtrusive enhancement in the initial design adds selective
disclosure as an optional, additional layer in the interface. The
Initial prototypes for the system used common web prototype accomplishes this by first creating a page with all
application techniques and technologies. MySQL database interface elements available and disclosed. Figure 2 shows a
tables were defined to collect and store reported information. screen capture for such a page.
Scripts written in PHP provided a web-based front end to the
database, allowing users to enter data and view collected
information. The web-based user interface also provided a
means for system administration, including defining regions
and countries, managing users, and supporting data backups.
A key aspect of the user interface design was its reliance on
unobtrusive enhancement[6]. This strategy creates a layered
user interface, where successive layers provide more
functionality, but are only invoked if the user (in particular, the
user's web browser) supports the required technologies. For
example, many of the input pages consist of a number of
individual data elements. To make it easy for the user to focus
on the information in manageable pieces, the interface
provides selective disclosure--information not relevant to the
immediate task is hidden or de-emphasized so that it does not
distract the user. This approach has significant benefits to the
user compared to alternative strategies (such as making each Fig. 2. When the page loads initially, all sections are fully shown. If the
section a separate page or screen), as users are able to decide browser does not support javascript, the page remains displayed in this form.
for themselves what information is relevant to the task. For The interface then attempts to execute javascript to add
example, a user may wish to refer to other inputs when selective disclosure. If the web browser does not support
supplying a particular value. If the other inputs exist on javascript (or if the user has disabled that option), then the
separate pages, then the user would be forced to jump back page remains as in Figure 2. Although there is no selective
and forth between multiple pages. With selective disclosure, disclosure, the interface remains fully functional. When
283
javascript is available, however, selective disclosure national blood center in Lusaka, Zambia as well as in Kitwe,
immediately activates and the new functionality is available to the center for the Copperbelt region. While in Zambia, the
the user. Figure 3 shows a screen capture for the same page team also conducted the study with data managers in Dar-es-
after the javascript executes. Note that in most cases the Salaam, Tanzania, over the telephone. We next describe each
javascript execution is fast enough that the user will never of the three test sites in more detail.
even notice the initial interface of Figure 2. For most users, the
Zambia National Center, Lusaka. The evaluation
only interface of which they will be aware is the interface of
conducted at the ZNBTS consisted of 6 participants in total,
Figure 3.
some were assigned tasks as country managers while others
were assigned tasks for provincial managers. The participants
consisted of a physician, biomedical scientists, financial
manager, information technology manager and data managers.
The participants were given the web address to the web tool
and were asked to complete the tasks. Two observers from
Georgia Tech were available to administer the evaluation and
provide help only if absolutely necessary. None of the
participants had prior experience with the website so they
were allowed to navigate around the website to determine how
to use the features. Internet connectivity available was via
satellite and evaluators used laptops with wireless
connections.
Zambia Copperbelt Center, Kitwe. In Kitwe, there was
dial-up internet connectivity, although it had not been
functioning for some months. After several attempts during the
first hour of our visit, the problem was discovered and fixed,
Fig. 3. If javascript is able to execute in the user's browser, it activates
selective disclosure and enables the user to hide or reveal sections making it possible for the evaluation to happen. There were 3
independently. participants consisting of a physician and two data managers
who were all familiar with the Excel-based tool. The dial-up
This design illustrates the benefits of unobtrusive
enhancement. Users without the more advanced technology connection had to be reset twice during the entire evaluation.
(in this case, browser support for javascript) still have access Tanzania National Center, Dar-es-Salaam. The usability
to a serviceable interface. The majority of users who do have study for the Tanzania data managers was conducted over the
the necessary technology, however, receive a more advanced phone. The observers called in from Lusaka, Zambia and went
user interface. through the tasks on the task list with each participant
individually; two participants had a copy of the task list by
email and were asked to indicate by speaking aloud how they
V. EVALUATION
were navigating the system. Fiber-based internet connectivity
A. Methodology was available in Tanzania which made for significantly faster
navigation.
Prior to the implementation of the web-based system, there
were several meetings between the CDC and the team from B. Study Design and Analysis
Georgia Tech. These meetings helped determine the The usability study consisted of 3 components: a pre-trial
specifications of the system including the design of the user- survey, an observed trial where data managers were asked to
interface, the preferred programming language and hosting perform a set of tasks and a post-trial survey.
and management of the website.
Pre-trial survey. A survey in form of a questionnaire was
To evaluate the effectiveness of the tool, we conducted a conducted prior to the trial of the web-based tool to assess the
series of studies involving the end-users. We selected a subset expectations of the levels of ease or difficulty of using the
of the PEPFAR countries to use for evaluation. All the data web-based tool. The survey was conducted via the web by
managers in the National Blood Transfusion System in the using a commercial web-survey tool that provided information
PEPFAR countries are currently expected to use the Excel about the date the survey was taken, the IP Address and
based reporting system and therefore already had some aggregate results from all participants etc. By conducting a
understanding of what kinds of data would be requested. web-survey, we were also able to assess the response time and
Zambia was chosen as the lead country for deployment and the availability of internet access. There were 9 participants in
evaluation due to the progress in using the Excel based tool the pre-trial survey, all from the Zambia NBTS; 7 from the
and their overall blood safety effort. The team from Georgia Lusaka and 2 from Kitwe. The pre-trial survey link was sent
Tech spent approximately two weeks, July 27 to August 8, out by July 21, 2008 and all surveys except one were taken
2008, in Zambia studying the local setting, usability of the between July 22 and July 27. 2008, prior to observers’ arrival
system, network characteristics and gathering qualitative in Zambia. All the participants had some prior knowledge of
feedback. The usability study component was conducted in the the Excel-based tool. The questions in the survey are shown in
284
Table I. Each participant was asked to rate the ease of each results indicate that user expectations of the difficulty of using
task on a Likert scale of 1-5 when appropriate, with 1=very the web based tool were above the actual difficulty
easy, 2=easy, 3=neither easy nor difficult, 4=difficult and encountered while using the tool. There was a reasonable level
5=very difficult. Questions 4 and 5 were given out to aid in of ease while working with the web-based tool. Participants
qualitative analysis. were also able to find some typographical errors and provide
concrete suggestions for changes.
Table I. Pre-trial Questions
The overall survey results indicate that the system was not
1 You have forgotten your password and need to log into very difficult to adapt to and use. Figure 4 shows the average
the web-based system. How hard or easy do you expect difficulty levels pre- and post-trial for each site. From Figure
to find this task? 4, it is apparent that the difficulty experienced by users in
2a You are only responsible for reporting some of the data, Zambia was significantly lower than what they anticipated.
while other users report the remaining data. You need to
enter the data for which you are responsible without Pretest Tanzania Zambia
disturbing the other data. How hard or easy do you
expect to find this task? 3.0
2b At the close of a deadline, you have only partial
information to report. How hard or easy do you expect to 2.4
find this task?
3 Previously entered data is wrong, and you must provide
1.8
the correct information. How hard or easy do you expect
to find this task?
4 What are some of the advantages you expect from a 1.2
web-based system compared to the old spreadsheet tool?
5 What are some of the disadvantages you expect from a 0.6
web-based system compared to the old spreadsheet tool?
0
Trial and post-trial survey. Discount usability techniques Task 1 Task 2 Task 3
were used to evaluate the usability of the system and the
Fig 4. Pre- and post-evaluation comparisons show tasks were as or easier than
interfaces. This evaluation provided a means of obtaining expected.
insights about usability problems and possible solutions. A
task list was used to obtain quantitative and qualitative The results from Tanzania, show less ease in using the
measurements of the ease of use of the tool. When participants system; this effect may be attributed to the additional
completed all tasks, they were asked to rate the ease of use of difficulty of having to evaluate the system while on the phone
the system on the same rating scale as in the pre-survey. This with the observers. Also, the evaluators may have felt some
evaluation occurred immediately after each participant pressure to accomplish the tasks quickly because the observers
completed all tasks, during the observers visit to Zambia. The were waiting for them to complete each task while holding on
list of tasks to be completed by each evaluator is shown below the phone line.
in Table II. In table III, the average difficulty for each task is given.
Table II: Evaluation Task List Overall, the pre-trial results are higher than the post-evaluation
results, this supports the notion that the interface was well-
1 Your username is country/province@blood-safety.org designed, to the extent that such a self-evaluation can assert.
and the password is the word "password''. Use the The "Export" task, not mentioned in the pre-trial survey was
system to change the password to "country/province''. perceived to be more difficult than the other tasks; it involved
2a Enter the regional information for the current quarter. transferring data from the system to the local computer in a
standard format (XML) which may have been unfamiliar to
2b You are entering data for the current quarter for your many evaluators.
country/province. You only have data for all the non-
Table III. Mean Scores for Selected Tasks
numeric fields. Enter and save this part of the data.
3 There is an error in the HIV prevalence field. Please
Task Pre TZ ZM ZM-NTL ZM-CB
correct it.
4 Export the country/province data to an XML file and 1 2.56 2 1 1 1
view the data using Excel.
2 2.22 2.5 1.17 1.33 1
C. Results
3 2 2 1.29 1.67 1
Analysis of the results obtained from this evaluation, shown
below in table I and figure 4, revealed that most evaluators Export - 3 2.5 3 2
rated most of the tasks between scale level 1 and 2. These
285
D. Qualitative findings of the network infrastructure1 . Because the same infrastructure

was used for all Internet access and, thus, lengthy page load
Utility and maintenance. Respondents to the pre-trial
times were common for nearly all web sites, participants were
survey felt that the web-tool would:
somewhat conditioned to expect this delay. Nonetheless, their
• Result in more timely reporting. dissatisfaction levels suggested that it would be worthwhile to
• Be quicker and less tedious. investigate opportunities to reduce the page delay.
• Be more cost-effective. The revised design significantly reduces the perceived
• Reduce the number of errors. initial loading time through the use of Asynchronous
• Be more user-friendly than a spreadsheet. Javascript and XML (AJAX) technology[7]. With AJAX, a
• Permit easier correction of data entry errors. web page can continue to interact with the server after it is
initially loaded without requiring explicit action on the part of
• Make it easy to share data between users.
the user. By using AJAX, the initial contents for each page can
All respondents were strongly of the opinion that a web- be substantially reduced so that the page loads more quickly.
based tool would be considerably superior to a spreadsheet- While the user views or interacts with the page, it continues to
based report. retrieve additional information from the server. By combining
AJAX functionality with selective disclosure, the revised
Security. On one hand, some users felt data security is a
design provides quicker initial page loading in a manner that is
concern since the data is no longer locally stored and
seamless for users.
vulnerable to malicious attacks; on the other hand, keeping
data on a server insures against data loss in the case of Figures 5 and 6 illustrate the steps involved in loading a
problems with local computers. In any case, it was clear that page using the approach of the revised design. The server first
users were concerned about how secure the data would be. supplies an initial web page that is almost completely empty.
However, the individual sections on the page are marked as
Functionality. While the original purpose of the tool was to
hidden. Even though they contain no content, users do not
allow for data aggregation for the purpose of generating
perceive a problem because the sections are hidden.
PEPFAR-mandated reports, data managers were quick to see
that having a real-time aggregate picture would greatly
enhance their ability to manage blood collection and
utilization. They asked for data analysis, such as historical
trends and comparisons of data from different regions to be
included in the functionality of the system.
Network constraints. Given the state of network
infrastructure in Zambia and other participating countries, the
potential pitfall of relying heavily on the internet was noted by
several data managers. Would pages load too slowly? Would
the system be able to handle connectivity going up and down?
While Dar-es-Salaam has a fiber-based internet connection,
there is no fiber connecting Zambia to the rest of the world.
Thus, all internet traffic goes via satellite, imposing a strong
constraint on bandwidth, and perhaps even more significantly,
on latency. Indeed we did not notice much difference in
response times between Kitwe, with a dial-up connection, and
Lusaka, with wireless connectivity and a dedicated satellite Fig. 5. The web server initially supplies only the skeletal framework for the
connection. full page; however, since the framework contains minimal content the network
can transfer it very rapidly and present the user with the appearance of a full
interface.
IV. REVISED DESIGN As soon as the initial page has loaded, it begins using AJAX
requests to retrieve the actual content for each section. During
Usability testing of the initial prototype indicated that, in
this time, however, the user is still viewing the initial page and
general, the application and its interface were suitably easy to deciding how to proceed. In many cases, by the time users
use. None of the tasks were rated as difficult or very difficult take their first action by selecting to show one of the sections
by any user. Comments made by participants during the study,
on the page, the content will have been retrieved from the
however, indicated that one of the most unsatisfying aspects of server and can be displayed immediately. When that is not the
the system was the time required for the initial loading of each
case, a progress bar tells the users that the interface is acting
web page. The main cause of the lengthy delay was the quality on their requests.
1While on site in Zambia, the developers were able to make measurements of network bandwidth and latency from several locations over 24-
hour periods. The tests measured performance from a local computer in Zambia to the prototype web site operating by a commercial web
hosting company in New York. The tests showed available bandwidth ranging from 12 to 57 kb/s to New York and 88 to 116 kb/s from New
York. Mean latency ranged from 5.1 s to 6.8 s.
286
use. Because the database captures all inputs to the system as

new data insertions (rather than updates to existing data), there
is no need to lock database access to avoid concurrent updates.
Such updates are problematic for any Internet-based
application, and would be especially challenging in an
environment with sporadic connectivity. The system's
approach resolves conflicts during report generation rather
than database modification. These features ensure that the
system meets the needs of both the sponsoring agency and the
local health officials.
V. DEPLOYMENT TIMELINE
An initial prototype for the blood safety system was created

Fig. 6. While the user decides which section to view, the network is able to during the spring of 2008. Field trials and evaluations took
complete the transfer of the hidden content. When the user choses to reveal place during the summer, and the revised design was created
that content, it can be immediately displayed. in the fall.
The timeline in figure 7 compares the perceived and actual Beginning in the first quarter of 2009, the system will
page loading times for the initial and revised designs. Even transition into production in 14 countries which are part of the
though it actually takes longer for the entire page and all its PEPFAR program. Discussions are currently underway with
content to load in the revised design, the perceived load time the World Health Organization to deploy a version of the
is less. system worldwide.
non-AJAX design
VI. DISCUSSION
AJAX design Overall survey results indicate that a web-based system was
highly desirable and the system developed here was easy to
adapt to and use. This finding is in line with other experiences
with web-based tools for health projects in the developing
time
world[8]. A very significant observation from our survey is the
t1 t2 t3
need to incorporate local users in the preliminary design
initial page full page is hidden elements
framework can loaded; user are transferred; phases of a project. The original goal of the system was to
be displayed; perceives page page loading is serve as a data collection tool for CDC data managers in
user perceives loading as fully complete Atlanta; while conducting surveys, we discovered a gap in the
page loading as complete
complete
stakeholder perspectives as the local users perspective had not
previously been incorporated in the design. This highlights the
Fig. 7. Although the AJAX design requires longer to load because of extra necessity for human-driven design and research practices as
content, it allows the web browser to present a functioning page sooner than noted by Brand & Schwittay [9]. Local data managers stressed
the non-AJAX design; to the user, this difference gives the perception of faster
loading to the AJAX design. their preference for a tool that would also help in blood
management by analyzing and tracking recorded data. The
One consequence of the revised design is that it requires final design incorporates perspectives of both stakeholders; the
javascript. There is no way to make the page fully functional CDC data mangers and local users. This reflects several
when the user's browser does not have javascript support, so dimensions of Information and Communications Technology
unobtrusive enhancement is not possible. Based on for Development (ICT4D) as outlined in work by Tongia et al.
interactions with prospective users, however, it was [10] by addressing the original goals of the CDC’s Excel-
determined that improving page loading times was more based data entry system while also meeting the needs of the
desirable than supporting non-javascript capable browsers. In local users in the developing countries.
the future, should support for non-javascript browsers be
necessary, a separate user interface could be developed. We highlight two aspects of our design that give us
confidence in its long-term sustainability and might serve as a
Another key element of the revised design directly model for similar projects in the future.
incorporated requests from the blood safety staff in the remote
locations. In many cases, local staff wished to collect and track Data flow constraint. The data entry requirements of the
information in addition to that required by the CDC. Also, system should match the data collection, i.e., if the system
reporting periods were sometimes more frequent than the expects too much data, it will be hard to maintain and impose
quarterly reporting required by the CDC. By supporting an overhead on data collection. In technologically advanced
collection of local information at locally-defined reporting countries, there is often an abundance of easily available data
periods, the system provides a much greater incentive for its and thus it is possible to deploy systems that use vast amounts
287
of data collected rapidly. For a system to be sustainable, support HIV treatment in rural Haiti,” BMJ (Clinical Research
however, its data requirements should match the local data Ed.), vol. 329, Nov. 2004, pp. 1142-6.
flow, even if this necessitates a less complex system.
[3] J. A. Blaya, S. S. Shin, M. J. Yagui, G. Yale, C. Z.
Flexible design. The second most important aspect of our Suarez, L. L. Asencios, J. P. Cegielski, and H. S. Fraser. “A
system design, closely related to the first, is its flexibility, both web-based laboratory information system to improve quality
in data entry and analysis. Web-based systems are, in general, of care of tuberculosis patients in Peru: functional
designed to minimize the constraints inherent with variable requirements, implementation and usage statistics.” BMC Med
rates of data entry. As we observed with the spreadsheet based Inform Decis Mak. 2007;7:33.
data collection system, the potential for errors seemed to rise
[4] Lani Marquez. Are Web-Based Platforms to Support
as the numbers of users, and the numbers of copies in
circulation, increased. The web-based tool is less vulnerable to Quality Improvement Feasible in Developing Countries?
Experiences from Ecuador, Honduras, Nicaragua, and
these errors: delays between entries will not have as great an
impact on the overall quality of submissions. Because of this, Rwanda, 2007. http://www.qaproject.org/news/ISQUA07/
ISQua2007%20Marquez%20handout.pdf.
users will be more confident in the database, and more likely
to use it as a decision-informing tool. Our design goes further [5] World Health Organization, "Blood Transfusion Safety",
in that data managers can choose the frequency of entering 2004. http://www.who.int/bloodsafety.
data to match the frequency at which they receive data. For
example, in Lusaka we observed that blood collection data [6] Jeremy Keith, Bulletproof AJAX. New Riders Press.
arrived weekly (by fax) from each of the other centers. On the February 2007.
other hand, blood utilization data was not obtained so [7] Jesse James Garrett. "Ajax: A New Approach to Web
regularly and a month-long window was more realistic. The Applications". 2005. http://www.adaptivepath.com/ideas/
system allows users to choose views of the data for analysis essays/archives/000385.php.
and management. The views of interest depend on the local
context; e.g. in Zambia blood types have roughly the same [8] Jennifer Manne and Walter Curioso (2007).
distribution across the country, but the incidence of malaria “Technology and health care in the developing world.” Student
(and thus the need for blood) varies by both time of the year BMJ, December 2007. http://student.bmj.com/issues/07/12/
and region. Therefore, reports that show geographic and life/436.php
seasonal trends in demand for blood would be more relevant [9] P Brand, A Schwittay. The Missing Piece: Human-
than reports that indicate blood type distributions. Driven Design and Research in ICT and Development. In
Blood safety monitoring is an important special case of the Proceedings of the International Conference on Information
more general problem of monitoring and tracking large data and Communications Technologies and Development, May
sets which are updated frequently. The latter has many 2006.
possible applications in health, emergency relief, water and [10] R Tongia, E Subrahmanian. Information and
other basic resources, etc. In all these cases a web-based Communications Technology for Development (ICT4D) – A
distributed monitoring and aggregation solution seems worthy Design Challenge? In Proceedings of the International
of exploration, in spite of network constraints. We believe that Conference on Information and Communications Technologies
the insights gained from our development of a blood safety and Development, 2006.
monitoring tool could be applied more generally.
ACKNOWLEDGMENT
We are grateful to the Zambian National Blood Transfusion

Service, Tanzania National Blood Transfusion Service, the
Centers for Disease Control and Prevention in Zambia and
Atlanta, and the US President's Emergency Plan for AIDS
Relief. We thank Sridhar Basavaraju, Michael Best and
Gregory Abowd for helpful comments on earlier drafts of this
paper.
REFERENCES
[1] World Health Organization, "Fact Sheet No. 279: Blood

Safety and Donation." 2008. http://www.who.int/mediacentre/
factsheets/fs279/en/.
[2] H.S.F. Fraser, D. Jazayeri, P. Nevil, Y. Karacaoglu, P.E.
Farmer, E. Lyon, M.K.C.S. Fawzi, F. Leandre, S.S. Choi, and
J.S. Mukherjee, “An information system and medical record to
288
Dimensions of IT Literacy in an Arab Region: A

Study in Barkha (Oman)
Sherif M. Aziz
result of the adoption and integration of ICT at home, work,

Abstract— The transformational potential of the Information education and recreation [3].
and Communication Technology lies in the ability of the people
to make the best use of it. In fact, they need to be e-ready to To move towards a digital society, there needs to be a synergy
benefit from what is known as the digital society. The rapidity of of technology infrastructure, a skilled population, and an
technological advances is, in fact, pushing many developing
overall enabling environment. The enabling environment
countries digitally far behind. However, as of now, it is well
accepted that the digital divide is not merely about technology should not only have technological components and cost-
deprivation, but is also about many inter-related social, related incentives, but also aspects such as motivation and a
educational, cultural, political and economic issues. sense of purpose among the citizens.
Understanding its subtle nuances in the context of different parts
of the developing world is extremely important to address the In Oman, soon many government-to-citizen interfaces will be
problem in a meaningful way. This paper is about an action web based or through other electronic means. An Ubar Portal
research done in the district of Barkha in the Sultanate of Oman, is designated to be the main gateway to electronic services
through a pilot IT literacy campaign. The study reveals that, for offered by the public sector. Citizens will access the
evolving a digital society in an Arab region, issues such as
government via the Ubar Portal that links to other portals and
linguistic literacy and information literacy are far more
challenging than the much publicized computer literacy. The web sites hosted by the Government entities and enables
study also concludes that Oman is at the threshold of access to seamless execution of e-services. This portal is named after
technology but has challenges with regard to cognitive access, the ancient Omani city of Ubar, the “Atlantis of the desert”
content access and an enabling environment. and a main trading gateway into Arabian Peninsula. It is
envisaged that anytime, anywhere access to the portal can
Index Terms— Access to ICT, digital divide, IT Literacy, occur through multiple channels such as the web and mobile
Oman devices [2].
However, for the citizens to be able to derive benefits out of

I. INTRODUCTION all these developments, they need to be ‘ready’. In other
words, there is an urgent need for an overall societal
preparedness and citizen e-readiness to make the whole effort
I NFORMATION and Communication Technology (ICT) has
pervasively intruded into our everyday life. Perhaps, the
most recent picture of this pervasiveness is demonstrated by
successful and beneficial to the society. To achieve this, there
are many essential components, one of which is the
the aggressive way in which most governments are information technology (IT) literacy among the masses. Apart
automating their service deliveries and citizen interfaces from providing ICT infrastructure, IT literacy is equally
through what is popularly known as e-government. important to alleviate what is popularly known as the ‘digital
divide’.
The Sultanate of Oman, like all its neighbours, is pursuing a
national e-government initiative since the year 2002, along Information technology is now taught in all schools in Oman
with a wider Digital Oman strategy [1]. The country’s as a separate subject from grade one to grade ten in basic
Seventh Five Year Development Plan (2006-2010) lays education. In grade eleven, the ICDL (International Computer
emphasis on upgrading the information technology sector by Driving License) is a required course and each student has the
implementing the national strategy for Oman’s digital society opportunity to learn IT skills [2].
[2]. The Information Technology Authority of Oman, which
spearheads these initiatives, provides a definition for ‘digital Nevertheless, considering the fact that nearly one-third of the
society’ as a modern, progressive society that is formed as a population is beyond the usual ages of formal education 1 ,
there is a need for alternative means of achieving basic citizen
readiness to benefit from the ongoing digital society
initiatives. Perhaps, some non-formal means of achieving IT
The pilot IT literacy campaign part of this study was fully supported by
the Middle East College of Information Technology, Muscat.
1
Sherif M. Aziz is with the Public Establishment for Industrial Estates, Estimated from the data in the Statistical Year Book- 2004 published by
Sultanate of Oman (e-mail: Mohammed.Sherif@peie.om). the Ministry of National Economy.
289
literacy would yield better reach and coverage in the shortest tested questionnaires, a follow-up telephone interview of the
possible time. participants, and focus group exercise to evolve the national
strategy.
In the above background, a group of volunteers under the The focus group consisted of a homogenous group of ten
leadership of the author, and with the support of an persons who had equally fair idea about the IT literacy
educational institution, carried out a pilot IT literacy campaign campaign. They had sufficient understanding about the
in the district of Barkha in Oman. This campaign was planned concepts of e-government and had basic awareness about the
as an action research, which provided ample insights for Digital Oman initiative.
evolving a national strategic framework for achieving IT
literacy. The IT literacy campaign as such is not a new concept.
However, the locale under consideration was not very familiar
Objectives of the research: with such non-formal and non-governmental interventions.
Preliminary assessment of the situation demanded an action
The overall objective of the research was to evolve a basic research, which follows an ‘intuitive proactive’ approach as
strategic framework for a national IT literacy campaign, with against a ‘rational reactive’ one [4]. In the first case,
a view to enhance citizen e-readiness in the context of the practitioners who know or think they know, what needs to be
Digital Oman initiative. done, implement an intervention program first and then see
how well it is progressing. In the latter approach, the
Specific objectives: researcher examines what is occurring with a specific focus on
some known problem and then draws up a program to react to
To design and carry out a pilot IT literacy campaign in a semi- what has been discovered.
urban region as a test case through an action research.
II. THE REAL ACCESS
To evaluate the lessons learned, and other analytical findings
resulting from the aforesaid objective, in the background of The information (technology) era demands a higher degree of
relevant international thinking and literature on the issues of literacy and complexity of skills to function effectively in
access to ICT, IT literacy and digital divide. what is known as the “digital society”. An OECD adult
literacy survey [5] reported that, across 20 countries, one in
To draw inferences from the aforesaid stages, and to evolve a four adults who participated in the study do not possess the
basic strategic framework for a national IT literacy campaign. necessary skills to manage in today’s world. Findings point to
large disparities in the average level and distribution of
Scope of the research and its limitations: literacy skills, both within and between countries. The report
pointed out the need for policies to be directed at the work
With regard to the IT literacy campaign, there was a larger place and family settings so that an overall societal
community of interested parties such as the Wali 2 and his preparedness for the information age is possible. Lack of
office, local community leaders, schools where the initial literacy skills does cause difficulties for living, working and
campaigns were conducted and, of course, the national even survival [6]. It is not the work place alone, which
government. Therefore, from concept to implementation, demands such skills. Even one’s own healthcare demands
subtle socio-political issues had to be handled carefully to skills to access and use appropriate information.
maintain the interest and motivation of all the key actors. In
such a background, apart from coping with the inherent Being information-literate has been defined as "having the
methodological complexity, the very management of the ability to recognize when information is needed, then to be
campaign, data collection, and the research in general able to locate and evaluate the appropriate information and
demanded limiting the IT literacy campaign to a pilot action use it effectively"[7]. These abilities have always been
research in a limited geographical region. The pilot literacy important to lead a successful life. However, as of now, IT
campaign also had to be short to maintain the morale and skill has become an important component of information
motivation of the volunteers. In the light of the above, the literacy, given the fact that the channels and quantum of
field study was limited to the Wilayat of Barkha. information have multiplied phenomenally on account of the
rapid technological advancements in the field of information
Methodology: and communication technology.
The question of digital divide
A mixed methodology was adopted for conducting this
research. This included a literature survey, selection of Digital divide is a term that emerged in the 1990s as an
appropriate theoretical framework, structured and unstructured expression of concern about the distribution of access to and
interviews, action-research through an IT literacy campaign, a benefits from advances in communications and information
survey among the participants of the campaign using pre- infrastructure. Digital divide, in simple terms, is the
differentiation between those who have access to IT and
2
Wali is the administrative head of a district, which is known as the digital information (particularly, the Internet) and those who
Wilayat
290
do not [8]. But, with rapid technological advances, technology The foregoing, however, appears to be idealistic, and is more
itself is a moving target, thereby enhancing the rate of further close to a developed country situation. In a developing
divide between what could be called as the ‘information country context e-readiness requires to address the issues of
haves’ and the ‘information have-nots’. basic literacy, poverty, health, and other social aspects first. In
any case, digital divide is increasingly a matter of universal
The British Council’s info@UK website glossary defines concern, as the personal computer and the Internet have
digital divide as the gap between those who can effectively become critical to economic success and personal
use new information and communication tools and those who advancement. Therefore, understanding its subtle nuances in
cannot. This divide can be the result of one or many different the context of the developing world is extremely important.
factors. Being a member of a socially excluded or
disadvantaged group, or lack of necessary skills to participate The Access for Opportunity Framework
are all important factors [9].
ICT for development has traditionally focused more on
The average OECD country, which has 11 times the per capita technology infrastructure issues. It is true that, for most
income of a South Asian country, has 40 times as many developing nations, creating physical networks is very much
computers, 146 times as many mobile phones and 1036 times beyond their means. However, there are a few other cash-rich
as many Internet hosts [10]. North America is home to only 5 developing nations such as those in the Persian Gulf region,
percent of the world’s population, but represents 28 percent of which are examples to the argument that technology
the world’s Internet users. The Middle East with about 4 infrastructure is a necessary, but not a sufficient condition for
percent of the world’s population has only 1.4 percent of the access to various aspects of a digital society.
world users [11].
It is in the above context that the Access for Opportunity
The foregoing picture is with regard to the existing disparities. Framework, which is detailed in the United Nations Global e-
The rapidity of technological advances is, in fact, pushing the Government Readiness Report 2004 [15], becomes relevant to
‘have not’ countries further behind. But the disparity with study the situation of developing countries such as the
respect to e-readiness is not merely due to technological Sultanate of Oman. This approach emphasizes the aspect that
components. As of now, it is well accepted that the digital the end goal should be access for opportunity, rather than
divide is not merely about technology deprivation, but is also access to ICT.
about many inter-related social, educational, cultural, political,
and economic issues. An ITU opinion poll had shown that The Access for Opportunity Framework lays out the
77.3 percent of those covered by the survey citing poverty as a parameters of real access. The salient points of this concept
major barrier towards integrating to the information society, are:
followed by lack of education (76 percent), and lack of • Physical access to ICT is only the first step towards
infrastructure (72.8 percent) [12]. The UN report [13] states: building real access.
• Access must be blended with relevant and culturally
As countries progress in employing ICTs appropriate content for onward transmuting into
for development, the challenges knowledge
underpinning inequality in access have • The blended knowledge is processed and utilized to
shifted from connectivity issue to create opportunity for economic and social
encompass a wide array of economic, empowerment
social, cultural, and language
barriers…The issue of a digital divide is This framework goes further to outline the Model of Access-
essentially one of a disparity in real Acceleration, which maintains that:
access which is inequality in both
physical access to ICTs and the ability, Technology infrastructure needs to reach
know-how and the culture to use the some threshold level in a given nation
technology well. for access to start accelerating, but only
as long as other access-supporting
The Centre for International Development at Harvard economic, social, educational, and
University defines an e-ready society as one that has the cultural elements are in place [15].
necessary physical infrastructure (high bandwidth, reliability,
and affordable prices); integrated current ICTs throughout Thus, a nation, which is advanced in terms of educational,
businesses (e-commerce, local ICT sector), communities cultural and social support structure, is in a better position
(local content, many organizations online, ICTs used in when the technology infrastructure reaches the threshold
everyday life and also taught in schools); the government (e- access point.
government); strong telecommunication competition;
independent regulation with a commitment to universal The Access-Acceleration Model is based on the basic premise
access; and no limits on trade and foreign investment[14]. that there is a threshold level of `real access’, which when
291
attained, will allow the subject to enter a state of accelerated implies, a felt need for information, a sense of purpose, and
access. The threshold level of access comprises ICT the ability to use information, all of which put together means
infrastructure, penetration of technology, government the realization of opportunities and empowerment due to
leadership, education directly supportive of technology, information availability.
culture of technology, and global language of technology
[15].This model goes further to make the following Access But at least the process of seeking information is a multi-
Framework definitions: dimensional technology issue, one dimension being the basic
or even rudimentary ability to use the related technology.
Access threshold : A static cut off point where the mix of However, the Internet, which is currently the ubiquitous
technological, economic and social systems begins to blend means of global information sharing, is still substantially a
together to provide a synergy which allows positive feedback text-based technology. Therefore, all the aforesaid aspects
cycles of technology for the utilization of knowledge, which become significant only if the person is fundamentally literate
allow for greater economic and social opportunity. Beyond the in a language, which is used in the technology driven
access threshold a country enters the accelerated access phase. information network or environment.
Accelerated Access: A dynamic changing optimal mix of Therefore, being IT literate would have three almost
technology, educational skills, economic and societal hierarchical components viz. (a) basic literacy in a language
conditions, which come together at any given point in time to (b) technical literacy of being able to use the related
produce threshold access, which in turn leads to real access. equipment or computer literacy and (c) being information
literate as explained above (see Fig.1).
Real Access: The equilibrium level of access, whereby an
individual has the required availability of technology,
educational skills, culturally appropriate and relevant content
in his/her language of choice at an affordable cost.
The UN report [15] states that the vast majority of countries Information Literacy
have not yet reached real access or even a threshold level of
ICT infrastructure needed to enter the acceleration stage, and ---------------------------------
most are lagging in the other critical access elements.
A detailed study, which will consider a possible maturity Computer Literacy
process to real access, was beyond the scope of this research.
Therefore, in this study, the concept of real access is viewed ---------------------------------
in a very simplistic manner as an equilibrium of availability of
technology (basic telecom and internet access), education and Linguistic Literacy
skills, culturally appropriate and relevant content in the
language of choice at an affordable cost, and the motivation
(sense of purpose) to use.
IT Literacy redefined
Fig.1 Hierarchical components of being IT Literate
The information age is largely characterized by the
technological innovations that have had far reaching
In regions such as the Persian Gulf, and therefore in Oman,
consequences in the way people work, communicate and do
the first component itself is a matter of concern in that the
business. In fact, the convergence of telecommunication,
abundance of information available in the so called world-
broadcasting and computers has had significant impact on the wide web itself is not in the ‘language of her/his choice’,
society, which is comparable with the previous major societal which is Arabic. On the basis of the above concept, it can also
changes such as the industrial revolution or the advent of the be said that being information literate also indicate a general
printing press. As with the advent of the printing press, the maturity with regard to the human capital, not necessarily in
technological age requires the development of higher levels of terms of mere educational qualifications, but essentially the
skills in the society [16]. It is not enough to be literate in the positive impact of education and technical skills to lead better
sense of being able to read and write in some language. If a quality of life. In other words, such societies will have an
person is to survive in the information age he needs to be ‘information maturity’ to make use of the information and
‘information literate’, which, simply put, means that he has knowledge around them to lead a better life.
the skills to seek, understand, and use information 3 . But it also
3
The American Library Association Presidential Committee on needed and have the ability to locate, evaluate and use effectively the needed
Information Literacy 1989 defines Information Literacy as: “ To be information (ALA, 1989 p.1)
information literate, a person must be able to recognize when information is
292
The foregoing is the other conceptual framework, which has (c) The Internet Bus: The Internet bus was used for promoting
been evolved and used in this research. the campaign, for providing follow-up support to the already
trained participants, for providing basic training to small
III. IT LITERACY CAMPAIGN groups assembling in souks 4 , and to reach out to the interior
villages.
Citizen e-readiness is a basic necessity for any digital society
initiative to be successful, and IT literacy is one of its primary The volunteers were divided into three groups – (a) publicity
components. Perhaps, for quite some time, it could have been group (b) trainers group, and (c) follow-up group. The overall
taken for granted that such IT literacy will come through the approach was that of ‘training of trainers’ within the
formal education system. However, at least for the time being, community, which was expected to result in a chain reaction
this will be inadequate considering the fact that about 34 and speedy coverage.
percent of the Omani population has already crossed the usual
age of schooling [17]. Therefore, there is a necessity for some Lessons learned during the campaign:
non-formal means of making the population IT literate.
However, since this is an issue of the masses, a hit and try Many observations were recorded, discussed on a daily basis
approach can result in a strategic catastrophe. Therefore, such and corrective measures taken. The major issue was that the
a non-formal intervention is best implemented after a pilot campaign could not attract significant number of men and
project so that, based on the lessons learned; a nation-wide women above the age of 30 years, though it was primarily
strategy can be evolved and implemented in the most effective aimed at attracting people who have already passed the usual
manner. age of formal education.
It is in the above background that the IT literacy campaign at Therefore, the Internet bus was deployed to the souks to
Barkha was organized as an action research. attract aspirants from higher age groups. Even with sufficient
cajoling by the volunteers, older citizens were not able to
Overall organisational set-up: relate themselves to this activity. By the second week of the
campaign it was more or less clear that the campaign has to
The campaign had to be legitimized by governmental reach out to the villages in a more intrusive way to attract
patronage. This was obtained from the district administration. older women and men.
This campaign had the support of the Middle East College of
Information Technology, Muscat, by way of providing student Converting the campaign into a family affair – a minor
volunteers and a fully equipped Internet bus. breakthrough:
A core team headed by the researcher, five volunteer leaders, The whole campaign actually worked on the concepts of
and three technical support staff was formed with full voluntarism and local community participation. In order to
autonomy to plan and organize the campaign. About 60 break multi-dimensional (social, cultural and gender issues)
volunteers actively participated in the various activities of the hurdles in reaching out to the household level, the volunteers
campaign. were encouraged to take the Internet bus to the villages, where
she/he can impart basic awareness training to mostly people
Design of the campaign: who are her/his relatives and neighbours, all to be done by
her/him without any outsider being there. This approach
The IT literacy campaign was structured in the following generated more active participation from women, but still
manner: could attract only a very few above the age of 40 years.
(a) Two days class-room sessions: This included introduction Survey among the participants:
to the era of computers, creating a sense of purpose for
gaining IT literacy, and creating awareness about the A survey was conducted among the participants of the IT
opportunities in the information era and in an e-government literacy campaign. A pre-tested questionnaire was
environment. This was followed by general familiarization administered to all those who attended the campaign. First set
with computer. Trained volunteers gave lectures quoting real of data was collected prior to imparting training and at the
life situations. Subsequently, small groups were formed to time of registering for the campaign. The participants were
provide hands-on training. asked to fill and return the second part of the questionnaire
after the training. Subsequently, 45 days later, a telephonic
(b) Walk-in-and-use support for the next two days: Two interview of the participants was done to cross-check certain
volunteers from the neighbourhood of each venue were made data as well as to get some supplementary information such as
available for about three hours per day to provide further actions taken after the campaign.
helping hand to those who have already attended the
campaign on the previous two days. 4
Old market place with very narrow streets, typical of the erstwhile Arab
commercial centers
293
TABLE II
The empirical data obtained from the aforesaid survey as well MOTIVATION FOR ATTENDING THE CAMPAIGN
as from the published documents were used to analyse the
issue of citizen e-readiness in Barkha, in the background of Sl. No Motivational Number % to total
the UN’s Access for Opportunity framework and in the factor participants
context of the Digital Oman initiative. 1 Acquire computer 64 31.2
skills
Details of the analysis and findings of this primary data are 2 To get a decent 74 36.1
given under the Study Findings in section IV. job
3 Can’t live without 55 26.8
computer skills
IV. STUDY FINDINGS
4 Upgrade the skills 20 9.8
Though the campaign had targeted attracting more of those
who have passed the age of formal education, most of the 5 Others 10 4.9
participants (55.1 percent) were from the age group of 16-25 Note: Figures are not mutually exclusive.
years, followed by the age group of ‘up to 15 years’ (Table I).
A rough comparison of this with the age distribution in the Education
national population reveals that the campaign could not attract
all the segments of the local population representatively, the Majority (30.1 percent) of the participants had education up to
maximum distortion being at the two upper age-groups. the 12th class, followed by those with 10th class (14.9 percent)
Incidentally, the first age group seems to be properly and 11th class (16.6 percent) education (Table III).
represented.
TABLE I Nearly 60 percent of these participants with higher secondary
AGE DISTRIBUTION education did not have computer skills. In fact, no correlation
Corresponding was found between the level of education and computer skills
Age representation or degree of awareness about Digital Oman or e-government.
Sl. Number of % to
Group in the national
No Participants total TABLE III
(years) population (%)
* EDUCATIONAL BACKGROUND
1 Up to 74 36.1 34 Sl. No Level of Number of % to total
15 schooling participants
2 16-25 113 55.1 22 (class)
3 26-40 14 6.8 25.2 1 12 62 30.2
4 More 4 2.0 18.8 2 11 34 16.6
than 40 3 10 44 21.5
Total 205 100.0 100.0 4 7 to 9 51 24.9
Note:* Approximate figures computed based on the national population
5 Up to 6 14 6.8
statistics, sourced from the Oman Census Survey Report 2003. The total Total 205 100.0
population is 2.34 million with more than half a million expatriates.
From the above observations, it is inferred that the formal
Apparently, computer is still seen as a modern tool meant for school education in Oman up till now has not made significant
the younger generation. This aspect, seen along with the fact contribution towards general IT literacy. Further, about 34
that more than 30 percent of the participants had higher percent of the Omani population is already outside the scope
secondary education (see Table III), indicate that many of of formal education. Therefore, the non-formal means of an
them were job seekers trying to acquire more computer skills. IT Literacy campaign was an appropriate intervention.
This is also evident from the higher representation (55.1
percent) from the relatively more active job seeking age group Literacy
of 16-25 years.
An analysis of all the hierarchical components of being IT
The expressions of motivational factors for attending the literate (see Fig 1) has been attempted here. The Barkha
campaign reinforce the aforesaid inferences (Table II). survey revealed that 87.8 percent of the people could read and
write Arabic (Table IV). This is fairly a high level of literacy
rate in a developing country. Many Arab countries are likely
to have such high Arabic literacy, as reading the Holy Quran
is almost a must from the religious point of view.
294
matrimony are all relevant contents. The relevance of these

TABLE IV contents also depend on the maturity of the services offered by
LITERACY the different actors in these sectors, be it in the private or
public sector. For instance, an online shopping portal is of no
Nature of Literacy No. of persons % to total significance if the associated online payment and the delivery
including all (including mechanisms are not intact, or if the legal framework is not
family members family mature enough to support e-commerce.
of the members)
participants * The above observations should be seen along with the fact
Arabic (Read) 1471 93.5 that in the Barkha sample only 27 percent had basic computer
Arabic (Read & Write) 1381 87.8 skills. This is out of a sample population of predominantly
English (Read) 889 56.5 youth with nearly 60 percent having higher secondary
Basic Computer Skills 427 27.0 education.
Note: *Total number of family members covered – 1573
From the point of being IT literate, the last component is to
The published national statistics show about 80 percent become ‘information literate’ (see Fig.1). From the survey of
literacy rate among Omanis [18]. Obviously, this is their those who attended the campaign to acquire IT literacy, it is
ability to read and write in the local language, which is not possible to assess the status of their ‘information literacy’.
Arabic. In fact, the figure of 87.8 percent obtained from the Such a survey needs to be carried out differently among a
Barkha sample is almost in line with the national figure. The larger cross-section of the population. However, a sum total
survey also revealed that only 56.5 percent could read of all the above observations points to the possibility of weak
English. Those who had computer skills were even less – 27 information literacy among the population, no matter whether
percent. it is computer-based or not.
The English reading ability in the national population can be Access-for-Opportunity framework – A revisit in the
much below what is reflected in the Barkha sample, as this context of Oman:
sample had a poor representation of people above the age of
40 years. Further, a substantial number of participants of the The UN Access-for-opportunity framework detailed under the
campaign had higher secondary education as well. section REAL ACCESS talk about the importance of a perfect
blend of technological, economic and social systems to
The opportunity to use one’s own language on global provide a synergic situation for ‘real access’. Considering the
information network such as the Internet determines the extent various dimensions of this approach, a slightly modified
to which one can participate in the knowledge society. conceptual framework was evolved to include some finer
However, an estimated 87 percent of documents on the elements such as ‘motivation or sense of purpose’ among the
Internet are in English [19]. The UN Report [20] confirms that masses to make the ‘use of ICT a way of life’ (Fig.2). Such
English continues to dominate the cyberspace material. subtle nuances can look trivial but are often critical
components of many development issues involving a society.
Only 1.4 percent of the world online users are Arabs, whereas In this modified conceptual framework, the elements of
they represent 4.6 percent of world population. This is in spite ‘physical access’ and ‘affordability’ are considered together
of the fact that the Internet use in the Middle East grew 219 along with a via-media solution of ‘public access’ points, all
percent during the period 2000-2004. This is in sharp contrast of which put together can provide a synergic effect in
to the situation of 39.4 percent of world online users using enhancing the ‘access to technology’.
English content, where as only 5.4 percent of the world
population are English speaking [20].
Thus in the case of Oman, though Arabic literacy is high, the PHYSICAL ACCESS
Electricity
issue of ‘linguistic literacy’ has two aspects to tackle in order CONGNITIVE ACCESS
Telephone
Computer
to enhance IT literacy: Education
Internet
IT Literacy
(a) the language access barrier, with a substantial chunk of REAL

Public
Access
World Wide Web content being in English. CONTENT ACCESS
ACCESS AFFORDABILITY
Per Capita income

Culturally appropriate Competitive tariff
(b) limited web-content in Arabic, though all governmental Relevant
Language of choice
sites are primarily in Arabic language. ENABLING
ENVIRONMENT
Awareness
The foregoing lacunae are apart from the issue of ‘relevance’ Motivation/Sense of Purpose
of the web-content. What is relevant is related to the local

culture and other socio-economic conditions. In most Fig. 2. Modified Conceptual Framework for Real Access
developed countries, health, education, recreation and
295
Physical Access U.K 1.7 2.5 2.1 Well

Ahead
Primarily, ICT development is built upon the
telecommunication infrastructure. The UN e-Government In the entire Barkha sample only 31.7 percent of the
Readiness Report [21] shows strong linkage between greater respondents had a computer at home and only 16.6 percent
telecommunication access and higher states of e-government had internet connectivity at home. A closer look through a
readiness. However, in the access-for-opportunity framework, cross tab of the availability of computer and internet at home
this alone is not a satisfying condition for ‘real access’. revealed that, out of the 65 respondents who had a computer
at home, only 34 had internet connection as well. In other
But the aspect of ‘physical access’, which is a critical words, nearly half of those with computers at home did not
component of ‘real access’, comprises the availability of have internet connection (Table VII). Further analysis of the
telephone, computer and the internet. As of now, the need for reasons for not having internet connection revealed that only 2
and availability of some form of electricity is often taken for out of the 31 persons (6 percent) had expressed money as the
granted. The Barkha survey, for instance, revealed that 99.5 constraint for not taking internet connection. 3 persons (10
percent of the households covered by the survey had percent) expressed that their families feared cultural erosion.
electricity connection. The other related data are shown in Remaining 26 persons (84 percent) did not have any specific
Table V. reason, in a way indicating that they weren’t either aware or
did not have a real sense of purpose in having an internet
TABLE V connection. Most households procured computers, because
PHYSICAL ACCESS COMPONENTS some family members’ education demanded one at home, or
because they got inspired by neighbours, especially during
Sl. No Description Persons per technology festival promotions.
Published Barka
Data * Survey TABLE VII
1 Telephone 11.9 10 AVAILABILITY OF COMPUTER VS INTERNET AT
2 Computer 26.7 24.4 HOME
3 Internet 14.1 47.6
Note: * See UN Global e-Government Readiness Report 2004, p.96 Availability of internet at
home Total
Except in the case of internet penetration, the Barka survey
NO YES
figures more or less matches with the published data. A
comparison of these factors in the case of Oman with those of Availa
bility NO 140 0 140
a select few countries reveals that Oman is not at a
disadvantaged position in the Asian and Middle East Region of
(Table VI). comp
uter YES
TABLE VI 31 (47.7%) 34(52.3 %) 65
PHYSICAL ACCESS COMPONENTS – A COMPARISON at [ 18.1%] [ 100] [ 31.7%]
(PERSONS PER TECHNOLOGY) home
Total 171(83.4%) 34 (16.6%) 205

Country Telephon Compute Interne Remark
e r t s with Note:( ) indicate % to row total; [ ] indicate % to column total
respect
to Oman Another interesting observation was about the gender
Oman 11.9 26.7 14.1 - distribution of the participants of campaign. 58 percent of the
participants were women. Further, among the participants who
Singapor 2.2 1.6 2.0 Far attended the campaign with some basic idea about computers,
e Ahead majority (64 percent) were women. Here, it may be noted that
India 25.1 138.9 62.8 Far the male-female ratio in the national population is 1.02:1.
Behind
Syria 8.1 51.5 77.5 Ahead in Some of the foregoing observations thwart the general
Tele. perception that the conservative cultural values and gender
lines. Far inequalities are the deterrents to social advancements in this
behind in Arab region. On the other hand, the predominant picture that
other is emerging is that of general lack of a sense of purpose of
aspects having these facilities at home, as compared to material
Bahrain 3.8 6.2 4.0 Well constraints to own them. A general perceptional data obtained
Ahead from all the participants further strengthens the above
China 6 36.2 21.7 Behind argument (Table VIII).
296
TABLE VIII
GENERAL PERCEPTIONS ABOUT COMPUTER AND Affordability
THE INTERNET (AMONG ALL THE PARTICIPANTS)
Sl. Description Yes No Remarks The earlier analyses indicated that affordability or financial
No capability to own and access computers cannot be a matter of
1 I knew what 101 104 *Nearly 50 % did serious concern in the context of Oman. Oman with a per
computer and not have basic capita income of about US$ 13000 has roughly the 60th
Internet are awareness position among all the countries 5 . There are promotional
2 Was not 3 202 98.5% was schemes offering internet connection and personal computer
interested in interested; very together.
computers positive signal
3 Learning 31 174 Only 15 % However, for at least a small segment of the population,
computer is considered it affordability, especially in relation to the money required for
difficult difficult; positive their mundane priorities in life, could still be a problem. But
4 Money is my 37 168 Only 18% such people could be facilitated through public access points.
constraint indicated financial In the years to come, as more Internet service providers come
constraints into play, the internet and telecom tariffs are likely to be more
5 Internet 17 188 Only 8.3 % competitive.
spoils our thought so;
culture against the usual Public Access
perceptions about Private cyber-cafes are the only form of public internet access
the Arab region points available in this country. These are mostly limited to
6 Computer is 3 202 Very positive towns. There has been no significant attempt either from the
not a government or any social institution to provide public internet
necessity access points, at least to encourage the use of the internet.
7 Knows about 91 114 OK The concept of self-employed IT kiosks are under government
internet consideration, essentially to function as an interface with the
8 Know Gov 40 165 Poor awareness IT illiterates, once e-government services are rolled out. In
services fact, maintaining a public internet access point will be cheaper
delivery than maintaining even a small tree in this desert land. The
through question of physical access as explained earlier, affordability,
Internet and the possibility of public access points throughout the
9 The above 34/40 6/40 Positive within country are a set of components of the ‘real access’
will benefit those who were framework that need to be considered together.
me aware of From the above perspective, Oman can be considered to have
e-gov. reached an appropriate threshold level with respect to access
10 Going 29/40 11/40 Signs of lack of to technology (Fig. 3). This ‘access to technology’ would
personally to confidence in encompass physical access and affordability.
gov. offices is electronic services
better
11 One day I 28/40 12/40 Positive
will be forced PHYSICAL ACCESS
to use Has reached the threshold level
12 Ever visited a 53 152 Poor PUBLIC

ACCESS
website
Will be cheaper
13 Did you often 17/53 36/53 Poor; only 32% than maintaining
a small tree in
AFFORDABILITY
get what you appears satisfied REAL ACCESS
the desert
look for from Per capita income-60th

rank among countries;
these Reasonable telecom
tariff in the GCC region;
Survey do not indicate
websites finance constraints;
Promotional schemes to
14 Knows about 20 185 Poor awareness own PCs exist.
Digital Oman
15 Any step 25 180 Not enough
taken after initiative? Fig.3. Access to Technology in Oman
campaign
5
This is as per the PPP (Purchase Power Parity) Method. As per the Atlas
Note: Incidentally, 60 percent of the participants had higher secondary Method the per capita income of Oman is US$7830, with almost the same
education. global ranking.
297
Cognitive Access has strict filtering of pornographic material; and practically do

not have any other control mechanism.
Only 7.7 percent of the national population has post-
secondary education [18]. It was also shown earlier that in the
Enabling environment - Motivation or sense of purpose
sample population covered by the survey, the school
education has not contributed significantly towards IT
19.5 percent of the participants knew about e-government in
Literacy among the masses.
general, mostly through media reports about global
developments. Only 9.8 percent had heard about Oman e-
The overall picture obtained from the analyses under the sub-
government initiatives. Out of the 40 participants who were
headings Education and Literacy is that of a poor cognitive
aware about e-government, 29 persons still felt that going to
access situation, with weakness at the levels of linguistic
offices personally is better to ensure speedy processing. A
literacy, computer literacy and information literacy, to
look at Table VIII reveals that the responses to most aspects
participate and benefit from a world-wide information
connected with the individual and his capabilities are positive
network.
in nature (descriptions corresponding to Sl. no. 1 to 7). Others,
wherein government has to play an active role in popularising
In the absence of sufficient cognitive access, the digitization
or reaching out with the correct message with regard to e-
of government processes, which is part of the Digital Oman
governance, the people need to gain better awareness and
movement, is unlikely to have much impact on the quality of
confidence levels. But the fundamental attitude of the people
life of the wider cross-section of the Omani society.
is positive enough to mature into an information society,
provided the related government interventions are appropriate,
Addressing the issue of linguistic literacy can have two
timely and motivating.
possible approaches:
The ‘real access’ issue in Oman
(a) A long-term strategy of enhancing Arabic
content
All the above findings put together, it can be concluded that,
(b) A short-term solution of popularizing the
among the various components of `real access’, the Omani
learning of English so that what is available in
population is almost at the threshold of access to technology,
English in the world-wide-web is readily usable,
but are yet to reach an advantageous position with regard to
and relevant
cognitive access, content access, and the required sense of
purpose to partake and benefit from the digital society era
Even computer literacy can be considerably enhanced through
(Fig. 4).
a multi-pronged approach of:
(a) Strengthening computer education at schools and PHYSICAL

colleges ACCESS
Access to
(b) Non-formal means such as IT literacy technology has
reached the
campaigns, done with a missionary zeal. threshold level
COGNITIVE ACCESS
Poor; Efficacy of
However, attaining information literacy is a complex issue, education not reflected
in Barka sample
PUBLIC
ACCESS
Will be
which can be achieved only through a long drawn societal cheaper than
maintaining a
transformation at various levels. Language of one’s choice, small tree in
the desert
relevance of the contents, and the possibilities to directly CONTENT ACCESS AFFORDABILITY
benefit from such electronic interfaces are all factors, that Very limited culturally
REAL
ACCESS Percapita income-
motivate the population to be information literate. appropriate content and
language of choice.
60th rank among
countries;
But no apprehensions Reasonable telecom
about cultural erosion. tariff in the GCC
region; Survey do
Content Access not indicate finance
constraints;
ENABLING
Promotional
ENVIRONMENT schemes to own
PCs exist.
All the above analyses also reveal a situation of very limited Poor awareness ; some
awareness of global
‘Arabic’ and ‘culturally appropriate’ content. Most developments. Not
aware and confident
about local initiatives. No
government web sites, which are mostly in Arabic, do not evidence of motivation or
sense of purpose.
provide significant interactive or transactional benefit to the
users.
The survey among the participants of the campaign revealed Fig.4. Assessment of ‘Real Access’ Situation in Oman. Grey
that among those who have visited one website or the other boxes indicate challenges.
(i.e. 26 % of the Barka sample) only 32 percent felt that they
got what they were looking for.
Majority (91.7%) did not have any apprehension about
cultural erosion due to the Internet. This country, of course,
298
V. KEY CONCLUSIONS As is evident from the figure, what has been proposed is a
multi-pronged and holistic strategic framework, with the
This section tries to highlight the key conclusions of the study. following broad considerations:
The predominant picture that is emerging out of this study is a. There need to be a thrust in maximizing social
that the access to technology cannot be considered as a inclusion, which does not necessarily mean
deterrent to citizen e-readiness in Oman. However, the issue making everybody IT literate; but it would
of cognitive access capabilities need to be addressed from simply mean maximizing social inclusion
different angles and a non-formal means of training (such as through a multi-pronged approach, including the
the IT literacy campaign) is imperative to alleviate possible use of intermediary mechanisms.
digital divide. b. Those above the ‘job seeking’ age group have to
be reached out through mobile training facilities
Some of the specific conclusions are: in a sustainable and friendlier manner.
c. An appropriate communication plan is a
a. The issue of ‘linguistic literacy’ has two problems to prerequisite to create the right type of awareness
tackle. Firstly, the inability to use the relatively huge about all what is happening in the name of
English content available in the World Wide Web. digital society.
Secondly, though Arabic literacy in Oman is high, d. Attempts shall be made to involve Sheiks 6 as
the web-content in Arabic is limited. champions who would take the ownership of IT
b. Therefore, for the masses to be IT literate and derive literacy initiatives at the local level.
meaningful benefits of a digital society, the language
access barrier may be considered as a fundamental It may be emphasised that, even to achieve IT literacy and
deterrent. thereby e-readiness among the people, the overall
c. Addressing the issue of linguistic literacy requires a environment should be conducive and encouraging to create a
long-term strategy of enhancing Arabic content as sense of purpose among the people to gain these new skills.
well as a short-term solution of popularizing the This can be achieved if the related government interventions
learning of English. can bring demonstrable benefits that will touch upon the
d. The possibility of weak information literacy among quality of life of the average citizen.
the population has to be considered while evolving
digital society strategies.
e. The findings indicate that affordability or financial Barka model campaign at Recognise the need for social change and
capability to own and access computers cannot be a schools; Separate training rooms
for men and women; Reach out
preparedness; Maximize social inclusion;
Minimise gender disparities while achieving
matter of serious concern in the context of Oman. to the village neighborhoods and
family gatherings through mini-
targets; Use of ICT shall become a way of life.
Consider that Oman is at the threshold of access
internet bus; More public access to technology – if public access points and more
f. There is a general lack of sense of purpose for having points – some free, some as IT POLICY promotional schemes to own computers are
kiosks as self-employment National level policy; encouraged, access to technology is a non-
access to computers (say, at home), as compared to ventures; One or two full-fledged
internet buses, pre-arranged
Declared as an `IT
literacy mission’;
issue; Government’s main focus shall be on
enhancing cognitive access through a multi-
other material constraints to own them. connectivity and long duration
stay at each location as per a
Corpus development
fund; Empowered
pronged development strategy
planned and publicized supervisory body

g. The survey among the participants revealed that program
conservative cultural values and gender inequalities APPROACH
Focus on cognitive access
cannot be considered as deterrents to social TOOLS and enabling environment
IT Touch upon quality of life
advancements in the information era. It may, Adult education at
schools LITERACY aspect
Mostly Non-formal education
Internet Bus
however, have some effect on the rate of progress. Public access points Achievable targets against a
timeframe
IT kiosks
h. Oman can be considered to have reached an
appropriate threshold level with respect to access to Awareness training to government officials, Training of trainers approach; Enhance
community leaders (includes Sheikhs and elected ACTORS English and computer skills at schools;
technology. members) and key private entity officials, as a National Government
Local Government
Generate a sense of purpose and
prelude to mass campaigns- shall aim at motivation through demonstrable projects;
i. There is a poor cognitive access situation, with transforming their attitude and their way of dealing
with the public in the information era; Define their
Private entities
People
Immediate steps to enhance Arabic web
content – in terms of quantity, relevance
Community leaders
weakness at the levels of linguistic literacy, computer roles in the national IT literacy mission; Multi-stake
holder situation- source support from the private Volunteers
and reliability; Enhance trust
confidence in local web content and e-
and
sector and the academia; Incentives to community Academia services; Quality standards for government
literacy and information literacy. volunteers websites; At least one IT literate person in
every household in two years.
VI. RECOMMENDATIONS
Fig.5. Strategic framework for a national IT literacy campaign
In the light of the above findings, a focus group exercise was

conducted.
REFERENCES
The findings based on the analyses of the published data, field [1] http://www.ita.gov.om/ITAPortal/eServices/eoman_strate
data collected during the IT literacy campaign, and the focus gy.aspx
group exercise have been combined to evolve a strategic
framework for a national IT Literacy campaign (see Fig. 5). 6
Tribal chief
299
[2] ESCWA, National Profile of the Information Society in

the Sultanate of Oman. July 2007. Available from:
http://www.escwa.un.org/wsis/reports/docs/Oman-07-
E.pdf (Accessed 07 April 2008)
[3] http://www.ita.gov.om/ITAPortal/Info/FAQ.aspx
[4] E.C. Wragg, An Introduction to classroom observation.
London: Routledge, 1994.
[5] http://www.literacytrust.org.uk/database/internat.html
[6] E. Gayner, Back to basics: the role of preparing young
people for the information society. Available from:
http://www.emaraldinsight.com/0090-7324.htm
(Accessed 20 June 2005)
[7] American Library Association, Presidential Committee
on Information Literacy: Final Report, 1989.Available
from:
http://www.ala.org/ala/mgrps/divs/acrl/publications/white
papers/presidential.cfm (Accessed 18 June 2008)
[8] http://www.ntia.doc.gov/ntiahome/digitaldivide/
[9] http://www.britishcouncil.org/fr/ism-info@uk-
glossary.htm
[10] The World Bank, Information and Communication
Technologies: A World Bank Group Strategy.
Washington, D.C.: Global ICT Department, April 2002
[11] UN, Global e-Government Readiness Report 2004:
Towards Access for Opportunity. New York: Department
of Economic and Social Affairs, 2004, p. 107.
[12] http://www.itu.int/osg/spu/wsis-themes/survey/index.html
of Economic and Social Affairs, 2004, p.8.
[14] http://www.cid.harvard.edu/cr/gitrr_030202.html
of Economic and Social Affairs, 2004, p.76.
[16] E. Gayner, Back to basics: the role of preparing young
people for the information society. Available from:
http://www.emaraldinsight.com/0090-7324.htm
(Accessed 20 June 2005)
[17] Ministry of National Economy, Statistical Year Book.
Muscat: MoNE. 2004.
[18] Ministry of National Economy, Oman Census Survey
Report. Muscat: MoNE, 2003
[19] K. Taglang, “Content and the Digital Divide: What do
People Want”, as quoted in UN Global e-Government
Readiness Report 2004: Towards Access for Opportunity.
New York: Department of Economic and Social Affairs,
2004.
of Economic and Social Affairs, 2004, p.107
[21] _____, Global e-Government Readiness Report 2004:
of Economic and Social Affairs, 2004, p.94
300
Emergency Communication and System Design:

The Case of Indian Ocean Tsunami
R. Chen, J. Coles, J. Lee, and H.R. Rao
Abstract—On December 26th, 2004, the largest natural disaster

in recent recorded history took the lives of over 225,000 people
from over 40 different nations and displaced millions more. In an
increasingly global environment, these disasters are no longer
isolated phenomena and must be responded to with a global
perspective. The number of casualties from Indian Ocean
Tsunami is a testament to the failure of humans in mitigating
large scale emergency incidents and it is the consequence of
ineffective information sharing and communication among key
stakeholders. Using Activity Theory as the theoretical lens, we
explore the communication phenomena in emergency response Fig. 1. Overview of Emergency Communication
and we identify the major challenge facing communication
practices. The paper further explores the design, implementation, The 2004 Indian Ocean earthquake took place at 00:58 UTC
and management of advanced information technologies to address on December 26 2004. The earthquake was measured 9.0 on
the communication challenges.
the Richter scale 1,000 km away from the Andaman Islands. It
Index Terms—Emergency services, communication, caused a powerful tsunami that killed more than 225,000
information technology, globe people in eleven countries, including Indonesia, Sri Lanka,
India, and Thailand. The Indian Ocean Tsunami was regarded
as one of the deadliest natural disasters in history.
I. INTRODUCTION Communications in the emergency response of tsunami was
a failure. During the early hours of December 26th, 2004, such
E mergency management is “the process of gathering
resources and acting upon the problems immediately
during and after a critical incident” [1]. Emergency incidents
effective communication was not demonstrated on many fronts,
both national and international. Indian soldiers on the islands of
may be natural or man-made. Limited information, Nicobar and Andaman sent warnings to their mainland
unpredictable disaster development, short time window, and counterparts after getting hit a full two hours before it struck the
high operation complexity renders the management of Indian coastal region. These warnings went unheeded and no
emergency incidents a challenging task [2, 3]. In order to evacuation orders were issued for Sri Lankan or Indian coastal
mount an effective response to an emergency situation, it is regions resulting in the loss of tens of thousands of lives. The
critical that the key actors communicate effectively and timely head of Thailand’s Meteorological Service refused to issue a
so as to exchange task critical information and form tsunami warning despite the looming danger because he didn’t
collaborative response activities. While communication has want to cause unneeded panic [4]. News of the true damage
been increasing recognized as the key to success emergency done to Aceh Indonesia was not known for hours after the
response, the practices in the field are far from satisfactory and initial impact possibly due to cut phone lines and poor military
failures are frequently documented. Fig. 1 presents a brief communication channels [5].
illustration of emergency communications involving the The goal of this research is to uncover the dynamics in
variety of stakeholders. emergency communication and to recognize the major
challenges and barriers. More importantly we discuss the
available technical instruments and propose their design and
management to meet these challenges for effective inter-agency
communication. The Indian Ocean Tsunami provides an
Manuscript received September 21, 2008. This work was supported in part
by the National Science Foundation under Grant 0809186.
opportunity to observe and analyze how the information and
Rui Chen is with Medaille College, NY 14216 USA (phone: 716-880-2161; communication technologies impact on the socio-economic
e-mail: rui.chen@medaille.edu). development in the world. Despite of the plethora of prior
John Coles is with State University of New York at Buffalo, NY 14260 USA
literature on the response to Indian Ocean Tsunami, little work
(e-mail: jbcoles@buffalo.edu).
Jinkyu Lee is with Oklahoma State University, OK 74106 USA (e-mail: has been done to analyze the phenomena using a theory driven
jinkyu.lee@okstate.edu). approach. As a consequence, the knowledge thus gained is
H. Raghav Rao is with State University of New York at Buffalo, NY 14260 segregated and a complete understanding is missing. Through
USA (e-mail: mgmtrao@buffalo.edu)
301
the lens of Activity Theory, we identify the key issues email) to exchange task critical information (e.g., chemical
pertaining to the socio-technical systems involved in release) in an attempt to build common operating picture of the
emergency communication. This paper thus contributes to the status quo of the ongoing incident and mitigation. This
improvement in information system to enable and support inter-agency communication is influenced by the rules (e.g.,
emergency communication in large scale disasters. standard operating procedure –SOP), community (e.g., both
This paper is organized as follows. In the subsequent section, professional communities and municipality communities), and
we review Activity Theory that is used in this paper as the division of labor (e.g., operations, logistics, and planning) in
theoretical foundation to explore emergency communication. emergency management realm [16].
Next, we apply Activity Theory in emergency management Instrument
using the case of tsunami disaster. We then elaborate the Email, Phone
information technology design and management to cater to the Real Time
requirements and challenges raised. We conclude this paper Information Sharing
with limitations and future studies.
Subject Activity Object
II. THEORETICAL FOUNDATION Emergency Communication Situational
With the advancement in modern information technologies, Responder Awareness
information systems designated for emergency management
have been practiced and they have proved to play an important
role in facilitating the incident management [6-8]. To design
and implement effective systems for emergency
communication, it is important that we understand the set of Rule Community Division of Labor
design requirements. The development of communication Standard Government, NGO, Task & Task
operating volunteer, business, Assignment
systems requires systematic approaches to elicit and analyze the procedures news media, victims
internal elements, structure, and relationships of core
Fig.2. Application of Activity Theory in Emergency Communication
communication elements. In this paper, we use Activity Theory
to guide the requirement engineering process in
communication system development [9, 10]. An approach III. COMMUNICATION IN EMERGENCY MANAGEMENT
driven by Activity Theory represents a method that has gained In this section, we employ Activity Theory to analyze
increasing attention in recent years [11-13]. Activity Theory emergency communication in an attempt to understand the
provides a lens to analyze the computer-supported activity of a challenges and requirements for emergency communication
group or organization and to study the design of artifacts for system design.
individuals and organizations.
Activity Theory suggests that human activity is directed A. Emergency Communication in General
toward a material or ideal object, mediated by artifacts or In accordance with Activity Theory, the subjects in
instruments, and socially constituted within the surrounding emergency communication are the first and second emergency
environment [14]. Activity can be understood as a systemic responders. They may include fire department, police
structure with various activities that are collated or extended department, emergency medical services (EMS), hazard
away from the core activities [15]. The subject is the active materials workers (Haz-Mat), etc.
element of the process and can be either an individual or a The subjects actively communicate with each other to
group. The object transformed by the activity can be an ideal or achieve situational awareness. Biros et al define situational
material object. The transformation process is enabled and awareness as “the decision-makers’ moment-by-moment ability
supported by instruments (physical or logical). The instrument to monitor and understand the state of a complex system and its
provides the subject with the experience historically collected environment” [17]. During emergency, the completeness and
by his/her community. During the interaction, subjects accuracy of decision makers’ situational awareness is crucial to
internalize and/or externalize their cognitive schemes and their their abilities in comprehending the disaster facts, hazards and
understanding of the relationship between themselves and the risks, and mitigation capabilities, which “piece together an
external objects, instruments, surroundings, etc. Activity accurate of reality” [18].
Theory also considers interaction as one critical aspect and The communication activities function on the basis of
suggests that interactions are the driving force in human technical and non-technical instruments such as existing forms,
interaction and system design. The interactions may also exist guidelines, and alert systems. These instruments are likely to be
inside the subjects, objects, instruments, and their interactions. developed under the supervision of the individual actors and
In Fig. 2, we present the application of Activity Theory in with propertied designs in terms of taxonomy, notation, and
emergency management. That is, the emergency responders ways of implementation. For example, the fire companies in
(e.g., firefighter, police, and emergency medical service neighboring counties may adopt different instruments (e.g.,
personnel) use the communication systems (e.g., phone and alert systems and reporting forms) in managing disaster
302
response. These heterogeneous and incompatible instruments involved as is the news media. The four issues that plague
are typically involved in a large scale incident response where private citizen involvement are as follows. These four issues
multiple subjects are called in. are exacerbated by the involvement of news media.
Emergency communication is subject to influences from 1. Easy numbness to warnings
rules and division of labor. The rule systems embody the norms 2. Dramatization of information given
and conventions adopted in emergency response community 3. Mistrust of centralized government
and their management practices. Example may include 4. Demand for constant information especially immediately
responders’ attitudes toward risk taking and risk sharing. before and after a disaster occurs
Division of labor, on the other hand, refers to the individual News media descends upon any significant event quickly
domain expertise and tasks typically performed by response and in an unfiltered manner, however, they are also the primary
subjects. Based upon the rules in place and the division of source of information that private citizens turn to in emergency
labor, subjects may behave differently in their communication situations. For this reason it is crucial for the news media to be
behaviors. Additionally, response community impacts the viewed as an ally and not an enemy in emergency response
emergency communication patterns. General speaking, information gathering and dissemination. A relationship built
emergency community include government agencies, on mutual trust in critical to the effective utilization of local and
non-governmental organizations (NGOs), independent international media and should be closely guarded [23].
volunteers, local businesses, news media, and victims [19]. Table 1 summarizes a list of typical challenges to emergency
Government agencies are often the actors looked to in times communication. We elaborate an example challenge.
of disaster, thus it is critical for open and functional Table 1. Emergency Communication and Challenges
relationships to be established prior to a disaster in order to
facilitate effective communication. With the variety of Challenges to Effective Communication and Action
Emergency Response
governmental levels and the red tape that must be sorted Tasks Data Time
Scope of
Lack of Information
through, it is also critical for the intra-governmental Decisions
Quality Pressure Trust Complexity
Impact
communication channels to be aided by the early establishment Creation of a Common
X X X X
of relationship and trust [20, 21]. Without these additional Operating Perspective
channels the government paths are dysfunctional and often Formal Decision
X X X
Making Procedures
avoided by other actors when possible. The issue of agency Collection of Relevant
tribalism between local government sanctioned response X X X X
and Effective Data
agencies adds a new dimension of complexity to the situation. Correct Interpretation
X X X
In the United States, an example of this comes up often between of Data
Appropriate Actions
the individual fire companies and the police departments due to Identified and Executed
X X
issues in funding and jurisdiction. Coordinated Response
X X
Some governmental actors can be highly effective in Effort
emergencies by nature of their position or the information and
resources to which they have access. Local elected officials are
The quality of data which a disaster generates can easily be
a great avenue of communication to the general populace and
overwhelming for three major reasons: the varying lack of
are most effective when they have the crucial relationships
quality information, the time consuming task of accurate data
established with potential resources in the region and have the
interpretation, and the differentiation of pertinent compared to
trust of the general populace. Another great resource are the
additional extraneous information [24]. The process of finding
government monitoring agencies such as the Pacific Tsunami
quality data sources often involves trust-bonds created prior to
Warning Center (PTWC) and the US Geological Survey
the disaster [25]. Thus it is crucial for emergency managers to
(USGS), which were the agencies on the front line of tsunami
be open to sources that could provide additional information
detection and information dissemination in the early hours of
but should also not be too accepting of all information so as not
the Indian Ocean Tsunami [4].
to be overwhelmed with the task of accurately interpreting the
Other resources that can often provide crucial data or
pertinent elements of data. Once the seemingly pertinent data is
resources are the local scientific community, private for-profit
extracted, the issue of deciding whether the new information is
organizations, and private not-for profit organizations. Local
immediately relevant is crucial in making informed and
colleges and universities can be often be a great source of
effective decisions. When the data quality is low, emergency
crucial information both prior and in response to emergencies
management is hindered as decisions are often be short-sighted
due to the intricate and invasive approach to problem solving
or ineffective in the larger response as a whole [20].
on a regional and international scale [22]. Private organizations
are great resources for both equipment and manpower B. Case of Tsunami
mobilization. From factory workers to church goers, the We discuss in this section the major communication failures
effective training and utilization of local labor is critical to a in the response to tsunami incident. Our discussion is focused
speedy and effective recovery [4]. on subject and community in Activity Theory as these are
Whenever there is a disaster, local private citizens are
303
deemed as the major failure in emergency communication. communication activities, it is necessary that one determines
Emergency responders, subjects, along Indian coast did not the scope of decision impact (i.e., parties he/she needs to reach)
equip systems capable of modeling and predicting tsunami [24]. By identifying these critical individuals, the reduction in
scales and impacts. Immediately after an earthquake, computer travel time and increase in accuracy of information between
models can calculate how fast the waves will travel, as well as actors will lead to more effective responses [21]. This broader
their amplitude. In an interview, Tsunami expert Tad Murty perspective gives the basis for a much stronger international
mentioned that he had persuaded the Indian government to response instead of a solely national response. In the case of the
build a fully functional tsunami warning system but was told tsunami, this would have significantly strengthened the early
that there were not sufficient financial resources [26]. A system response and preparedness of other nations had India, Thailand,
like this might had saved hundreds of lives. As a matter of fact, Indonesia, and the United States communicated their
26 countries in the Indian Ocean were notified of the tsunami knowledge both with each other and the other nations affected
within 15 minutes after the earthquake took pace. India was in a timelier manner. However, the scope of decision impacts
unfortunately not among them and thus missed the opportunity was hard to decide by the emergency mangers who tried to
to detect the incident. gather and disseminate information during the response to
Emergency responders also experienced difficulties in sense tsunami. As a consequence, responders were not aware of what
making of the incident and how it unfolded. In spite of a lead information should be passed on to what parties. Without a
time of almost six hours, no early warning could be issued formal central framework and method of collecting information
before tidal surges lashed the Indian coasts. The main reason centrally and disseminating, it is difficult for actors to receive
was the absence of systems to translate common knowledge or even where to start when attempting to acquire the
into scientific information. The response communities were not information necessary for an effectively mounted response [19,
familiar with the phenomenon of tsunami and had neither 22].
proper prior experience nor trainings. Using Activity Theory (AT) framework, Table 2 summarizes
Even after the information was collected interpreted, it was the observations of Tsunami response communications along
difficult to pass the information from one point to another. A with the problematic issues raised. In the subsequent section,
great majority of communication infrastructures failed in the we propose the set of solutions.
wake of Tsunami, just at the very time when they are most
needed. At Maldives Island of Sri Lanka, telecommunications Table 2. Activity Theory Informed Analysis on Tsunami Response
infrastructure was based on a microwave terrestrial backbone Communications
network. When Tsunami hit, all public telecommunication AT Construct Problems Consequence
services to the 13 atolls (163 inhabited islands) were Lack of input Failure in detecting the event
interrupted, causing the complete breakdown of the Subject Lack of subject Failure in interpreting the
communications [27]. matter knowledge raw data
Disconnections among response communities also lead to
Lack of
the devastation of tsunami incident. First, communication Failure in information
Community international
breakdown took place inside the domestic response sharing among countries
collaboration
communities. The IAF administration of India, for example,
was informed that the Car Nicobar air base was hit by tsunami Lack of Failure in information
an hour before it reached Tamil Nadu. However the first Instrument communication exchange and sharing
message was not published from the Indian Meteorology infrastructure between any two points
Department to the government until 41 minutes later. This
example reveals the lack of standard operating procedure in the Lack of standard Failure in domestic response
Rule
local government. Standard operating procedure prescribes the operating procedure information supply chain
control and flow of emergency information among the
stakeholders. Without it, emergency information will not flow IV. EMERGENCY COMMUNICATION SYSTEMS
among the involved agencies smoothly and the information The information system discipline has progressed with new
supply chain will break down. systems designs and management practices developed. For
Breakdown also took place among international response example, Hale develops a layered crisis communication
communities. The Pacific Tsunami Warning Centre of the architecture (CCA) that enumerates the communication
United States successfully recorded the earthquake in the functionality requirement for emergency information systems
Indian Ocean and warned that “There was no threat of tsunami [18]. The layers include connectivity, data-validation, filtering,
to the US area of responsibility.” While the warning would be value interpretation, organizational memory, and group process
useful to alert the countries along the Indian coast, the warning layers. Michalowski et al develop a palm-based mobile system
did not fully reach the Asian response community which could Mobile Emergency Triage (MET) [28]. In this section, we
had saved thousands of lives. discuss the technologies that facilitate emergency
Considering the broad areas affected, emergency communication. As in Table 3, the discussions are driven by the
communication activities were difficult to initiate. To initiate analysis of principal constructs of Activity Theory and their
304
interactions. United States, for example, are launching a locally-targeted

emergency alerting service featuring real-time tsunami
Table 3. AT-Driven Communication System Design warnings with evacuation routes. RAINS' Connect & Protect
Designs driven by AT Constructs service captures NOAA/National Weather Service's tsunami
AT Construct Problems Design Components warnings when issued, and immediately sends localized alerts
Lack of input Early warning systems via computers, pagers and cell phones, to local citizens
Knowledge management responsible for public safety. It will dramatically increase the
Subject Lack of subject
systems, intelligence
matter knowledge speed and reach of the warnings within a community.
processing
B. Knowledge management systems
Lack of During emergency response, individuals, teams, and
International emergency
Community international
communication organizations share and apply knowledge as they process
collaboration
information, make decisions, and act on existing knowledge
Lack of Improved infrastructure [30, 31]. Knowledge may be categorized in two distinct forms:
Instrument communication management for mobility, explicit knowledge which is easy to communicate and can be
infrastructure reliability and redundancy codified and tacit knowledge which is inextricably woven with
the experiences and situational contexts [32, 33].
Lack of standard Rule-based interoperable Knowledge management systems such as knowledge
Rule
operating procedure information supply chain repository provide vital support to help the decision makers
analyze and understand the facts. As manifested by the
Designs driven by Interactions between AT Constructs Tsunami case, responders may not be familiar with all types of
Subject –n– Community Grid computing incidents and the approaches in mitigation. The lack of training
Instrument –n– community Open source approach and expertise may account for failures as such. To this end,
knowledge management systems may be adopted to equip the
A. Early Warning Systems responders with the right information when incident first
strikes. There exist a few public knowledge repositories for
Early warning systems are important to disseminate incident
emergency response such as the Responder Knowledge Base
(e.g., tsunami) occurrence to the stakeholders in the disaster
(www.rkb.us). Private knowledge repositories are available in
affected regions. The significance of early warning for
most commercial-off-the-shelf (COTS) emergency
emergency management has been emphasized in major
management systems such as DisasterLAN (disasterlan.com),
international agendas including the Barbados Plan of Action
E-Team (eteam.com), and WebEOC (esi911.com). Systems as
for Small Island Developing States, the Johannesburg Plan of
such retain domain-specific knowledge that complements the
Implementation, and the G8 summit in Gleneagles as well as
knowledge repositories maintained by individual teams.
major environmental agreements such as the UN Framework
Convention on Climate Change and the UN Convention to C. Intelligence Processing
Combat Desertification [29].To improve the alert system, it is During emergency, the completeness and accuracy of
important to leverage the existing systems as well develop new information shared and exchanged among responders is crucial
warning systems. to their abilities in comprehending the disaster facts, hazards
A lesson learned in the response to tsunami is that the and risks, and mitigation capabilities, which “piece together an
existing warning systems function below optimal performance. accurate of reality” [18]. System designs such as information
International Tsunami Information Centre (ITIC), for example, infusion are advised in this regard. As noted above, however,
monitors the tsunami activities in the Pacific and disseminates
the information quality in emergency communication is
tsunami information and warning messages to well over 100
typically low where errors are frequently observed. For
points scattered across the ocean. Unfortunately ITIC was
example, Chen et al suggest that information on emergency
confined to its member countries before the Indian Ocean
Tsunami; this led to non-member countries such as India did incident is likely to be tainted and prone to mistakes [34].
not receive warning information in a timely manner. A series of Eye-witness accounts of the scene are often biased by their
consultation meetings led to a general agreement by comprehension processes, background, recollection, and
implementing agencies and donors that all countries affected by verbalization skills. As a consequence, triangulation-based data
the 24 December 2004 tsunami and other countries at risk to validation [35, 36] is an indispensable component for
tsunami should be involved in this project for the establishment emergency response information systems. The validation
of tsunami early warning system in the Indian Ocean Region usually relies on multiple and independent sources to validate
[29]. an event, fact, or conclusion. Designs using data correlation
New warning systems are also in need to augment the and mining techniques have proved to be adequate solutions
existing capabilities of warning systems and to utilize modern [37, 38]. Reference databases are also advertised for
computing technologies. Oregon and RAINS (Regional employment as they enhance the system competence in logic
Alliances for Infrastructure and Network Security) in the reasoning and inconsistency detection [39].
305
Field interviews with emergency responders find that channel accessibility, and system adaptability. Take
incident information may expand rapidly during the course of connectivity reliability for example. During incidents, “wired”
mitigation; the increasing level of response enactment [40] communication infrastructures are likely to be destroyed by the
accumulates new information on incident, responder, resource, disaster. It is important that the response community plan and
operation, and environment from all stakeholders. This implement technologies to remedy the communication
multi-dimension information cross interrelate with each other breakdowns. In the case of tsunami, the lack of communication
and together match the high “requisite variety” [41] of the connectivity in the most tsunami-affected Indian Ocean
decision problems in extreme events. Literature on Cognitive countries negatively affected some national and local level
Processing Capacity (CPC) suggests that individuals are activities. Options may include (1) mobile systems to deploy
limited in cognitive resources and incapable to allocate them communication channels in the field, including satellite phones,
optimally for problem solving [42, 43]. In the face of portable radio systems, WAN network, and mesh networks; (2)
information of large volume and high complexity, emergency radio-communication systems such as maritime, aeronautical,
managers are prone to be “overloaded.” The mental stresses and radio determination services; and (3) redundancy plans
resulted by time pressure, perceive risks, and concerns on with backup channels to support system breakdown or overload
“public image” further intensify the cognitive overload. To this due to request and command flooding. The existing emergency
end, tactics such as information filtering are recommended by system literature has identified and evaluated a set of solutions
prior emergency studies so as to reduce the amount of irrelevant which include increased technology investment, partnership of
data, organize related messages into cohesive and coherent sets, public/private network, load balancing scheme, priority
and prioritize message sets according to level of importance [8, telecommunication services, and redundancy design [18,
18, 34, 39]. In addition, research on Human Computer 53-56].
Interaction (HCI) suggests that cognitive overload may be Information assurance, on the other hand, refers to
reduced through design schemes in interface structure, information security and privacy. Security is a key element to
information packaging, and information accessibility [44, 45]. emergency communication as long as sensitive information is
present; it continues to be focal concern for channel design
D. Communication Infrastructure Management
considering the increasing awareness of terrorist attacks [57].
Timely communication among responders (i.e., “subjects”) System designs such as encryption, decryption, and intrusion
relies on the establishment of robust and efficient detection may suffice the security requirement. Privacy is a
communication infrastructures. Before the Indian Ocean more recent issue facing emergency management nowadays.
Tsunami, many of the countries in the region did not have a Existing regulations such as HIPPA (Health Insurance
well-organized disaster management system except for tropical Portability Accountability Act of 1996) require that privacy
cyclone-prone countries such as India and Bangladesh. information including personal identity and medical history be
National disaster management offices in most counties were carefully handled [58].
very weak, and there were few established infrastructures
(channels and platforms) for emergency communication. E. International Emergency Communication
Modern emergency management employs a variety of To mitigate large scale incident such as tsunami, emergency
communication channels including LAN, WAN, ad-hoc communication may easily span over multiple countries and
wireless network, mesh network, 800M Hz radio, and satellite districts. The presence of a coordinated communication plan
phone [46-48]. During any typical emergency operation, during emergency is pivotal to allow task critical information
hundreds of task critical updates, briefings, reports, requests, and intelligence travel across the boundaries of response
queries, and orders are circulated inside the incident communities both domestic and international.
management organizations [49]. These channels enable Regional coordination of emergency communication should
real-time information sharing and communications, first be strengthened through activities of working groups on
establishing a “common operational picture” to keep all the mitigation, preparedness and response in order to ensure
decision makers on the same page [50, 51]. A common integration of communication channels such as early warning
operational picture is a single identical display of relevant systems into national and regional mitigation, preparedness and
operational information shared by more than one actor. response capability building efforts within a multi-hazard
Through information transfer, responders working on varying framework [29]. In addition, communication coordination may
aspects of the mitigation tasks are able to synchronize their be established through both country and community level
visions on the focal incident, smooth out potential approaches to ensure those good practices, including hazard
inconsistencies, and synthesize a complete “shared mental and vulnerability assessment, organizational strengthening,
model” [52]. Information sharing mechanisms such as “publish community participation, warning system operation, capacity
/ subscribe” system are recommended for this regard. building, evacuation planning, and the design and construction
Emergency communication infrastructure is subject to of shelters, can be shared and exchanged in a coordinated
threats on performance, information assurance, and manner to avoid duplications, confusions, and overlaps.
communication interoperability. Channel performance Communication among international organizations is subject
measures the communication capacity, connectivity reliability, to interoperability challenge. The interoperability issue is
306
pronounced for large scale emergency management (e.g., offered cost-effective alternatives to help the responders. Grid
tsunami) because the technologies adopted by participating computing and grid storage has recently been employed in
organizations from varying countries and regions are in general emergency response thanks to the rapid growth in network
incompatible for reasons ranging from the ability of local bandwidth. The computing and storage resources of individuals
agencies to fund availability to the lack of unified guidelines and volunteer organizations can be pooled to boost the
for software and hardware [59-62]. To this end, it is extremely information analysis capabilities of the local responders. These
important that the emergency response community as a whole systems have been used recently to study weather, pollution,
coordinate to develop consistent data standards, transmit and remediation, etc [68].
protocols, homogenous devices, compatible application Telecommunication software and systems, instruments,
interfaces, and congruent regulations [16, 63-66]. A number of function to support communication activates when incidents
efforts have been observed national wide with as to establish strike. The development of systems as such may be improved
the communication interoperability solutions. These examples through the aids from the community. Open source approach
include the National Information Exchange Model offers the opportunity for technology innovations and
(www.niem.org), Emergency Data Exchange Language improvements in disaster communication system designs. This
(www.comcare.org/edxl.html), Common Alert Protocol approach is also known as Free and Open Source Software
(www.incident.com/cap), SAFECOM initiative (FOSS) development and community mechanism, where
(www.safecomprogram.gov), and Federal Enterprise collaborative designs are achieved through inputs from the
Architecture (www.whitehouse.gov/omb/egov). Each of them entire community. Example FOSS systems include the Sabana
addresses a portion of the interoperability challenges and more disaster system [69].
efforts are needed before a full-fledged solution may be
achieved. Without coordinated efforts, emergency V. CONCLUSION
communications are likely to break down when they travel The damage caused by the tsunami was unprecedented, but
across the boundaries of communities, both domestic and the lessons learned are crucial to creating communication
international. That is, relief teams from the world will not be frameworks that can withstand such buffeting uncertainty. By
able to communicate using their individual systems during identifying the key characteristics that can cause
response to large disaster such as tsunami, ultimately slowing communication in emergencies to be ineffective or to be
down the response progress. entirely void, future researchers and practitioners will have a
F. Rule-base Information Supply Chain frame of reference from which to approach the issue of
emergency communications [20].
Standard operating procedure (SOP) prescribes the rules to
In this paper, we employ Activity Theory to study
control and direct emergency information among stakeholders.
communication in emergency management and explore the key
Unfortunately many developing countries do not have
dimensions of issues. Taking Indian Ocean Tsunami as an
systematic SOP and they may learn from the developed
example, we identify and discuss the pitfalls in the
countries. Emergency management in U.S. is guided by the
communication practices for incident response. The authors
National Incident Management System - NIMS [67]. NIMS
further explore the issues in information system design and
facilitates the emergency communication in the following
management in an attempt to address the pending challenges in
ways: (1) it constructs a clear organizational hierarchy that
emergency communication. New knowledge gained will allow
directs the information flows; (2) it standardizes the format and
effective pro-active measures to be taken to mitigate and
content of information input and output; and (3) it establishes
respond more effectively to future disasters by recognizing the
the protocols of reporting, meeting, and cross-boundary
key players in an emergency, effectively facilitating
adjustments. And thus, SOP supports the information to flow
international cooperation, turning academic research into
smoothly among the agencies and form an emergency response
practical application, and avoiding pitfalls in areas needing
information supply chain. To this end, rule-based
greater research.
communication systems that share and exchange information
To further improve emergency communications, it is critical
following SOPs are desirable to the emergency responders.
that future studies extend the proposed design and management
G. More Designs driven by Interactions between AT factors frameworks to learn from and connect with the broad spectrum
Activity theory underlines the importance of interactions of information and communication technologies and
between factors as an important driving force for activity development (ICTD) applications.
system development.
Emergency responders, subjects, analyze the incident facts REFERENCE
before they can share it with the other agencies. While
numerous initiatives have been started to improve information [1] S. Y. Shen and M. J. Shaw, "Managing Coordination
processing, local agencies are typically limited in the systems in Emergency Response Systems with Information
that they can offer to build and manage. To this end, the broad Technologies," in Tenth American Conference on
community of public/private firms and the general public has Information Systems, New York, 2004, pp.
2110-2120.
307
[2] R. Chen, R. Sharman, R. H. Rao, and S. Upadhyaya, [18] J. Hale, "A Layered Communication Architecture for
"Coordination in Emergency Response Management," the Support of Crisis Response," Journal of
Communications of the ACM, vol. 51, pp. 66-73, Management Information Systems, vol. 14, 1997.
2008. [19] D. Alexander, "World Disasters Report 2005: Focus
[3] E. L. Quarantelli, "Problematical Aspects of The on Information in Disasters," 2006.
Information/Communication Revolution for Disaster [20] L. K. Comfort, "Crisis Management in Hindsight:
Planning and Research: Ten Non-Technical Issues Cognition, Communication, Coordination, and
and Questions," Disaster Prevention and Control," Public Administration Review, 2007.
Management, vol. 6, pp. 94-106, 1997. [21] S. Corbacioglu and N. Kapucu, "Organizational
[4] J. Telford, J. Cosgrave, and R. Houghton, "Joint Learning and Self-Adaptation in Dynamic Disaster
Evaluation of the International Response to the Indian Environments," Disasters, vol. 30, pp. 212-233, 2006.
Ocean Tsunami: Synthesis Report," Tsunami [22] N. Martin, "The Asian Tsunami: An Urgent Case for
Evaluation Coalition2006. Improved Government Information Systems and
[5] R. Samarajiva, "Policy Commentary: Mobilizing Management," Disaster Prevention and Management,
Information and Communications Technologies for vol. 16, pp. 188-200, 2007.
Effective Disaster Warning: Lessons from the 2004 [23] N. Kapucu, "Collaborative Emergency Management:
Tsunami," News Media Society, 2005. Better Community Organizing, Better Public
[6] B. Van de Walle and M. Turoff, "Emergency Preparedness and Response," Overseas Development
Response Information Systems: Emerging Trends and Institute, 2008.
Technologies," Communications of the ACM, vol. 50, [24] M. W. Seeger, T. L. Sellnow, and R. R. Ulmer,
pp. 29-31, 2007. Communication and Organizational Crisis. Westport,
[7] M. Turoff, "Past and Future Emergency Response CT: Praeger Publishers, 2003.
Information Systems," Communications of the ACM, [25] J. M. Kendra and T. Wachtendorf, "Elements of
vol. 45, pp. 29-32, 2002. Resilience after the World Trade Center Disaster:
[8] T. X. Bui and S. R. Sankaran, "Design Considerations Reconstituting New Year City's Emergency
for A Virtual Information Center for Humanitarian Operations Centre," Disaster, vol. 27, pp. 37-53,
Assistance/Disaster Relief Using Workflow 2007.
Modeling," Decision Support Systems, vol. 31, 2001. [26] S. Jain, A. Malik, and R. D'Souza, "26 Countries
[9] Y. Engestrom, "Activity Theory and Individual and Alerted in 15 Minutes, India not One of Them," in
Social Transformation," in Perspectives on Activity Indian Express Mumbai: Indian Express Newspapers
Theory, R. M. a. R. P. Engeström, Ed. Cambridge, Ltd, 2004.
UK: Cambridge University Press, 1999, pp. 19-38. [27] T. A. o. M. (TAM), "Report on the Failure of the
[10] Y. Engestrom, Learning by Expanding: An Telecommunications Services as a result of the Asian
Activity-Theoretical Approach to Developmental Tsunami on 26 December 2004,"
Research. Helsinki: Orienta-Konsultit, 1987. Telecommunications Authority of Maldives
[11] L. Uden and A. Kumaresan, "Usable Collaborative (TAM)2005.
Email Requirements Using Activity Theory," [28] W. Michalowski, S. Rubin, R. Slowinski, and S. Wilk,
Informatics, vol. 31, pp. 71-83, 2007. "Mobile Clinical Support System for Pediatric
[12] V. Kaptelinin and B. A. Nardi, Acting with Emergencies," Decision Support Systems, vol. 36,
Technology: Activity Theory and Interaction Design. 2003.
Cambridge, MA: MIT Press, 2006. [29] D. C. Pattie, S. Dannenmann, and Y. Hagiwara,
[13] A. Chaudhury, D. N. Mallick, and H. R. Rao, "Web "Evaluation and Strengthening of Early Warning
Channels in E-Commerce," Communications of the Systems in Countries Affected by the 26 December
ACM, vol. 44, p. 99, 2001. Tsunami," in Information Systems for Crisis Response
[14] L. S. Vygotsky, Mind and Society. Cambridge, MA: and Management, 2008.
Harvard University Press, 1978. [30] C. Ryu, Y. J. Kim, A. Cahudhury, and H. R. Rao,
[15] O. W. Bertelsen and S. Bodker, "Activity Theory," in "Knowledge Acquisition via Three Learning
HCI Models Theories, and Frameworks: Toward A Processes in Enterprise Information Portals," MIS
Multidisciplinary Science, J. M. Caroll, Ed. San Quarterly, vol. 29, pp. 245-278, 2005.
Francisco: Morgan Kaufmann, 2003, pp. 291-324. [31] T. S. Raghu, A. Chaudhury, and H. R. Rao, "Business
[16] R. Chen, R. Sharman, N. Chakravarti, H. R. Rao, and Process Change: A Coordination Mechanism
S. Upadhyaya, "Emergency Response Information Approach," Knowledge and Process Management,
System Interoperability: Development of Chemical vol. 5, pp. 87-98, 1998.
Incident Response Data Model," Journal of the [32] I. Nonaka, "A Dynamic Theory of Organizational
Association for Information Systems, vol. 9, 2008. Knowledge Creation," Organization Science, vol. 5,
[17] D. P. Biros, M. Daly, and G. Gunsch, "The Influence pp. 14-37, 1994.
of Task Load and Automation Trust on Deception [33] J. C. Spender, "Strategy Theorizing: Expanding the
Detection," Group Decision and Negotiation, vol. 13, Agenda," in Advances in Strategic Management, P.
2004.
308
Shrivastava, A. Huff, and H. Dutton, Eds. Greenwich, Coordination: Lessons from the World Trade Center
CT: JAI Press, 1992, pp. 3-32. Response," Center for Technology in Government,
[34] R. Chen, R. Sharman, H. R. Rao, and S. Upadhyaya, 2004.
"Design Principles for Critical Incident Response [49] R. Chen, R. Sharman, H. R. Rao, and S. Upadhyaya,
Systems," Information Systems and E-Business "Design Principles of Coordinated Multi-incident
Management, vol. 5, 2007. Emergency Response Systems," in IEEE
[35] J. Douglas, Investigative Social Research Newbury International Conference on Intelligence and Security
Park, CA: SAGE Publications, 1976. Informatics 2005, Atlanta, GA, 2005.
[36] E. J. Webb, D. T. Campbell, R. D. Schwartz, and L. [50] J. J. Carafano, "Preparing Responders to Respond:
Sechrest, Unobtrusive Measures. Chicago, IL: The Challenges to Emergency Preparedness in the
Rand-McNally, 1965. 21st Century," Heritage Lectures, vol. 812, 2003.
[37] H. Chen, D. Zeng, H. Atabakhsh, W. Wyzga, and J. [51] S. F. Midkiff and C. W. Bostian, "Rapidly-Deployable
Schroeder, "COPLINK: Managing Law Enforcement Broadband Wireless Networksfor Disaster and
Data and Knowledge," Communication of the ACM, Emergency Response," in The First IEEE Workshop
vol. 46, pp. 28-34, 2003. on Disaster Recovery Networks (DIREN '02) New
[38] H. Chen, J. Schroeder, R. Hauck, H. Ridgeway, H. York, NY, 2002.
Atabakhsh, H. Gupta, C. Boarman, R. K, and A. [52] T. X. Bui, "Towards a Theory of Shared Mental
Clements, "COPLINK Connect: Information and Model in CSCW," in 1992 Workshop on CSCW, 1992.
Knowledge Management for Law Enforcement," [53] D. Berge, The First 24 Hours: A Comprehensive
Decision Support Systems, vol. 34, pp. 271-285, 2003. Guide to Successful Crisis Communications.
[39] M. Turoff, M. Chumer, B. Van de Walle, and X. Yao, Cambridge, MA: Basil Blackwell, 1990.
"The Design of A Dynamic Emergency Response [54] B. S. Manoj and A. H. Baker, "Communication
Management Information System (DERMIS)," Challenges in Emergency Response,"
Journal of Information Technology Theory and Communications of the ACM, vol. 50, pp. 51-53,
Application, vol. 5, 2004. 2007.
[40] K. E. Weick, "Enacted Sensemaking in Crisis [55] NCP, "Priority Telecommunications Services,"
Situations," Journal of Management Studies, vol. 25, Washington, DC: Department of Homeland Security,
1988. National Communications System, 2006.
[41] K. E. Weick, K. M. Sutcliffe, and D. Obstfeld, [56] M. Turoff, S. R. Hiltz, M. Bieber, B. Whitworth, and
"Organizing for High Reliability: Processes of J. Fjermestad, "Computer Mediated Communications
Collective Mindfulness," in Research in for Group Support: Past and Future," in HCI in the
Organizational Behavior. vol. 21, B. M. S. a. L. L. New Mellennium, J. Carroll, Ed.: Addison Wesley,
Cummings, Ed. Greenwich, CT: JAI Press, 1999, pp. 2001.
81-123. [57] H. Chen, F. Y. Wang, and D. Zeng, "Intelligence and
[42] J. Sweller, "Cognitive Load During Problem Solving: Security Informatics for Homeland Security:
Effects on Learning," Cognitive Science, vol. 12, Information, Communication, and Transportation,"
1988. IEEE Transactions on Intelligent Transportation
[43] B. Britton and A. Tesser, "Effects of Prior Knowledge Systems, vol. 5, pp. 329-341, 2004.
on Use of Cognitive Capacity in Three Complex [58] t. Congress, "Health Insurance Portability and
Cognitive Tasks," Journal of Verbal Learning and Accountability Act of 1996," in Public Law 104-191,
Verbal Behaviour, vol. 21, 1982. T. Congress, Ed., 1996.
[44] Y. J. Kim, R. Kishore, and G. L. Sanders, "From DQ [59] BJA-DOJ, "NIEM Executive Briefing," D. o. Justice,
to EQ: Understanding Data Quality in the Context of Ed., 2007.
E-Business Systems," Decision Support Systems, vol. [60] J. Fedorowicz, U. J. Gelinas, J. L. Gogan, M. Howard,
forthcoming, 2003. M. L. Markus, C. Usoff, and R. Vidgen, "Modeling
[45] R. Chen, "Consumer's Initial Acceptance of Physical Barriers to Interorganizational System
E-Commerce Website: A Contingency Approach," in Implementation Success," International Journal of
AMCIS 2007, Keystone, CO, 2007. Information Technology and Management, vol.
[46] P. Anderson, "Information Technology: Where is It in forthcoming, 2007.
the Coordination of Emergency Services," in Asia [61] J. L. Gogan, C. B. Williams, and J. Fedorowicz, "Fatal
Pacific Police Technology Conference, 1991. Flaws in Information Sharing," International Journal
[47] G. E. G. Beroggi and W. A. Wallace, "Real-Time of Technology, Knowledge and Society, vol. 1, pp.
Decision Support for Emergency Management: An 93-100, 2005.
Integration of Advanced Computer and [62] C. B. Williams, J. L. Gogan, and J. Fedorowicz,
Communications Technology," Journal of "Public Safety and Cross-Boundary Data Sharing:
Contingencies and Crisis Management, vol. 3, p. 18, Lessons from the CapWIN Project," IEEE Computer,
1995. vol. 38, p. 28, 2005.
[48] S. S. Dawes, T. Birkland, G. K. Tayi, and C. A.
Schneider, "Information, Technology, and
309
[63] D. Aylward and E. Jones, "Data Interoperability:

Sharing Information for a Safer America." vol. 2006:
COMCARE Emergency Response Alliance, 2006.
[64] DHS, "Fact Sheet: Achieving First Responder
Communications Interoperability," in Department of
Homeland Security Press Washington, DC, 2005.
[65] D. Frale, "Emergency Interoperability Consortium
Announces Agreement with Department of Homeland
Security to Promote Data Sharing During
Emergencies," in PrimeZone Media Network, 2005.
[66] T. Harrison, J. R. Gil-Garcia, T. A. Pardo, and T.
Fiona, "Learning about Interoperability for
Emergency Response: Geographic Information
Technologies and the World Trade Center Crisis," in
The Thirty-Ninth Annual Hawaii International
Conference on System Sciences, Hawaii, 2006.
[67] DOS, "National Incident Management System,"
Department of Homeland Security, Washington,
DC2004.
[68] IBM, "World Community Grid," 2008.
[69] P. Currion, C. De Silva, and B. Van de Walle, "Open
Source Software for Disaster Management,"
Communications of ACM, vol. 50, pp. 61-65, 2007.
310
Empowering Muslim Youth through Computer

Education, Access, Use: A Gender Analysis
Farida Khan and Rehana Ghadially
Abstract – In the present information society, technical Rehana Ghadially is with Department of Humanities and
education has acquired immense importance as the Social Sciences, Indian Institute of Technology
Information and Communication Technologies (ICTs) hold Bombay, India (e-mail: rehana@hss.iitb.ac.in) .
potential for bridging socio-economic divides and
empowering the marginalized such as women and minority
groups. This paper explores the access and use of
computer/Internet and examines how these affect
empowerment levels among young men and women. This Even as India is a leading destination for
paper considers the psychological, social, educational and outsourced IT and IT enabled work that promises
economic benefits following from computer education and
employment opportunities, there are several divides within
usage of computer and Internet technology. Data was
collected from 155 young girls (N=82) and boys (N=73) from
the country that leave the disadvantaged groups out of the
three computer training centers in Mumbai. Statistical tests development radar. The digital divide operates across
such as t-tests, 2-way ANOVA and chi-squares were geographic location (rural-urban), class (rich-poor), gender
computed to compare male and female subjects on (male-female) and language (English-vernacular) [1].
empowerment and ownership/access and use. The figures for While development effort in India has considered one or
ownership and home Internet connection were low for the the other of these factors, they overlook the marginalized
entire sample. Computer training centre and cyber café are status of minorities such as Muslims. The Muslim
important points of access for females and males respectively. minority is disadvantaged both educationally and
Further, it is found that young women report higher gains
economically, thus driving them off the information
from computer learning and technology use, hence, computer
education can be a key gender equalizer. In light of the highway [2, 3, 4]. In addition, they face societal barriers
above, policy measures to widen access and provide such as limited awareness and discrimination [5]. Cost and
subsidized training are suggested. language barriers compound the constraints further
limiting the extent of computer and Internet use. Women
Index Terms – access, computers/Internet, empowerment, are doubly disadvantaged as they face an additional gender
gender barrier [6]. A coverage of ICT and India reports that ICT
skills and access provide an opportunity to leapfrog and
I. INTRODUCTION bridge the disparities in information, education and income
[7]. Hence, we postulate that equipping the young
Today’s world is shaped by availability of members of the Muslim community, especially its women
information and ability to communicate, both of which are with the technical skills and access to the technology will
enabled through the rapid expansion of Internet and result in a more balanced development of the country.
extensive use of computers. While there have been
developmental efforts to increase the access and use of In a developing country like India, ownership of
Information and Communication Technologies (ICTs) in personal computers and Internet penetration is limited at
general, there is still a marked difference in the technology 2.5 and 4.2 per cent respectively [8]. Most users depend on
access across different groups, resulting in the digital shared access points, predominantly at educational
divide. Access considers equality of access to resources, in institutions and workplaces. Three-fourths of Internet
this case to computer and Internet technology. The uneven users in the country depend on community access points
distribution of ICTs across the world as well as within such as cyber café, telecenter, and information kiosk, with
societies gives rise to a digital divide; that results into two cyber cafes as the predominant public access model in
groups; namely, the information rich, those who have urban areas [9]. It is estimated that there are around 50,000
access to abundant information and the information poor, cyber cafes used by almost 70 percent of Internet users
who lack such an access. [10]. However, the cost of surfing Internet at cyber cafes
restricts the usage to the middle and high income groups
Manuscript received September 22, 2008. [11]. In addition, research indicates that even when access
Farida Khan is with Department of Computer Science and is available, men are advantaged over women in the extent
Engineering, Indian Institute of Technology Bombay, of computer usage [12]. In light of the above discussion on
India (phone: 91-022-25764966, e-mail: disparity in the access and use of the technology, the
farida@iitb.ac.in). present study seeks to explore the different access points
for the two sexes and assess gender differences in
computer ownership, Internet connection at home, points
of access and extent of computer/Internet use.
1
311
Education especially technical education is sidelining the other facets of empowerment. This paper
instrumental in expanding capacities and improving attempts to provide a more complete view of
employment opportunities, thus resulting not only in empowerment by considering the psychological, social,
economic but also personal and social empowerment such educational and economic gains flowing from computer
as enhancing confidence levels and social status [13]. learning, technology access and usage.
Considering that computer skill is of immense importance
in the technology driven society, this paper considers the
gains resulting from having acquired basic computer
education. Unlike the much emphasized economic gains,
Objectives of the study
the present paper adopts a multidimensional perspective
and considers four different facets of empowerment ---
psychological, social, educational and economic. In • To explore gender differences, if any, in
addition, it explores the access and use of computer and computer ownership, Internet connection, points
Internet among men and women and examines how these of access and extent of computer/Internet use.
affect the different dimensions and levels of • To assess the empowerment flowing from
empowerment. computer education for males and females.
• To explore differences, if any, in empowerment
In recent times, research in gender and among males and females with different levels of
development studies has focused on the concept of computer/Internet use.
empowerment [14]. According to their theoretical
background and parameters of assessment, empowerment II. METHODOLOGY
is understood by different researchers in different ways.
Those from the field of sociology focus on empowerment Participants of the study comprised of 155
of the marginalized groups or society as a whole, and trainees (82 females and 73 males) from Mumbai, enrolled
assess it in terms of political activism, advocacy and in a one year diploma in computer applications and multi-
networking. Those from the field of economics focus on lingual desktop publishing offered by the Ministry of
empowerment of the individual, group or economy as a Human Resource Development, under the National
whole and assess it in terms of improved income levels. Council for Promotion of Urdu language (NCPUL)
Experts in gender studies consider women’s empowerment scheme 1 . The course requires the trainees to complete a
per se and elaborate on how altered gender relations would diploma in Urdu language along with the computer course.
benefit women and assess it in terms of increased income While theoretically open to anyone interested in the Urdu
opportunities, greater participation in decision making, language, it is particularly the low income Muslim young
social networking and advocacy. On the other hand, those who are drawn to it because it is highly subsidized. Hence,
from psychology and management studies concentrate on a large section of Muslim boys and girls who are well
psychological gains and characterize empowerment in versed with Urdu language enroll for the course.
terms of personality (self-efficacy, internal locus of Computation from the demographic profile revealed the
control), motivational (feeling enabled) and cognitive following information about the sample. The average age
(meaning, competence, self-determination, impact) factors of the subjects was 21.34 years (males = 20.46 years;
[15, 16, 17]. For the present study, a cognitive perspective females = 22.31 years) and their educational level varied
of psychological empowerment is adopted. Hence, from higher secondary (45.93 per cent) to undergraduates
psychological empowerment is construed as --- (22.67 per cent) and college graduates (31.40 per cent).
meaningfulness (commitment to the task), Competence Majority of the subjects (69.77 per cent) studied in Urdu
(self-confidence), Self-determination (choice) and impact medium schools, while less than one-third (26.74 per cent)
(making a difference). had English as the medium of instruction.
In the present information society, computers and Majority (73.2%) of the sample belonged to low
Internet are the pervasive technologies that hold potential (monthly family income of less than Rupees 5000/-; US 1$
for equalizing the disparities in information, education and = 48.70 Rupees) and low middle income group (monthly
income [18]. With the growth of Information Technology family income of Rupees 5000/- - 10,000/-). 47.67%
enabled services (ITES) in India, a plethora of computer fathers were running their own business, 33.14% engaged
related jobs have opened up. Comparison of work- in service and 19.19% had retired. 87.2% mothers were
participation rate of the ITES with other service sectors homemakers, while 12.79% engaged in home-based work
indicates that the field is more accommodating of women
[19]. Research has highlighted the personal (self
confidence) and social gains (status, increased 1
For a detailed description of the scheme, refer Rehana
connectivity) that come from ICT education [20, 21]. Ghadially and Farida Umrani (2004) IT Education: Initiatives among
However, most of these studies are focused on women; Mumbai Muslims”, i4D Information for Development, February.
hence a gender comparison does not follow. In addition,
the literature is skewed towards the economic benefits thus
312
or service. Participants report low educational levels of Permission was ought from the heads of three
parents as nearly one-fourth (24%) of the fathers and half NCPUL computer training institutes, to conduct the study.
(44%) of the mothers had completed only primary Data was collected across twelve visits by the senior
education. author from a group of 10-15 subjects, on completion of
ten months of computer training.
The tools for data collection included
standardized psychological scales to measure
empowerment, a questionnaire to gauge access and use of
computer/Internet and a demographic profile. Computer Data Analysis
ownership was estimated by one question each for
computer and Internet connection at home. Point of access
Chi-squares and independent t-tests and were computed to
was assessed by one question where subjects selected the compare male and female subjects on ownership/access,
places of access from a list provided to them. Extent of
use and empowerment. Two 2-way ANOVAs were
computer and Internet use was estimated in terms of the
computed to gauge the influence of gender and differential
number of hours subjects spend on a computer/Internet in level of computer and Internet use on the different
a week. In addition, a personal profile was estimated to get
dimensions of empowerment.
a few demographic details.
III. RESULTS
Psychological empowerment was assessed by an
adapted version of Spreitzer’s 12-item scale [17]. It has
four subscales – meaningfulness (value of the task), The study gathered information on computer
competence (efficacy), self determination (choice) and ownership, Internet connection at home and points of
impact (difference made). A sample item from the access. It was found that 14.8 per cent subjects had a
psychological empowerment scale is, ‘Learning computers computer at home. This included 12.2 per cent females
has given me considerable opportunity for independence and 17.8 per cent males (χ2 =1.38, ns). 6.0 per cent of the
and freedom in how I do my daily tasks.’ The reliability sample (5.0 per cent females and 6.8 per cent males, χ2
co-efficients (internal consistency) range between 0.61- =1.58, ns) had Internet connection a home. These findings
0.72. Only overall empowerment score is considered for indicate that although the difference between male and
analysis in the present study. For social, educational and female subjects’ computer ownership and Internet
economic empowerment, the scales were designed by the connection is not significant, there is a trend in favour of
researchers. Social empowerment was measured by a 4- males.
item scale that tapped four aspects --- increased status,
social comparison, being with the times and keeping in The different places from where subjects
touch with friend and/or relatives. A sample item from accessed the technology beyond class hours at the training
social empowerment scale is, ‘Computer education makes centre were explored. The computer institute where the
me feel more up to date and current.’ Educational subjects got their training and cyber cafe emerged as the
empowerment was measured by a 3-item scale that most important points of access followed by home
covered three aspects, information source on (own/relative/friend/neighbor) and workplace. A small
courses/colleges/universities, accessing online resources group of subjects did not have access to computers beyond
and preparation of class reports and presentations. A class hours. A gendered view of these findings indicates
sample item is, ‘Learning computers has opened new ways that females and males differ significantly on points of
to find information about different access (χ2 = 20.19, p<0.001). While the computer training
courses/colleges/universities for me.’ Economic institute emerged as a major access point for females (χ2 =
empowerment was assessed by a five-item scale that 10.88, p<0.001; females 57.3 per cent and males 27.4 per
tapped the following five areas --- opening of new job cent), cyber café is more important for males (χ2 = 4.26,
opportunities, application in family business, starting a p<0.05; females, 12.2 per cent, males, 38.4 per cent). The
computer related entrepreneurial venture, earning from other points of access --- home and workplace --- did not
home and managing family responsibilities with a job. A show any significant gender difference. 8.5 per cent
sample item of economic empowerment scale is ‘Learning females and 4.1 per cent males did not access computers
computers has opened new job opportunities for me.’ The beyond class hours (Table 1). The findings on the points of
reliability co-efficients (Cronbach’s alpha) of the social, access indicate that computer training institutes and cyber
educational and economic empowerment scales were 0.66, cafes are the major access points for females and males
0.72, and 0.77 respectively. Each of the empowerment respectively. Though not significant, the trend indicates
scale asked the subjects to indicate the extent to which that more males have access to home computers and twice
they agree that learning computers has resulted in the as many females as compared to males have no access to a
following outcomes on a 6-point rating scale ranging from computer at all.
1 = disagree to 6 = highly agree. The scales were pilot
tested and in line with participants’ feedback, Hindi
translation was provided along with the English version.
313
High 31.0 35.6 26.8 0.32 7.1 12.3 2.4 4.45*

use
χ2
df=2 7.94** 32.15***
**
p<0.01, * p<0.05
TABLE 1
Points of Access to Computers for Males and
Females
Points of Total Males Females Gender A similar classification was done with regard to
Access (T) (M) (F) χ2 Internet use. More than half of the sample did not use
df =1 Internet at all, one-third surfed Internet for less than 5 hours
Computer 43.2 % 27.4 57.3 10.88*** and a small number surfed for more than 5 hours in a week.
institute Female and male subjects differed significantly with regard
Cyber café 24.5 38.4 12.2 4.26* to the extent of Internet use (χ2 = 32.15, df = 2, p<0.001).
The two sexes differed significantly on each of the three
Home 23.2 27.4 19.5 0.44 categories of Internet use --- no use (χ2 = 17.39, df = 1, p <
0.01), low use (χ2 = 11.08, df=1, p < 0.01) and high use (χ2 =
Workplace 2.6 2.7 2.4 0 4.45, df = 1, p < 0.01). 80.5 per cent females reported not
using Internet at all, while the corresponding figure for males
Do not 6.5 4.1 8.5 1.6 was 35.6 per cent. 17.1 per cent females and 52.1 per cent
access males report low Internet use; while 2.4 per cent females and
χ2 20.19*** 12.3 per cent males fall in the high Internet use category
df =4 (Table 2). Thus, Internet use is significantly low for females
as compared to males. Comparing computer and Internet
*** usage of the two sexes indicated that the disparity in
p<0.005, *p<0.05 technology usage is more marked in case of Internet than the
computer, thus pushing the women off the information
In addition to the above, the extent of computer highway.
and Internet use was also measured. On the basis of the
number of hours in a week subjects worked on computers, In addition to the above, the study estimated the
beyond the class hours at the institute, they were classified levels of empowerment of male and female trainees. An
into three groups --- no use (not at all), low use (less than 5 estimate of composite empowerment was calculated by
hours) and high use (5 hours or more). More than one-fourth adding scores on the psychological, social, educational and
of the sample did not use computers at all, two-fifth reported economic empowerment dimensions and dividing it by four.
low use and less than one-third were in the high use To test whether the male and female subjects differ
category. The gender view indicates that female and male significantly on empowerment, independent sample t-tests
subjects differ significantly with regard to the extent of were computed. The two sexes did not differ significantly on
computer use. The number of females and males in no any of the dimensions of empowerment --- composite (t = -
computer use category differed significantly (χ2 = 7.71, df = 1.40, df =153), psychological (t = -0.93, df =153), social (t = -
1, p<0.01). Even when the women use computers beyond 1.10, df =153), educational (t = -1.53, df =153) and economic
class hours, the extent of use is limited as compared to men. (t = -1.27, df =153). Although no significant differences were
However, this difference is not significant. Thus, the obtained on the different dimensions of empowerment, the
findings provide evidence for the disadvantage of women trend in all cases was for women to experience more
participants (Table 2). empowerment than men. Basic computer education results in
relatively higher composite empowerment for females (mean
TABLE 2 = 30.58, SD = 3.82) as compared to males (mean = 29.75, SD
Extent of Computer and Internet Use for Males = 3.55). Similarly, females reported higher psychological
and Females (mean = 58.83, SD = 9.46) and social (mean = 21.44, SD =
Computer Use χ2 Internet Use χ2 2.99) empowerment than males (mean = 57.51, SD = 8.05;
df=1 df=1 mean = 20.90, SD = 3.05, respectively). Following the trend,
T M F T M F females reported relatively higher educational (mean = 16.17,
No 27.1 16.5 36.6 7.71* 59.4 35.6 80.5 17.39** SD = 2.83) and economic (mean = 25.89, SD = 4.33)
use % * empowerment as compared to males (mean = 15.53, SD =
2.27; mean = 25.04, SD = 3.93 respectively).
Low 41.9 47.9 36.6 0.38 33.5 52.1 17.1 11.08**
use
314
The interaction effect of gender and The trend of the scores indicated that females
computer/Internet use on the empowerment was assessed by with computer use reported higher empowerment than males
means of a two-way ANOVA (Tables 3 and 4). The combined with computer use as well as females and males with no
effect of gender and computer use resulted in a significant computer use (Table 5).
difference in the composite [F (1, 151) = 8.02, p<0.01],
psychological [F (1, 151) = 4.14, p<0.05], educational [F (1, 151) =
9.62, p<0.015] and economic [F (1, 151) = 6.12, p<0.01]
empowerment. In other words, females reporting computer
use obtained significantly higher scores on all aspects of
empowerment as compared to males.
TABLE 3
Results of Two-Way ANOVA Assessing the TABLE 5
Effect of Gender and Computer Use on Empowerment
Dependent Source SS df MSS F t-test Scores of Male and Female Ss with Different Levels
Variable of Computer Use on Empowerment
Composite Gender* 101.17 1 101.17 8.02**
empowerment Computer use Emp Computer Use Internet Use
(CE) Error 1905.34 151 12.62 M F t M F t
Psychological Gender* 306.63 1 306.63 4.14* df=111 df=61
empowerment Computer use CE 29.61 31.71 - 29.63 31.70 -2.09*
(PE) Error 11189.03 151 3.22***
Social Gender* 25.04 1 25.04 2.82 PE 57.39 61.07 -2.43* 57.23 61.19 -1.64
empowerment Computer use
(SE) Error 1340.46 151 8.87 SE 20.85 22.04 -2.22* 20.98 21.56 -0.70
Educational Gender* 58.17 1 58.17 9.62**
empowerment Computer use EDE 15.43 17.02 - 15.19 17.00 -2.82**
(EDE) Error 913.115 151 6.05 4.26***
Economic Gender* 101.84 1 101.84 6.12** ECE 24.75 26.69 -2.61** 25.11 27.06 -2.03*
empowerment Computer use
(ECE) Error 2512.89 151 16.64
**p<0.01;*p<0.05 (one-tailed)
**
p<0.01, *p<0.05
A similar computation was done with respect to
TABLE 4 gender and Internet use. The interaction effect of gender and
Results of Two-Way ANOVA Assessing the Effect of Internet use resulted in a significant difference in educational
Gender and Internet Use on Empowerment empowerment [F (1, 151) = 4.40, p<0.05)]. The trend of the
mean scores indicates that females with Internet use reported
Dependent Source SS df MSS F higher empowerment than males with Internet use as well as
Variable females and males with no Internet use (Table 5). The mean
CE Gender* 21.69 1 21.69 1.59 scores of females (mean = 31.70, SD = 2.91) and males
Internet use (mean = 29.63, SD = 3.59) on composite empowerment were
Error 2065.48 151 13.68 significantly different (t = -3.22, p<0.005). The mean scores
PE Gender* 99.42 1 99.42 1.27 on psychological, social, educational and economic
Internet use empowerment of the two sexes reporting computer use were
found to be significantly different (t = -2.43; p<0.05, -2.22;
Error 11791.48 151 78.09
p<0.05, -4.26 p<0.005, -2.61; p<0.01). Females reporting
SE Gender* 0.02 1 0.02 0.002
computer use was significantly higher than males on each of
Internet
Error 1391.486 151 9.21 the dimensions (Table 5). Thus, technology use provides
EDE Gender* 28.90 1 28.90 4.40* maximum benefits for women.
Internet use
Error 990.60 151 6.56 IV. DISCUSSION
ECE Gender* 11.80 1 11.80 0.68
Internet A. Computer Ownership/Internet Connection, Access,
Error 2607.009 151 17.26 1.27 Computer/Internet Use
*
p<0.05
315
The study assessed computer access and usage of the had primary education or were illiterate), parents may not
technology in terms of family ownership of computer, be aware of the potential benefits of Internet.
having an Internet connection at home and the various
points of access to computer/ Internet. Computer Besides private access, the different points of
ownership and home Internet connection for the sample is access were explored. It was found that significantly
considerably lower as compared to ICT penetration in higher number of females access computers at the training
Mumbai where the study was conducted. The city ranks institute. Access at the institute does not require cost
first in PC penetration with 32 per cent household owning investment, but does not provide Internet access either.
a PC [22]. Research indicates that Mumbai and Delhi Further, the women will be left vying for access after
together constitute close to 70 per cent of the total installed completion of the course. Cyber café is significantly more
computers in Indian homes and 46 per cent of the important point of access for males than females. This
computer owners have an Internet connection [12]. Thus, difference may be explained by three reasons --- nature of
even though the figures of computer ownership and use, cost and socio-cultural attitudes. Males seek to use the
Internet connection for the sample under study are better Internet, which the computer training centre does not
than the national average (2.5 and 4.2 % respectively [8]), provide. In urban India, cyber cafes are the predominant
they are much lower than the figures for Mumbai city. The public access model for Internet usage. Even though the
sample consisting of college students and fresh graduates, cost of surfing in the cyber cafes has reduced from Rupees
have no personal disposable income to purchase a 50 to Rupees 20 per hour in the last decade, this is still out
computer. Besides 80 per cent of the subjects in the study of reach of the low income groups. Young girls may
belonged to low income group and hence cannot afford a hesitate to use family resources for surfing the Internet,
personal computer or Internet connection. In addition, due thus curtailing their visit to cyber cafés. Third, socio-
to low education levels of the parents, there is lack of cultural restrictions add to the disadvantage of young
awareness about the potential of the Internet. Hence, even women as families impose sanctions against cyber cafes as
when they purchase a computer to support education of these are characterized as boys’ hangouts and lack female
their children, less than half of the sample has an Internet attendants [26]. Besides, families may be more willing to
connection. Besides, technology access at home is provide financial support to the boys for accessing
adversely affected by space constraints. Households in technology, thus placing them at an advantage as far as
Mumbai utilize an average of 2.9 square meter of floor access at cyber café is concerned.
space per person, one of the lowest in the world [23].
Given the low socio-economic levels of the sample and the Workplace does not emerge as an important
pigeon-hole like houses of Muslim ethnic enclaves of access point as the sample consists predominantly of
Mumbai, the disadvantage for the young members of the college students and fresh graduates. Only six participants
community with regard to computer ownership and (4 men, 2 women) were working in part time jobs such as
Internet connection follows. giving home tuitions, office assistant, etc. Thus, despite
possessing the technical skills, the women from the current
The two sexes did not differ significantly on sample have lower access to the technology as compared
computer ownership and home Internet connection. This is to their male counterparts. Limited Internet connection at
because majority of the sample belong to low income home and cyber café usage seriously disadvantages the
groups and are financially dependent on their families. women as research indicates that Internet provides women
Hence, both boys and girls were affected by economic access to information, greater participation in decision-
constraints. However, the trend indicates that males are making and networking, thus empowering them [18].
slightly advantaged as compared to females. There is lack
of gender segregated data on computer ownership, hence Subjects were asked the number of hours they
stretching a point, a study on difference in family used computers/ Internet in a week, besides the regular
computer usage is considered. A study on computer usage
class hours at the training institute. On the basis of their
in India indicates that 82 per cent of males compared to responses, the usage was classified into three levels --- no-
only 16 per cent of females are the primary users of family use, low-use and high-use a week. A similar classification
computer [12]. Besides, Indian families tend to make
was done with regard to Internet use. Two issues are
higher investments in education of the boys compared to apparent with regard to computer/ Internet use; one, the
girls due to different role expectations [24]. As computer disparity across technology --- computer v/s Internet; and
is identified as an educational tool, it is likely that families
gender --- males v/s females. While a quarter of the
would purchase computers for their sons rather than subjects fell in no-use category for computer use, with
daughters. In addition, males identify the home computer regard to the Internet, this rose to 60 per cent. One-third of
as personally owned; while females identify it as
the subjects reported high computer use but only 7.1 per
belonging to the head of the family [25]. Cost barrier is cent report high Internet use. These figures bear testimony
more prominent in case of Internet connection as
to the fact that computer use and Internet use are distinct
compared to computer ownership, in terms of recurring categories and merging them into a single unit could be
monthly payments. In addition, due to low education misleading. While using computers requires only access to
levels (one-fourth of the fathers and half of the mothers
the machine, for Internet use the cost gets multi-fold due
316
to investments in modems, optical fibers, charges of paragraphs.

Internet service providers and computer maintenance. This
may explain the lack of Internet usage. Although the two sexes do not differ significantly on all
measures of empowerment, the trend in all cases is for
The other reasons for limited Internet use are the females to be more empowered than males. This is
non-availability of subsidized public access points, limited supported by research that indicates that though
English language competency and socio-cultural attitudes. modernization and technological development initially
Though cyber café is the main access point in urban India, disadvantaged women, ICT promises a technological U-
the cost of usage is beyond the reach of low income turn [29]. Findings of this study are strengthened by
groups. Research indicates that one-third of Internet users evidence from a qualitative study of the impact of
in Indian cities have an income above Rupees 25,000/- , computer education that indicates that more women than
while only 4.4 percent of the Internet users have an men experience higher boost in self-confidence and secure
income below Rupees 5000/- [11]. Access at cyber cafes is computer related jobs [30]. It is likely that since South
beyond the reach of low economic groups. As majority of Asian women in general are socio-economically deprived,
the sample are poor, they have little disposable income their expectation from something new will be higher than
that could be used for Internet access. Research indicates males who have better chances of being exposed to these
that there is a strong relationship between the use of the technologies before hand.
Internet and ability to speak English [9, 27]. Surveying a
group of students from Gujarat, it is observed that students Analyses of the participants’ response on
educated in the vernacular medium at school struggled at psychological empowerment indicate that females
college where the medium of instruction is English [28]. expressed slightly higher meaning (mean =14.91, SD
As 70 per cent of the sample in the present study had Urdu =2.76), self-determination (mean = 14.83, SD = 3.18) and
as the medium of instruction in school, their limited competence (mean = 15.11, SD = 2.57) as compared to
English language skills probably affected their Internet males (mean = 14.78, 14.01, 14.48; SD = 2.32, 2.59, 2.58
use. Although considerable web content is now available respectively). In other words, computer education resulted
in a variety of Indian languages; a cursory review of the in creating a sense of new possibilities for involvement,
Urdu websites by the first author in 2007, revealed that independence and self-confidence for women. On the
they focused on literature, poetry and news rather than other hand, males (mean = 14.23, SD = 3.00) reported
education and employment. Hence, tangible gains do not slightly higher impact than females (mean =13.97, SD =
follow for the subjects surfing these websites. 3.28), indicating that computer education provided them
possibilities for enhanced control over their environment.
B. Computer Education, Use and Empowerment Thus, the psychological gains for females and males focus
on micro and macro levels respectively. Women reported
In addition to computer access and use, the study relatively higher gains on all aspects of social
focused on analyzing gender differences in empowerment empowerment --- feeling contemporary, enhanced status,
levels following from technical skills training. It was social comparison, and connectivity --- as compared to
found that basic computer education yields similar males. This is in line with research that women perceive
empowerment gains for males as well as females. There is family gain, communication and social gain as the first,
no research comparing men and women on empowerment second and fifth most important benefit of personal
flowing from computer education hence, the explanations computer use [20]. In addition, the social connectivity
offered are speculative in nature. Lack of gender aspect is supported as women are responsible for
differences on empowerment could be due to four reasons. maintaining family and kin relations [31].
First, the sample consisted of young, college going or
fresh college graduates from a metropolitan locale on the Women report relatively higher educational
verge of entering the job market. The homogenous nature empowerment than men with higher scores on information
of the sample makes for similar sources of empowerment on courses/colleges/universities, accessing online
such as marketable skills. Second, empowerment as journals/books/reports. This is corroborated with the
assessed by the scales focuses on the cognitive level and research that when females use computers, they do so
reflects the subject’s estimate of the possible gains from mainly for educational purposes; unlike males, who use it
computer education. It does not assess the tangible for general purposes or playing games [11]. More than
benefits flowing from technical skills, thus overlooking one-third females have completed college; hence search
constraints in the real world that affect the actual benefits for educational information follows. Males score slightly
of computer education. Third, the item pool of the higher on the use of computers to prepare reports and
different empowerment scales is narrow and specific to presentation, may be because unlike the women, more than
this sample. It is likely that if a more diverse measure was half of the men are in college. Females reported slightly
adopted, the two sexes would indicate differential higher scores on four of the five aspects of economic
empowerment levels. Fourth, men and women score empowerment --- job opportunities, managing family
differentially on the items on the scales, thus canceling out responsibilities with work, earning from home and
the gap. This will be discussed in the following assisting in family business. There is a general perception
317
in urban India that a combination of college degree and avenue to become economically independent. Even though
computer skills enhances the possibility of employment. the advantage of females in social empowerment is not
Since a higher number of women participants had already significant, the trend is in the same direction. Media
completed college, they felt better equipped to find a job. advertisements of personal computers in India, project the
The economic gains for females focus on the potential of women computer user as a modern, westernized
IT for combining family with work, a gain which ensures individual; cues that highlight the social status and
maintenance of traditional values and yet being modernity associated with this technology [36]. A
productive. On the other hand, males reported a higher computer being associated with masculinity, its usage
possibility of setting up personal business. This may be brings the females on an equal social footing, enhancing
because belonging to business families (47%), males may feelings of being contemporary and moving with the
be socialized to start their own business. Besides, families times.
could be more willing to provide them the financial capital
due to future role expectation. The interaction effect of gender and Internet use
was found to result in significant difference on educational
Looking at computer/Internet use and empowerment, with females using Internet obtaining
empowerment, it was found that significantly more highest mean empowerment scores as compared to their
females (more than one-third) fell in the no-use category male counterparts. This is not surprising as research
of computer use, than males (one-fifth). Similarly, indicates that when women used technology, they do so
significantly more females (more than 80 per cent females) for educational purposes unlike males who use it for
than males (one-third) report no Internet use. Although general purposes or playing games [11]. The sample
both computer and Internet use was relatively limited for included higher number of females who had completed
females, they were more disadvantaged with regard to college; hence they are likely to explore opportunities for
Internet use. This is a cause of concern, as most of the continuing education using Internet. Thus, young Muslim
research that make a case of ICT-led empowerment of women derive educational benefits out of using Internet.
women, focus on the Internet technology and its ensuing
benefits like networking, political activism and e- With regard to other aspects of empowerment, it
commerce [14, 32]. Women’s low Internet use in this was found that there was no significant interaction effect
sample is in line with previous findings. In a survey of of gender and Internet use. As far as psychological and
Macau residents, it was found that Internet users were social empowerment is concerned, the trend is in favor of
more likely to be males [33]. In a survey of Indian Internet females. This is in line with previous research that Internet
users, it was found that females start using cyber cafes, a use increases self-confidence, facilitates networking and
common Internet access point in urban India later; use expands income generation capacity [14]. There is limited
them less often and for shorter duration per session [11]. comparative evidence on the differential benefit to males
and females from computer and Internet usage. In light of
Gender and extent of computer/Internet use this, the present research fills an important gap by
significantly influence empowerment levels. Findings providing empirical evidence for the benefits of computer
indicate that with the exception of social empowerment, education and technology usage for empowering the two
the interaction effect of gender and computer use is sexes. However, more research with gender as an analytic
significant for composite and the other three kinds of variable is necessary before a clearer picture emerges.
empowerment --- psychological, education and economic.
The mean empowerment scores of males and females V. CONCLUSION AND POLICY
indicate that females reporting computer use are more RECOMMENDATIONS
empowered than males with computer use as well as
females and males without computer use. Thus, the
Five major conclusions can be drawn from the
relevance of computer use for empowerment of females study. First, computer ownership and Internet connection
cannot be overstated. It is in line with the research that is very low for the sample compared to figures for
emphasizes psychosocial, educational and economic Mumbai city. Hence, the minority poor as other urban
benefits of computer access and use for women [34, 35]. A poor are on the wrong side of the digital divide. Second,
study of six projects of the Information for Development computer training centre is the major access point for
Program (infoDev) found that computer skills increased females, while cyber café emerges as the chief access
self-esteem and promoted self-confidence of the young point for males. The primary access model in urban India
women participants [18]. Similarly, a study on novice is cyber café and is underused by women. Related to this,
women computer learners found that computer skills the third conclusion is that when women do have access to
resulted in increased self-esteem and enhanced status [20]. computers and the Internet the extent of usage is limited as
The sample of the present research comprised of young compared to males. Thus, the poor urban minority women
college students; hence, educational benefits of using are off the information highway. Fourth, technology use
computers like preparing class assignments or empowers females more than the males, thus making a
presentations seem relevant. Further, females using strong case for ensuring sustained access and usage. The
computers understand that this technology offers an fifth important conclusion is that computer education is a
318
key gender equalizer as unlike other technologies [1] K. Kenniston, “The Four Digital Divides,” in Bridging
women’s engagement with new technologies open’s new the DigitalDivide: Lessons from India, K. Kenniston
possibilities and empowers them as much as the men. and D.K. Das, Eds. New Delhi, India: Sage, 2003.
[2] Sachar Committee Report, “Social, Economic and
In light of the above conclusions, policy Educational Status of the Muslim Community of India,”
recommendations are suggested. In order to widen access Prime Minister’s High Level Committee, Government
points and encourage technology use amongst those at the of India, 2006.
edge of the information society, community access points [3] A. Razaack and A. Gumber, “Differentials in Human
based on the tele-center model should be set up in urban Development: A Case Study of Empowerment of
locales with public-private partnerships. Government can Muslims”, Programme of Research on Human
provide tax benefits to business houses investing in such Development, New Delhi, India: National Council of
endeavors. Besides, to improve women’s access, cyber Applied Economic Research, 2003.
cafés can reserve special timings, provide separate space
for girls, appoint female support staff and open women [4] A. Shariff and A. Razzack, “Communal Relations and
only cyber cafes. This will help the young women to Social Integration,” in India Social Development
overcome the inhibitions in visiting these places for Report, Council for Social Development. New Delhi,
technology access. In addition, civil society organizations India: Oxford University Press, 2004.
can create awareness about potential benefits of [5] O. Khalidi, “Ethnic Group Recruitment in the Indian
technology use and facilitate higher familial support for Army: The Contrasting Cases of Sikhs, Muslims,
technology use by women. Gurkhas and Others,” Pacific Affairs, vol. 74, no. 4,
pp.
In order to maximize tangible gains from 529-552, 2001.
[6] Z. Hasan,. and R. Menon, “Unequal Citizens: A Study
technical skills the computer training center has to be
diversified. The center can liaison with the local small and of
medium enterprises to provide placement support to its Muslim Women in India”, New Delhi, India: Oxford
University Press, 2004.
trainees. It can serve as an intermediary to secure micro-
credit to its students, with special concessions for girls. [7] R. Heeks, “Social Outsourcing: Creating Livelihoods”,
These finances can be used for purchase of a computer and Information for Development (i4d), September, pp. 17-
19, R. (2006.
setting up computer-related business such as a cyber café.
Further, it can assist the trainees to form a self-help group [8] IMRB “International Market Research Bureau,” 2005.
to pool their resources and secure credit for computer- Available: http://www.imrbint.com
[9] A. M. Haseloff, “Cyber cafés and their Potential as
related income generation. Urban Muslims being largely a
Community Development Tools in India,” The
petty business community, their business spirit may assure
Journal of Community Informatics, vol. 1, no. 3, pp.
sustainability. What is required is the initial thrust from an
53-65, 2005.
external agency.
[10] A. Pasricha, “To Get Online, Most Indians go to
Cyber cafés,” VOA News Service, 2004. Available:
It is likely that the impact of these policies can be http://apiap.blogspot.com/2004_10_01_apiap_archive.htm
delimited due to societal conditions such as discrimination l
and prejudice. Hence, fair disbursement of micro-credit [11] A. Haseloff and R. Ghadially, “Gender, Cyber cafés
and sensitization to overcome discrimination will help to and Internet Access: Bridging the Digital Divide?” in
widen the gains flowing from technology. These measures Urban Women in Contemporary India: A Reader, R.
will enable the Muslim youth to compete on an equal Ghadially, Ed. New Delhi, India: Sage, 2007, pp. 270-
footing and lead to their mainstreaming in India’s 283.
economic growth. [12] A. Venkatesh, “Computers and New Media
Technology in Indian Households: Based on a Study
As far as directions for future research are of Eight Major Cities in India,” Centre for Research
concerned, a more comprehensive study can yield on Information Technology and Organizations, 2000.
representative results. The findings of the study are limited Available:
in generalizability as data was not collected from locations http://www.crito.uci.edu/noah/paper/IndianSurvey1.pdf
other than Mumbai and across socio-economic class. A [13] S. Huyer, “Gender, ICT and Education,” 2003.
comparative sample from other minority groups may Available: http://www.wigsat.org/engenderedICT.pdf
provide better insight into the standing of Muslim youth [14] S. Mitter, “Globalization, ICTs and Economic
on the information highway. Empowerment: A Feminist Critique,”
GenderTechnology and Development, vol. 8, no. 1, pp.
5–29, 2004.
[15] J. A. Conger and R. N. Kanungo, “The Empowerment
REFERENCES Process: Integrating Theory and Practice,” Academy
of
Management Review, vol. 13, pp. 471–482, 1988.
319
[16] D.L. Corsun and C.A. Enz, “Predicting Psychological [28] V. Ramanathan, “Ambiguities about English:
Empowerment among Service Workers: The Effect of Ideologies and Critical Practice in Vernacular-
Support-Based Relationships” Human Relations, vol. medium College Classrooms in Gujarat, India,”
52, no. 2, pp. 205-224, 1999. Journal of Language, Identity, and Education, vol. 4,
[17] G. M Sprietzer, “Psychological Empowerment in the no. 1, pp. 45–65, 2005.
Workplace: Dimensions, Measurement and [29] A. Goyal, “ICTs and the Technological U-turn for
Validation,” Academy of Management Journal, vol. Women,” in Urban Women in Contemporary India: A
38, pp. 1442– 1465, 1995. Reader, R. Ghadially, Ed. New Delhi, India: Sage,
[18] N. Hafkin, “Are ICTs Gender Neutral? A Gendered 2007, pp. 255 – 269.
Analysis of Six Case-studies of Multi-donor ICT [30] F. Umrani, “Computer Adoption among Mumbai
Projects,” Background paper to UN/INSTRAW Virtual Muslim Youth: Empowerment and Impact
Seminar Series on Gender and ICTs, 2002. Assessment”, Unpublished Dissertation, Indian
Available:http://www.un- Institute of Technology Bombay: India, 2007.
instraw.org/en/docs/gender_and_ict/Hafkin.pdf. [31] R. Ghadially, “Women’s Friendship,” in
[19] G. Kelkar, “Women in the Digital Era: Gender and Encyclopedia of Women and Islamic Cultures, S.
Culture,” Paper presented at the International Joseph, A. Najmabadi, J. Peteet, J. Siapno and
Conference on Women in the Digital Era: J.Smith, Eds. Netherlands: Leiden, EWIC, pp. 194-
Opportunities and Challenges. Organized by 195, 2003.
Annamalai University and Software Technology Parks [32] S. Huyer and M. Carr, “Information and
of India, at Chidambaram, India, 2004. Communication Technologies: A Priority for
[20] F. Umrani and R. Ghadially, “Empowering Women Women,” Gender, Technology and Development, vol.
through ICT Education: Facilitating Computer 6, no. 1, pp. 85-100, 2002.
Education,” Gender, Technology and Development, [33] W. H. Cheung, “Internet Adoption in Macao,”
vol. 7, no. 3, pp. 359-377, 2003. Journal of Computer Mediated Communications, vol.
[21] N. Hafkin and N. Taggart, “Gender, Information 7, no. 2, 2001. Available:
Technology, and Developing Countries: An Analytic http://jcmc.indiana.edu/vol7/issue2/macao.html
Study,” Academy for Educational Development, for [34] S. Huyer and T. Sikoska, “Overcoming the Gender
the Office of Women in Development, Bureau for Digital Divide: Understanding ICTs and their
Global programs, Field Support and Research, US Potential for Empowerment of Women,” Virtual
Agency for International Development, 2001. Seminar Series on Gender and ICTs, United Nations
[22] Dataquest, “Momentous years of Indian InfoTech”, International Research and Training Institute for the
Dataquest, vol. 18, no. 1, 25-28. Advancement of Women (INSTRAW), 2003.
[23] A. Bertaud, “Mumbai FSI Conundrum: The Perfect Available:
Storm: The Four Factors Restricting the Construction http://www.uninstraw.org/docs/gender_and_ict/Synth
of New Floor Space in Mumbai,” 2004. Available: esis_Paper.pdf
http://alainbertaud.com/AB_Files/AB_Mumbai_FSI_ [35] N. Primo, “Gender Issues in the Information
conundrum.pdf Society,” United Nations Educational, Scientific and
[24] K. Chanana, “Interrogating Women’s Education: Cultural Organization (UNESCO) Publications for the
Bounded Visions, Expanding Horizons,” New Delhi, World Summit on the Information Society, 2003.
India: Rawat Publications, 2001. Available:
[25] R. Heimrath and A. Goulding, “Internet Perception http://portal.unesco.org/ci/en/file_download.php/2505
and 61f24133814c18284feedc30bb5egender_issues.pdf
Use: A Gender Perspective,” International Journal of
Human-Computer Studies, vol. 55, pp. 919-938, 2001. [36] R. Ghadially and S. Ranganathan, “Gendering
[26] F. Umrani and R. Ghadially, “Economic Impact of IT Computer Marketing: A Study of Print Media
Education among Muslims,” i4D Information for Advertisements in India,” Asian Women, vol. 22, no.
Development, 2006. 1, pp. 69-96, 2006.
[27] V. Vehovar, Z. Batagelj and K. Lozar, “Language as
a Barrier,” 1999. Available:
http://www.isoc.org/inet99/proceedings/3i/3i_3.html
320 1
eServices Provisioning in a Community

Development Context Through a JADE MAS
Platform
Mamello Thinyane, Alfredo Terzoli, Peter Clayton
Department of Computer Science, Rhodes University
{m.thinyane, a.terzoli, p.clayton}@ru.ac.za
Abstract—A growing proliferation of ICT4D interventions leveraging new agriculture, improving human capital and sus-
has necessitated the exploration of innovative solutions for the taining the rural environment [1]. The adoption of ICT in rural
provisioning of eServices in rural, marginalized communities. communities is a phenomenon with pronounced variability and
The challenges currently faced in these interventions include:
situating the developed applications within the cultural and one which is still not thoroughly understood [2]. It is therefore
ethnographic context of the target communities, integrating still not very clear why some rural communities adopt ICT and
greater levels of granularity and flexibility within the applications why others are sometimes even resistant to the integration of
for increased context sensitivity, handling the intermittence and ICT into their lives. However, research has found that some
instability of supporting infrastructural services. These are the of the leading reasons for rural communities adopting ICT are
challenges that we address in the context of ICT4D intervention
undertaken in a rural community in the Eastern Cape province more social than economic. These communities adopt ICT for
of South Africa. We explore the design and implementation of a improvement of their well being as far as access to information
Multi-Agent System (MAS) for this community as a platform for is concerned and as far as facilitating communication in the
provisioning of context-sensitive eServices, and highlight some communities, and to that end the widely used applications of
observations with regards to the applicability and adequacy of the Internet in these communities are email applications and
the solution.
search engines [3].
Index Terms—eServices platform, ICT for development, eth- ICT is indeed providing opportunities for dealing with rural
nocentric applications, JADE MAS
poverty and in many ways it is challenging the traditional
paradigms of doing business, delivering services to citizens,
I. I NTRODUCTION and running societal institutions [4]. The dynamics associated
A number of factors are fueling the adoption of ICT as an with ICT in development are two fold: on one side its able
enabler towards community development, poverty alleviation to leap-frog community development based on the ability of
and the elimination of the digital divide. One of these is the the society to synergistically embrace the technology (e.g. the
realization of the benefits that are afforded to a community Asian economies: Hong Kong, Taiwan, Singapore) and on the
from participating in the global information and knowledge other side, ICT can further the marginalization, participatory
society, and the other is the sometimes mis-applied observation exclusion from global economies and retardation of commu-
of the penetration of ICT and technology in the developed nities that are not able to integrate technology into their life-
first world countries. The latter factor often leads to the systems [5].
technological determinism flaw, in which the ICT adopters Technology, in and of itself is not a panacea for the underde-
incorrectly assume that the technology that works in the first velopment woes of communities, it is however a pre-requisite
world countries will necessarily work in third world contexts. for social development in this day and age [5]. Technology is
The third world ICT4D interventions are undertaken in a not a target towards community development and social well-
context of pronounced socio-economic challenges that are being [6], but rather a tool for facilitating the achievement of
often not prevalent in first world contexts. They are also a desirable features in a society: well being, health, peace,
undertaken in contexts of extreme rurality characterised by and communality. To a large extent human activity depends
lack of infrastructure (roads, electricity, telecommunications), on information and therefore a synergestic interaction of
lack of service provisioning (health, education, government technology and information leads to a competitive advantage
services), and high levels of marginalization (information, for societies [5].
economic). One of key contributions to understanding the dynamics
It is therefore generally within this landscape of rurality that associated with the adoption of ICT in rural communities
ICT interventions are explored to facilitate development and and in development contexts is made in the discipline of
enhancement of life in rural communities. This environment ethnocomputing. One of the primary tenets of ethnocomput-
of rurality presents unique challenges and opportunities. Five ing is the realization of the culture specific influences on
of these challenges that have been identified by Drabenstott computing and subsequently on the Internet [7]. The aspects
and that are critical in shaping the rural economic outlook of computing that are universal and that are aligned with
are: tapping digital technology, encouraging entrepreneurs, the general human nature, are easily transplantable from one
321 2
community to another. Aspects that are cultural however, need only at a few of the locations in the community, no centers of
to be considered, adapted and validated for adequacy within governmental services provisioning, a small under-resourced
the environment where they’re being implemented [8]. ICT clinic and poor access to information resources for the schools
for development should therefore take into consideration the and the community members at large.
ethnographic considerations and expressions of a community The initial objective of the project was to implement a
to avoid the technology determinism flaw that has plagued prototype of an eCommerce portal in support of the en-
many ICT4D projects [9]. trepreneurship activities already prevalent in this community.
The ability to adapt and localize ICT solutions into the dif- This has been revised and enlarged to establish a distributed
ferent contexts as far as the cultural expressions, the language, community communication and knowledge platform. Over the
the metaphors, the themes, the interaction modalities, and three years that the project has been running a number of
the associated knowledge are concerned, is becoming a key activities have been undertaken and different infrastructure
requirement for ICT4D interventions. This requirement factors setup:
directly into the sustainability and the successful integration of
the solution into the life-system of the communities. Another 1) Community rapport: One of the key achievements in
requirement is for the solutions to address issues of infrastruc- the project has been establishing a working relationship
ture intermittence and unstable deployment environment that with the community as far as exploring ICT based
are typical in rural marginalized communities. This fact of solutions. The buy-in from the different stakeholders
intermittence necessitates that measures and mechanisms are (the headman, the schools, the ICT ’champions’) in the
integrated at all the levels in the service provisioning stack, community provides the essential support that ensures
to be able to recover from interruptions, and to tolerate faults a synergy out of working together with the community.
in the supporting infrastructure. The other constraint as far as It is on the basis of this relationship that the project is
handling faults and interruptions in the network is concerned transitioning into a Living Lab where the community
is the lack of technical expertise in these communities. As members become the active (and key) participants in
far as possible the services should automatically recover and undertaking the implementation of solutions.
self-heal without the need for a technical expert’s intervention. 2) Training: Prior to the introduction of the project in
In this paper, we discuss a solution that we have im- this region, there was minimal computing literacy in
plemented particularly to address these requirements in ICT the community. We have successfully run a number of
applications. The next section introduces the project that we training sessions, with the schools as the centers for
are undertaking, highlighting the different eServices applica- training. The training that we’re undertaking is based
tions being developed for the community. We then discuss on the openICDL curriculum and one of the training
the design factors that should be taken into consideration objectives is ’confidence’ not ’competence’. What this
for context-sensitivity, followed by the detailed discussion of means is that we aim to get the community confident
the Multi-Agent System (MAS) we have developed. The last enough to learn on their own and thus encouraging
sections provide some observations that have been made with a culture of life-long learning in the community. The
regards to the adequacy and applicability of the platform in schools have become the points of access to the ICT
development contexts. solutions for the community because they’re some of
the few places where there’s electricity. The schools
II. ICT4D I NTERVENTION IN SA are also traditionally the centers for learning within the
The ICT4D landscape in South Africa is one in community, and as such we have managed to tap into
which there’s a growing effort from the government, non- that inertia to establish a training system in which we
governmental organizations and academic entities in under- ’train the trainers’ who then continue training the other
taking community development interventions based on ICT people in the community who are not as yet computer
solutions. It is within this context that a joint ICT4D project literate. We have currently setup ’labs’ in five schools
between Rhodes University and University of Fort Hare was in the region ranging in size from 2 computers to 15
initiated in 2006, with support from government and industry, computers.
towards exploring ICT-based interventions that can be under- 3) WiMAX backbone: We have setup a WiMAX network
taken in rural marginalized communities. between the schools where we have established comput-
ing labs. WiMAX is a suitable solution in this case due
to the landscape of the region, and the distances between
A. Overview and Current Status the different points. It also provides adequate bandwidth
The site of this project is a deep-rural, marginalized re- for the services that are provided within the network.
gion of Dwesa, which is located in the former homeland of 4) VSAT connection backhaul: The link to the Internet
Transkei. Dwesa has an estimated population of about 15000 is via VSAT, which is setup in one of the schools
distributed in 2000 households. Dwesa is predominantly a and shared between the rest of the schools through the
subsistence farming community that characterizes the majority WiMAX network. This also provides a link back into the
of rural communities in South Africa and other developing network for remote management and troubleshooting.
regions. Typical of many rural communities, there is a lack 5) VOIP telephony: An Asterisk PBX has been setup at
of infrastructural services in Dwesa: gravel roads, electricity one of the schools to provide VOIP telephony services
322 3
for the whole community. In one way, this allows the be afforded by porting this service onto the MAS eService
community and the schools to collaborate a lot easily platform.
with each other, and in another way it also allows for the 3) eHealth service: One of the primary objectives of the
provisioning of voice based services for the community eHealth services platform is the preservation and aggregation
which is a relevant alternative in a community where of the local traditional health knowledge in Dwesa. This is the
there are people who are illiterate. knowledge that has been developed over the years and that is
The established infrastructure and activities undertaken thus available in the community. The portal also provides accessi-
far provide the basis on which to explore the provisioninig of bility to other sources of medical knowledge. While the portal
different, relevant network services for this community. is aimed for use by the average members of the community,
it can also specifically provide the extra knowledge resources
for the clinics in Dwesa.
B. Associated eServices 4) Indigenous Knowledge portal: One of the reason for
This section highlights a few of the key eServices that the slow adoption and integration of ICT into the lives of
are being integrated into the network in Dwesa. This is rural communities is the lack of relevant knowledge on the
therefore not an exhaustive list of all the different services Internet. This portal aims to alleviate this by allowing for
deployed. Some of the ones not discussed here include: a the community members to author, create and record the
local Wiki which the community can populate and access, knowledge that is relevant and local to their region, to establish
access to offline versions of Wikipedia and Project Gutenberg, and enforce authorization permissions on ’their’ knowledge,
eJudiciary service portal for the community, a community and to have access to other knowledge that is available of the
help-desk system. These are the services that will be ported portal. This of all the services requires the greatest levels of
onto the JADE MAS eService platform to allow for a greater flexibility and context-sensitivity due to the heterogeneity of
contextualization and sensitivity to the local environment. the underlying knowledge, the diversity of user profiles and
1) eCommerce service1 : The Dwesa community has a high preferences, and the multiplicity of end-user devices that are
economic potential due to the fact that there is a nature utilized.
reserve and arts and craft entrepreneurship activities in the
region. There’s therefore the eco-tourism and cultural-tourism III. D ESIGN C ONSIDERATIONS
potential that can be explored and activated through the The development of a MAS eService platform has to address
eCommerce portal. The current arts and craft entrepreneurship a number of challenges. Of particular significance are the
is also currently operating within the geographical constraints challenges that pertain to rurality, processing of indigenous
of Dwesa and the eCommerce portal opens it up to wider knowledge, keeping abreast of technological developments
international markets. A basic functional eCommerce portal around web 3.0, and also ensuring ethnocentricity and context
has been setup and is operational. sensitivity of the developed solution. The design has to take
This portal however has the limitations that it is not into consideration a new way of structuring and layering the
extensively contextualized to the community. The aspects service platform to achieve the following goals:
under consideration for integration in the subsequent versions 1) Provision of an end-user device agnostic interface to the
include: implementing an ontologies based back-end for a underlying knowledge - provision for handling hetero-
semantic processing of the underlying eCommerce knowledge; geneous device requests
allowing for alternative interfaces into the portal which would 2) Allowance for varied interaction modalities with the
be more appropriate for the community (e.g. voice-based users
interface for some of the illiterate arts and craft entrepreneurs, 3) A context-senstive and ethno-centric knowledge plat-
and a J2ME mobile phone interface which would make the form - through encapsulation of local knowledge and
portal more accessible for the majority of the users); and a the emulation of local knowledge system dynamics.
more integrated handling of different associated multi-media This is towards achieving a seamless mobility between
content on the eCommerce portal. knowledge exchange in the real world and knowledge
2) eGovernment service2 : Government service provision- exchange in the virtual space (on the knowledge plat-
ing is very minimal in Dwesa. This eService portal aims to form)
facilitate the ease of access to government services. To a 4) Handling of multimedia knowledge and media transcod-
large extend this portal interfaces and aggregates the availing depending on the capabilities of requesting devices
able government services to provide a one-stop shop for the 5) A future-proof platform that embraces the developments
community. The activities that can be undertaken through this in knowledge engineering technologies and service ori-
portal include: downloading application forms (grants, birth ented architectures
certificates), accessing information released by the govern-
ment, reporting matters to the police, and engaging in an IV. T HE AGENT-BASED P LATFORM
open discussion forum around government related matters.
The similar improvements to the eCommerce portal will also Application development platforms and architecture are
constantly evolving to accommodate the needs within the
1 http://www.dwesa.com software development domain. Evident characteristics in the
2 http://www.dwesa.com/egov nature of applications that are being developed, is that there’s
323 4
Figure 1. A distributed JADE MAS platform
Figure 2. Multi-host eService platform MAS container
an increasing heterogeneity and therefore a need to operate

in a context of multiplicity of protocols, languages, technolo-
gies and programming paradigms. Application development
agent platform with containers distributable across different
architectures are also harnessing the developments as far as
hosts (Fig 1). Fig 2 is a depiction of the eServices platform
distributed networking is concerned, and this has resulted
MAS through a Remote Management Agent (RMA) showing
in the implementation of service composition architecture
the Main Container on one host (thepentagon) which contains
implemented on the web, and also in the implementation of
most of the platform agents including the AMS and the DF,
services that incorporate the development from the semantic
and also another container located on another host (thebox)
web community.
and containing the Multi-Modal Interaction Agent (MMIA).
The provisioning of an architecture for distributed systems
in heterogeneous usage contexts is one that has been pursued
extensively in literature [10, 11]. One architectural paradigm
that has been explored extensively is the web services archi-
tecture. The other perspective to problem solving in hetero-
geneous environments is through agents. Agents are different B. Platform Agents
from web services in that they are [12]: problems solvers; pro-
active; goal-oriented; context-aware; and autonomous. One of
the predominantly implemented agent standards is the IEEE This eServices platform has been realized as a system of
Foundation for Intelligent Physical Agents (FIPA) standard, independent agents that collaborate and communicate for the
the other common standard is the ”Multi Agent Facility” achievement of the common goal. The agent architecture fits
(MAF) standard which is provided by the Object Management within the PIASK architecture to provide decoupled, modular,
Group (OMG) [13]. units (agents) of computation that exist within a community
[15]. We developed the PIASK architectural pattern as concep-
tual layering of aggregated application functionality for multi-
A. JADE MAS modal, multimedia application deployment in heterogeneous
JADE is the abbreviation of Java Agent DEvelopment contexts. PIASK provides a distinct categorization of function-
framework and is a middle-ware and a software framework ality into presentation, interaction, access, social networking
for the development of agent based applications. JADE is and knowledge base layers [16].
developed by Telecom Italia Lab (TILAB) and is written
entirely in Java [14]. JADE implements the FIPA agent stan- 1) Hyper-Text Transfer Protocol Agent (HTTPA): Access to
dard and provides a platform that can be distributed across the PIASK platform is facilitated by the HTTPA access layer
different platforms and machines. Within the JADE MAS agents whose role is to implement the specific transport level
platform, agents exist within containers, which can be located protocols for communication with end-user devices. The ac-
on different hosts. Therefore to be part of community, agents cess layer agents provide the ability to handle the multiplicity
have to exist in containers that join a single Main Container of the end-user devices in an extensible manner and in a way
to form a distributed system that appears as a single platform. that is separate from any other operational structures of the
All the agents that are associated with a Main Container form platform.
a single agent platform. The Main Container is different from The HTTPA handles request from web browsers but also
other containers in that it holds two platform specific agents: handles requests from the openVXI VXML browser, which
the Agent Management System (AMS) and the Directory runs as an Asterisk application [?]. The HTTPA in essence
Facilitator (DF) [14]. provides a web server service that receives requests from any
The eServices platform MAS is implemented as a single HTTP compliant browsers.
324 5
The MMIA is also responsible to managing the user sessions

on the platform. This is maintained through an in-memory list
of the current sessions. For every session on the platform,
the MMIA queries the Social Networking Agents (SNA) to
determine the relevant users details, and stores this in the
session list. The details that are queried from the SNA include,
the IM address of the user and the online status, the telephone
extension of the user, and their email address. This information
Figure 4. Multi-Modal Interaction Agent
is utilized by the MMIA to determine the best channel of
interaction with the users. For example, if a user is logged
onto platform through a VOIP agent, and requests a download
of a certain non-audio file, the MMIA agent will determine
that the best channel to send the file through is via an email.
As mentioned already, the MMIA is the primary agent that
implements procedural logic within the eServices platform.
This it does by defining various contexts of operation to
which each request is forwarded. The following table (Table
I) gives a summary of the different contexts that have been
defined within the platform and which can be extended to
accommodate for the definition of various other network
services.
MMIA description
context
login.piask this is the context for handling login
requests
logoff.piask this handles the request for logging
off
know.piask handles all the requests for knowledge
Figure 3. Hyper Text Transfer Protocol Agent on the platform
update.piask handles the updating of knowledge on
the platform and adding new
Besides handling the HTTP requests from web browsers, knowledge
do.piask handles requests to perform platform
HTTPA also handles openVXI VXML browser requests from related operations - send an email,
Asterisk. This provides an alternative access from Asterisk and call an extension, notify a user of a
also a richer set of constructs available within VXML mark-up message, etc.
upload.piask handles uploading of content
language which are not available in AGI scripting. register.piask provides the logic for new user
2) Multi-Modal Interaction Agent (MMIA): The MMIA registrations
interaction agent plays a central role in the implementation delete.piask handles deleting of content
of application service logic within the knowledge platform. Table I
It provides the logical interface between the platform users MMIA CONTEXTS
and the underlying platform knowledge services, and between
the different access layer agents and the knowledge base layer
agents. While these contexts are not an exhaustive specification of
The key modules within the MMIA (Fig 4) collaborate to all the functionality implementable in the platform, they form
provide an end-to-end logical handling of the user requests. the core of the procedural logic for operation on the underlying
The request handler module is the first to receive the request knowledge base repositories, which facilitate the provisioning
from the access agents. This module first unpacks the message of different services.
payload to strip off internal inter-agent platform messages. The MMIA also implements the basic handling of different
These messages encapsulate system information such as User interaction modalities with the platform. The MMIA interacts
Session IDs, request context, and variables required for the directly with the access layer agents to receive the requests
processing of the request. from the users. These multiple channels, existing as the various
The AAA module is primarily responsible for the tasks access layer agents, provide the interaction input and also are
of: authenticating the users on the eServices platform, by available as output channels for communication with the users
interfacing with the Social Networking Agent which is aware 3) Social Networking Agent (SNA): The social networking
of the different platform users’ profiles and authentication agent plays the pivotal role of positioning the eServices (and
credentials; determining and enforcing authorization for vari- knowledge) platform within the social context and environ-
ous requests on the underlying knowledge in the system; and ment by integrating the existing social dynamics within the
finally for maintaining basic audit logs for the requests and community. The role of social networking agents is pivotal
processes that are handled on the platform. in the knowledge platform towards the overall objective of
325 6
requests that are sent from the MMIA to the knowledge base
layer are sent directly to the KBA, which processes the request,
to determine the appropriate agent to forward the request to.
The KBA therefore provides an interface and an aggregation of
the knowledge handling agents. The following are the modules
within the KBA that enable it to achieve its role of managing
the interaction with the rest of the knowledge layer agents (Fig
7):
Figure 6. Knowledge base layer agents
achieving both a context sensitive and culture sensitive plat-

form. The platform’s interface with the human agents (which
in one role is the user of the system, but in another sense also
an active participant in the life of the community of agents)
is primarily through the social networking agents
The SNA interacts primarily with the MMIA agent from
which it typically receives requests associated with the users
of the platform. One of such interaction is the request from the
MMIA to process user authentication, or to determine contact
details or the specific profile information of the users.
Figure 7. Knowledge Base Agent modules
The request analyzer receives a natural language request

Figure 5. Social Networking Agent from the MMIA (possibly initially from the XMPPA agent -
which implement XMPP Instant Messaging interface into the
The SNA encapsulates two key modules, the request pro- platform. The XMPPA is not discussed in this paper), which
cessor module, and the user’s profile module (Fig 5). The gets stripped down into basic tokens by the NLP Analyzer.
user profile module directly interfaces with the underlying The search engine identifies the associated ontology, the
knowledge base layers to query the user profile ontologies. subject within that ontology, and where available the required
4) Knowledge Base Agent (KBA): One of the core layers predicate. The request dispatcher then sends the formatted (by
of the platform is the knowledge base layer, which is the the request writer) request to the knowledge agents. Once
predominant end-point of most requests that are handled on a response is available from the knowledge agents, it get
the platform. The different eService portals decompose down normalized to a standard content format, and then send back
into the different units of knowledge that have to be accessed to the requesting MMIA.
and that users interact with. 5) Media Presentation Agent (MPA): The MPA primarily
While most of the knowledge that is handled is encapsulated handles the heterogeneity of the content that is communicated
in ontologies, there is also a large amount of information that with the platform. MPA implements a number of renderers in
is stored in other types of knowledge bases: legacy storage order the achieve the transcoding of content between different
systems, knowledge encapsulated in folksonomies, DBMS media formats and mark-up languages. The MPA is also
based knowledge, and native file system based knowledge. The responsible for handling interface localization in terms of
required flexibility and applicability on the knowlege platform, allowing for customization of themes, and the translation of
necessitates the implementation of agents that naturally handle platform interface strings into specific languages.
the different knowledge bases (Fig. 6). The role of the presentation agents is therefore to render
The Knowledge Base Agent (KBA) acts as the manager content at two levels: the device level and the user level. At
agent for the rest of the knowledge base layer agents. All the the device level, the content is formatted according to the
326 7
part of the FIPA specification. A communication protocol

is implemented on top of this message system, for ex-
changing platform specific messages between the agents. The
FIPA specification provides a number of features to facilitate
communication between agents and for differentiating the
messages processed: sender specification, conversation ID,
replyWith string and message performative [18]. Within the
platform, these features are used for fine grain identification
of specific messages between the agents (Fig 10).
Figure 8. Media Presentation Agent
requesting device capabilities (e.g. mark-up language, screen-

sizes) and at the user level, the content is rendered according
to the user preferences (e.g. UI language, aesthetics, layout,
colour).
The themes engine is utilized specifically for web browser
based requests. The engine utilizes Cascade Style Sheets (CSS)
as the theme definition language for the web pages that are
rendered by PIASK.
The localization engine is specifically responsible for trans-
lation of UI components into different languages. The engine
utilizes the standard GNU Gettext Portable Object (.po) files
which are loaded in memory (into a HashMap) when the MPA
agent is initialized. The in-memory storage of the translations Figure 10. Platform message exchange
allows for faster processing of requests, by eliminating the
file read and string comparison operations on the .po file. The The critical communication channels between agents are
utilization of the .po files also provides the added benefit of encrypted within platform. One of such links is the MMIA
being easily editable by humans, allowing for easy localization to SNA channel. The integrity, confidentiality and accuracy of
of the platform. this channel messages is important for the following reasons:
1) Authentication on the system is executed through mes-
C. Agent Interaction sages passed from the MMIA to the SNA since the SNA
The execution of end user requests is handled by different is the only agent that has access to user’s credentials and
agents that communicate with each other and exchange mes- preferences.
sages. An example interaction is depicted in Fig 9 in which an 2) Authorization on the platform, which is undertaken by
access layer agent (HTTPA) receives a request from a user and the MMIA, is done in dialog with the SNA which has the
then forwards the request to the MMIA. The MMIA handles relationship types (see section x) that define the different
the request and determine the next path of execution, in this authorization levels.
specific case, the request is of a user authentication on the 3) New user registrations are also handled through this
platform, and so the MMIA sends the request to the SNA channel by the MMIA.
agent. The response from the SNA agents is send back to The platform utilizes an asymmetric encryption based on the
the HTTPA, which then contacts the MPA agent that it is RSA algorithm, with a 1024 bytes key size. The encryption
associated with in order to get the content rendered according public keys are pre-shared between the agents, and the as-
to the requesting device capabilities and the requesting user’s sociated Cipher objects are initialized during the agent setup
preferences. sequence. The encryption occurs as a last stage before the
content is added to the ACLMessage, and the decryption is the
first process after the content has been read from the incoming
message.
V. O BSERVATIONS
The following preliminary observations have been made,
Figure 9. Example platform interactions around the applicability and the adequacy of this eService
platform in Dwesa:
The JADE MAS platform inherently provides an Agent 1) The distributed nature of multi-agent systems provides
Communication Language (ACL) messaging mechanisms as an architecture closely aligned to the infrastructural
327 8
situation on the ground in Dwesa (and in most rural communities. The developed solutions have to be implemented
marginalized communities). The resources in Dwesa are with a high level of flexibility to allow for situating within
distributed between different communities who share different cultural and environmental contexts. The services
and cooperate with each other in a symbiotic manner. deployed on these platform also have to be relevant to the local
The eService platform can easily be distributed depend- communities and to integrate the local (indigenous) knowledge
ing on the availability of the resources that the agents and systems dynamics with these communities.
have to access. Thus the agent that provides or handles In addressing these challenges in a ICT4D context is a
connectivity to the Internet, can be located on a gateway community in South Africa, we have developed a JADE based
host in one school, and an agent that interfaces with the MAS that provides an architectural support for the provision-
Asterisks PBX can be co-located with the PBX. ing of eServices for this community. The observations made
2) The eService platform is robust and fault-tolerant. The thus far allude to the applicability of such a platform in the
reality of rural communities is that the availability of provisioning of context sensitive, ethnocentric eServices.
different network nodes is not always guaranteed, some-
times due to the underlying infrastructural constraints R EFERENCES
(e.g. lack of electricity). The fault-tolerance within the [1] Drabenstott, M., “New Policies for a New Rural America”, International
eService platform is established through a mechanism Regional Science Review, vol. 24, no. 1, pp. 3, 2001
in the JADE MAS that allows for migration of agents [2] Mitchell, S. and Clark, D., “Business adoption of information and com-
munications technologies in the two-tier rural economy: some evidence
between different hosts on a network, and the automatic from the South Midlands”, Journal of Rural Studies, vol. 15, no. 4, pp.
initialization and termination of agents available on the 447–455, 1999
platform. The platform is therefore able to self-heal in [3] Malecki, E.J., “Digital development in rural areas: potentials and pit-
falls”, Journal of Rural Studies, vol. 19, no. 2, pp. 201–214, 2003
cases where agents get terminated when a host goes [4] Bhatnagar, S., “Information Technology and Development Foundation
down. There is also the ability for the replication of and Key Issues”, Information and Communication Technology in Rural
the Main Container within the JADE platform, which development: Case Studies from India. World Bank Institute, pp. 1–12,
2000
allows for the continuance of a MAS life in the case of [5] Castells, M., “Information technology, globalization and social develop-
one Main Container being terminated. ment”, United Nations Research Institute for Social Development, 1999
3) The eService platform has been validated for functional [6] Hietanen, O., “The global challenges of e-Development - from digital
divides towards empowerment and a sustainable global information soci-
adequacy through a load testing experiment, in which ety”, in Seminar of global perspectives of development communication,
the platform was observed to provide a linear scaling University of Tampere, 2004
in performance for handling users’ requests. The perfor- [7] Tedre, M., Sutinen, E., Kahkonen, E. and Kommers, P., “Ethnocomput-
ing: ICT in cultural and social context”, Communications of the ACM,
mance degradation, under heavy usage loads, was also vol. 49, no. 1, pp. 126–130, 2006
observed to be gradual. [8] Thinyane, M., Dalvit, L., Terzoli, A. and Clayton, P., “The Internet in
4) The interaction modalities implemented within the eS- rural communities: unrestricted and contextualized”, in ICT Africa -
Nepad Council, Addis Abba - Ethiopia, 2008
ervice platform provide a crucial interface that has been [9] Thinyane, M., Terzoli, A. and Clayton, P., “Transitions towards a
lacking through a number of the solutions that have knowledge society: Aspectual pre-evaluation of a culture-sensitive im-
been implemented in rural communities where there’s an plementation framework”, IFIP – Learning to live in the knowledge
society, July 2008, Vol. 281, pp 271-278
especially higher levels of illiteracy. One such interface [10] Cybenko, G., Gray, R., Khrabrov, A. and Wu, Y., “Information theoretic
is the voice interface to the eService platform, provided principles of agents”, in Proceedings of the CIKM - Workshop on Intelli-
through the available VOIP telephone infrastructure in gent Information Agents, Third International Conference on Information
and Knowledge Management (CIKM 94), 1994
the community. Not only is the voice interface crucial [11] Leymann, F., “Web Services: Distributed Applications without Limits”,
for the illiterate people, it is also import due to the nature Business, Technology and Web, Leipzig, 2003
of the knowledge that is available in these communities [12] Payne, T. R., “Web Services from an Agent Perspective”, IEEE Intelli-
gent Systems, vol. 23, no. 2, pp. 12-14, 2008
and also that the majority of these communities have [13] Leszczyna, R., “Evaluation of agent platforms”, European Commission,
traditionally audio based communication mediums (e.g. Joint Research Centre, Institute for the Protection and Security of the
the African drum). Citizen, Ispra, Italy, Tech. Rep., June, 2004
[14] Bellifemine, F., Poggi, A. and Rimassa, G., “JADE–A FIPA-compliant
These observations position the implemented eService portal agent framework”, Proceedings of PAAM, pp. 97–108, 1999
to provide a solution that meets the needs and the requirements [15] Thinyane, M., Dalvit, L., Slay, H., Mapi, T., Terzoli, A. and Clayton, P.,
“An ontology-based, multi-modal platform for the inclusion of marginal-
that are prevalent in rural, marginalized communities. These ized rural communities into the knowledge society”, Proceedings of the
are preliminary observations with will be validated through ACM SAICSIT conference on IT research in developing countries, pp.
the extensive usability testing and usage profiling. 143–151, 2007
[16] Thinyane, M., Dalvit, L., Terzoli, A. and Clayton, P., “Towards a Model
of an Ontology Based, Multi-Modal and Multimedia Knowledge Portal
VI. C ONCLUSION for Marginalized Rural Communities.”, in Proceedings of IEEE Infor-
mation Communication Technologies International Symposium, Fez -
The ICT4D research and implementation landscape is one Morocco, 3rd - 5th April,
[17] Eberman, B., Carter, J., Meyer, D. and Goddeau, D., “Building
faced with unique challenges which necessitate new and voiceXML browsers with openVXI”, Proceedings of the 11th interna-
innovative solutions. We have highlighted challenges which tional conference on World Wide Web, pp. 713–717, 2002
are associated with: the fact of cultural diversity in which [18] FIPA, ACL, “Message Structure Specification”, Foundation for Intelli-
gent Physical Agents, 2000
the ICT4D interventions have to be undertaken, and the in-
termittence of the infrastructure services in rural marginalized
328
Extending the Technology-Community-

Management Model to Disaster Recovery:
Assessing Vulnerability in Rural Asia
Arul Chib A.L.E Komathi
Nanyang Technological University Nanyang Technological University
ArulChib@ntu.edu.sg KOMA0004@ntu.edu.sg
Abstract—The recent increase in natural disasters has a 15.3% internet penetration rate [17], growing from 11.8%
significant impact on the lives and livelihoods of the poor in in 2006 [5] to over 500 million internet users [17]. One
Asia. The spread of information communication should note there is a disparity in access to these
technologies (ICTs) in this region’s rural areas suggests the technologies, as captured in the digital divide debate.
potential of technologies to enhance recovery efforts. While
Further, urban dwellers have more economic, educational
many ICT initiatives have been implemented to aid disaster
management, from providing early warning to immediate and technological familiarity with these technologies
relief, there exists a gap in the theoretical understanding of compared to the rural population [18]. Despite this divide
the role of technologies in disaster recovery and in technology access and use, the use of ICTs in disaster
rehabilitation. We propose a conceptual framework for warning, mitigation and management shows the benefits
understanding the implementation of ICTs in recovery of technology diffusion in rural areas.
operations, drawing attention to vulnerability reducing An examination of the literature suggests that ICTs are
potential of the initiatives. We review theories on ICT use in being used beyond merely facilitating early-warnings
disaster management, and propose the Extended about impending disasters. The role that ICT systems,
Technology-Community-Management model focusing on
which range from mature technologies such as radio,
vulnerability assessment for the design and implementation
of ICT programs for development in rural areas. We television, and land-line telephony, to advanced modern
illustrate this model using case studies from ICT technologies such as Global Positioning Systems (GPS),
deployments in post-disaster Asia, particularly India, Geographic Information System (GIS), Very Small
Indonesia, Sri Lanka and China, and suggest implications Aperture Terminal (VSAT), and cellular phones and
for theory and practice. satellite communications, play at various stages of
disaster management has been recognized by academics,
Index Terms—Asia, disaster management, ICTD, governmental agencies, civil society groups, non-
rehabilitation, vulnerability. governmental organizations (NGOs), and voluntary
welfare organizations in the development arena [19]-[22].
Communication technologies provide access to
I. INTRODUCTION information that is a vital form of aid in itself. “Disaster-
Globally, more than 250 million people are affected by affected people need information as much as water, food,
natural disasters every year, and the intensity and medicine, or shelter” [23], particularly in the post-disaster
frequency of catastrophes have steadily been increasing in context, from immediate relief to long-term rehabilitation
the last decade [1]. Asia is amongst the most affected and efforts. This suggests the potential of ICTs in various
vulnerable regions, with disasters such as earthquakes, operations immediately following a disaster, from
floods, tsunamis, cyclones, and droughts killing enabling prompt information dissemination to relief
thousands of people each year, while millions have lost agencies and affected communities, to livelihood
their homes, properties, livelihoods and families [2]-[4]. rebuilding initiatives in the recovery phase. Researchers
The increase in the proportion of people living below the have focused on the use of these technologies for disaster
poverty line, from 30% to 50% in post-tsunami Indonesia, response in the mitigation and preparedness stages [24]-
suggests the impact of calamities on the poor, especially [29], with less emphasis placed on the scope of ICTs in
those living in less developed Asian countries [5], [6]. the rehabilitation stage with the objective of exploiting
These marginalized groups face the challenge of coping the long-term and continual benefits of technologies,
with, and recovering from the effects of disasters as particularly in rural communities.
overall development and economic growth suffer major There is a need to establish a canon for best practice
setbacks in the aftermath [7]-[10]. [7]. This study aims to bridge the existing theoretical gap
The recent spread of information communication by proposing a conceptual framework to guide the design
technologies (ICTs) in Asia [11], [12] offers more people and implementation of ICT for development (ICTD)
access to technologies that can to aid in disaster programs, specifically in post-disaster management.
management efforts [13]-[15]. There has been a steady In this paper, we critically examine and categorise
rise in the penetration, accessibility, and use of ICTs in existing theories in disaster management and ICT. We
Asia. India, in particular, continues to be one of the integrate the theoretical implications of this framework in
fastest growing major mobile telecom communication relation to the Technology-Community-Management
markets in the world, with annual growth rate of 91% [5]. (TCM) model [18], [30]. We propose an extended model
With China’s 253 million internet users [16], Asia has a
329
and illustrate it using primary data from India and structure of the steps that need to be taken at each level,
Indonesia, and cases from China and Sri Lanka. the stages theories chart an overview of the levels through
which ICT users and disaster operation managers
successively progress. The longitudinal approach was
II. REVIEW OF MODELS AND FRAMEWORKS initially adopted by the eight socio-temporal stages of
We conducted a structural understanding of existing disaster framework [31]-[33] that proposed the
conceptual frameworks on disaster management (DM) classification of the disaster management efforts
and ICT usage in order to assess their theoretical impact according to specific outcome strategies, such as warning,
and practical effectiveness to technology in disaster impact, rescue, and recovery. This argument was further
recovery (see Fig. 1). The ICT-DM framework identifies expanded in models, such as the six-phase disaster cycle
specific factors that are emphasised consistently in these [34], [35], [4], three stages of disaster [36], three time-
studies, to draw two broad categorizations – Procedure- dimensions of rehabilitation efforts [14], and strategic
Based and Community-Oriented theories. Procedure- disaster management cycles [37], that elaborated the
Based theories cover the temporal segmentation of relationships between specific temporal parameters and
disaster management and ICT initiatives, and illustrate objectives guiding these efforts. Relating to practical
the management structures involved. Community- situations, the influence-impact model [8] emphasized the
Oriented theories provide insights into practically importance of accurate, timely, and complete information
relevant issues involving the community, and suggest outflow for basic level ICT impact to occur. Furthermore,
impediments to productive technology influence. These the ICT step change table [25] identified three levels
categories are further divided to guide the systematic within projects, where the informational needs of people,
review of concepts within each category, and to aid such as knowledge of technology, literacy level, and skill
analysis in recognizing the commonalities between capacity of users, determine the hierarchal levels of ICT
theories. The objective of the review is to identify gaps in projects. We argue that while stages theories are useful in
theory and focus, and to propose a theoretical model that highlighting the hierarchy of actions and measures
fills this research gap. necessary prior to, during, and after disaster contexts;
they fail to specify potential barriers to the efforts
implemented at each stage, particularly in ICT
deployment. It is difficult to draw distinct boundaries in
processes where users, especially those affected by
disasters, possess varying, and often overlapping
informational needs, and capacities to process
information. In these cases, the boundaries between
stages can be fuzzy, and stage-related behavior can be
concurrent [38].
Systems theories present factors affecting information
flow in recovery efforts, arguing that ICT providers in
disaster contexts have to be critical of the information
content disseminated to affected people. They draw
Fig. 1. Review of Disaster Management and ICT Models/ Frameworks. attention to the management of recovery projects by
pointing out ways in which governments and NGOs make
critical investments in ICT projects with the support of
A. Procedure-Based Theories
the private sector. This illustrates the essential financial
Procedure-Based theories describe the processes roles external bodies play in ensuring the economic
involved in effective management of disaster and ICT livelihood of disaster affected people. Systems theories
projects by pointing out the key stages in planning such as the CARICOM structure [39] and disaster
recovery efforts, as well as identifying the stakeholders management consortium [40] identified stakeholders
and management implications to technology involved in decision-making and coordination of disaster
implementations. The theories emphasize the importance operations, while the optimal disaster information flow
of informational flow in influencing economic and model [39] and the technical conceptual system model
psychological rehabilitation in recovery. While these [41] map the roles information plays in disaster
theories are useful in understanding enabling factors, they management. We suggest that systems theories can
largely fail to address the importance of community expand beyond coordination of information and
involvement, or identify barriers to success for ICT interactions by drawing attention to the action-oriented
initiatives in post-disaster situations. information exchanges that highlight the combined efforts
We sub-divide Procedure-Based theories – stages and of stakeholders. An action-oriented strategy relates to the
systems theories. While stages theories classify linear management style adopted by technology managers who
processes and define their scope, systems theories look beyond profit making, and encourage community
demonstrate what needs to be done, by whom, and the level participation in the management of ICT
methods to use for optimal information flow. deployments. We argue that the deliberate attention to
Stages theories illustrate the management of disasters in community involvement and ownership in running
a longitudinal approach, where both disaster-based and programmes would expand the boundaries of systems
ICT-specific frameworks craft the different stages as theories to look beyond traditional organizational styles in
mutually exclusive levels of impact. By mapping a governing ICT initiatives.
330
Together, the stages and systems theories suggest that could increase community vulnerabilities in post-disaster
ICTs can be best capitalized in the recovery stage, where situations. Among the initial debate [46]-[49], it is
emphasis is on the psychological and physiological suggested that the design of development programs
rehabilitation of affected communities. Most stages should focus on decreasing vulnerability of communities
theories contend that ICT rehabilitation efforts, in the to disasters and their negative consequences [50].
areas of physical and social rehabilitation, and health and Supplementing this, the capacities and vulnerabilities
educational rehabilitation, should focus on multi-year analysis [46] provided disaster managers a framework for
programs. This emphasizes the need for well-planned understanding and reducing vulnerabilities, and suggested
long-term recovery efforts to ensure that people regain an approach for the design and evaluation of development
their economic and psychological balance. Further, these projects. Similar to the pressure and release model [44],
models suggest that fostering of economic livelihood and access model [44] proposed that reducing
should be based on recognizing principles of social vulnerabilities involves increasing capacities of the
inclusion and gender equity, especially those involving affected people at the individual, group, and societal
fishing and farming communities. On the whole, levels, arguing that well-organized and cohesive
procedure-based stages and systems theories describe the communities can withstand or recover from disasters
linear dimensional design and practical management better than those divided by issues of race, religion, class,
aspects useful in planning and implementation of ICTs in caste, gender, ethnicity, or age. Reference [51] criticized
disaster situations. the sustainable livelihoods approach [52] framework for
failing to acknowledge socio-cultural determinants that
B. Community-Oriented Theories
influence poor people’s access to economic-enhancing
Various studies have argued that particularly at the resources and livelihood assets. This model emphasizes
recovery stage, plans need to be understood, accepted, the impact of socio-cultural factors on technology
and implemented at the local level to be sustainable, and adoption and use, critical in determining the equal access
to achieve intended outcomes on communities in postof an individual, group, or community to ICT resources.
disaster situations [42], [43]. Community-Oriented We suggest that vulnerability theories need to examine
theories emphasize various aspects of community factors impeding community livelihood, access, and
involvement in disaster management; essentially, developments, and recognize ways to improve the overall
elaborating the existence of susceptibility factors that community participation in vulnerability reduction.
impact recovery efforts aimed at affected people. While The ICT skills and discretionary slack [53] framework,
the theories connect various practical applications a capacity-oriented theory, highlighted the importance of
recognising community needs and wants in disaster capacity building in vulnerability reduction by proposing
recovery, they generally ignore the technical demands of a correlation between ICT skill level, and availability of
these recovery efforts, where over-emphasis had been time. It fails to account for the fact that the factors of skill
placed upon providing a range of applied methods to and time are affected by both psychological and
enhance community well-being instead of elaborating on physiological determinants of users. This would affect the
the know-how of running these recovery initiatives. adoption, use, continuation, and termination of ICT
We split Community-Oriented theories – vulnerability, deployments. For instance, discretionary time available
and capacity theories. Vulnerability theories describe could be affected by conditions such as the amount of
various aspects of vulnerability that both exist and arise in training sessions offered, and cultural, gender-based and
post-disaster contexts. Conversely, capacity theories look social restrictions. The psychological component
at ways to improve the capabilities of people in order to affecting ICT skill level could include individual literacy
optimize ICT usage and benefits. levels, attitudes towards ICT adoption, and overall
Affecting people with varying magnitudes, willingness to learn. Overall, this theory suggests that
vulnerability has been defined as the “characteristics of a information processing capacity and access to
person or group, and their situation that influence their information is dependent on the socio-cultural
capacity to anticipate, cope with, resist, and recover” determinants as well as the mental state of individual to
from the impact of a disaster [44]. With insufficient absorb and use the information.
capacity to cope, the poor are most vulnerable in the We contend that the vulnerability and capacity theories
aftermath of disasters [9], [24]. Vulnerability theories highlight the mutually dependent relationship between
analyze the role ICTs play in recovery efforts aimed at reducing vulnerabilities and increasing capacities in
reducing vulnerabilities of the rural poor, by increasing enhancing overall community involvement in post-
capacities and focusing on livelihood enhancing disaster ICT use. However, we think community
development efforts. The theories discuss aspects of involvement in ICT initiatives need to go beyond mere
informational, socio-cultural, and psychological recovery participation, and explore opportunities for community
that significantly impact economic vulnerability. In this empowerment through ownership, and active contribution
study, we look at five inter-related vulnerability theories; in the management of ICT efforts directed towards
(1) development-disaster framework, (2) capacities and recovering the livelihoods of the rural poor.
vulnerabilities analysis, (3) pressure and release model,
(4) access model, and (5) sustainable livelihoods
approach. III. EXTENDED TECHNOLOGY-COMMUNITY-
The development-disaster framework [45] highlighted MANAGEMENT MODEL
the importance of information in assessing vulnerabilities,
arguing that the lack of, or poor access to, information The examination of disaster management and ICT
literature allowed us to develop a framework for the
331
design and implementation of ICTs in disaster recovery.

Procedure-Based theories raised key components in the
planning and management of technology operations by
illustrating various management styles and approaches
involving key stakeholders in the financial planning and
execution of ICT efforts. The Community-Oriented
concepts provided insights on the potential of
communities to participate in critical livelihood
enhancing initiatives brought about by ICTs. We found
that the prevailing concept of vulnerability needs to be
considered in technology implementations. This analysis
draws our attention to the main aspects to take into
consideration in ICT planning, namely the community
Fig. 2. Extended Technology-Community-Management (Extended
involvement, the management components, the overall TCM) Model.
design of technologies, and evaluation of existing
vulnerabilities. The physiological and psychological vulnerabilities are
Encapsulating these aspects, we propose the use of the micro-level dimensions involving the physical and mental
Technology-Community-Management (TCM) model well-being of affected persons, or a specific community.
[18], [30] in this study. The TCM model proposes that the It is the extent to which people are susceptible to post-
three key intersects of ICT characteristics - technology, catastrophic emotional stress and trauma, as well as their
management, and community, will lead to sustainable access to depleted health infrastructure and resources.
ICT for development (ICTD) interventions. Reference This cognitive variable influences people’s outlook on
[30] argued that the technological design is the life, beliefs, and motivations to persevere in regaining
combination of software and hardware components, while their normal lives.
management of a project requires an understanding of the The informational vulnerability deals with the access to
financial requirements, establishment of key partnerships, and availability of information within affected
and a regulatory environment. A key ingredient to success communities. The lack of informational resources,
is the community involvement and participation in ranging from personal documents, books and critical data,
programs. The Community component of the TCM model to opinion leaders and professional experts, affects the
was sub-divided into dimensions of basic needs of the capabilities of people who are dependent on these sources
community, modes of ownership of ICT investments and of information. The informational vulnerability among
profits, and training of community users both in the use rural people is further augmented by the low literacy
and in technology management. Here, financial and social levels and lack of relevant technological skills necessary
sustainability are dependent on the replicability and to enable the learning and processing of information.
scalability of projects. The TCM model presents a The economic vulnerability is triggered by the loss of
comprehensive illustration of the issues surrounding ICT livelihood activities and equipments to financially support
implementations in rural areas, condensing the key households and sustain economic growth in rural villages.
conditions affecting each dimension, and its value to From the destruction of houses, fishing boats and nets,
overall sustainability for effective long-term and crops, to the loss of breadwinners and savings, the
implementations. In disaster recovery, technology and its economic impact of disasters pose a significant problem
management prompt to community’s need to regain a to recovery in Asian rural communities. The destruction
sustainable livelihood and to acquire new skills to to the physical infrastructure, in terms of the breakdown
improve their capabilities. However, our analytical of telephone wiring and electricity supply, amplifies the
framework leads us to believe that the theory is economic burden for overall development in villages.
inadequate in its failure to examine the issue of The socio-cultural vulnerability of communities is
vulnerability, crucial in this study on disaster recovery. determined by the social structure and values of society
Consequently, we improved on this limitation by that define human relationships in communities. These
integrating the TCM model to key factors and variables hierarchies affect access to resources and assets, and
relating to vulnerability in disaster recovery that were decision-making power of people, established by gender,
emphasised consistently in the theories reviewed. We age, race, religion, caste, and class egalitarianism within
extend the TCM model, and propose a new framework – communities.
the Extended Technology-Community-Management The proposed Extended TCM model states that
(Extended TCM) model (see Fig. 2), that identifies the physiological and psychological enrichment, information
four dimensions of vulnerabilities influencing technology optimization, economic resuscitation, and socio-cultural
implementation among the rural poor; reconciliation can be achieved when paralleled with ICT
physiological/psychological vulnerability, informational interventions aimed at recovery. Here, it is crucial to first
vulnerability, economic vulnerability, and socio-cultural examine and assess the scale of each dimension of
vulnerability. We argue that effectively designed ICT for vulnerability in order to critically determine the roles
development (ICTD) programs have the potential to technology, community, and management focused ICT
reduce the magnitude of these vulnerabilities in disaster initiatives can play in improving the lives of marginalized
recovery. communities. We suggest that technology programs
implemented encompassing objectives to reduce existing
vulnerabilities, distinctive to each disaster affected
332
community, will foster sustainable development beyond VI. FINDINGS

disaster recovery. We found that the implemented ICTD projects
attempted to reduce existing vulnerabilities, either as
over-arching objectives or as supplementary activities.
IV. RESEARCH FOCUS The design of technology programs from the perspective
In this paper, we propose the Extended Technology- of social sustainability required the addressing of inherent
Community-Management model to guide the design and vulnerabilities of communities. Doing so allowed for the
implementation of ICT for development (ICTD) adoption, use, and maintenance of technologies towards
initiatives in disaster recovery, and illustrate the model overall development goals. Technological, management
using cases from disaster-affected communities in Asia. and community participation were critical factors in
Specifically, we examine cases from disaster affected addressing key vulnerabilities. We discuss in-depth
rural areas in India, Indonesia, Sri Lanka, and China. To technological design elements of hardware and software,
do so, we look beyond the individual characteristics of community needs and issues of ownership and training,
technological design, management perspective and and ICTD management variables of regulation, financials,
community participation, and examine their relationship and partnership structures.
with the four key dimensions of vulnerability.
A. Physiological/Psychological Vulnerability
In post-tsunami Tamil Nadu, India, computers
V. METHODOLOGY purchased by the NGO partner, World Vision, aimed to
build familiarity with technology and ultimately translate
The Asian rural ICTD cases, used to highlight the into educational benefits for children. However, the use
Extended TCM model, were gathered from both primary of the same technology can alleviate the emotional
and secondary sources. The primary fieldwork conducted distress of psychologically vulnerable communities.
in Banda Aceh, Indonesia, and Tamil Nadu, India from Children are prone to trauma from disasters, particularly
July 2005 to February 2006, included semi-structured when coping with the loss of their families and friends.
interviews with representatives from relief agencies and Placed in child-friendly spaces, donated and maintained
NGOs, and members of affected communities. These by community resources, these devices exposed children
interviews were complemented with first-hand data to games while introducing them to basic computing. The
collected by one author, based as a research consultant at multimedia games eased them back into routines of play,
the Asia Tsunami Response Team of World Vision to and effectively reduced their stress level [54].
record and evaluate the ICT strategies and plans of World The well-being of communities is greatly enhanced by
Vision. ICT projects designed with an in-depth understanding of
Supplementary secondary data were gathered from community needs. In the coastal regions of Tamil Nadu
various case illustrations. Cases illustrated rehabilitative where many fishermen lost their lives and livelihood
measures initiated in the coastal regions of India, mainly assets to tsunami, 200 GPS navigation tools were
using GPS and computers [54]. The Nanasala projects deployed in 12 fishing villages with the primary objective
[55] and Govi Gnana system project [56] in Sri-Lanka, of regaining economic livelihoods via improved catches.
and the Jayagiri project [57] in Indonesia illustrated the This electronic device assisted the fishermen as a
function of ICTs, such as internet and radio, in compass in distant waters by showing distance and travel
empowering community livelihoods, and facilitating directions, thereby assisting the fishermen to navigate
technology adoption. The midwives project [58] easily and more efficiently. From a vulnerabilities
illustrated the use of mobile phones in responding to perspective, the project helps fishermen overcome fear of
social issues of increasing child mortality rate in the sea. Besides increasing their confidence to fish, the
Indonesia. These pilot ICTD interventions, typically psychological health of fishermen was improved when
executed by NGOs, or funded by local governments, were the saved time was spent interacting with family and
mainly focused at rehabilitating rural populations affected friends [54].
by the 2004 tsunami, except for the Chinese study on
drought-hit rural villages. The study in rural China [59] B. Informational Vulnerability
presented the potential of ICTs in enhancing education in In post-disaster villages, technology enhanced
rural China, while emphasizing the role of community capabilities of communities in obtaining updated
involvement for successfully implemented projects. information in information-constrained environments.
A combination of quantitative and qualitative research Traditionally, local farmers faced exploitation due to the
methods was used in the secondary studies. These in- lack of access to updated market prices. Middlemen
depth interviews, participant observations, documentary capitalized on this factor to gain high profits by usurping
analysis, and surveys were conducted with NGO program much of the value-added to the basic product. The Govi
officers, fisher-folk, local technology users, teachers, and Gnana system project aimed to reduce rural farmers'
community and government leaders. While some cases exposure to price volatility by providing updated crop
demonstrated the use of ICTs, such as GIS, Internet, information about trading prices [56]. It enabled buyers
radio, and multimedia technologies to purely disseminate and sellers of agricultural produce to view live transaction
disaster-related information to people, we note that cases prices at various trade stalls in their own markets as well
selected for this study focus on initiatives involving as prices in other areas of the country, made accessible at
community-level participation and use of technologies. telecenter terminals and kiosks, and broadcasted via the
telephone and internet. Thus, the reduced information
333
differential provided better economic resources to rural Indian fishing communities, with reported livelihoods
farmers. increasing from 20% to 50% [54]. Group ownership of
A mobile communications system was established in the devices ensured maintenance of the equipment. The
Banda Aceh, Indonesia, as a potential means of long-term environmental impact of technology usage was
improving health services, where comparatively poor mitigated by inviting local government officials from the
maternal and child mortality rates worsened after the fisheries department to impart responsible fishing
tsunami [60]. The primary objective of mobile phone practices.
usage by rural midwives was to provide low cost means The effects of poor informational infrastructure can be
of improving the handling of complicated obstetrical devastating on the available educational resources in
cases, thus lowering rates of pre-natal and maternal affected areas. In addition to the destruction and damage
mortality. However, a key information gap identified in of scholastic infrastructure, there is a crucial impact on
the rural healthcare system was the lack of maternal human capital, in terms of the loss of teachers and
health data by urban-based obstetricians. To rectify this parents. For children, ICTs are a source of learning
vulnerability, a Short Message Service (SMS) software opportunities that help in building critical technological
system was developed to allow midwives to upload skills. With the primary objective of providing
critical health information to an online database, thus educational opportunities, World Vision equipped two
improving access, as well as enabling the rural midwife to secondary schools in affected-villages of Tamil Nadu,
consult senior nurses and doctors using mobile phones for India, with computers, providing lessons integrated with
particular cases [58]. the school curriculum. In addition to computer-oriented
Offering trainings to aid greater adoption of technology skills, children learnt Tamil, English, and other academic
can make a significant difference to the information subjects [54]. It was essential to provide an outlet for
accessibility of rural community, beyond mere access to graduates to utilize their skills in an economically
the technology. In drought-hit Yellow Sheep River, productive manner. Business training, including
China, the “Town and Talent Project” equipped the introduction to productivity software, was provided to
agricultural village’s e-commerce centre with 25 young adults to improve their job prospects in the
computers. Online courses equipped farmers with increasingly technologically-savvy Indian workforce.
relevant computer skills for better retrieval and usage of
D. Socio-Cultural Vulnerability
agricultural information [59]. A community centre in
Jayagiri, Indonesia promoted community radio and Community pressures are seen most directly in the
multimedia as educational tools to encourage people- existing networks of power within the social system. The
centered development in disaster recovery. The training traditional power of the higher castes (social power), the
series, directed at capacity-building of centre managers middlemen (economic power), and males (gender power)
and volunteers running the community radio and limits the opportunities available to those groups with less
programmes, aimed to improve technological skills for power. Introducing ICTs in such biased environment
creating program content [57]. could work in the favour of the haves, and likely further
Finally, informational capacity building was greatly marginalize the have-nots. Tactfully designed ICTD
enhanced by community management of technology in initiatives have the potential to champion the cause of
recovery efforts. The collective-ownership of facilities development without widening existing social
can promote the dissemination and sharing of information inequalities. Recognizing this, the M. S. Swaminathan
within communities. The "Town and Talent Project" in Research Foundation, an Indian NGO, established
Yellow Sheep River introduced internet-enabled knowledge centers in Dalit (lower caste) villages [61];
computers to enrich educational resources in the local while World Vision trained women Self Help Groups to
school. Collaborating with a local company, the teachers take ownership of the ICTD projects within their
were trained to set up the scholastic network. These communities [62], [63]. The World Vision multimedia
teachers were the central information providers on centres in most villages claimed no socio-cultural
computer usage to both their peers and children [59]. The restrictions on access, with no separate timings for young
Nanasala telecenter project in Sri Lanka was managed by girls, or lower castes individuals, allowing a co-mingling
individual entrepreneurs within the community to provide of different sections of society [54].
internet access to rural villages [55]. The initiative In some situations, the socio-cultural issues
exposed rural people to internet, computer, and telephone surrounding communities may indirectly affect people’s
technologies with free training sessions to gain perceptions on their capacity to use technologies. In rural
knowledge of technology use and online retrieval of classrooms in China, elder teachers were reluctant to take
information on new agriculture practices and techniques, on computer usage for fear of not being able to
and information on a variety of health issues. effectively learn and adopt computer skills,
predominantly possessed by younger users. Training
C. Economic Vulnerability courses helped staff overcome this technophobia, and
Information communication technologies have the grasp basic E-literacy in daily teaching practices [59].
potential to empower communities to regain economic Indonesian midwives in the Banda Aceh mobile-phone
stability in the aftermath of disaster. However, beyond the project initially perceived the ICT-based health practice
economic outcomes, the ability of communities to engage as a monitoring system; rather than a tool to assist their
in relatively expensive technology usage, and to manage work. This negative perception, triggered by existing
their economic expectations, need to be considered. The hierarchical class structure within the health
use of GPS and fish-finders led to income generation in infrastructure, could have encouraged them to revert to
334
the prior, more inefficient, paper-based system for followed by designing initiatives, is an iterative process
reporting maternal health [64]. Another criticism was the involving evaluation and reconfiguring of project design,
unfamiliarity with text-based technology. To overcome and will necessarily be a long-term phenomenon. The
these barriers, joint training seminars were held with durations of reaching this goal will vary depending on
doctors, health coordinators and rural midwives. several intrinsic and extrinsic conditions affecting ICTD
Secondly, a JAVA-based applet was developed that programs at different stages of execution. First, it is
functioned with scroll and select functionality, rather than necessary to determine a methodological approach to
using the traditional text-based, and possibly elaborate on the criteria for assessing the vulnerabilities,
cumbersome, SMS. Constructive responses to the determining the order of priority for tackling the
suggestions and feedback from midwives increased obstacles, and gauging the effectiveness of measures.
credibility of the initiative, which subsequently Secondly, beyond limiting the analysis of psychological
encouraged the use of ICT beyond existing social class- vulnerabilities to disaster affected rural poor, the
based barriers between rural midwives and hospital-based assessment should observe other extrinsic conditions such
doctors. as the mental barriers of ICTD program managers and
policy makers in related domains. Preconceptions of the
marginalized communities’ learning capacities, adoptive
VII. DISCUSSION AND CONCLUSION abilities, enthusiasm levels, and financial capabilities to
An examination of Asian cases reveals that the use and manage technology could impede their
deployment of ICTs in disaster recovery needs to take commitment to implement and sustain ICT initiatives.
into consideration the vulnerabilities impacting social We should note that the identified variables of
sustainability. This study extends the theoretical literature physiological and psychological vulnerability,
on disaster management and ICT, by addressing a gap in informational vulnerability, economic vulnerability, and
research on the best practices for ICT deployment in socio-cultural vulnerability, are not mutually excusive
disaster recovery. The TCM model proposed that ICTD variables. The cases highlight the existence of
program design should include critical variables overlapping vulnerabilities, and those that are sometimes
involving communities, including needs, ownership, and in opposition to one another. ICT programs focusing on
training, in addition to the external impacts of technology reducing economic susceptibility, such as the use of GPS
and management. We extend the TCM model to include system by fishermen, also have a positive impact on the
four key vulnerability dimensions. We recognize the psychological well-being of users. Educational programs
limitations of frameworks with clearly defined for rural midwives involving cell-phone training
boundaries, and discuss the inter-relationships of addresses the psychological fear of technology, while
vulnerabilities and ICTD program deployment simultaneously reducing the informational barriers
characteristics of technology, its management and the present. However, vulnerabilities can contradict each
community involvement. The case studies illustrate real other when the attempt to reduce one barrier magnifies
world complexities, in terms of barriers to the negative impact of another. For example, ICTD
implementation and considerations for long-term program programs that aim to enhance information flow among
success. communities through the usage of mobile phones could,
A critical distinction needs to be made between core in the long-run post-subsidy, negatively affect the
development objectives and vulnerabilities, as these may financial capacities of communities which struggle to
not coincide, and may even be antithetical. Vulnerabilities maintain and continue usage of the adopted technology.
are inherent factors inhibiting communities from We have previously critiqued the stage-based approach
accessing, adopting, and realizing the benefits from of extant theories; yet we acknowledge that the
technology implementations. We believe these barriers establishment of clear stages can provide a guideline for
need to be addressed in addition to core development ICTD program assessment, particularly for practitioners.
objectives, such as those recommended by the United It is possible that different vulnerabilities may take
Nation’s MDG [65]. The choice of development priority at certain stages of implementation, leading to
objectives can influence the selection of beneficiaries, specific TCM configurations. The extended TCM model
their control and use of technology, leading to an does not incorporate hierarchies in management flow,
amplification of existing vulnerabilities in communities. community participation, or technology diffusion,
For instance, the management of rural initiatives, largely suggesting their relationship to vulnerabilities occurs in a
led by village leadership councils, could lead to cooperative manner.
privileged access to technologies, in turn furthering the The communication infrastructure within a community
sociological imbalance within society. Further, the consists of both old and new ICTs, as well as established
emphasis on education in post-disaster societies, often and evolving social networks. This study focused on
placed on basic primary education, could ignore the needs implementations involving a single technology, aided
of adolescents and young adults, raising issues of age largely by the fact that most ICTD interventions are
discrimination to technology access. We believe that designed as such. We propose that researchers examine
project designers need to recognize the existence of dynamic communication and information environments,
vulnerabilities within the social system, and take steps to themselves embedded within fluid social networks, rather
mitigate them, in addition to focusing on overall project than one technology at a time.
objectives. Future research can build on the extended TCM model
Realistically, project managers have to recognize that as a theoretical framework to guide research in the design
the process of identifying existing vulnerabilities, of ICTD programs, and to test the significance of
335
vulnerabilities beyond disaster recovery. We propose that [Accessed: Aug. 15, 2008].
[9] B. K. Paul, “Disaster relief efforts: an update,” Progress in
the vulnerabilities assessment approach can and should be
Development Studies, vol. 6, pp. 211–223, 2006.
applied to varied contexts such as education, healthcare, [10] United Nations Development Programme, ICT and human
agriculture, and conflict management. Subsequent studies development: Towards building a composite index for Asia. New
in the area of disaster recovery could examine how the Delhi: Elsevier, 2004.
[11] International Telecommunication Union, ICT and
TCMV model can examine the impact of individual
telecommunications in the least developed countries. Geneva:
vulnerability dimensions to technology, management, and International Telecommunication Union, 2006.
community variables. Alternatively, research could focus [12] United Nations Development Programme, “Community-based
on specific components within the TCM model, such as networks and innovative technologies: New models to serve and
empower the poor,” United Nations Development Programme,
community training, within the Management-Community
2005. [Online] Available:
intersect, and examine its relation to the four identified http://www.undp.org/poverty/docs/ictd/ICTD-Community-
vulnerabilities. Here, the aim would be to explore Nets.pdf [Accessed: Sep. 10, 2008].
thoroughly the barriers that exist and arise within specific [13] N. Gunawardene and F. Noronha, Communicating disasters: An
Asia Pacific resource book. Bangkok: United Nations
components, and suggest solutions to overcoming them.
Development Programme Regional Centre, 2007.
There is no doubt that calamities have a [14] M. Jansz, “Disaster recovery and ICT in Sri Lanka: The day after,”
disproportionately negative impact on the poor. However, Information for development, vol. 3, pp. 10–22, Jan. 2005.
the rapid proliferation of ICTs, particularly in Asia, has [15] C. Wattegama, “ICT in disaster management,” 2007. [Online]
Available: http://www.apdip.net/apdipenote/16.pdf [Accessed:
the potential to engender positive recovery to the lives
Sep. 15, 2008].
and state of those affected. We believe well-designed [16] China Internet Network Information Centre, “Statistical survey
ICTD programs have the power to increase the capacities report on the internet development in China: Abridged edition,”
of communities, while simultaneously reducing the China Internet Network Information Centre, 2008. [Online]
Available:
vulnerabilities towards technology adoption. Finally, rural
http://www.cnnic.cn/download/2008/CNNIC22threport-en.pdf
communities can witness long-term social sustainable [Accessed: Aug. 09, 2008].
progress. [17] Internet World Statistics, “Asia marketing research, internet usage,
population statistics and information usage,” Internet World
Statistics, 2008. [Online] Available: from
REFERENCES
http://www.internetworldstats.com/asia.htm [Accessed: Sep. 15,
2008].
[1] Department for International Development, “International day for [18] A. Chib and J. Zhao, “Sustainability of ICT interventions: Lessons
disaster reduction,” Department for International Development, from rural projects in China and India,” in
2007. [Online] Available: Communicating for social impact: Engaging communication
http://www.dfid.gov.uk/news/files/disasters-day.asp [Accessed: theory, research, and pedagogy, L. Harter and M. J. Dutta, Eds.,
Aug. 15, 2008]. ICA 2008 Conference Theme Book, Hampton Press, in press.
[2] Centre for Research on the Epidemiology of Disasters, “Annual [19] H. E. Hudson, “Universal access to the new information
disaster statistical review: The numbers and trends 2007,” Centre infrastructure,” in Handbook of New Media: Social Shaping and
for Research on the Epidemiology of Disasters, 2007. [Online] Consequences of ICTs, L. A. Lievrouw and S. M. Livingstone,
Available: Eds., London: SAGE, 2002, pp. 368–384.
http://www.emdat.be/Documents/Publications/Annual%20Disaster [20] C. K. Prahalad, The fortune at the bottom of the pyramid:
%20Statistical%20Review%202007.pdf [Accessed: Aug. 10, Eradicating poverty through profits. Philadelphia: Wharton School
2008]. Publishing, 2004.
[3] Commonwealth Telecommunications Organization, “Using ICT [21] United Nations Conference on Trade and Development,
for effective disaster management forum: A summary report on Information and communication technology development indices.
sessions and panel discussions,” Commonwealth Geneva: United Nations Publications, 2003.
Telecommunications Organization, 2006. [Online] Available: [22] S. Yodmani and D. Hollister, “Disasters and communication
http://www.lirneasia.net/wp- technology: Perspectives from Asia,” in Second Tampere
content/uploads/2006/07/CTO%20Report%20Uning%20ICT%20f Conference on Disaster Communications, 2001. [Online]
or%20DM%20June%202006.pdf [Accessed: Aug. 14, 2008]. Available:
[4] United Nations Economic and Social Commission for Asia and the http://www.reliefweb.int/telecoms/conference/cdc2001/dh.pdf
Pacific, “ICST-enabled natural disaster management in Asia and [Accessed: Aug. 15, 2008].
the Pacific region,” United Nations Economic and Social [23] J. Walter, ed., World disasters report 2005: Focus on information
Commission for Asia and the Pacific, 2005. [Online] Available: in disasters. Geneva: The International Federation of Red Cross
http://stdev.unctad.org/unsystem/cstd/escap8.doc [Accessed: Aug. and Red Crescent Societies, 2005, pp. 11.
15, 2008]. [24] M. I. Ahmed and E. C. Lentz, “Enhancing the livelihoods of the
[5] International Telecommunication Union, “ICT statistics database,” rural poor through ICT: A Knowledge Map,” 2007. [Online]
International Telecommunication Union, 2008. [Online] Available: http://www-
Available: http://www.itu.int/ITU- wds.worldbank.org/external/default/WDSContentServer/WDSP/IB
D/ICTEYE/Indicators/Indicators.aspx [Accessed: Aug. 15, 2008]. /2008/11/26/000333037_20081126002013/Rendered/PDF/466300
[6] OneWorld South Asia, “Making MDGs disaster proof,” OneWorld NWP0Box31esh0Country0Study111.pdf [Accessed: Aug. 01,
South Asia, 2008. [Online] Available: 2008].
http://southasia.oneworld.net/Article/making-mdgs-disaster-proof [25] A. Chetley, “Improving health, connecting people: The role of ICT
[Accessed: Aug. 15, 2008]. in the health sector in developing countries. InfoDev framework
[7] Centre for Research on the Epidemiology of Disasters, “An paper,” 2006. [Online] Available:
analytical review of selected data sets on natural disasters and http://www.asksource.info/pdf/framework2.pdf [Accessed: Jul. 15,
impacts,” Centre for Research on the Epidemiology of Disasters, 2008].
2006. [Online] Available: [26] K. Delgadillo and R. Borja, “Learning lessons from telecentres in
http://www.emdat.be/Documents/Publications/TschoeglDataSetsR Latin America and the Caribbean,” 1999. [Online] Available:
eview.pdf [Accessed: Aug. 15, 2008]. http://www.idrc.ca/telecentre/evaluation/nn/16_Lea.html
[8] A. F. Gorp, “ICT as an enabler for humanitarian relief? An [Accessed: Aug. 30, 2008].
assessment of communication and information flow in inter- [27] T. Eisensee and D. Stromberg, “Disaster relief,” The Quarterly
organizational context,” presented at the annual meeting of the Journal of Economics, vol. 122, pp. 693–728, 2007.
International Communication Association, New York, 2005. [28] A. Hovanesian and A. Cox, “ICT for tsunami recovery: Best
[Online] Available: practices and lessons learned,” presented at the E-Asia
http://test.scripts.psu.edu/users/a/f/afv103/ICA2005.pdf Conference, Putrajaya, Malaysia, 2007.
336
[29] O. Takizawa, “Using ICT in disaster mitigation,” National [48] A. Kreimer and M. Zador, Colloquium on disasters, sustainability
Institute of Information and Communications Technology News, and development: A look at the 90s. Washington, D.C: The World
vol. 344, pp. 1–2, 2004. Bank, 1989.
[30] S. Lee and A. Chib, “Wireless initiatives for connecting rural [49] A. Lavell, “The impact of disasters on development gains: Clarity
areas: Developing a framework,” in Participation and media or controversy Latin American social science faculty and the
production: Critical reflections on content creation, N. Carpentier network for the social study of disaster prevention in Latin
and B. De Cleen, Eds., ICA 2007 Conference Theme Book. America,” presented at the International Decade for Natural
Newcastle, United Kingdom: Cambridge Scholars Publishing, Disaster Reduction Programme Forum, Geneva, 1999.
2008. [50] L. Pyles, “Community organizing for post-disaster social
[31] R. R. Dynes, Organized behavior in disaster. Lexington, development,” International Social Work, vol. 50, pp. 321–333,
Massachusetts: Heath Lexington Books, 1970. 2007.
[32] L. Palen and S. B. Liu, “Citizen communications in crisis: [51] F. Ellis, Rural livelihoods and diversity in developing countries.
Anticipating a future of ICT-supported public participation,” in London: Oxford University Press, 2000.
Proceedings of the Special Interest Group on Computer-Human [52] C. Ashley and D. Carney, Sustainable livelihoods: Lessons from
Interaction (SIGCHI) Conference on Human Factors in early experience. London: Department for International
Computing Systems, San Jose, California, 2007, pp. 727–736. Development, 1999.
[33] J. Powell, An introduction to the natural history of disaster. [53] T. Newholm, K. Keeling, P. Mcgoldrick, L. Macaulay, and J.
Baltimore: University of Maryland, Psychiatric Institution, 1954. Doherty, “The digital divide and the theory of optimal slack,” New
[34] L. Buzna, L. Issacharoff, K. Peters, V. Rosato, and D. Helbing, Media and Society, vol. 110, pp. 295–319, 2008.
“How can future ICT enhance disaster relief and recovery?” [54] Center for Media Studies, Final Report: Evaluating ICT-4D
presented at the Information Communication Technology-Future project of World Vision India. India: World Vision India, 2007.
and Emerging Technologies Workshop, Dresden University of [55] F. J. Proenza, “ICTA Nanasala Program: Field visit report,” 2005.
Technology, Dresden, Germany, 2007. [Online] Available:
[35] Caribbean Information and Communications Technology Virtual http://siteresources.worldbank.org/INTEDEVELOPMENT/Resour
Community, “Facilitating effective disaster management in the ces/Telecenter.doc [Accessed: Aug. 15, 2008].
Caribbean: A response from the Caribbean Information and [56] infoDev, “Sri Lanka-Govi Gnana system project,” 2007. [Online]
Communications Technology Community,” 2004. [Online] Available: http://www.ict4rl.info/CaseStudies/GGS?v=fau
Available: [Accessed: Aug. 15, 2008].
http://www.devnet.org.gy/documents/Caribbeandisasterbrief- [57] United Nations Educational, Scientific and Cultural Organization,
Final.pdf [Accessed: Aug. 15, 2008]. “UNESCO develops community multimedia centre for education
[36] D. Basu, “Disaster relief: A compendium of learnings from in Indonesia,” United Nations Educational, Scientific and Cultural
engagements in Afghanistan, Iraq, Liberia, Iran, Sudan, Organization, 2008. [Online] Available:
Guatemala, Indonesia, Sri Lanka, Pakistan, Lebanon,” NetHope, http://www.unescobkk.org/index.php?id=6800andtx_wecknowled
2006. [Online] Available: gebase_pi1[tt_news]=1215andtx_wecknowledgebase_pi1[backPid
http://nethope.org/doc/NetHope%20Disaster%20Relief%200906.p ]=6771andcHash=96224fb29a [Accessed: Aug. 15, 2008].
df [Accessed: Aug. 15, 2008]. [58] A. Chib, Information and communication technologies for health:
[37] U. Sastrokusumo, “ICT for countermeasures and strategies Aceh Besar midwives mobile-phone project final report, tech.
against emergencies and disaster crisis: A suggestion to build report, World Vision Indonesia Tsunami Response, Singapore,
Asia-Pacific pilot project,” presented at the Second Joint Project 2008.
Team Meeting for Establishing a Disaster Management Support [59] J. Zhao, The Internet and rural development in China: The socio-
System in the Asia-Pacific Region, Bangkok, Thailand, 2006. structural paradigm. Bern: Peter Lang AG, 2008.
[38] L. Palen, S. Vieweg, J. Sutton, S. B. Liu, and A. Hughes, “Crisis [60] A. Chib, M. O. Lwin, J. Ang, H. Lin, and F. Santoso, “Mobiles
informatics: studying crisis in a networked world,” in Proceedings and midwives: Improving healthcare communications via mobile
of the Third International Conference on E-Social Science, Ann phones in Aceh Besar, Indonesia,” Asian Journal of
Arbor, Michigan, 2007. Available: Communication, vol. 18, pp. 348–364, 2008.
http://www.cs.colorado.edu/~palen/palen_papers/palen- [61] United Nations Centre for Regional Development, “Information
crisisinformatics.pdf [Accessed: Aug. 23, 2008]. and communication technologies for human security in local
[39] Caribbean Community and Common Market, “Caribbean ICT development,” Regional Development Dialogue, vol. 23, pp. 65–
disaster management network,” 2004. [Online] Available: 88, 2002.
http://www.devnet.org.gy/documents/Final_ICT_and__DM__Prop [62] A. Chib, “Information communication technologies for
osal_Nov_15.pdf [Accessed: Aug. 15, 2008]. development: Cases from tsunami-affected communities in
[40] B. R. Balachandran and A. Haer, “Corporate involvement in Indonesia and India,” 2006. [Online] Available:
disaster management: A conceptual framework,” presented at the http://www3.ntu.edu.sg/sci/sirc/download/Arul_presentation%20re
Workshop on Disaster Management: Disaster Preparedness port.pdf [Accessed: Aug. 20, 2008].
and Mitigation, New Delhi, India, 2004. [63] Environmental Planning Collaborative, Tsunami evaluation
[41] Y. S. Alghtani and S. Al-Zahrani, “A conceptual framework for coalition: The international community’s funding of the tsunami
information network system to exchange information to control emergency and relief,” 2006. [Online] Available:
accidents in Saudi Arabia,” presented at the Saudi Technical http://www.un.org.in/untrs/reports/local_response_india.pdf
Conference and Exhibition, Saudi Arabia, 2006. [Accessed: Aug. 29, 2008].
[42] Associated Programme on Flood Management, Social aspects and [64] A. Chib, Information and communication technologies for health:
stakeholder involvement in integrated flood management. Geneva: Midwives with mobile-phone project in Aceh Besar baseline
World Meteorological Organization, 2006. report, tech. report, World Vision Asia Tsunami Response,
[43] Earthquakes and Megacities Initiative, “Disaster risk reduction in Singapore, 2007.
metropolitan regions through the cluster cities project,” [65] United Nations, The Millennium Development Goals Report 2005.
Proceedings of the 2005 Americas Cluster Cities Meeting, Bogotá, New York: United Nations, 2005.
Columbia. Available: http://www.emi-
megacities.org/upload/CCP_America_2005_10_Meeting_Report.p
df [Accessed: Aug. 15, 2008].
[44] P. Blaikie, T. Cannon, I. Davis, and B. Wisner, At risk: Natural
hazards, people's vulnerability and disasters. London: Routledge,
1994.
[45] F. C. Cuny, eds., Disasters and Development. New York: Oxford
University Press, 1983.
[46] M. B. Anderson and P. J. Woodrow, Rising from the ashes:
Development strategies at times of disaster. Paris: Westview Press,
1989.
[47] A. Kreimer and M. Munasinghe, Managing natural disasters and
the environment. Washington, D.C: The World Bank, 1991.
337
Featherweight Multimedia for

Information Dissemination
Gerry Chu, Sambit Satpathy, Kentaro Toyama, Rikin Gandhi, Ravin Balakrishnan, S. Raghu Menon
Abstract— Featherweight multimedia devices combine audio include printed material, radio and television public service
with non-electronic visual displays (e.g., paper). Because of their announcements (PSAs), and verbal instruction. These methods
low cost, customizability, durability, storage capacity, and energy are certainly viable, but suffer from several shortcomings,
efficiency, they are well-suited for education and information such as requiring literacy to comprehend textual printed
dissemination among illiterate and semi-literate people. We
present a taxonomy of featherweight multimedia devices and also
material, the relatively high cost of producing, distributing and
derive design recommendations from our experiences deploying consuming radio and television PSAs, and the arguably higher
featherweight multimedia in the agriculture and health domains cost of verbal instruction by humans who might not even
in India. We found that with some initial guidance, illiterate deliver a consistent message over time. Of these, only verbal
users can quickly learn to use and enjoy the device, especially if instruction provides the possibility of interactive exchange
they are taught by peers. with the consumer.
In this paper, we explore the notion of featherweight
Index Terms—Audio, illiteracy, information dissemination
multimedia” – various combinations of electronic audio
devices with non-electronic visual displays – for interactive
I. INTRODUCTION
multimedia information dissemination (Fig. 1). Featherweight
A major thrust towards bringing information and

communication technologies to underserved populations
is the development of low-cost computational devices such as
multimedia devices require minimal power, are more rugged,
and are dramatically cheaper than low-cost computers or
feature-rich mobile phones. Yet, they integrate many of the
the XO laptop from OLPC [1], Classmate PC [2], and even a best elements of traditional techniques with the enhancement
potential $12 laptop [3]. While these initiatives and other of interactivity. Our contribution is not in the technology per
factors such as increased penetration of mobile phones will se, but rather in 1) mapping out the design space of
undoubtedly have tremendous impact in providing the global featherweight multimedia, 2) exploring factors affecting their
poor with better communication tools and access to use for information dissemination to illiterate or semi-literate
information, the cost of these devices remain prohibitive to the populations using our experiences deploying them, and 3)
2 billion+ people worldwide who live on less than $2 per day. reporting the reactions of the target population to initial usage
Beyond monetary cost, illiteracy remains a major hurdle. of several featherweight multimedia variants. Our preliminary
Conservative estimates of illiteracy suggest that there are over experiments suggest that non-literate users with little formal
one billion illiterate people worldwide [4] for whom even a education can quickly understand how to use featherweight
free, connected computer is useless given that some level of multimedia, but that social support is required to maximize
literacy and basic education is typically required to operate effectiveness.
them; recent attempts at text-free interfaces [5] hold the
promise of partial accessibility to computing by the illiterate,
but still don’t address the complexity of the device.
While it is certainly worthwhile trying to grapple with the
cost and literacy constraints of “full-fledged” technology for
many applications, there are other applications that may only
require a little bit of technology in order to add significant
value. One such broad application area is the information
dissemination that government and non-profit organizations
do to improve the education and health of people. In these
populations, traditional information dissemination methods
Manuscript received September 22, 2008.

G. Chu, S. Satpathy, and R. Balakrishnan were with Microsoft Research
India (email: gerrychu | ravin @dgp.toronto.edu, sambit407@gmail.com).
K. Toyama, R. Gandhi, and S. Raghu Menon are with Microsoft Research Fig. 1. Village residents using a talking book
India, “Scientia”, 196/36 2nd Main, Sadashivnagar, Bangalore, 560 080, India
(phone: +91 (80) 6658-6000; email: kentoy|riking|t-raghum@microsoft.com).
R. Gandhi is additionally with Digital Green II. RELATED WORK
R. Balakrishnan is with the University of Toronto Recently, there have been several initiatives adapting
specific featherweight multimedia devices to the development
338
field. These evolved from devices that were originally created Public advertising on billboards, as well as research on
for the developed world. In addition, featherweight devices more interactive public ambient displays have demonstrated
can trace their design histories to other multimedia initiatives designs intended to engage casual bystanders in information
that use audio and video in development. exchange [21][22]. A featherweight multimedia instantiation
of this concept might involve billboards with phone numbers
A. Existing featherweight multimedia devices in development
that passersby can call for additional prerecorded information,
Books of Hope is an organization that creates talking books or paper posters with an audio player, a design that we explore
in South Africa [6]. Each 16-page book is held within an later in this paper.
electronic device that has 16 buttons. The user turns to a page,
each of which has an icon. The user presses the button labeled C. Related work in development
with the same icon, triggering the appropriate caption to play There have been two notable audio interfaces for
aloud. Each book holds 5 minutes of audio, and with two developing countries. Tamil Market uses a spoken dialog
AAA batteries, 200-500 plays can be achieved [7]. In an system that allows illiterate people to query the system by
evaluation [8], a control group of grade 7 students attended a voice to hear crop prices spoken aloud [23]. Similarly,
workshop about mental health, while an experimental group Pakistani community health workers can call HealthLine, ask
received talking books. Pre- and post-tests determined that it verbally for information, and hear medical information over
both groups learned the same amount but a large percentage of their mobile phone [24]. Both require backend computers to
students who received the talking books showed them to other provide speech recognition, hence imposing a fairly
students (80%), to their families (80%) and to church significant computational and cost requirement.
members (43%), pointing to possible viral dissemination A system using visual codes and mobile phone cameras,
benefits of cheap portable featherweight multimedia. such as that explored by Parikh et al. [25] could also be used
Talking books such as LeapPad [9], WhizKid [10], and to trigger audio playback.
PowerTouch [11] are educational toys for children. They are Work on text-free user interfaces [5] has focused on
operated by overlaying a piece of paper or workbook onto a replacing text in computing interfaces with audio prompts that
touch or pen-sensitive screen. The device can identify the are played when the user mouses over a graphical interface
current page, so that when elements on the page are pressed, element. The focus of this work has been on making
the device plays a recorded caption. There was an announced conventional computers accessible to the illiterate population,
initiative in 2004 to deploy the LeapPad in Afghanistan [12], and hence the usual cost and power requirements of standard
but nothing has been published about the deployment since. computation remains. Featherweight multimedia can be
However, the PicTalk device, which uses LeapPad characterized as being text-free user interfaces without an
technology, is being used to teach Indian schoolchildren [13]. electronic visual display.
Students using PicTalk to learn English performed 40% better The concept of integrating paper and audio is supported by
on post-tests than a control group learning using a traditional work by Medhi et al. [26]. They tested five different
curriculum. representations (video, photo, animation, static drawing, and
The Literacy Bridge project has created an interactive text) of medical conditions with and without audio captions on
digital audio player/recorder costing roughly US$5-$12 and Indian slum residents. They found that users best understood
are deploying it in Ghana [14]. Although they call it a drawings that are combined with audio.
“Talking Book” the focus is on the audio and not on the paper Video might be considered as the “heavyweight” version of
book component. featherweight multimedia. Like featherweight multimedia, it
has both audio and visual electronic displays. Video has been
B. Devices similar to featherweight multimedia
used to teach Indian schoolchildren [27] and has also been
Audio guides (e.g., [15]) are widely used in museums and shown to be effective in teaching illiterate users to use a job-
galleries. They play an audio clip when a numeric code is search computer application [28]. The Digital Green project
typed in. An interesting element of these sorts of audio uses video to teach sustainable agricultural techniques to
interfaces is the spatial decoupling of the visual printed media Indian farmers [29]. It has been shown to be six times more
from the audio device. effective than conventional agricultural extension workers.
Similar to the aforementioned talking books, Despite these successful examples of video for information
communication aides are used by people with developmental dissemination in developing countries, video requires an
disabilities who cannot communicate verbally [16]. expensive television and DVD player that depend on electrical
More advanced and expensive audio/paper devices such as power that can be unreliable in many development settings.
the Audio Notebook [17] and the Pulse SmartPen [18] record
audio as the user writes in a notebook. Tapping on a word III. DESIGN SPACE
plays the audio that was recorded when the word was written.
At first glance, our definition of featherweight multimedia –
The Daisy Consortium is the publisher of a file format for
the combination of an electronic audio device with some non-
talking books meant for the blind or those with learning
electronic visual display – connotes a relatively simple range
disabilities [19]. The Dolphin company creates players and of possible instantiations. On closer investigation, however,
authoring tools for talking books in the Daisy format [20]. we find a rich design space worthy of careful delineation.
339
TABLE 1. COMPARISON OF ELECTRONIC AUDIO PLAYERS

LOW-END SMARTPHONE CD PLAYER PORTABLE TALKING CUSTOM GREETING-CARD
MOBILE AUDIO/MP3 BOOK EMBEDDED AUDIO CHIPS
PHONE PLAYER (LEAPPAD) ELECTRONICS
DEVICE COST $10-50 $200-500 $10-$50 $10-$200 $15-$50 $0.05-$100+ $0.05
CONTENT High High Medium High High Medium Medium
CUSTOMIZABILTIY
CAPACITY High High Medium High Medium Medium Low
AUDIO QUALITY Medium Medium High High Medium Medium Low
DURABILITY High Medium Medium High High Medium Medium
POWER Medium Low Low High Medium High High
EFFICIENCY
USABILITY Medium Medium Medium Medium High Medium Medium
UBIQUITY High Low Medium Medium Low Low Low
SENSORS Low High Low Low Low Low Low
in the types of batteries used and whether or not they can be
A. Electronic audio player
recharged and/or replaced easily. Some devices use readily
The audio component of featherweight multimedia can be available replaceable batteries, some are more specialized and
sourced from various readily available technologies including can only be replaced by a technician, while others can only be
CD players, portable MP3 players, mobile phones, museum recharged by plugging into mains power. Depending on the
audio guides, and embedded audio devices, including audio usage locale, the type of battery might be a crucial deciding
chips (as can be found in audio greeting cards [30]), embedded factor in choice of a featherweight multimedia device.
MP3 players, and other custom audio electronics. These vary Usability. Given the focus on information dissemination in
along several interesting dimensions, summarized in Table 1: low-education target audience, usability is arguably even more
Device cost. This can vary considerably, ranging from just a crucial than in more mainstream technology. For example,
few dollars at wholesale prices to several hundred dollars for many cheap audio players have surprisingly complex
state-of-the-art equipment. An embedded audio chip of the interfaces that require the user to decipher instructions on a
sort found in greeting cards cost just several cents, and one tiny LCD screen, thus negating the positive value of their
could embed many of these chips into a single featherweight inexpensive cost. We believe that for most usage scenarios,
multimedia device. At the other end of the spectrum, museum ultra-simple interfaces with just a few buttons (start, stop,
audio guides can cost thousands of dollars. rewind) or with one button per audio caption are generally
Content customizability. CD/DVDs cost just a few cents to best. A numeric keypad can added if random access to content
manufacture with content, while content on MP3 players can is provided through numeric codes.
easily be uploaded from a computer. However, the distribution Ubiquity. Mobile phones have staggering penetration in
costs of getting those CD/DVDs to the user or providing a many parts of the developing world. In contrast, more
geographically distributed user base with access to a content- specialized embedded systems might be cheaper but might
serving computer to update their MP3 players needs to be ultimately not be viable given their specialized nature.
accounted for. Mobile phones offer the highly flexible solution Sensors. Smart phones and custom embedded devices might
of wireless access to content that may not even be stored on have cameras, RFID readers, and other sensors that can enable
the phone itself, although costs for such access can vary automatic access to the appropriate audio clip depending on
considerably across regions. Many embedded devices only the visual content, whereas simpler sensor-free devices would
allow for a one-time write of the content; where they do allow rely on the user to select audio clips for a given visual display.
updating of content, the costs are roughly similar to that of
updating an MP3 player although updating an embedded B. Non-electronic visual display
system typically requires more specialized and complex tools This can be considered along several key dimensions:
than the mainstream ones available for MP3 players. Physical form factor and portability. The display can range
Capacity. Capacity ranges from 5 seconds for some from small and highly portable to larger fixed installations.
embedded electronics to days with MP3 players. Example form factors include sheets of paper, greeting cards,
Audio quality. This ranges from high-quality audio in CD brochures, books, posters, and even billboards.
and MP3 players to somewhat lower quality audio on mobile Cost. This can range from essentially free (e.g., hand-
phones to tinny audio in greeting card style embedded devices. scribbled content on a sheet of paper), to a few cents (printed
Volume also varies from personal levels in embedded devices paper), to a few dollars (printed books), to thousands of
to sharable levels in MP3 players with external speakers. dollars (billboards). The high cost of a billboard seemingly
Durability. With the possible exception of CD/DVD media detracts from our goal of ultra-low cost information
that is prone to scratching, the remaining options tend to rely dissemination, but the display’s cost can be amortized over the
on solid-state memory and electronics that are roughly number of people that might use it. Indeed, a printed sheet of
equivalently robust. However, the housing for the electronics paper priced at just a few cents that is only used once by one
could make a difference, with greeting-card style embedded person could cost more in aggregate than an expensive
devices being relatively fragile compared to an MP3 player. billboard that disseminates information to thousands of people
Power. Most of the devices surveyed have low power over a longer period of time.
consumption requirements. However, they vary considerably
340
Electronic integration. Normal visual displays for video was replaced with a paper poster that illustrates the
featherweight multimedia are non-electronic. However, more agricultural technique (Digital Green staff assured us that the
sophisticated versions might be inexpensively tagged to enable audio was understandable without the original video). Since
tighter integration with the associated electronic audio player. video is more visually engaging, we wanted to observe how
For example, RFID or optical tags might enable a talking book featherweight multimedia fares in comparison.
to determine which page is currently active, or optical codes Our explorations were conducted in a village in the state of
on a poster could be used to index into auditory content on a Tamil Nadu, India. Attendance at the mediated meetings (Fig.
camera-equipped audio player. 2), which take place in the evenings at a private home three
times a week, stayed relatively constant at approximately 20
IV. USAGE EXPLORATIONS
village residents throughout the meetings that we directly
We explored the viability of several forms of featherweight observed. The village comprises about 50 households or 160
multimedia via usage explorations in different field settings. individuals (50 men, 50 women and 60 children). Only 10% of
These were not intended to be formal experiments or extensive the population could read or write.
deployments, but rather initial forays. We experimented with
two different domains (agriculture extension and healthcare),
five device configurations, and five qualitatively different
preliminary investigations. The investigations were intended
to shed light on several questions we felt were critical to
determine the overall promise of featherweight multimedia:
Q1. User literacy and education requirements: A major
motivation for featherweight multimedia is that it relies on
audio rather than text for communication. As such, we expect
the technology to be usable by low-literacy populations. Is this
in fact the case? Are there other cognitive barriers to use?
Q2. User engagement: In contrast to more full-fledged
multimedia systems that provide a rich set of highly engaging
interactive content, will the simpler and minimally interactive Fig. 2. Mediator teaching agricultural information to village residents using
content provided by featherweight multimedia be sufficiently featherweight multimedia – poster with audio player
engaging that users will actually feel compelled to access
A. 1st Prototype: Poster + Custom Audio, Mediated
information through such devices?
Q3. Social support requirements: Ideally, the device is Our featherweight multimedia device for this experiment
usable by low-literacy populations on their own without any was a custom audio device (Fig. 3, left) that was meant to be
mediation from trained helpers or support from peers, but in hung alongside a poster. When pressed, each of the seven
reality, it is highly likely that some amount of social support buttons on the right play an audio file stored on a removable
will be required. To what extent is social support required for SD card. These audio clips are captions corresponding to
featherweight multimedia to be effective? mediator-drawn illustrations on the poster. On the left is a
button that stops playback, below which is an on/off switch.
V. EXPLORATIONS IN AGRICULTURE EXTENSION For ruggedness and portability, the device is equipped with a
handle and is covered with cardboard. A µmp3 player [31],
Our first exploration of featherweight multimedia was in the
powered by three AA batteries, drives external speakers. The
agriculture extension domain. We partnered with the Digital
cost of this custom device was approximately US$150, with
Green project [29], which seeks to teach targeted sustainable
nearly two-thirds going to the cost of the mp3 player.
agricultural techniques to small and marginal farmers in India
using video clips of their peers learning those same
techniques. Their human-mediated video approach has shown
higher adoption rates of the agricultural techniques as
compared to traditional methods of dissemination such as
paper posters or extension officers conducting 1-on-1 lessons.
The requirement of a TV and DVD player, however, poses
challenges in terms of cost, portability, and electrical power.
Interestingly, Digital Green’s own assessments show that a
version of their system where a poster is used in place of the
TV/DVD can achieve much of the gains, though at a lower
cost-effectiveness rate.
Instead of the TV/DVD, we experimented with several
kinds of featherweight multimedia, and also with the devices
in mediated and non-mediated contexts. In all cases, the audio
was copied from video clips used by Digital Green, while the Fig. 3. Custom audio player (left). MuVo V100 (right). Scales are not equal.
341
the sessions were again mediated.

At the time of our research, eight Digital-Green-style Observations and Feedback: Despite some usability issues,
meetings were conducted with this device (with no video). this combination of device, poster, and mediation worked very
Usage of the device varied from taking up almost all of the well. In fact, Digital Green has been actively continuing the
hour-long meeting to almost none on several occasions. We use of this device for five months, with over 50 meetings
attended four meetings and spoke with the mediator twice conducted to date. They report that adoption rates have been
outside of the meeting time. steadily increasing using our featherweight-multimedia
Observations and Feedback: Overall, both the mediator and solution, and that the rates are coming close to adoption rates
the audience received the device well. We found that the with TV and video content. This, together with other benefits
audience did not get restless even when listening continuously of the device such as portability, cost, and less reliance on
to longer clips of 5 minutes or more, although this is power grid (Fig. 4), makes featherweight multimedia a
surprising in light of previous research [29]. Half the audio powerful contender for this scenario.
clips on the SD card were played at each meeting. Although The experiment has not been without problems: even after
people did say they were tired of hearing the same content several meetings, the mediator had not mastered the bi-
over and over again, this repetition has been shown to increase directional switch. One cause of the difficulty is the fact that
understanding. One village resident said that she could listen the switch is functionally overloaded: when flicked, it
to the same audio two or three times before getting bored, and switches between tracks, but when held, it fast-forwards or
in fact needs to hear it that many times to fully understand the rewinds. This suggests that controls should be modeless,
content. So, the featherweight device appeared to serve as a especially to accommodate users new to electronics.
reasonable alternative to video.
Interestingly, no one else besides the mediator operated the
audio player, or expressed interest in doing so. It was seen as
something only the mediator should operate.
Meeting participants pointed out that since the device was
custom-built by the authors, when it broke (for one meeting), a
normal repairman could not fix it. They also commented that
the device looked ugly. Perhaps the most significant design
flaw was the device’s lack of a pause button. Pressing stop
“rewound” the audio to the beginning of the particular audio
clip. For this reason, the stop button was rarely used. On the
positive side, people from neighboring villages heard about
the meetings and asked the mediator about them with interest. Fig. 4. Featherweight multimedia working during a power outage!
B. 2nd Prototype: Poster + Off-the-shelf Audio, Mediated C. 3rd Prototype: Talking Book, Non-Mediated
Based on the feedback from the first iteration, we In this third exploration, we look at a more integrated
abandoned the custom audio player and switched to a featherweight multimedia solution: talking books. These
commercially available one (Creative Technologies MuVo devices are relatively cheap, have had some traction in
V100: Fig. 3, right). The commercial device is more durable, developed countries for childhood education, are built robustly
is capable of playing more than 7 audio files (10+ hours), and because they are designed for children, and provide access to
has pause/rewind/fast-forward capability. The total cost of the audio and visuals via a single integrated package.
system was US$38. Being a low-end device, the audio We repurposed a VTech WhizKid [10] device for
player’s screen cannot show the name of more than one audio agriculture information dissemination. The WhizKid can
file at a time. In addition, it does not have playlists and does operate in a standalone fashion, or it can be hooked up to a
not support titles in the local language. The audio player laptop via USB port, in which case the WhizKid tablet and
allows grouping of audio files by folder but the user interface pen are treated as a mouse by the operating system. In the
for switching between folders is difficult to use. latter mode, it is possible to customize the content and
Therefore, we created a numbering scheme where the title interaction completely, so that an inserted page of our own
of each audio clip consisted of two numbers separated by a design can be mapped to audio files we choose. (Should this
hyphen. The first number is a code for the topic, for example solution work, the intent would be to work with the
(1=azolla, 2=vermicompost). The second number is the clip manufacturer to produce agriculture content for the device in
number within the topic. A printed guide was given to the its standalone mode.) The retail cost of the device without the
mediator explaining the code. Navigation is accomplished by laptop is approximately US$25. The WhizKid has another
skipping forwards and backwards through the list of tracks by benefit in that its robust, colorful design is less intimidating
flicking a bi-directional self-centering spring-loaded switch. than a regular PC. During our studies, the laptop was hidden
Since this is cumbersome with a large number of audio clips, so as to not intimidate users. Screen shots from the video were
we only loaded a few topics at a time onto the audio player. used for visual content (Fig. 5), and relevant portions of audio
Posters were designed similar to the first experiment, and were copied from the videos.
342
During the post-experiment feedback session most of the

users were able to answer the factual questions on the
agricultural processes (mean of 75% answers correct for N=4).
This confirmed that most users had been able to retain the key
pieces of information that they had heard.
Crowds immediately gathered when a participant used the
device. Many of the bystanders often prompted the primary
user. Also, there were a number of instances when people who
listened to the audio information enthusiastically participated
in the post-experiment discussion. One user said that she
would be willing to share the device if the device was left at a
common place in the village and remarked that it will be quite
Fig. 5. An audio-augmented page for a talking book
useful on days she missed the Digital Green video-screening
sessions. This suggests that featherweight multimedia can be
We walked around a village on two different days, for a complementary to other forms of information dissemination.
total of six hours, to various households and requested
residents to try out the device in their homes. Besides the
second author, a translator and a mediator were also present.
Each user was given 20 minutes of free use to work with the
device, and was allowed to continue if he/she wished. We
provided no up-front instruction, but if the user failed to
engage with the device at all after five minutes, the mediator
or the translator provided gentle and low-grade assistance.
A feedback questionnaire was administered verbally that
asked for the user’s background information and opinions
about the device. For the second day (N=4), we also
administered a verbal test and feedback session after each use,
to evaluate how much of the presented information the user Fig. 6. A user navigating through the talking book for information.
had learned and to hear comments about the device. D. Discussion
Observations and Feedback: In spite of some initial
These explorations revealed some preliminary insights
hesitation, most of the villagers who tried the device were
relative to our study questions.
comfortable with it and could use it with little or no assistance.
In terms of literacy and necessary user background,
Those participants who took the pre- and post-tests showed
featherweight multimedia appears able to bridge challenges of
increased knowledge of the content presented.
illiteracy. Audio playback seems sufficient for this purpose,
Over six hours spread over two days, we were able to find
and the addition of even static visual images helps anchor
ten people (5 male, 5 female) willing to try the device to learn
discussion in mediated sessions. This is not a surprise, but it is
agricultural content (Fig. 6).
assuring that there were no unexpected surprises with respect
Individuals were very hesitant to press anything at first. The
to adults with little formal education being able to make sense
elapsed time between when an audio clip had finished playing
of these devices. Many of the participants in this phase of the
and the next selection was chosen was initially as long as 10
study had also been previously exposed to agricultural content
seconds. With more use, this reduced to about one second.
on video from Digital Green, yet none explicitly expressed a
Many users repeated aloud certain pieces of audio information
negative comparison with video.
after they heard it spoken from the device.
The simple interface of featherweight multimedia also
Older users were slower to learn the device and slower to seems to reduce barriers for the first-time user. Users related
learn the content – they listened to all the clips at least twice. well to the use of local content and illustrations depicting local
One user kept the pen pressed down while the audio was surroundings and people, enabling viable user engagement. In
playing and never figured out that tapping would suffice. several cases, both in mediated and non-mediated use, it was
On the first day, we had black-and-white printouts of the seen that users repeated pieces of information to re-affirm
video. Color turned out to be essential for some of the audio their knowledge. This process of repeating the narration may
descriptions, and so on the second day, we returned with enhance retention and assimilation of information. For us, it
color-printed sheets. confirms that the users were paying attention in the first place.
Users did not experiment very much. They never pressed Featherweight multimedia also seems sufficiently engaging
the Pause button unless it was pointed out to them. They only so as to enable effective information dissemination, both under
played the audio in sequence, even while listening to the audio mediated and non-mediated circumstances.
captions for the second time to review the material. When In mediated sessions, audiences were not only willing to sit
questioned later, users said that they feared that they might through uninterrupted audio playback of five minutes or more,
spoil the device. in many cases, they requested repeated playback of the audio.
343
In non-mediated situations, users expressed initial findings that audio annotations enhanced comprehension of
trepidation handling the device. One remarked that he was drawings provided further justification to our intent to add
afraid he might spoil the device, and others visibly hesitated in audio to these posters.
early interaction, until their confidence grew. In spite of the
initial reluctance and limited direct guidance, most users were
able to engage meaningfully with the device after just a few
minutes of exposure to the device.
During post-experiment feedback sessions, several people
expressed a desire for shorter audio clips that summarized
longer content. On the other hand, longer audio clips engaged
a greater number of passive users, attracted more passers-by,
and generated more discussion because of the clips’ longer
playing time. Whereas longer clips sustain the attention of a
larger group of people, shorter clips cater to an individual’s
short attention span.
With respect to the need for social support, a human
facilitator is essential, at least in the initial introduction of the
device, regardless of the device’s perceived ease-of-use. In all
of the mediated sessions, the audience fully took in the content Fig. 7. Post-operative do’s and don’ts poster. We added numeric codes to turn
it into a “talking poster” when coupled with an audio player.
without concerns about the device. In contrast, users in the
non-mediated sessions needed prompting and encouragement A. Technology
to use the device at all. Based on our agriculture participants’ difficulties with the
The need for such mediation, however, appears to decrease MP3 player interface, we decided to explore a different device
with time and also with group interaction. When crowds for this setting. We wrote a program that displays a large
gathered in non-mediated sessions, the actions of the primary numeric keypad on a HTC Touch smartphone’s touchscreen
user were influenced by people loitering around. They either (Fig. 8). As numbers are typed in, they appear in the textbox at
voiced common requests or prompted him when the user was the top of the screen.
stuck, serving as encouragement. Such voluntary peer
guidance further alludes to the effectiveness of social support.
VI. EXPLORATIONS WITH HEALTHCARE INFORMATION

In this second exploration, we sought a more institutional
setting in which featherweight multimedia devices might be
used. We partnered with Sankara Eye Hospital, located in
Coimbatore, Tamil Nadu, India. It provides free eye-care Fig. 8. HTC Touch running featherweight multimedia application.
surgeries to 700-1000 patients per week, funded by the much Unnecessary buttons below the touchscreen are covered.
smaller segment of patients who can afford to pay. Sankara
On the poster, numeric codes are printed in each box (Fig.
identifies non-paying patients by conducting 3-5 weekly
7). When one of these codes is entered via the numeric
outreach screening camps in villages. The patients requiring
keypad, the appropriate audio caption plays in the manner of
eye surgery are bused to the hospital, where they receive free
museum audio tour guides. No “enter” or “play” button is
treatment, room and board over several days, followed by a
needed, since the numeric codes are all the same length (in our
bus trip home. Two of the authors spent a week visiting
case, two digits). There is also no “clear” button. If an
Sankara Eye Hospital to find opportunities for deploying
unrecognized code is entered, both the textbox and the
featherweight multimedia. We observed screening camps in
program’s numeric buffer are cleared. These design decisions
rural areas and shadowed patients around the hospital.
were made to simplify the user interface as much as possible.
After these visits, we decided to focus on improving
Sankara’s informational poster on post-operative care, which
few patients were paying attention to despite the crucial nature
of the information. This information is also read aloud by a
nurse to the patients during discharge, indicating compatibility
with featherweight multimedia.
The poster is 0.9m x 0.6m in size. It explains ten things
patients should do on the left-hand panel, while on the right, it
lists ten things they should not do (Fig. 7). The images on the
poster are static cartoon drawings along the lines
recommended by Medhi et al. [26], although the poster
Fig. 9. Ward at Sankara Eye Hospital where we conducted our observations.
additionally includes text captions in Tamil. That paper’s
344
B. Iterative Explorations
We conducted four sessions of testing with the device, each
with slightly different usage scenarios and with a different
group of patients, in a ward with hundreds of beds (Fig. 9).
Around 200-230 patients were registered in each ward, but
occupancy ranged from 70-230 at any time. Tests were
conducted in the early evening when patients were most likely
to be present. A sampling of 75 patients over the four sessions
showed that the majority were illiterate; the few who were
literate often had difficulty reading text due to poor eyesight.
All tests were conducted by one of the authors and a
translator. We tried to interview anyone who interacted with
the device or poster after they had finished.
1) First Session: Regular Poster, Non-Mediated
To establish a baseline, we set up a non-augmented poster,
and sat where we could observe people viewing the poster.
Fig. 11. Augmented poster set up with audio device and speakers
Over the course of a 90-minute observation, we estimate that
150-200 people (non-unique) passed by the poster. Of those, First, the translator used the device in an animated,
only 7 people looked at the poster (Fig. 10); the rest passed by demonstrative manner for three minutes, gesturing to the
without a glance. We questioned 5 of the 7, and among them, numbers on the poster and theatrically pressing the onscreen
only one was able to answer simple questions about the keyboard. Three patients gathered around, but after the
content of the poster. Among the rest, one person even asked translator finished demonstrating, the patients started to leave.
us whether it was permitted to sleep on the operated side, We spoke to one of the three, and he was able to explain the
despite the fact that the poster expressly prohibits this. purpose of the device and the correspondence between the
poster, the printed numbers, and the resulting sounds. When
prompted to try the device for himself, however, he demurred,
saying that he was afraid to break the device.
During the second two hours, we played a pre-recorded
Tamil-language audio prompt, “Learn about eye care by
pressing the numbers,” every time a person passed or glanced
at the poster, simulating a motion-sensitive trigger. The audio
prompt never once achieved its intended goal, and the patients
always left without trying the device.
Occasionally during the two hours, the translator again
either demonstrated the device himself or actively encouraged
patients to try it, and managed to get a total eleven people to
try the device. Every time the poster was used, a crowd of 2-
12 people gathered (Fig. 12), but would then disperse when
the person operating the device stopped.
Fig. 10. The regular poster, with no audio annotation: This is a rare patient
who is actually taking the time to look at the poster. Our translator is the
gentleman in white; on the left, two women are passing by without showing
any interest.
2) Second Session: Audio Poster, Occasionally Mediated

We set up the poster with numeric codes in the same
location as in session one, but this time we put the audio
device in front of the poster on the table (Fig. 11), with
speakers positioned behind. The backlight on the device was
left on throughout the session.
We ran this session for 3 hours, during which time a total of
230 non-unique patients passed by the poster (we estimate 60-
70 unique patients). Because it was immediately apparent (as Fig. 12. A brave patient interacts with the poster (seen from the back) while a
crowd looks on.
with the first session) that very few patients would interact Nine of those who tried the device were interviewed. All
with the poster or the device unprompted, we tried a number but two had either glanced at the poster or had been in a crowd
of different things throughout the session to encourage them. of people watching the device being used, then had been
taught and encouraged to use the poster by the translator.
345
The other two were instances of patients teaching other about the device and the health information. One of the men in
patients how to use the device. In one instance, a young man the group was a retired 70-year-old teacher. Five minutes later,
in his mid-20s who was accompanying one of the patients the teacher was gesturing animatedly to the assembled people
used the device through encouragement by the translator. He around him, demonstrating the device, but keeping control of
then taught another man of the same age, who was also there it. The translator then told the teacher to encourage others to
to accompany a patient. He used the device for several use the device rather than to use the device himself as a tool to
minutes. One of them then came back an hour and fifteen teach the content. The teacher then moved to different parts of
minutes later to continue interacting with the device. the ward, teaching other groups and having discussions with
In the other instance, a man and a woman stood in front of them (Fig. 14). Most of the time he did not relinquish control.
the poster. The woman was about to use the device, but the Often, a member of the group pointed to a number on the
man tried it first. After each audio caption is played, they brochure, which the teacher then typed in. A few times, other
nodded approval together. The woman then took over, and a people were observed pressing the numbers themselves.
crowd of six gathered. She pressed and listened to a few Thirty-five minutes after the teacher got control, he returned
captions. Another woman joined her, and the first woman the device and the brochure to us, and reported that he had
systematically pressed all the numbers, taking 6 minutes. After completed teaching all 30 patients who were awake!
each caption was played, the first woman made sure that the
second woman understood the content.
There were at least two people who were not able to
understand the use of the device even after the translator’s
explanation.
For this group of patients, active human mediation seems
necessary even to get to a point where patients will try the
device, and even then only a small minority will do so.
However, more people expressed curiosity and stayed to listen
if someone else was operating the device. The audio prompt to
invite usage (without mediation) was a failure.
3) Third Session: Audio Poster, Mediated
In this session, we asked a nurse to explain the usage of the
audio poster to all patients in the ward over the PA system.
Immediately afterwards, a crowd of about 12 patients gathered
around the poster, and the nurse demonstrated to that group.
Fig. 13. Translator (standing) teaching a group of patients how to use
Individual patients in the crowd then tried using the device smartphone with brochure
assisted by the nurse. Within eight minutes, the crowd
dispersed. A second PA announcement an hour later did not
result in significant sustained interest.
4) Fourth Session: Audio Brochure, Peer-Mediated
For the fourth session, we tried a more intimate form factor
and a different kind of mediation. Since patients spend a lot of
time sitting on their beds waiting for treatment or recovering,
we decided to take the content to them instead of having them
come to the poster. To this end, we created an audio brochure,
which is simply the poster shrunk onto A4-sized paper that
could be passed around. All words and numbers remained
legible, and the audio-guide device remained the same.
On the first day of this experiment, we taught one of the
male patients how to use the device. The patient was asked to
use the device for himself, teach others to use the device, and
then to pass the device on. The patient took on his Fig. 14. Patient (a retired teacher) teaching other patients
responsibility eagerly, perhaps overly so. Over the course of
10 minutes, he demonstrated the audio brochure to 11 patients, We interviewed 10 people (besides the teacher) who had
letting them hear a few captions before moving on to another either tried the device or learned from the teacher. For the
patient, without relinquishing the device. most part, they reported that they felt that the information was
The next day, the translator taught a group of several very important and were able to recall some key points. There
patients to use the device and pass it along (Fig. 13). After a were no negative responses, and no one said that they had
few minutes, the group taught a group of women sitting across difficulty in using the device. One woman even expressed a
from them how to use the device. Two women were then desire to buy the device, but said that unfortunately she did not
observed sitting at the feet of the men and having a discussion have enough money on her!
346
C. Discussion can be easily modified to create devices of varying sizes, form

The social framework that we found to work in institutional factors (from A5 to poster sizes), and costs. The kit would
healthcare settings involves (1) approaching groups, (2) using consist of a board with an array of buttons that trigger audio
a portable form factor (in this case the brochure/smartphone), playback. Sheets of paper could be overlaid onto this board
(3) providing small-group tutorials with instructions to the with playback icons positioned over the underlying buttons.
This kit would cost US$20-25 for an A4-sized "audio tablet"
group to teach others. In some cases, one member of the group
with a 1GB SD memory card. However, the kit would cost
will take the initiative and take charge of teaching others. We
only US$0.50 to make an A5-sized “audio card” that contains
make several further observations: 10-15 seconds of audio content.
First, technical manipulation of the interface was not a
problem for most of the patients (except for two) who VIII. CONCLUSION
experienced the device. In fact, even patients who did not
directly interact with the device, but instead observed others In this paper, we presented a broad exploration of
featherweight multimedia – combinations of electronic audio
operating it, were able to understand what it did and to
and static visuals – that can be put together for significantly
associate the audio with the visual information.
lower cost than even very low-cost PCs, but can nevertheless
Second, user engagement varied greatly depending on the
provide a rich, interactive multimedia experience.
person and on the circumstances of device introduction. Most Within the contexts of agriculture extension in rural villages
patients seemed reluctant, even afraid, to use the device, and healthcare information dissemination in an eye-care
particularly when it was placed in a public location with a lot hospital, we tried exploratory studies with five different
of patient traffic, even if they understood how to use the configurations of featherweight multimedia devices, all with
device. Patients said that they were fearful of breaking the illiterate and semi-literate users. Over 75 people directly
device. This is consistent with previous findings on interfaces interacted with the devices, and more than 150 people
for the illiterate, where inexperience with electronic devices participated in formal or informal learning sessions in which
creates a barrier for first-time users [28]. These fears were not they observed the use of the devices.
allayed by audio prompts from the device. The results from this preliminary exploration suggest that
On the other hand, when given private tutorials, some users most illiterate users can become immediately comfortable with
would all but monopolize the device, appointing themselves the simple interfaces provided by featherweight multimedia,
both teacher and technology expert for their peers. Between although fears of breaking the technology and such can create
these two extremes, there was a range of engagement patterns barriers to casual usage. At the other extreme, some users,
– one point of note was a tendency for crowds to gather especially when given short, private tutorials on usage, seem
around a person who was willing to operate the device, and to delight in using these devices and showing them off to
then to disperse when the person stopped. peers. We also note that, consistent with much of the literature
Third, it seems clear that human mediation of some kind is on technology for development, that social support for the
technology is an essential component of impactful usage.
essential for the device to be used, at least initially. The public
poster itself was largely ignored, and the device was used only
IX. ACKNOWLEDGEMENTS
when our translator encouraged people to try it. On the other
hand, peer tutorials were very effective in encouraging use. Thanks to Joice Sister, M. Saravanan, and Dr. R.V. Ramani
Fourth, the brochure was more effective than the larger at Sankara Eye Hospital; Mahalaxmi and Srikantamurthy at
poster. This may have been because patients perceived it less Digital Green/GREEN Foundation; S. Sivaraju, Ajay Khanna,
like hospital property, or because it was easier to pass around. Kalai Arasi, Meera Lakshmanan, and Avinash Setty for
Finally, we observed several usability problems with the translating; and Sid Vishwananthan & family, Leah Findlater,
and Gursharan Singh.
smartphone application, but once patients were taught how to
overcome the problem, they continued using the device. One
X. REFERENCES
patient slid his finger between numbers while another pressed
[1] One Laptop Per Child. [Online]. laptop.org. [Accessed Aug 31, 2008].
too hard. There were also mis-registration problems with the
[2] Classmate PC. [Online]. classmatepc.com. [Accessed Aug 31, 2008].
touchscreen, causing numbers to be mistakenly pressed twice [3] International Team Aims to Develop a $12 Laptop During MIT Summit.
or the adjacent number being pressed instead. Most of these [Online]. chronicle.com/wiredcampus/article/3219/international-team-
issues would be remedied by using a large, physical numeric aims-to-develop-a-12-laptop-during-mit-summit. [Accessed Aug 31,
2008].
keypad instead of a touch-screen virtual keypad. [4] Lourie, S. (1990). World literacy: where we stand today - One billion
non-literates. Editorial, UNESCO Courier. July 1990.
VII. FUTURE WORK [5] Medhi, I., Sagar, A., Toyama, K. (2006). Text-free user interfaces for
illiterate and semi-literate users. Proc. ICTD. p. 72-82.
We hope to better understand how to integrate [6] Books of Hope. [Online]. booksofhope.com. [Accessed Aug 24, 2008].
featherweight multimedia into development-focused programs [7] Q & A about Speaking Books. [Online].
booksofhope.com/pdf/Q%20&%20A%20about%20Speaking%20Books
and organizations, as well as to measure their educational
%20.pdf. [Accessed Aug 24, 2008].
impact in comparison to other forms of media. [8] SADAG’s school based curriculum evaluation. [Online].
We are also exploring a possibility of a low cost “kit” that booksofhope.com/research/SADAG%27s%20School%20Based%20Curr
can be used to construct featherweight multimedia devices in iculum%20Evaluation.pdf. [Accessed Aug 24, 2008].
[9] LeapFrog. [Online]. leapfrog.com. [Accessed Aug 24,2008].
which the audio and visual content are customizable. This kit
347
[10] VTech WhizKid . [Online]. vtechwhizkid.com. [Accessed Aug 24, [23] Plauche, M., Nallasamy, U., Pal, J., Wooters, C., Ramachandran, D.
2008]. (2006). Speech recognition for illiterate access to information and
[11] Fisher-Price PowerTouch [Online]. fisher- technology. Proc. ICTD. p. 83-92.
price.com/us/powertouch/default_flash.asp. [Accessed Aug 24, 2008]. [24] Sherwani, J., Ali, N., Mirza, S., Fatma, A., Memon, Y., Karim, M.,
[12] HHS to provide new interactive book of health information to women of Tongia, R., Rosenfeld R. (2007). HealthLine: Speech-based access to
Afghanistan and their families. [Online]. health information by low-literate users. Proc. ICTD. p.. 131-139.
hhs.gov/news/press/2004pres/20040803.html. [Accessed Aug 24, 2008]. [25] Parikh, T., Javid, P., Sasikumar K., Ghosh, K., Toyama, K. (2006).
[13] PicTalk. [Online]. edindiasolutions.com/index-4.html. [Accessed Sep Mobile phones and paper documents: Evaluating a new approach for
20, 2008]. capturing microfinance data in rural India. Proc. CHI. p 551-560..
[14] Literacy Bridge. [Online]. literacybridge.org. [Accessed Aug 24, 2008]. [26] Medhi, I., Prasad, A., Toyama, K. (2007). Optimal audio-visual
[15] Tour-Mate. [Online]. tourmate.com. [Accessed Aug 24, 2009]. representations for illiterate users of computers. Proc. WWW. p. 873-
[16] Prentke Romich Company. [Online]. prentrom.com. [Accessed Aug 24, 882.
2008]. [27] Digital StudyHall. [Online]. dsh.cs.washington.edu/info/papers.html.
[17] Stifelman, L, Arons, B., Schmandt, C. (2001). The Audio Notebook. [Accessed Aug 24, 2008].
Proc. CHI. p. 182-189. [28] Medhi, I., Toyama, K. (2007). Full-context videos for first-time,
[18] Livescribe. [Online]. livescribe.com. [Accessed Aug 24, 2008]. illiterate users. Proc. ICTD. p.140-148.
[19] Daisy. [Online]. daisy.org. [Accessed Aug 24, 2008]. [29] Gandhi, R., Veeraraghavan,R., Toyama, K., Ramprasad, V. (2007).
[20] Dolphin. [Online]. yourdolphin.com. [Accessed Aug 24, 2008]. Digital Green: Participatory video for agricultural extension. Proc.
[21] Tang, A., Finke, M., Blackstock, M., Leung, R., Deutscher, M., Lea, R. ICTD. p. 21-30.
(2008). Designing for bystanders: reflections on building a public digital [30] Talking Greeting Card. [Online].
forum. Proc. CHI. p. 879-882. simaproducts.com/products/product_detail.php?product_id=632.
[22] Vogel, D., Balakrishnan, R. (2004). Interactive public ambient displays: [Accessed Aug 31, 2008].
Transitioning from implicit to explicit, public to personal, interaction [31] µMP3 Playback Module. [Online].
with multiple users. Proc. UIST. p. 137-146. roguerobotics.com/products/electronics/ump3. [Accessed Aug 24,
2008].
348
ICT Governance in Higher Education:
Case Study of the Rise and Fall of Open Source
in a Gulf University
Sofiane M. Sahraoui
friendliness, availability of support, and the validity of its

Abstract—Open source software is a natural fit in higher business model [2]. Yet it is at the application, hence visible,
education as both are based on an ‘open science’ model of
knowledge development, and collaborative learning. It level that the FOSS model offers its greatest promise not only
also helps bridge the gap between knowledge production through idiosyncratic solutions that cater to the specific
and its use whenever academics become involved with requirements of university and educational computing, but
open source development contrary to the proprietary also by providing a model for collaborative learning and
model where software is produced within the confines of capacity building [3]. Whilst resistance to FOSS, beyond the
commercial vendors R&D labs. However in higher bottom of the software stack, is thought to be a universal
education, open source has not broken through the bottom phenomenon, the problem is compounded in the developing
of the software stack where it is confined to fulfilling world with governance issues. In the absence of proper ICT
mostly system computing requirements. At the governance in developing countries’ higher education
application and visible level, open source adoption is institutions (HEIs), FOSS adoption is left tributary to the
hampered by a number of factors chief amongst these is whims of the power-play between users, IT departments, and
the lack of reliable ICT governance structures. The university administrators with the outcome rarely in favor of
problem is exacerbated in developing country contexts FOSS and if so hardly sustainable over the long term.
where governance in general is subjected to the power- There has been little attention devoted to studying FOSS in
play between several actors of the higher education scene. HE let alone in developing countries’ HE [4]. The few studies
The governance conundrum faced by open source that exist mostly focused on inhibitors and facilitators of
adoption in developing countries is illustrated through a FOSS in the university environment and few of these dealt
detailed case study of an open source project failure at with the context of developing countries. A landmark
GNU, an American-style university in the Gulf region. contribution by David and Shapiro [4] examined the extent to
The rise and fall of GNU open source project is discussed which FOSS is being created and used within emerging
within a general framework of ICT and open source countries’ HE sector. However, hardly any study exists to
governance in developing countries higher education depict vividly the process of FOSS adoption and its
sector. Lessons are drawn from the case to recommend a subsequent becoming in an emerging or developing country
FOSS strategy for development. context. This paper purports to do just that by chronicling the
ill-fate of an initially successful FOSS project in a Gulf
Index Terms—Case Study, Developing Countries, Gulf, Higher university. In the specific context of this case, ICT
Education, ICT Governance, Open Source Software governance, or lack thereof, is thought to be the root cause for
FOSS failures. Details of the ‘rise and fall’ of a specific FOSS
project, help illustrate the general governance conundrum
I. INTRODUCTION faced by ‘innovators’ in a developing country context.
he global higher education (HE) scene has seen an increase The paper is structured as follows: section II places FOSS
T in the adoption and use of free/open source software
(FOSS). However this presence tends to be restricted to
within a development discourse and section III introduces ICT
governance in HE. Section IV illustrates the natural fit of
FOSS within HE with a corollary treatment of developing
the bottom of the software stack; mainly the infrastructural
components of university computing that are not visible to the country contexts. Section V chronicles a case study depicting
users [1]. On the application side, the debate does not seem to the failed adoption of FOSS in a Gulf university environment.
have evolved beyond the technicalities of FOSS user- In the final two sections, lessons are drawn to help foster the
adoption of FOSS in developing countries’ university
environments.
Manuscript received September 22, 2008.

II. OPEN SOURCE FOR DEVELOPMENT
S. M. Sahraoui is an Associate Professor of Management Information

In developing countries where traditions of free knowledge
Systems at the American University of Sharjah, UAE. He is currently visiting pursuits are rather restricted by the lack of resources and the
with the University of Tunis LARODEC research laboratory. (phone: 216-23- inadequacy of governance systems, proprietary software
554-255; e-mail: ssahraoui@aus.edu).
349
further limits the opportunities for skill development and belonged to Africa and the Middle East with many not acted
knowledge transfer [4]. FOSS thus presents developing upon [9].
countries with an immeasurable opportunity to plunge into The following section depicts the process of ICT governance
open and localized sources of knowledge [3] and muster the in HE in the context of Open Source adoption.
rudiments of knowledge creation and assimilation beyond the
confines of the poorly endowed local environments. FOSS III. ICT GOVERNANCE IN HIGHER EDUCATION
was thus hailed as the most promising avenue for developing As ICT becomes ubiquitous permeating all facets of
countries to achieve ‘technological self-determination’ [5]. university computing including teaching and learning, it
However evidence points that developing countries are failing becomes primordial to take an institution-wide approach to
to harness this great potential of FOSS for development [6]. identify computing requirements and plan their adoption and
The beneficial effect of FOSS is thought to materialize in dissemination. In this respect, ICT governance provides the
developing countries only if (1) basic problems of Internet set of responsibilities and practices that are exercised to
access are resolved which is far from being the case as “the provide a strategic direction, evaluate achievements, manage
average African university has bandwidth capacity equivalent risks, and generally ensure that institutional mechanisms are in
to a broadband residential connection available in Europe, place to implement the ICT strategic plan [10]. Strategic plans
[and] pays 50 times more for their bandwidth than their encompass the campus management of enterprise systems and
educational counterparts in the rest of the world” [7]; (2) a services with an assorted strategy for software development
critical mass of FOSS programmers is nurtured given the lack and acquisition [2]. Planning covers administrative
of skilled personnel available to develop hence support FOSS computing, academic computing, teaching & learning and
systems [5] 1 and (3) appropriate governance structures are research.
enacted to follow through with FOSS-friendly policies. The Universities are characterized by a divide between the
first two factors are infrastructural and as such can be resolved administrative and the academic with lingering conflicts
through relevant investments. The third factor, which is between the two. This conflict contributes to the ills of ICT
governance-related is more critical, and could thwart any governance in HE. For administrative computing, faculty
attempt to steer clear of the current situation of committees, whenever they exist, are called upon to offer their
underachievement. input into a university computing strategy without a real
In Brazil for instance, over a third of policies in a sample of consequence when systems are implemented, as faculty are
HEIs required the procurement of FOSS software whenever left out of the loop for such decisions. The adoption of
available, but in the absence of appropriate governance expensive administrative systems is usually perceived by
structures to ensure due implementation of such policies, there academics as a big opportunity cost because it takes away
is little accountability for their frequent violations [4]. resources from the core academic mission of the university
Moreover and in countries that are considered as leaders in IT [11]. For academic computing which directly impacts
in the developing world, India being a case in point, FOSS teaching and learning and more generally the direct working
was not found to play a major role in HE organization and environment of faculty, the stakes for defining an ICT strategy
delivery [4]. To the contrary, Indian HEIs dedicated a are more significant. ICT components that are dealt with in
proportionally higher amount of their IT budget to software this regard include e-learning and related solutions, computer-
and license fees in comparison with other countries like based teaching, the integration of ICT in the curriculum, and
Croatia and Bulgaria which are economically better off [4]. the general organization of IT services throughout campus
India’s involvement with FOSS seems rather situated within including hardware and software support and resources
the realm of global software outsourcing with a limited available for faculty-initiated IT activities in teaching and
knowledge creation and spillover effect. In their survey of a research. The faculty is usually involved in the elaboration of
number of HEIs located in 8 developing and transition an academic computing strategy through different channels of
economies, David and Shapiro [4] found that the number of university governance.
FOSS courses offered was very limited, hinting that FOSS is It is at the level of adoption and implementation that
more of an industrial and business phenomena than any conflicts arise as decision-making shifts to administrators
promise of a new FOSS-inspired learning model premised on especially for enterprise applications like LMS and learning
a more egalitarian redistribution of knowledge production portals. The disempowerment of faculty within HEI
between North and South. In contrast, Glance et al. [8] survey governance structures can be accentuated through a
of FOSS use in Australia, New Zeland, and the UK found it to procurement system that generally does not involve them [4].
be well integrated at all levels of university operations be it Furthermore, the adoption of innovative uses of ICT for
administration, teaching, laboratory, or research. This teaching and learning, though prescribed in ICT plans, can be
suggests the important role of overall ICT governance in a blocked for budgetary reasons heightening the frustration of
country in the uptake of FOSS initiatives. Indeed, faculty and the demise of ICT strategies. Lack of evaluation
Government policies or lack thereof could be a further and accountability mechanisms contributes to the impunity in
hindrance to FOSS adoption in the developing world. Out of a violating the ICT strategic plan and undermining the related
total of 275 FOSS government initiatives in the world, only 8 governance structures.
Though the above seems to depict a dysfunctional
university environment, it is very close to the reality of HEIs
1
The Global Desktop Project by UNU-IIST seeks to grow the number of in developing countries. Worse yet, in many countries,
FOSS programmers in East Asia (cf. http://www.iist.unu.edu/globaldesktop/)
350
structures of shared governance that would enable faculty block against open source adoption beyond the infrastructural
participation in the elaboration of ICT strategies are level [22]. These factors are examined below within the
sometimes inexistent. Shared governance being distributed context of HE. First, we highlight the natural fit between
conveys a loosening of control from the center; an idea that is FOSS and HE as they are both underlined by an open learning
resisted in environments accustomed to centralized decision- model.
making [12]. Decisions about ICT for administrative as well
A. Open Learning Model
as for academic purposes are then taken unilaterally by
administrators who are sometimes only remotely connected to The most prominent examples of FOSS initiatives in
HEIs. In his study of HE reform in Tunisia, Sahraoui [13] universities consists of teaching FOSS tools like Linux, PhP,
illustrates how strategic planning for change in that country’s Perl, and MySQL [17]. Educational portals like uPortal and
HE system was conducted centrally at the level of the ministry learning management systems (LMS) like Moodle and Sakai
of HE with a token involvement of faculty within ad-hoc have also been emerging as new foci of FOSS academic
structures that were defined outside of the existing structures initiatives [21]. As of late, the development of the Kuali
of faculty governance. Within such governance framework, financial system as an enterprise management solution for
initiatives to embed open source in teaching and learning HEIs seems to be porting the open source tide to the
become tributary to the power-play of forces between different administrative side of academe [22]. Class wikis, wherein
stakeholders in the ICT governance game. Similarly and students and professors alike jointly construct learning
though FOSS might be extensively used within developing artefacts within a Wikipedia-like environment are becoming a
country contexts, it is generally in the absence of clear dominant collaborative learning tool and have been integrated
software policies and at best within the framework of IT within commercial LMS like Blackboard for instance [23].
‘neutral’ policies [4] but rarely within the mandate of pro- Prominent institutions like MIT and Harvard use similar
FOSS ICT policies. platforms to enable open access to teaching and research
The next section deals specifically with Open Source resources [23]. Projects of a more global nature are also being
governance in HE highlighting the unique fit between FOSS developed with open source tools including providing online
and HE. textbooks to the developing world [24]. For a comprehensive
list of FOSS tools used in education, see [3]2 .
Open source has been making headways in HE through
IV. HIGHER EDUCATION AS A NATURAL HABITAT FOR OPEN leading infrastructure products like Apache, Linux, MySQL,
SOURCE SOFTWARE Firefox, and Tomcat. In the US, a third of HEIs use some
form of open source infrastructure software although some are
At a time when organizations embark upon IT infrastructure not aware that they are doing so [14]. However its impact
upgrades and within increasingly austere budget environments beyond the bottom of the software stack has been more
and Internet-based open computing platforms, FOSS has mitigated [1]. Moreover few academic institutions seem to
emerged as a promising alternative to proprietary solutions. have a strategy to integrate open source in their academic
Whilst the cost factor and the probing quality and prominence offerings so as to ensure students’ familiarity with an
of open computing products should logically tilt the balance to increasingly dominant reality in the business world [22]. On
the advantage of open source over proprietary solutions [14], the administrative side of computing, few FOSS alternatives
[15], IT infrastructures remain largely proprietary [11]. The are being considered despite the greater concern and
cost, reliability and security implications for organizations dissatisfaction with proprietary solutions [26]. The support
with mainly proprietary platforms are significant, yet limited factor especially for enterprise applications is usually invoked
consideration seems to be devoted to alternative open-source first against FOSS adoption although some research found
platforms [2]. Moreover the response to the ‘IT paradigm support for FOSS to be equivalent or better than for
shift’ engendered by open source computing remains largely proprietary solutions [24]. Doubts about the legitimacy or
limited to the bottom of the software stack [16], spearheaded credibility of the development model of FOSS are also
mainly by the increase in server installations of the Linux routinely expressed. This has a double cost, the first financial
operating system which stands at roughly 20% of total incurred through the payment of licensing fees and expensive
installations and the Apache Web server which powers more upgrades, and the second is more fundamental and consists of
than two thirds of corporate websites [11]. Indeed and beyond the opportunity cost in foregoing the collaborative learning
this level, open source remains largely absent from the model of FOSS [24]. Even if one were to concede that the
‘application layer’ such as enterprise applications or TCO of open source solutions is as high or higher than that of
productivity tools [17]. Whilst many factors can be ascribed proprietary software, the advantage of FOSS over proprietary
to this including the gap between the technologists software is more fundamental and has to do with the open and
(developers) and the non-technologists (users) who hold collaborative learning environment afforded by FOSS and
different value-orientations regarding control and openness denied by the proprietary. Using the model of sourceforge.net
[17]; the risk attached to ‘entrepreneurial’ and individually- or one of its offshoots in the educational field, such as
centered modes of software development [18]; and the higher schoolforge, eduforge or any other similar virtual development
IT skill requirements of open source to configure and maintain environment, FOSS tools are amenable to transform the
software in the absence of conventional service contracts [19],
this paper contends that governance factors especially within
2
developing country contexts are by far the biggest stumbling A repository of FOSS for HE is downloadable from
http://www.iosn.net/education/software/.
351
classroom environment from a teacher-centered environment toolset but also as a development ethic akin to the
where knowledge is distilled by the instructor to a learner- universalism and openness of academic environments. An
centered learning environment where students develop skills added benefit for teachers is their reintegration within
and competencies beyond the mere acquisition of content communities of practice as many of the open source projects
knowledge [24]. The combination of access to the source for HE are being collaboratively developed across HEIs [14] 3 ;
code with the experiential learning within collaborative in contrast with their current isolation within the meanders of
development environments affords endless possibilities to functionalized teaching [24].
build learning artefacts that go beyond the mere use of The narrative below, depicting the failed adoption of open
software tools and could radically transform educational source at Gulf university is a vivid illustration of how the ills
practice for teachers and learners. of governance can bring to a halt, ambitious changes in the
“this marks the emergence of global science and knowledge curriculum. Further discussion of governance issues in HE
as a global public good that rest on an ethic of participation will be invoked as relevant details of the case unfold.
and collaboration based on the co-production and co-design of
knowledge goods and services…the role of nonmarket and V. OPEN SOURCE PROJECT IN A GULF UNIVERSITY
non proprietary production promotes the emergence of a new Gulf National University (GNU) 4 is a recently established
information economy that both depends upon and encourages American-style university in the Gulf and has quickly emerged
great individual freedom, democratic participation, as a leading institution of higher learning in the region. The
collaboration and interactivity” [27] Business School (B-school) offers degrees in the traditional
With open source, the development process is as important fields of management along with management information
as the outcome by enabling a higher form of learning that is systems (MIS). The MIS program in the B-school emphasizes
embedded within a set of social interactions and culturally the application of IT to business processes, and engages in
organized activities wherein students become knowledge service and research which serve the IT needs of the Gulf
builders and creators [24]. Likewise, “An academic region.
environment where FOSS is prevalent will encourage staff and The integration of open source into the MIS curriculum was
students to tinker and experiment with, and participate in the the embodiment of a ‘strategic direction’ in the MIS
development of FOSS that may eventually lead to innovative department to go open source on several fronts; research,
solutions.” [3] teaching and administration. However no overarching vision
The inherent ‘entrepreneurial’ characteristic of the FOSS or plan to adopt FOSS existed anywhere else in the university
model should make it more suitable to innovative cultures. although FOSS exists at the bottom of the software stack and
Such is supposed to be the academic environment. Hence, the many FOSS academic initiatives were in existence at the time
quest for open source in education seems ironic given that the the MIS FOSS project started. The department established an
academic community was the breeding ground for software open source interest group for this purpose, made up of
development before programmers knew the meaning of faculty, students and administrators. This followed the
‘proprietary’ [28]. Most code was indeed created within coalescence of research interests around the concept of open
academic-like environments before software engineering computing among many members of the department’s
became an industry in the early sixties [24]. It is only the faculty 5 . In general, the objective of GNU’s open source
aggressive drive for profitability in a booming software group was to serve as a knowledge gateway for open source
market that required the closure of the source code. In this research and practice and its subsequent dissemination in the
light, the discourse on intellectual property rights provided a region.
moral mantle to what is otherwise a dominantly commercial
drive. Once software development was evacuated from the A. ICT Governance at GNU
university and unto the corporation, a divide soon emerged Following a shared governance model of HE, GNU put in
between corporate software developers and its academic users. place nominal structures of collaborative governance to utilize
As commercial software developers have little appreciation for faculty input into its policy-making including about ICT.
HE software use [14], FOSS is amenable to bridge back this Though purely consultative, this model of shared governance,
gap and reconcile academe with its original vocation of articulated around a set of standing and ad-hoc committees,
creation and innovation, hence merging technology-producing could have enabled the active participation of faculty into its
with technology-using [24]. planning and governance processes. Indeed this worked at the
The cultural predisposition for open review and exchange beginning when the university needed faculty resources and
among peers in academe should make for the adoption of input to establish its basic processes of academic governance.
FOSS a natural fit [28] as it helps harness HE’s vast But as the university matured and the corporate model of
innovation capability [12]. Both rely on a model of ‘open governance took over, faculty role was gradually marginalized
science’ research collaboration for their knowledge/software especially in matters that were not perceived to be of their
creation [4, 29]. The current situation of HEIs which have
evolved as ‘technology consumers’ rather than ‘technology 3
Eduforge.org is a virtual collaborative learning and exploratory
innovators,’ contradicts the academic ethos based on academic environment designed for the sharing of ideas, research outcomes, open
freedom and the open dissemination of knowledge and source educational software, and tools within a community of learners and
information [3]. The efforts to adopt open source in academic researchers. They have many links to a range of FOSS initiatives in education.
4
Alias
institutions necessitate its integration in curricula as well as 5
The group coordinator chaired the open source track at a major
within the university environment as a whole, not only as a international conference held in the USA in 2004.
352
direct concern, chief amongst these administrative computing. activities of the university faculty development center where
Moreover academic computing and its corollary the university the potential and first results of the use of the FOSS
ICT strategy were to fall prey to the same tide of corporatism collaborative learning model were presented.
in university governance as the university administration The ‘Open Lab’ was installed in the basement of the B-
created shadow committees mirroring the existing faculty School building with a Linux server running Novell SUSE 7 , an
committees to perform strategic ICT planning. The Apache Web server, and a number of open source applications
‘administrative’ committees were responsible to work on basic including OpenOffice, MySQL and IBM Eclipse suite which
plans and submit them for examination to the university were used by students to fulfill the course requirements of the
administration, and hence to the faculty for consultation. It ‘Special Topics’ course and other courses where open source
did not take a long time for the faculty committees to cease to modules were introduced. However, no detailed pedagogical
function altogether and lose credibility as they were being plans were developed to integrate FOSS in the curriculum as
increasingly marginalized by a university administration that this was left to the discretion of faculty members.
started hammering the motto that ‘faculty belonged into the Following the initially successful, though not generalized,
classroom’ to justify its one-handedness in the management of introduction of open source modules within existing courses,
university affairs. The deterioration of the governance the thinking evolved into considering a full-fledged open
atmosphere in the university reflected at all levels with the source toolset for MIS students first, which would be
disempowerment of faculty trickling down to academic units generalized to business students later on. This toolset would
where Deans and appointed Chairs held absolute power. It is not be limited to enumerating possible solutions for various
worth mentioning that no formal ICT strategy was in place at course requirements but will dig deep into issues of open
the time and that ICT decisions including strategic ones were source-centered curriculum design. In other words, both the
taken haphazardly and without any significant faculty input. toolset and the FOSS collaborative learning model were to be
Hence, GNU and the B-School did not have clearly defined combined in reshaping the MIS program. Hence, a full-
strategies for using software either for administrative or fledged open source infused curriculum both in terms of
academic purposes. Although there had been a clear content and pedagogy was being contemplated. Feasibility
commitment to proprietary software platforms especially for studies illustrating the TCO benefit of using open source
administrative computing (i.e. ORACLE, SCT Banner), solutions rather than proprietary ones were elaborated for the
software used for educational purposes seemed to be driven by restructuring of the existing MIS program as well as the
ad-hoc requests from academic units. The B-School in creation of a new graduate offering in collaboration with other
particular had been toying with both WebCT and Moodle, a academic IT units in the university. This financial component
proprietary and open source LMS respectively. The university was accompanied with an integrated infrastructure proposal to
would later reverse its strategy by imposing Blackboard as a embed an open source-based learning environment in the B-
mandatory solution for course management. GNU open source School.
project was initiated within these governance premises. The general approach was to provide resources to those who
were willing to join the FOSS bandwagon and to make of
B. Open Source Initiatives at GNU
FOSS a dominant reality and a culture of excellence which
The key and initial elements of the project consisted of the everybody was willing to belong to. The project was boosted
implementation of an ‘Open-Lab’ FOSS competence by a university grant and the allocation of laboratory space for
incubator, integrating open source in the MIS curriculum by the open lab. However this support did not derive from a buy-
drawing from a FOSS toolset, hence building open source in into the project at the university or college level or from any
skills to be later deployed in the university environment and FOSS strategy implementation which did not exist hitherto.
subsequently in the market place. These efforts were in line Rather the two main proponents of the project at the time held
with the local HE ministry’s and university’s strategies for sway in university and college decision-making for reasons
training technically skilled graduates that are fully prepared unrelated to the project itself.
for the local marketplace although these strategies were not
explicitly stated in terms of FOSS skills. C. Project Methodology & Outcomes
On the curriculum side, Linux and other FOSS productivity Whilst project outcomes were defined for the MIS
tools were introduced within business foundation courses. department in the beginning, the plan was to roll them over to
The MIS department also offered a single shot (special topics) the rest of the B-School in a second phase and all of GNU at a
course dedicated to FOSS 6 wherein business-related aspects of later stage. The ultimate objective was to make of GNU a
open source were examined alongside a lab component showcase for the successful implementation of open source in
primarily based on software demos and tutorials. Moreover academe before advocating it outside. The deliverables that
the FOSS collaborative model was used to organize student were identified at the start were:
learning. Students were familiarized with virtual collaborative 1) A portal as a central reference point for FOSS initiatives in
environments as vehicles for student-driven knowledge GNU;
generation. Department faculty were invited on a weekly 2) State of the art open computing lab including the open
basis to partake in the collaborative learning experiment and source toolset and assorted training materials;
encouraged to consider implementing it in their own courses.
7
A pedagogical seminar was also held as part of the regular The Linux alliance program was first finalized with SUSE Linux AG
from Germany. In the meantime and before implementation started, Novell
took over SUSE and reneged on the alliance forcing a costly renegotiation of
6
MISXXX--Special Topics in MIS the alliance program.
353
3) Demonstrable prototypes of some components of the toolset met the student CLOSE project. Furthermore, the project
with applications; came short in terms of developing a full-fledged open
4) A lap-top based open source toolset for business students; computing program either at the undergraduate or graduate
5) Integrated program and course development using the levels. It became clear that such programs would not be
toolset and other open source components. In particular, adopted as alternatives to existing curricula. To take root,
two courses were fully deployed over Moodle.org, an changes of a similar extant would have to clear several
open source LMS, to show the viability of open source ‘governance hurdles’ including the all too-conservative
LMS in comparison with commercial ones such as college curriculum committee and the overtly politicized
WebCT or Blackboard. A workshop was also held for university curriculum committee. The absence of specific
GNU faculty, through the faculty development center, to policies for FOSS adoption left potential proposals open to
demo the course implementations on Moodle.org; arbitrary opposition from many parties.
6) Workshops and seminars with the participation of industry Though the project came short of achieving its overtly
as well as academic speakers. An open source research optimistic objectives, it could still be considered successful at
and industry forum was held in the spring of 2005 where this initial stage especially with the gradual infusion of open
the findings of this project were presented along with a source concepts and tools in the MIS foundation courses
survey of open source initiatives in the region and which were required courses for all incoming business
elsewhere. The e-government program of Bahrain which students.
was developed on top of an open computing platform was
E. Project Roll-Back
showcased for the occasion;
7) Establishment of a FOSS user group with the active A series of events coincided not only to stop development
participation of students and faculty all over GNU; and further achievements of the project outcomes but also to
8) Academic alliance programs with FOSS vendors including roll back the initial achievements to an extent that only
training, support, and certification for faculty, IT individual faculty initiatives to teach open source content
instructors, and students alike; within the foundation courses survived. The main reasons for
the downturn were the following:
An offshoot of the FOSS project consisted of the development
of a portal for collaborative student learning (CLOSE) 8 . The 1) The MIS Department Head who championed the open
project was designed by students as part of the special topics source strategy within the B-School moved to a different
course and won a regional e-biz award. Hence the overall institution in a different country and continent altogether.
FOSS project was initially quite successful but started running It became clear soon after that the open source project lost
into major problems at a later stage. Its gradual demise is its main sponsor who sat in the B-School’s Council of
discussed below: Chairs, hence defended the project’s claim for college
resources. The decision to reallocate the open lab to the
D. Initial Success graduate program was uncontested as it was taken by a
During the first two years of planning and implementation, body with no representation from the open source project
the project went full steam and many outcomes were achieved, group. This decision was officially motivated by the
some even ahead of time. The project was at first a major swelling of student numbers in graduate programs, which
success in its open lab component. For a fraction of the real were bringing in substantial revenue to the university, and
cost, the department of MIS at the B-School established a state the ensuing need to provide graduate students with better
of the art lab that helped GNU support open source education and even privileged access to learning resources.
and training. The hope was that in a second phase, the open 2) The establishment of a new testing center in the university
lab services would be offered outside of the B-School. brought along a renewed requirement to teach computing
Regarding the research and pedagogical aspects of the material that led to the ICDL certification; which is
project, several research works were generated and published primarily based on proprietary concepts and software 9 .
in academic and professional journals as well as international Yet again an undeclared ICT strategy was invoked to
conferences. However, a major collaborative piece on bring back proprietary tools into the curriculum. It is
“implementing FOSS as pedagogy” which was implemented worth mentioning that no consultation of any kind took
experimentally in the special topics course described earlier place to establish the testing center and subsequently to
never saw the light because the project quickly lost steam after require that students be trained in the proprietary skills
resources started being withdrawn from the project following that were necessary to acquire the related certification.
the departure of a key collaborator who had marshaled most of This led to a redesign of the basic introductory IT courses,
the initial resources. Similarly, a major outcome which was generally at the expense of open source content.
the development of an open source education and training 3) Within coordinated courses, some instructors had
portal, planned to be made accessible to GNU students also throughout expressed dissatisfaction with the inclusion of
aborted along with a proposal for a permanent elective course FOSS in the IT skill set taught to students. Their
on open computing. An obscure policy on university Web argument was twofold; first by committing students to
publishing was invoked by the coordinator of academic
computing to block any such portal going live. A similar fate
9
Despite its claims of ‘vendor neutrality’ in its descriptions of the skill
sets, training materials and test centers are almost predominantly Microsoft
8
CLOSE: Collaborative Learning for Open Source Education oriented.
354
FOSS skills, we were allegedly decreasing their vendor lacked a necessary strategic vision to forego
employability given the dominance of the proprietary in profits in the short term for bigger gains in the long term.
the corporate world, and second; there is a lack of good
didactical material for the delivery of open source At present, open source education in GNU is limited to few
content. However, it was always clear that some were software demos and a single class lecture within the
never really impressed with the open source phenomenon foundation MIS course. Some students and faculty elect to
and even found it lacking credibility and/or having little use popular FOSS tools in system design projects. Web
chance to forge ahead in the long term. Therefore, they Services and PhP for instance have become central to the
jumped on the first occasion of a shift in the balance of development of e-business solutions. However, these remain
power against the open source tide to help ‘shoot it isolated initiatives and not ones that are central to the delivery
down.’ of MIS education. On the pedagogical front, there are no
4) The university decided to adopt a standard and proprietary remnants of the FOSS model of collaborative learning.
platform for LMS, one based on Blackboard. The Ironically, the massive accreditation literature that the
decision was made based on a proposal by the university generates on a continuous basis keeps hammering
administrative IT committee. A last minute push by the the theme of student-centered learning without either planning
faculty IT committee gave a respite to Moodle courses for for it or creating the right governance system for it to emerge.
three years with an injunction to move to Blackboard at Worse yet, whenever a credible experiment came into
the end of that period. Resources were nonetheless existence, it was quickly neutralized.
withdrawn from Moodle, including server space, bringing
into a crisis the fifty or so courses that were hosted on the
Moodle server in the B-School and burdening the few VI. PROBLEMS OF ICT GOVERNANCE IN DEVELOPING
faculty volunteers who maintained the Moodle server. COUNTRIES
The three year period has just come to an end and the plug It is useful to relate the above failures to the ‘developing
has been pulled forever on the Moodle experiment despite country’ context of the project and especially its governance
the recurrent problems with Blackboard in contrast to the aspects:
greater stability of Moodle which is fast becoming the de
facto standard for virtual learning environments [30] 10 . 1) Being in a developing country context, the university or the
Whenever faced with such argument, the academic college lacked any form of significant strategic planning
computing coordinator and his staff invoked intellectual for IT; be it for academic or administrative computing.
property issues and other legal risks related to the use of This led to haphazard decision-making for IT adoption
Moodle. Yet again a major commitment of ICT resources and strategic initiatives. This applies to the inception of
for the long term was done in the absence of an explicit the FOSS project itself which was initiated within a
strategy and within the framework of a weak governance policy vacuum. Indeed if a strategic IT plan including
structure. provisions for FOSS in university computing and
5) Proprietary vendors became aware of the threat of open curriculum development were to exist, GNU FOSS
source especially when championed by competent and project would have survived. Furthermore, in the absence
knowledgeable insiders. As a result, they assailed GNU of clear plans and institutional mechanisms to ensure their
with attractive offers of academic alliance programs implementation, the university governance structure
which were very enticing, especially in view of would inevitably favor ad-hoc plans and projects of those
augmenting university graduates marketability. Since the in power, for as long as they are in power. In sum, most
unfolding of the open source project, GNU has entered governance structures in developing country contexts are
into strategic alliances with major proprietary vendors channels for autocratic leaders to extend their rule to
covering roughly every computing need of students. reach every corner of the organization. In other words,
Only the laptop initiative developed under the umbrella of powerful individuals supersede governance structures and
IBM ThinkPad University survived until last year yet with not the other way around.
a gradual marginalization of its open source component. 2) The main sponsors of the FOSS project failed to forge a
It is worth mentioning also that an academic alliance with common vision on open source across the B-School and
another open source provider which would have given a university-wide. The idea was mainly pushed through the
critical boost to the open source project in GNU fell B-School’s IT committee and was since the start
through because of its lack of competitiveness compared misconceived as being yet another IT project. Resources
to proprietary ones. This was used opportunistically to for the project were initially marshaled thanks to the
undermine the competitive claims of open source. The negotiation skills of the MIS department Chair rather than
alliance was sought in this particular case not because of through a strong discourse of legitimacy that should have
the need to access software that was anyways readily been aimed at the bulk of faculty in the college and
available in the public domain but mainly to provide a beyond. Better yet, such a discourse of legitimacy should
mantle of legitimacy to the open source project. The have imprinted governance structures themselves by
engendering ICT plans advocating FOSS as a viable
alternative if not a preferred course of action for academic
10
It is operating in nearly 50,000 sites including universities and other computing and IT curriculum development. Vigorous and
types of educational institutions (cf. http://moodle.org/stats/)
355
active leadership supporting a FOSS agenda at the higher VII. LESSONS FOR DEVELOPING COUNTRIES
level of university leadership is also a must [11]. When Though the above developments might sound fatalistic
faced with a roll-back in resources for the project, the and insinuate the uselessness of progressive efforts to
FOSS group could not invoke any such policy and had to adopt FOSS into a developing country HE context, the
yield to the otherwise much more ‘legitimate’ argument author remains adamant that such efforts could be
of learning requirements of graduate students for instance. successful if they adopt a governance-aware strategy.
In the absence of institutional mechanisms that would The problems encountered during the implementation of a
guarantee a progressive unfolding of the FOSS project, FOSS academic solution in GNU can be used to outline
the intrinsically anti-FOSS governance structure could such a successful strategy:
have been tamed only through a critical mass of pro-
FOSS power players at all levels of the university - Start at the governance level: It is more important to
decision-making structure. create sustainable governance structures for ICT
3) The FOSS group which started with a nucleus of faculty planning and implementation than to rush into
members from the MIS department soon expanded to specific projects within ill-defined structures. In the
include other members of the faculty especially from the case at hand, the two project champions should have
College of Arts and Sciences, including a sizeable number leveraged their ‘temporary’ power to create FOSS
from its Intensive English Program. It is interesting to conducive structures within the B-School first and at
note that most members were from the social sciences and a wider university level subsequently.
not from engineering or computer science as would be - Stress Open Learning: The FOSS model for teaching
expected. One possible explanation is the ‘militancy’ side was advocated more for its collaborative learning
of the project agenda which advocates a vendor-free and potential than its TCO and other technical
open environment for computing where the digital divide advantages. However beyond the conventional
is bridged and innovation is taken out of the corporate discourse opposing FOSS to the proprietary, a
mould. It is unlikely for similar ‘militant’ agendas to collaborative model of learning is expected to stir
develop in the overtly rationalist environment of stiff resistance amongst the conservative academic
computer and engineering sciences. However, it is establishment, which is well entrenched within the
precisely this perception that open source was being master delivery model. Yet again, the changes
pushed by ‘militant zealots’ in different quarters of the brought about the GNU project seem too radical and
university that probably accelerated its demise. Militancy too quick. A FOSS-driven model should rather be
is essentially antithetical to formal governance. The infused within the general discourse about change
outcome could have been more positive if this militancy towards student-centered learning. As accreditation
agenda was diluted within more acceptable discourses of and quality assurance efforts accelerate, FOSS should
curriculum change and through existing channels of be mirrored as central to the changes sought.
academic governance. The FOSS group itself, by its mere - Diffuse the battle with administrators: FOSS adoption
existence as a para-academic structure, posed a challenge should not be approached from a software adoption
to official governance. The FOSS group should have perspective but rather from the standpoint of
concentrated on lobbying for its own acceptation as a collaborative learning and reconciling academe with
legitimate player within the university governance the spirit of openness. There is a lot of confusion out
structure instead of situating itself in opposition to it. there with regards to the merits of FOSS and its long
4) Proprietary vendors spread fear, uncertainty, and doubt term viability. However, nobody can contest its open
(FUD, [11]) intimidating Academic Computing into learning capabilities. Administrators can oppose it
discontinuing the open source project. With the on technical/financial terms but will be reluctant to
consolidation of key proprietary offerings in the HE do so if it is presented as the core of a new learning
software market (exp. Blackboard & WebCT; Oracle & model.
PeopleSoft), proprietary vendors leverage is increasing - Build university-wide capacity and awareness: FOSS
[11]. This was most visible with the Moodle vs. initiatives should not necessarily start with MIS and
Blackboard battle. This is again a governance issue at computer science departments by insisting on
stake that materialized through a failure to uphold the programming and similar courses. Emphasis should
necessary control requirements in IT decision-making. It rather be on the use and integration of collaborative
has to do yet again with the developing country context of learning platforms across the curriculum. In the
the university. With the demise of open source, GNU project, a large number of faculty from the
proprietary vendors rushed to forge strategic deals that Intensive English Program came on board through
would ensure their perennial interests. It goes without the use of Moodle, the open source LMS.
saying that proprietary vendors wield a lot of influence - FOSS as Pedagogy: A critical problem that faced
over university procurement decisions in developing faculty adopting the FOSS toolset in the B-School’s
countries, precisely because of the lack of transparency in MIS department was the lack of a collaborative
governance processes [1]. learning pedagogy. Emphasis should be put as much
on developing the rudiments of such pedagogy as on
the toolset itself.
356
[19] S.K. Sowe, L. Angelis, I. Stamelos, Y. Manolopou-los. “Using

Repository of Repositories (RoRs) to Study the Growth of F/OSS
In essence, all of the above is feasible within a functional Projects: A Meta- Analysis Research Approach,” Third International
HE governance premised on faculty participation and Conference on Open Source Systems. Limerick, Ireland, 2007, pp: 147-
their supreme expert authority on matters of teaching and 160.
learning. HEIs in developing countries will not achieve [20] J. Kenny, “ Moodle nudges forward: Can any other learning platform
challenge Moodle’s dominance?”, in Guardian Education Supplement,
their development function for as long as they remain an January 8th 2008, p.10.
addendum to the state bureaucratic system. On the other [21] J. Cox, “Universities build Open Source Applications”, in Network
hand, private universities in the developing world have World, March 31st, 2008, 25, 13.
not amassed a critical mass of resources, either human or [22] C.N. Davidson, “We can’t ignore the influence of digital technologies”
The Chronicle Review, March 23rd, 2007, B20.
financial, to provide any significant alternative. The ICT [23] A.L. Foster, “Providing online textbooks to the developing world”
for development discourse in HE should not be centered Chronicle of Higher Education, 54, November 16, 2007.
on infrastructural requirements, which are still important [24] M.A. Peters, (2008) Globalization and the virtues of openness in higher
to fulfill, but highlight the central issue of HE governance education,” Global e-Journal. Available: http://global-ejournal.org
/2008/08/29/globalization-and-the-virtues-of-openness-in-higher-
based on autonomy, faculty governance, and ethics and education/
social responsibility. [25] A.H. Moore, “Lens on the Future: Open-Source learning”, in
EDUCAUSE Review, Boulder, Sep/Oct 2002, Iss. 2, Vol. 37, p.42.
REFERENCES [26] J. Marshall. “Negri, Hardt, distributed governance and open source
software,” Journal of Multidisciplinary International Studies, Vol. 3,
[1] R. J. Abel, “ Best Practices in Open Source in Higher Education Study No. 1, January 2006, pp. 1-25.
The State of Open Source Software.” March, Lake Mary, FL, the [27] R.T. Watson, M. Boudreau, P. York, M. Greiner, D. Wynn, “Opening
Alliance for Higher Education Competitiveness, Inc., 2006, pp. 1-47. the Classroom”, in Journal of Information Systems Education, Spring,
[2] P. Masson, “Barriers to the adoption of open source: Personal and Vol. 19, Iss. 1, 2008, pp. 75-86.
professional observations” World Campus, PennState University, 2007. [28] N. Bezroukov, “Open Source Software Development as a Special Type
Available: http://blog.worldcampus.psu.edu/index.php/2007/04/17/ of Academic Research” (Critique of Vulgar Raymondism),
[3] T. W. Tong. “Free/open source software: Education primer” FirstMonday, 1999, vol. 4(10).
International Open Source Network, UNDP, 2004. Available:
http://www.iosn.net/ education/foss-education-primer/fossPrimer-
Education.pdf
[4] P.A. David, J.S. Shapiro. “Higher education institutions and the global
role of free/libre and open source software.” Report on findings from
the FLOSSWORLD survey of developing and transition economies, 2007
Available: http://www.oii.ox.ac.uk/research/FLOSS_HEI_Report.pdf
[5] United Nations University. “Free Software in Developing Countries
Vital to Future Prosperity and Good Governance: UNU Technology
Experts.” Media Advisory, 2006 Available: http://www.unu.edu/media/
archives/2006/files/mre-iist-3-06.pdf
[6] V. van Reijswoud, A. de Jager. « Free and Open Source Software for
Development,” Polimetrica, Monza: Italy, 2008.
[7] K. Gakio “African tertiary institutions connectivity survey (ATICS),”
2006 Report, International Research Development Center, Canada.
Available: http://www.aau.org/renu/docs/ATICS2006.pdf.
[8] D. Glance, J. Kerr, A. Reid, “Factors affecting the use of open source
software in tertiary education institutions”, First Monday,Volume
9,No. 2, Feb 2004.
[9] J.A. Lewis “Government open source policies” Center for International
Strategic Studies, July 2008. Available: http://www.csis.org/media/csis/
pubs/0807218_government_opensource_policies.pdf
[10] S. Bacon, T. Dillon . “ The Potential of open source approaches for
education,” Futurelab, 2006, Available: http://www.futurelab.org.uk/
resources/documents/opening_education/Open_Source_report.pdf
[11] P.N. Courant, R.J. Griffiths, “Software and collaboration in higher
education: a study of open source software,” Available:
http://www.ithaka.org/strategic-services/oss/OOSS_Report_FINAL.pdf
[12] D.A. Wheeler. “Why Open Source Software / Free Software (OSS/FS,
FLOSS, or FOSS)? Look at the Numbers!” 2007. Available:
http://www.dwheeler.com/oss_fs_why.html.
[13] S. Sahraoui. “Implementation of a New Degree Structure in the Context
of Higher Education Reform in Tunisia.” Unpublished Master
dissertation, University of Oxford, 2007.
[14] J. Maguire, “Open source on the Brink”, in Enterprise Linux IT pp.1-2,
Sep 11 2003, Available http://www.Linuxenterprisenews.com/
perl/printer/22278/
[15] G. James, “Electronic Design Automation”, in Electronic Business,
Dec, 31 2005, 12, p.24.
[16] S.W. Van Rooij, “Perceptions of Open Source versus Commercial
Software: is Higher Education Still on the Fence?”, in Journal of
Research on Technology in Education, 39(4), 2007, 433-453.
[17] S. Yegulpap, “Open Source: When things fall apart”, in
InformationWeek, Manhasset, May 2008, issue 1184, p. 20.
[18] G. Hein, “Open Source Software: Risks and Rewards” 2004 Available
http://net.educause.edu/ir/library/pdf/ECR0405.pdf
357
ICTD State of the Union: Where have we

reached and where are we headed
Rabin Patra Joyojeet Pal Sergiu Nedevschi
rkpatra@cs.berkeley.edu joyojeet@berkeley.edu sergiu@cs.berkeley.edu
Department of EECS CIS, School of Information International Computer
University of California University of Washington, Science Institute,
Berkeley, CA USA Seattle, WA USA Berkeley, CA USA
Abstract— In this study we examine the history and growth of often were transnational expatriate technologists from the
ICTD since the 1990s. We underline the trends defining this developing world.
research field and examine the progress in research areas that By the mid 1990s, there was significant buzz on the role of
have come to dominate discussion in ICTD through a thorough the information technology boom in dramatic global change
literature review of the last decade of ICTD work. In order to [4, 5], and a first generation of ICTD literature discussing the
answer questions pertaining to the rigor, impact and significance
of ICTD, and to compare the expectations and perceived
specific nature of technology projects in development started a
achievements with respect to different development goals, we worldwide interest in the field [6-10]. Around this period,
interview 50 expert ICTD researchers and practitioners. We there was a dramatic rise in the number of ‘ICTD Projects’ –
analyze these results to understand stakeholders' opinions on the i.e., technology projects specifically aimed at creating
past performance of ICTD, both as an academic field and as an developmental outcomes for their recipients. This trend was
area of development practice, and identify defining ideas on the partly driven by a slew of research papers and policy
potential directions for the future of ICTD. This study is work in documents within international agencies [11-13]. As a result,
progress and we have continuing research in this area, the subset by the turn of the century, there were thousands of telecenters
presented here is rigorous and ready for wider discussion. 1 around the world, funded through various sources [14].
Index Terms— ICTD, developing world
By the early 2000s, engineers became interested in ICTD,
not just as a development agenda, but as an area that posed
I. INTRODUCTION
interesting research problems in their own fields of work [15-
T HE Scandinavians started it. It may be a lost piece of
trivia somewhere, but ICTD as we know it began with the
community computer rooms in Scandinavian villages.
19]. This followed the establishment of numerous academic
venues, both as part of existing established forums, and
independent venues specifically for the study and discussion
Starting with the first village (Fjaltring in Denmark), there
of ICTD.
were several small tele-cottages throughout Sweden, Denmark
and Norway by 1985, and it wasn’t until a few years later that
Following this early foundational scholarly and practitioner
Sri Lanka became the first developing country to get its
community technology access center [1]. work, more interesting follow-up work has been done on
emergent areas of ICTD including digital inequality [20-23],
The term ICTD started appearing in academia and industry on technology and sectoral development and macroeconomic
with minor alphabetical variances such as ICT4D, ICT4B change [24-26], on technology and urban change [10, 27], and
(billions), IT4D and so on by the mid 1990s. The years on the ‘potential’ of technology [12, 28, 29]. By the mid
preceding this were formative in the growth of interest in the 2000s, there was introspective work looking at the
subject around the world. The opening up of Eastern performance of the ICTD projects started in the 1990s in India
European economies coincided with phenomenal periods of and elsewhere, and also at the causes and outcomes of the
growth in China and India. The technology boom in the interest in technology and development [30-33]. More
United States featured a large pool of engineers from various recently, scholars have looked back at the growth of ICTD
parts of the developing world. The spillover economic effects through the years by tracing the various stages that ICTD has
of this international technology workforce ranged from moved through in this period [34].
remittances to home countries to the creation of new small and
medium-scale engineering companies [2, 3]. The public Our survey and research here hope to build on such work.
discourse of technology as being an engine of macroeconomic As stakeholders in an emerging field of study that is still in the
growth grew in strength rapidly as the early impacts of process of defining itself and carving a niche, it is critical that
globalization in the tech industry manifested themselves first
we look back at various points in our progress and review the
in the West, and soon thereafter in many parts of the
direction of our growth. Here, we present a fairly
developing world. The international faces of this phenomenon
comprehensive survey of what has happened in the field, and
use the opinion of significant voices from the field to support
1 the literature review, in the hope of shedding some light on the
This material is based upon work supported in part by the National
Science Foundation under Grant No. 0326582. general direction of ICTD work and its validity. We examine
358
the progress in dominant ICTD research areas, we identify the B. Sampling and Recruitment
areas that received more attention in previous research, and We had a selective sample of respondents from a few major
we compare the expectations with the perceived achievements ICTD-related online forums, in addition to other professors
in each of these areas. We also examine questions concerning and researchers that we contacted to fill out the survey.
the rigor, impact and significance of ICTD, and its
appropriateness in serving various development goals.
C. Sample description
II. APPROACH Our total sample included 50 respondents, who were asked
to select their areas of affiliation. We observe a fairly even
We begin our study by categorizing some broad domains of distribution of these areas, with a slight skew towards
ICTD research and practice, and by performing a high-level education and infrastructure. Overall, a higher number of
overview of the progress in each of these domains. We follow researchers than practitioners were represented in this survey.
this by presenting the results of our interviews with ICTD
experts on their opinions about research and practical work in TABLE I
RESPONDENT DESCRIPTION BY AREAS OF AFFILIATION
each of these areas.
Domain Area Research Practice
In addition to these discussions, we consider a few
recurring themes in our early conversations with interviewed Healthcare 11 3
researchers. The first emergent theme was the question of Education 17 7
‘Hope v/s Hype’ in the context of ICTD. This issue, has been Business 8 4
explored in the past [35-37], but continues to remains a Agriculture 8 3
niggling issue for most practitioners and researchers alike, Comm. & network infrastructure 16 4
who face up to it in their own work. A related theme that User interface design 11 0
emerged in our interviews has been that the role of various Governance 10 1
stakeholders. As ICTD moves from being an experimental Other 9 3
area towards mainstream development research and practice, it
is inevitable that we face hard questions about what the role of
government should be in funding ICTD projects, what kind of The respondents’ past academic disciplines (regardless of
ICTD projects are more likely to succeed and so on. Posing whether they are presently scholars or practitioners) were
these questions to our interviewees, we find an interesting fairly evenly distributed between science/engineering, social
variance of opinion on this issue. Finally we explore issues sciences, and hybrid areas such as design and information
about the interdisciplinary nature of ICTD – how far does it studies.
exist, and has it helped or impeded ICTD.
TABLE II
RESPONDENT BY AREAS OF PRIMARY ACADEMIC SPECIALIZATION
III. METHODOLOGY
The research consists of two parts – first, an extensive Academic Discipline Number of
literature review of the various projects in ICTD over the past Respondents
decade, and second, a survey of 50 researchers and Engineering 16
practitioners in the ICTD space. The two were done Information Studies 7
simultaneously – i.e. while we had some idea of what the Education 3
broad literature review would bring up, we had no idea of how Political Science 4
similar it would be to the results from our interviews. Design 2
City planning 2
Sociology 2
A. Instrument Design
Humanities 2
The interview schedule was created over three iterations on Media 2
the basis of feedback about the interview questions from International business and relations 2
researchers in our direct contact. The third version of the ICTD 2
questionnaire was an online hybrid with a mix of open-ended Development Studies 1
and close-ended questions, of which the former were created Anthropology 1
based on categories arrived upon in the first two iterations of Agriculture 1
the instrument. The final interview was anonymized and took Commerce 1
anywhere between 15-75 minutes to complete, depending on Physics 1
the interviewee. Environmental Science 1
Total 50
359
D. Caveats IV. FINDINGS

The findings are from a fairly small sample because we
wanted experts with several years in the field to comment on A. Healthcare
ICTD issues rather than a broad-based survey of a large The interest in using technology to widen access to
number of participants. Over half of our sample is represented healthcare and sanitation pre-dates the ICTD age. While
by people who have been active in ICTD for over 5 years and earlier initiatives focused on increasing access to specialists
are thus familiar with the landscape and the changes in ICTD and getting basic diagnosis to remote regions within the
and are qualified to comment on past and future issues. That developed world [38], recently, the UN Millennium
said, such a small and selectively sampled survey is prone to development goals (MDGs) have renewed the focus on ICTD
bias. This study is a work in progress, and while the results [39, 40]. There are four broad areas where active ICTD
presented here are rigorous and sturdy enough for discussion, projects tackle healthcare challenges. The first has been
there is always scope for increasing the sample size, and telemedicine [41], and here the focus has been on using long-
finding out through an iterative process what areas might need distance communication links to expand access to remote rural
better coverage in the questions. areas where there are no doctors. For-profit initiatives have
been deployed for specific kinds of diagnoses that can be
Finally, one other important caveat here is the regional
reasonably well managed remotely, such as ophthalmology
concentration of respondents. We find a very heavy skew
and dermatology [42]. 2 The second area of ICTD interest in
towards respondents currently based in the US and India, but
healthcare is information gathering especially for
this is explained by the fact that US is somewhat of a hub for
researchers and practitioners, and that a large fraction of the epidemiological research. There have been a number of
ICTD activity is located in India. This also seems to indicate projects focusing either on general population health
that academics from developing countries are going abroad to surveying, patient health monitoring, or healthcare aid impact
study ICTD in American or European Universities. assessment [43, 44]. A third focus of ICTD was driven by
expatriate doctor communities from developing countries who
TABLE III were interested in contributing back to their home countries
LOCATION OF RESPONDENT AND REGIONAL FOCUS through social initiatives. Some ICTD projects have used web
Country Physical Location where 2.0 technologies to connect doctors in the developing world
location of respondent with counterparts and experts in the developed world for
Respondent primarily active remote consulting on specialty care [45]. The fourth area of
in ICTD work ICTD in healthcare has concentrated on the use of technology
USA 25 4 in building low-cost medical diagnostic devices (such as
India 10 20 ultrasound, X-ray machines) and sensors. 3
Malaysia 2 1
Philippines 2 2
We asked respondents in the study to name the areas of
Barbados 1 1
ICTD and healthcare that they felt were of significance
Brazil 1 1
looking into the future. The top 5 responses, from a subset of
Ghana 0 2
Botswana 0 1 36 who commented on healthcare, are presented in the
Chile 0 1 following table:
Colombia 0 1 TABLE IV
Canada 1 0 AREAS OF FUTURE IMPORTANCE IN HEALTHCARE
Greece 1 0 Top 5 areas ranked by respondents Percent
Netherlands 1 0 Respondents
Nigeria 1 0 Medical records 57.8
South Africa 1 1 Supply-chain management 50.0
Spain 1 0 Tele-diagnosis and treatment 44.7
Sweden 1 0 Collection of epidemiological data 44.7
Switzerland 1 0 User interfaces 28.9
Uganda 1 0
Macedonia 0 1
Nigeria 0 1
No Specific Region -- 12 From the interview results, it stands out that supply chain
Total 50 50 management in healthcare and the design of user interfaces for
2
http://www.clickdiagnostics.com
3
Several of the Microsoft Research Digital Inclusion grantees in 2005
were doing some work in healthcare, at least two were specifically designing
low-cost diagnosis devices for developing countries.
http://research.microsoft.com/ur/us/fundingopps/RFPs/DigitalInclusion_2005
_RFP_Awards.aspx
360
healthcare work were both seen as very important areas for agriculture closely matched past work on agriculture within
future effort, though both of these are practically absent in our ICTD, and most respondents were in fairly good agreement
literature review of past work. Medical records, collection of about the perceived progress in these areas. What is worth
epidemiological data, and tele-diagnosis, all areas with noticing here is that ‘access to expert information’ which had
important past work were also viewed as very relevant by been one of the most important areas of concern in the early
respondents. In addition to these, a number of respondents days of ICTD takes a lower position, whereas market access
mentioned health education and emergency assistance for and information, despite the apparent challenges in making
locating medical facilities are potentially strong areas for the market information usable for actual transactions, remains a
future. There was surprisingly very little mention of low-cost top concern for ICTD experts.
diagnostic devices.
C. Education
B. Agriculture
Governments, philanthropic efforts, and private
From the earliest ICTD implementations in developing
corporations have all found the ICT in education space to be
countries, the apparent incongruity of computers in the rural
attractive right from the earliest days of ICTD
hinterland has been a key concern for researchers. We see
implementations. Many of the early ‘low-cost computing’
therefore that many early projects tried to increase the
projects such as the Hewlett Packard’s 4-4-1 computer, 4 and
relevance of computing in rural areas by providing
the Computador Popular were aimed at classroom use [58].
information on agricultural practices, market prices and
Computer aided learning projects have been among the largest
government schemes through telecenters [46, 47]. However,
and most prominent project categories within the ICTD space
persuading rural adults to be regular telecenter users has been
in the last decade, and today, it is arguably true that more poor
a challenge. As a result, besides agriculture, many telecenters
households have “access to technology” through a child in
have also focused on other services such as online assistance
school than through a kiosk or any other means of computer
to help small enterprises in villages, for instance, artisans that
access [59]. Impact of computers on learning is an area of
sell handicrafts on the Internet [48]. There has also been work
ICTD that has seen interest from mainstream economics as
on turning telecenters into points of purchase in supply chains
well [60]. Interface designers have been actively looking at
for rural produce [49].
innovative learning methods with computers and a lot of
interesting work has been done in shared computing [61],
Other kinds of ICTD projects in the rural agricultural
cellphones for game-based learning [62], and networked
market include the use of cellphones and PDAs in organic
systems for contextual classroom video instruction [63].
certifications for farmers [50], sensor networks in helping
In higher education, distance learning in developing regions
water management for rural areas [51], livestock management
has been an early area of interest within ICTD [64, 65]. While
[52], price information [53] and farming extension and
information sharing [54]. With the growing interest in micro- a significant chunk of the work with children has been India-
entrepreneurship, the use of technology to increase access to centric, we find a lot of work in higher education elsewhere,
microfinance has been very prominent in ICTD research. including the use of SMS for question/answers in university
Many projects have attempted to implement technological lectures [66] in Bangladesh, cellphones for interactive
solutions to assist on-the-ground rural microfinance operations learning [67] in the Philippines, and a one-mouse-per-desk
[55, 56]. Work in social sciences has examined whether the approach for lectures [68] in China.
ability of technological interventions in increasing the
efficiency of rural microfinance operations depends largely on TABLE VI
AREAS OF FUTURE IMPORTANCE IN EDUCATION
the scale of operations of the organization adopting the
Percent
technology intervention [57]. A survey of respondents on their Top 5 areas ranked by respondents
Respondents
opinion about the key areas in agriculture revealed the
following: Remote learning 31.4
Educational games 25.7
TABLE V Low cost computing 22.8
AREAS OF FUTURE IMPORTANCE IN AGRICULTURE Life-long learning 17.1
Online content 14.2
Top 5 areas ranked by respondents Percent
Respondents
Best practices and information sharing 56.7
There was far lesser agreement in the education space on
Market access and information 51.3
Supply chain management 45.9 the key areas of research among ICTD experts. Two
Sensors 21.6 interesting factors emerged in our discussions. First, that low-
Access to expert information 18.9 cost computing, which has attracted a lot of interest in recent
years, is seen comparatively as a less important area of
research than working with existing devices. This mirrors well
Unlike in the case of healthcare, the responses for
4
http://www.hp.com/e-inclusion/en/project/441_brochure.pdf
361
the general frame of discussion within existing ICTD The results in this category could possibly be seen as
literature, which has also largely taken the same position. The underlining the distinction between the interests of researchers
second important issue is that remote learning, in which we and those of the industry. Our sample, despite being open to
have seen some promising projects start in the last few years, researchers and industry, had a much higher research
is expected to continue to be an area of focus in the coming component. As a result, we see that low-cost infrastructure,
years. undeniably a major area of concern for ICTD, trumps the low
cost phones and devices, the area that shows more activity in
our literature review. However, this could also be seen as
D. Communications and Infrastructure indicative of what experts feel about the progress of low-cost
Among the earliest areas to get engineering scholars devices in the real world market. Although we list a large
interested in ICTD was the challenge of building and number of low-cost device projects here, only a small fraction
deploying novel high quality communication network of them have actually made any significant impact in the
solutions to connect low income regions with poor legacy market, and even some of the largest players abruptly left the
infrastructure. Studies suggested that communications market. It is also important that most of the research on low-
backbones could be the first form of infrastructure that would cost devices is on computers, and little or no work exists on
be affordable to deploy in the developing world [15] and that making low-cost mobile handsets, though a number of
this could in turn be a major driver for wider socio-economic researchers indicate that this is likely to change looking ahead.
growth [28]. Several projects looked at various inexpensive One potential reason for why work on cellular infrastructure
connectivity solutions both over the short and the long-haul has remained largely unaddressed in ICTD came from the
[69], and the use of low-cost WiFi for long-distance respondents. Several interviewees stated that since worldwide
connectivity became an important ICTD pursuit, with groups cellular coverage has grown so dramatically over the last few
from several parts of the world looking at the scope of this years, cellular coverage is seen as more of an industry
technology in real world deployments [16, 70-73]. Today, problem rather than an academic research area. Furthermore,
long distance connectivity using WiFi has made significant the closed protocols of dominant cellular technologies are not
progress showing test throughputs of upto 6 Mbps at almost open to experimentation, and thus harder for researchers to
400 km distances 5 and enabling functional field deployments work with.
for telemedicine applications [42, 74]. Research has also
focused on creating appropriate front-end infrastructure to E. Governance
work on low power and unreliable, or intermittently connected
networks [75, 76]. E-Governance was a major buzz area in the early days of the
Finally, the ‘low-cost computer’ has been one of the most ICTD, especially given a general subscription to the view that
important areas of work in technology infrastructure in ICTD, a lot of the problems of development are because of bad
and probably the area that has generated the greatest amount governance [77]. This was especially so in places like India,
of industry interest. We found over 50-projects in the past 10 which had seen sluggish growth in several sectors that were
years in this space including many such as Fulong Mini-PC 6, tightly controlled by the state, but in sectors where the state
and E-DUC 7, Sirius 8, and SofComp 9 that originated from kept itself off, growth came speedily and significantly [26].
research labs in the developing world alongside work by big Consequently, several projects emerged throughout the late
corporations such as Intel’s Classmate 10, and AMD with its 1990s to increase the use of technology in the processes of
PIC 11. governance, first in the developed world and eventually in the
TABLE VII developing world [78]. These included projects that migrated
AREAS OF FUTURE IMPORTANCE IN COMMUNICATIONS a number of state functions online as a way to reduce the
manpower cost of governmental transactions [79], enable e-
Top 5 areas ranked by respondents Percent payments [80] and e-voting [81], and presumably also reduce
Respondents
the scope for corruption. A number of projects also
Wireless/Low cost infrastructure 41.7 fundamentally changed certain state functions such as land
Low cost phones and devices 30.6 record maintenance [82].
Mobile phones & phone coverage 22.2
Community radio & TV 22.2 The status on ICTD in governance thus far has been fairly
VoIP 19.4 mixed. Although on one hand some studies have shown
generally positive feedback [83, 84], others have raised
serious questions about their impacts on the disempowered
[85, 86, 87].
TABLE VIII
AREAS OF FUTURE IMPORTANCE IN GOVERNANCE
5
http://radar.oreilly.com/2007/06/wifi-record-range-now-382-km.html Top 5 areas ranked by respondents Percent
6
http://www.lemote.com/ Respondents
7
http://www.e-duc.com/
8
http://www.fiveriverstech.com/sirius.htm Sharing of public information 47.3
9
http://www.ncoretech.com/mobilis/index.html Digitization of records (land, tax) 31.5
10
http://www.classmatepc.com/ Improved transparency, corruption reduction 26.3
11
http://50x15.amd.com/en-us/
362
E-payment and online retailing 23.6 limitations, voice recognition cannot be an effective
E-voting 7.8 replacement for other technologies – so while speech can
work well for small vocabularies and limited UIs, these tasks
can also be done equally effectively by inexpensive visual
The responses from our experts on governance again interfaces or numerical keypads. The applications where
mirrored fairly closely what we found in the literature review speech could really be helpful, for example in complex
– public information, digitized land and tax records were transactions, is also where the technical challenges are
viewed to be the most relevant research areas. There were also greatest.
a small number of respondents who felt that e-voting and GIS-
related work, both of which have been looked at only to a G. ICTD Stakeholders
limited extent, may be among the major areas for ICTD and One of the interesting questions around ICTD from its earliest
governance in the near future. days has been the role of various stakeholders in supporting
such projects. There were a large number of ICTD projects in
F. Design the 1990s, a period marked by the increasing privatization of
The User Interface (UI) design community has been among development funding [97] and a massive boost in Corporate
the major technical drivers of ICTD research. The basic Social Responsibility (CSR) funding, which in turn diverted
assumption with UI work in ICTD has been that most philanthropic funds from companies interested in human
technologies prevalent in markets today are designed with development towards ICTD projects [98, 99]. Because ICTD
high income, educated users in mind, and that re-designing projects are often at the crossroads of business, technology
computers and peoples’ interactions with computers could and human development, many initiatives have been
make technology more accessible for the poor and illiterate supported by large private corporations such as HP, Intel, or
[88]. Among the earliest ICTD projects – the Simputer [89] Microsoft, especially when there is an intersection of their
had an explicit interface design goal to make the computer a business interests with ICTD [100]. What is more interesting
simpler device to use, therefore with few alphabetical keys, is that ICTD projects have frequently been couched in a
and more audio-based interactions. In most of the low-cost discourse of ‘sustainability’ [101] raising questions on
computing projects that followed thereafter [90], including whether these projects should be market driven, and we have
most prominently the OLPC, appropriate UI design was a very seen some research in recent years suggesting that the state
important component [58]. should reconsider spending on ICTD projects, specifically
telecenters [59]. However, these questions are not easy to
Significant design work has also been done on building answer – the nature of development projects requires that a
visually enhanced interfaces for illiterate users [91] and range of stakeholders work closely together, and ICTD is no
speech based systems for agriculture [92] and healthcare [93]. different.
Other design work has looked across domains to redesign
existing devices for new application areas – such as In looking at how ICTD research can be divided up
innovating with small screens on mobile devices for between various stakeholders, we asked respondents to
systematic data collection for healthcare [94, 95], micro- comment on what they felt were important roles for industry-
finance services for the illiterate [55], and audio-visual based based research to play.
English-language education [96] delivered in local languages.
TABLE X
AREAS OF FUTURE IMPORTANCE FOR BUSINESS STAKEHOLDERS
TABLE IX
AREAS OF FUTURE IMPORTANCE IN DESIGN
Top 5 areas ranked by respondents Percent
Top 5 areas ranked by respondents Percent Respondents
Respondents
Microfinance and microcredit 41.7
Voice recognition and synthesis 57.1 Mobile commerce 41.7
Local language software 54.2 Supply chain management 16.7
Translation 20.0 Online commerce 13.9
Accessibility 17.1 Low cost sales devices 13.9
Illiterate-friendly interfaces 14.2
This is an interesting question which needs further

In the design area, spreading technology access to examination – on whether markets ought to decide what
populations without English (or other dominant) -language research agenda is better served by the industry, and where
literacy was a major concern for researchers, thus language academia should play a role. The surveyed experts seemed to
localization and voice recognition topped the list of future feel that most areas typically involving financial transactions
research directions. What is interesting here is that work in are best left to the industry. We asked respondents which
both these domain areas is extremely challenging and steps stakeholder does ICTD play for – does it serve a greater
forward have been slow thus far. Due to current technology development agenda or a greater business agenda – the
363
respondents were evenly split – a third believing the achieved that potential. Figure I presents a summary of our
development agenda was better served, another third believing results, with development goals ordered (from left to right) by
that ICTD worked better for business. the fraction of interviewees that believed ICTD can make an
impact (either “significant impact” or “some impact”) towards
We turn now to the respondents’ perceptions of where the that goal. For example, almost 100% of the survey takers
government’s role in ICTD ought to be, and as we find, most believed that ICTD has the potential to make an impact in
of the answers are fairly intuitive – regulate communications, education for all, while only about 60% believed the same
directly act in healthcare, education, governance, stay out of about eradication of hunger.
design and so on. An interesting finding was that very few of
the experts felt that the government should be directly owning The first thing to note it that most experts felt that ICTD
ICTD projects in agriculture, most respondents preferring a projects generally had potential, but that it had a significant
hybrid role instead. This is a fairly unexpected, because in the way to go in achieving it fully. This is reflected by the fact
earlier days of ICTD, the government’s role in e-agriculture that a large number of respondents chose not to answer
was seen as very important. This could perhaps tie in with the questions on the perceived impacts of ICTD thus far, but also
experiences with telecenters and e-agriculture run by state by the fact that few respondents believed that existing ICTD
agencies in the past decade, which have often suffered mixed work has made a significant impact. The areas where most
fortunes, usually weak. respondents felt there was potential for impact (either some
impact or significant impact) were in education, healthcare,
FIGURE I communications, and better governance. The area where the
ROLE OF GOVERNMENT IN ICTD
overwhelming majority believed the impact to be significant
100% No answer was in communication technologies. Other areas of
80% No role development such as sustainable environment and eradication
60%
Partner with non-profits or of hunger were comparatively seen as outside the scope of
40% for-profits ICTD. One remarkable trend we find is that 20% of the
Own and implement
20% projects directly respondents’ opinion was that ICTD had an overall negative
0%
Only play regulatory role
impact on poverty alleviation, and a small percent felt there
Communications
UI design
Agriculture
was a negative impact on governance as well. While in both

Healthcare
Education
Business
Energy
Governance
categories the proportion of respondents who felt that the

impact was positive was either equal or higher, this is
nonetheless a real concern that all of us engaged in serious
ICTD research need to ask ourselves – whether technologies
pose the serious risk of increasing inequalities.
V. COMMENTARY
FIGURE II
After analyzing ICTD research and practice through the lens COMPARISON OF RESPONSES ON POTENTIAL AND ACTUAL IMPACT OF ICTD
of individual research sub-areas, take step back to think about
the bigger issues in the field that appear as a theme throughout Potential Impact
the area. The first of these themes is the well-known debate on 100%
Negative
“hope v/s hype”, on expectations v/s reality that surrounds our 80%
No impact
field. Our literature review as well as conversations with 60%
No answer
experts on the matter suggests that the main concerns raised 40% Some
by existing studies were over whether ICTD stakeholders – 20% Significant
either implementers or funding partners – made incorrect 0%
Healthcare and
environment
Communication
governance
alleviation
equality
Education for
Eradication of
Sustainable
Gender
assumptions about the nature of the expected development

Poverty
technologies
sanitation
Better
hunger
outcome of some technology projects. Many reasons are

all
mentioned as being responsible for these effects, including the

changing nature of development goals around sustainability,
and changing stakeholders with a new breed of technologists
Impact Thus Far
entering the development fray [100]. The opinion of ICTD
experts further suggests concerns regarding the “hype” aspect: 100% Negative
when asked whether they believe ICTD was hyped, 52% of 80% No impact
our respondents responded affirmatively, and roughly 30% 60% No answer
Some
felt there wasn’t any hype. 40%
Significant
20%
0%
To further investigate the gap between what has been
Healthcare and
environment
governance
Communication
alleviation
equality
Education for
Eradication of
Sustainable
Gender
Poverty
technologies
expected from ICTD and what ICTD has already achieved, we

sanitation
Better
hunger
interviewed respondents on what they felt was the potential of

all
ICTD research to make an impact in achieving various MDGs

and on how far work in ICTD over the past few years had
364
To comment on the relative importance and research A look at these raises back the issue of the multi-
maturity of each of these areas, we also did a quick survey of disciplinary nature of ICTD. While the diverse nature of our
the publishing record in the past two iterations of the ICTD respondent profile is in itself an indicator of the range of
conference scholarly traditions involved in ICTD work, it is not clear how
much actual inter-disciplinary work is being done, and if so,
TABLE XI what is the exact nature of such collaborations. Within
ICTD CONFERENCE PAPERS ARRANGED BY THEMATIC AREAS
engineering, there is comparatively much greater interest in
Domain Area No. of Sub-topics ICTD from computer scientists, and this may be an outcome
papers of the nature of much ICTD research or that several of the key
Agriculture 7 Sensors (2) leading faculty members and senior researchers in ICTD are
Best practices / information
sharing (2) from a computer science background. We also find that most
Supply chains (2) publishing in ICTD even if interdisciplinary is usually
Market access / information (1) restricted to being either engineering oriented or in the social
Education 10 Literacy (2) sciences. Perhaps the most notable exception overall is design,
Novel input devices (2)
User-created content 2
which as a field had already started incorporating ethnography
Computer usage models (2) and other methods from anthropology. In fact, 81% of
Teacher training (1) respondents stated that ICTD itself as a field had helped bring
Educational games (1) multidisciplinary research to greater focus in academia as a
Communications 12 Wireless technologies (3)
whole. Although most experts tended to agree that the
Mobile phones (2)
(Technical) multidisciplinary nature of work was a good thing, one
DTN (2)
Disaster prevention and relief (1) concern raised by multiple participants was that of creating
Telecommunications (2) scholarly tenure track positions in ICTD. Although
Satellites (1)
Security (1)
information schools have been early in recognizing the
Communications 17 Telecenters (general) (8) importance of ICTD, the case for ICTD is yet to be made for
(Social Sciences) Low-cost computers (4) much of social science and engineering disciplines. A final,
Radio (2) concerning thought came in reference to the issue of rigor in
Disaster prevention and relief (1)
ICTD research. When asked if the existing ICTD research had
Governance 5 Telecenters (governance) (4)
been methodologically rigorous, less than a fifth of our
Digitized Records (1)
Healthcare 2 Medical education (2) respondents said yes, almost half the respondents felt the work
User interfaces (2) thus far hasn’t been rigorous enough.
Design 15 General User-centered design (3)
Input devices (3)
Illiteracy (2) VI. CONCLUSION
Voice recognition (2) ICTD has made strides since the 1990s, both the expanding
Accessibility (2)
Translation (2) pool of people: researchers, industry and state partners, and
Graphic interfaces (1) high quality new work is evidence of this. What has also
Stakeholders / 15 ICTD Business models (4) worked well for this growing field is that a lot of field projects
Microfinance (3)
Business of ICTD and research are now closer to maturity, and that we see a lot
Kiosks (2)
Models for ICT deployment (2) less work on pilot projects and a lot more work reporting on
Supply chains (1) the progress of initiatives and ideas. A fairly large body of
Social entrepreneurs (1)
Mobile phones (1)
ICTD researchers have now been in the field long enough to
Mobile phone commerce (1) think critically and retrospectively on progress in this work.
General ICTD 5 Methodology (3) We hope that this document will serve as another step in that
Gender Empowerment (2)
introspection.
REFERENCES 7. Howkins, J. and R.L. Valantin, Development and the Information

Age: Four Global Scenarios for the Future of Information and
Communication Technology. 1997: IDRC (International
1. Qvortrup, L., The Nordic telecottages : Community teleservice
Development Research Centre).
centres for rural regions. Telecommunications Policy, 1989. 13(1):
8. Roche, E.M. and M.J. Blaine, Information technology,
p. 59-68.
development and policy. Information Technology, Development
2. Saxenian, A.L., Silicon Valley's New Immigrant Entrepreneurs.
and Policy, EM Roche & MJ Blaine (eds.). Avebury: Aldershot,
1999: Public Policy Institute of CA.
UK, 1996.
3. Vertovec, S., Transnational Networks and Skilled Labour
9. Heeks, R., Technology and Developing Countries: Practical
Migration. COMPAS, Oxford, 2002.
Applications, Theoretical Issues. 1995: Frank Cass & Co.
4. Castells, M., Rise of the Network Society. 1996.
10. Madon, S., Information-Based Global Economy and
5. Negroponte, N., Being Digital. 1995, Cambridge, MA: MIT Press.
Socioeconomic Development: The Case of Bangalore. The
6. Bhatnagar, S.C. and M. Odedra, Social implications of computers
Information Society, 1997. 13(3): p. 227-244.
in developing countries. 1992: New Delhi: Tata McGraw-Hill.
11. Hanna, N., Exploiting information technology for development: a
case study of India. 1994: The World Bank.
365
12. Nulens, G. and L. Van Audenhove, An Information Society in 38. Perednia, D.A., Telemedicine technology and clinical applications.
Africa?: An Analysis of the Information Society Policy of the JAMA, the journal of the American Medical Association, 1995.
World Bank, ITU and ECA. International Communication Gazette, 273(6): p. 483-488.
1999. 61(6): p. 451. 39. Chandrasekhar, C.P. and J. Ghosh, Information and
13. Wehn, U. and R. Mansell, Knowledge societies. 1998: Published communication technologies and health in low income countries:
for and on behalf of the United Nations by Oxford University Press the potential and the constraints. Bulletin of the World Health
New York. Organization, 2001. 79: p. 850-855.
14. Oestmann, S. and A.C. Dymond, Telecenters - Experiences, 40. Sahay, S., Special Issue on “IT and Health Care in Developing
Lessons and Trends, in Telecenters: Case Studies and Key Issues. Countries”. EJISDC, 2001. 5: p. 0-1.
2001, Commonwealth of Learning: Vancouver. p. 1-19. 41. Brauchli, K., et al., iPath-a Telemedicine Platform to Support
15. Brewer, E., et al., The case for technology in developing regions. Health Providers in Low Resource Settings. Stud Health Technol
Computer, 2005. 38(6): p. 25-38. Inform, 2005. 114: p. 11-7.
16. Subramanian, L., et al., Rethinking Wireless for the Developing 42. Surana, S., et al., Deploying a Rural Wireless Telemedicine
World. Proc. of the 5th Workshop on Hot Topics in Networks System: Experiences in Sustainability. COMPUTER, 2008: p. 48-
(HotNets), November, 2006. 56.
17. Seth, A., et al., Low-cost communication for rural internet kiosks 43. Ibrahim, M., et al. Improving Healthcare Delivery in Developing
using mechanical backhaul. Proceedings of the 12th annual Regions: Coupling Output-based Aid and Mobile Technologies.
international conference on Mobile computing and networking, 2007. University of California Big Ideas 2007 Conference.
2006: p. 334-345. 44. Chandani, Y. and G. Breton, Contraceptive Security, Information
18. Raman, B. and K. Chebrolu, Experiences in using WiFi for Rural Flow and Local Adaptations: Family Planning in the Philippines
Internet in India. IEEE Communications Magazine, 2007. 45(1): p. and Morocco. EJISDC, 2001. 5(3): p. 1-18.
104–110. 45. Luk, R., M. Ho, and P.M. Aoki, Asynchronous remote medical
19. Jhunjhunwala, A., Unleashing Telecom and Internet in India. India consultation for Ghana, in Proceeding of the twenty-sixth annual
Telecom Conference at the Asia/Pacific Research Center, Stanford SIGCHI conference on Human factors in computing systems. 2008,
University, November, 2000. ACM: Florence, Italy.
20. Hoffman, D.L. and T.P. Novak, Bridging the Racial Divide on the 46. Ramamritham, K., et al., Innovative ICT Tools for Information
Internet. Science, 1998. 280(5362): p. 390-91. Provision in Agricultural Extension (December 2005). Information
21. Compaine, B.M., The Digital Divide: Facing a Crisis Or Creating and Communication Technologies and Development, 2006.
a Myth? 2001: MIT Press. ICTD'06. International Conference on, 2006: p. 34-38.
22. Hendry, J.D., Social inclusion and the information poor. Library 47. Kaushik, P.D. and N. Singh, Information Technology and Broad-
Review, 2000. 49(7): p. 331-336. Based Development: Preliminary Lessons from North India. World
23. Warschauer, M., Reconceptualizing the Digital Divide. First Development, 2004. 32(4): p. 591-607.
Monday, 2002. 7(7): p. 0-262. 48. Salvador, T., J.W. Sherry, and A.E. Urrutia, Less cyber, more cafe:
24. Singh, N., Information Technology as an Engine of Broad-Based Enhancing existing small businesses across the digital divide with
Growth in India, in The Future of India and Indian Business, P. ICTs. Information Technology for Development, 2005. 11(1): p.
Banerjee and F. Richter, Editors. 2002, Macmillan: London. 77-95.
25. Heeks, R., India's software industry. 1996: Sage Publications 49. Kumar, R., eChoupals: A Study on the Financial Sustainability of
Thousand Oaks, Calif. Village Internet Centers in Rural Madhya Pradesh. Information
26. Arora, A., et al., The Indian software services industry. Research Technologies and International Development, 2004. 2(1): p. 45-74.
Policy, 2001. 30(8): p. 1267-1287. 50. Patel, N. and T. Parikh. Designing a Farmer-centric Organic
27. Saxenian, A.L. Bangalore: The Silicon Valley of Asia? in Certification System. in ACM SIGCHI 2008, Workshop on HCI for
Conference on Indian Economic Prospects: Advancing Policy Community and International Development. 2008. Florence, Italy.
Reform,. 2000. Stanford, California: Center for Research on 51. Panchard, J., et al. COMMON-Sense Net: Improved Water
Economic Development and Policy Reform. Management for Resource-Poor Farmers via Sensor Networks
28. Jhunjhunwala, A., Can Telecom and IT be for the Disadvantaged? in Information and Communication Technologies and Development, 2006.
Rural Development, 1998. 17(2): p. 321–37. ICTD '06. 2006.
29. James, J., Low-cost computing and related ways of overcoming the 52. Heffernan, C., The Livestock Guru: Demand-led knowledge
global digital divide. Journal of Information Science, 2001. 27(6): transfer for poverty alleviation. Information and Communication
p. 385-392. Technologies and Development, 2006. ICTD'06. International
30. Heeks, R., Information Systems and Developing Countries: Conference on, 2006: p. 94-100.
Failure, Success, and Local Improvisations. The Information 53. Veeraraghavan, R., et al., Kiosk Usage Measurement using a
Society, 2002. 18(2): p. 101-112. Software Logging Tool. Information and Communication
31. Keniston, K., Grassroots ICT Projects in India: Some Preliminary Technologies and Development, 2006. ICTD'06. International
Hypotheses. ASCI Journal of Management, 2002. 31(1). Conference on, 2006: p. 317-324.
32. Madon, S., Evaluating the Developmental Impact of e-Governance 54. Gandhi, R., et al. Digital Green: A Participatory Digital
Initiatives: An Exploratory Framework. EJISDC, 2004. 20(5): p. 1- Framework to Deliver Targeted Agricultural Information to Small
13. and Marginal Farmers. in Second International Conference on
33. Donner, J., The Use of Mobile Phones by Microentrepreneurs in Information Technologies and Development, Dec 2007, IEEE
Kigali, Rwanda: Changes to Social and Business Networks. Conference Proceedings. 2007. Bangalore.
Information Technologies and International Development, 2007. 55. Parikh, T., K. Ghosh, and A. Chavan, Design studies for a
3(2): p. 3-19. financial management system for micro-credit groups in rural
34. Heeks, R., ICT4D 2.0: The Next Phase of Applying ICT for india. Proceedings of the 2003 conference on Universal usability,
International Development. COMPUTER, 2008: p. 26-33. 2003: p. 15-22.
35. Thomas, P., Bhoomi, Gyan Ganga, e-governance and the right to 56. Sathe, S.K. and U.B. Desai, Cell Phone Based Microcredit Risk
information: ICTs and development in India. Telematics and Assessment using Fuzzy Clustering. Information and
Informatics, 2008. Communication Technologies and Development, 2006. ICTD'06.
36. Antin, J. Cultural Assessment for Sustainable Kiosk Projects. in International Conference on, 2006: p. 233-242.
Information and Communication Technologies and Development, 57. Ratan, A. and M. Gogineni. Cost Realism in Deploying
2006. ICTD '06. 2006. Technology for Development. in Confronting the Challenge of
37. Keniston, K., IT for the Masses: Hope or Hype. Massachusetts Technology for Development: Experiences from the BRICS. 2008.
Institute of Technology (MIT), Cambridge, MA, 2003. University of Oxford.
58. Patra, R., Pal, J., Nedevschi, S., Plauche, M., Pawar, U. Usage
models of classroom computing in developing regions. in Second
366
International Conference on Information Technologies and 81. Riley, C.G., The changing role of the citizen in the e-governance
Development, Dec 2007, IEEE Conference Proceedings. 2007. and e-democracy e-democracy equation. Commonwealth Centre
Bangalore. for e-Governance, 2003.
59. Pal, J., Examining e-literacy Using Telecenters as Public 82. De, R. and C. Sen, The Complex Nature of e-Government Projects:
Spending. Second International Conference on Information A Case Study of Bhoomi, an Initiative in Karnataka, India.
Technologies and Development, Dec 2007, IEEE Conference Proceedings of Electronic Government, Third International
Proceedings, 2007. 2: p. 59-68. Conference, Zaragoza, Spain, Aug-Sep, 2004: p. 556–557.
60. Banerjee, A.V., et al., Remedying Education: Evidence from Two 83. Okot-Uma, R.W., Electronic Governance: Re-inventing Good
Randomized Experiments in India*. The Quarterly Journal of Governance. Commonwealth Secretariat, London, 2000.
Economics, 2007. 122(3): p. 1235-1264. 84. Narayan, G. and A.N. Nerurkar, Value-proposition of e-
61. Pawar, U.e.a. Multiple mice for retention tasks in disadvantaged governance services: Bridging rural-urban digital divide in
schools. in Proceedings of the SIGCHI conference on Human developing countries. International Journal of Education and
factors in computing systems. 2007. San Jose. Development using ICT, 2006. 2(3).
62. Kam, M., et al., Localized iterative design for language learning in 85. Benjamin, S., Bhuvaneshwari, R, Rajan, P. 'E-Titling' or an "Anti-
underdeveloped regions: the PACE framework. Proceedings of the Politics Machine" necessary to globalize Bangalore? in
SIGCHI conference on Human factors in computing systems, International Conference on Enhancing Land Registration and
2007: p. 1097-1106. Cadastre for Economic Growth in India. 2006. New Delhi: Centre
63. Wang, R., et al., The Digital StudyHall. Computer Science for Science, Development and Media Studies (CSDMS).
Department, Princeton University, Tech. Rep. TR-723-05, 2005. 86. Vasudevan, R., Changed Governance or Computerized
64. Feghali, T., M. Martin, and N. Sahyoun, An exploratory case study Governance? Computerized Property Transfer Processes in Tamil
of a Lebanese e-learning partnership model. World Review of Nadu, India. Information Technologies and International
Science, Technology, and Sustainable Development, 2006. 3(2): p. Development, 2007. 4(1): p. 101-112.
137-151. 87. Rodrigues-Filho, J. and N.P. Gomes, e-Voting in Brazil:
65. Colle, R.D., Building ICT4D capacity in and by African Exacerbating Alienation and the Digital Divide. Proceedings of the
universities.' International Journal of Education and Development 6th European Conference on e-Government.
using ICT, 2005. 1(1). 88. Tongia, R. and E. Subrahmanian, Information and
66. Islam, Y.M., et al., Mobile Telephone Technology as a Distance Communications Technology for Development (ICT4D)¿ A Design
Learning Tool. 2005, ICEIS. Challenge? Information and Communication Technologies and
67. Librero, F., Uses of the Cell Phone for Education in the Development, 2006. ICTD'06. International Conference on, 2006:
Philippines and Mongolia. Distance Education, 2007. 28(2): p. p. 243-255.
231-244. 89. Chandru, V., et al. The Simputer: Radical Simplicity for Universal
68. Moraveji, N., U. Pawar, and T. Kim, A Mouse on Each Desk: An Access. in Development by Design: ThinkCycle Open
Inexpensive Classroom Interaction Technique for Remote Collaborative Conference, Bangalore 2002. 2001.
Teaching. Microsoft Research Working Papers, 2006. 90. Harvey, F., Computers for the Third World. Scientific American,
69. Howard, J., C. Simms, and E. Simanis, What Works: n-Logue’s 2002.
Rural Connectivity Model. A Digital Dividend Study by the World 91. Medhi, I., A. Sagar, and K. Toyama. Text-Free User Interfaces for
Resources Institute, 2001. Illiterate and Semi-Literate Users
70. Mishra, S.M., et al., Economic Analysis of Networking in Information and Communication Technologies and Development, 2006.
Technologies for Rural Developing Regions. 1st Workshop on ICTD '06. International Conference on. 2006.
Internet and Network Economics, 2005. 92. Plauché, M. and M. Prabaker, Tamil market: a spoken dialog
71. Ramanathan, N., et al., Designing Wireless Sensor Networks as a system for rural India. Conference on Human Factors in
Shared Resource for Sustainable Development. Information and Computing Systems, 2006: p. 1619-1624.
Communication Technologies and Development, 2006. 93. Sherwani, J., et al. HealthLine: Speech-based Access to Health
72. Bhagwat, P., B. Raman, and D. Sanghi, Turning 802.11 inside-out. Information by Low-literate Users. in Second International
ACM SIGCOMM Computer Communication Review, 2004. 34(1): Conference on Information Technologies and Development, Dec
p. 33-38. 2007, IEEE Conference Proceedings. 2007.
73. Pentland, A.S., R. Fletcher, and A. Hasson, DakNet: Rethinking 94. DeRenzi, B., et al., E-imci: improving pediatric health care in low-
Connectivity in Developing Nations. 2004. income countries. 2008.
74. Zahedi, E., M.A.M. Ali, and M.J. Gangeh, Design of a Web-based 95. Ho, M. Smartphones for Output-Based Aid. in ACM SIGCHI 2008,
wireless mobile teleconsultation system witha remote control Workshop on HCI for Community and International Development.
camera. Engineering in Medicine and Biology Society, 2000. 2008.
Proceedings of the 22nd Annual International Conference of the 96. Kam, M., et al., Practical considerations for participatory design
IEEE, 2000. 2. with rural school children in underdeveloped regions: early
75. Seth, A., et al., An Architecture for Tetherless Communication. reflections from the field. Interaction Design And Children, 2006:
DTN Workshop, 2005. p. 25-32.
76. Fall, K., A delay-tolerant network architecture for challenged 97. Atkinson, A.B., New Sources of Development Finance. 2005:
internets. Proceedings of the 2003 conference on Applications, Oxford University Press.
technologies, architectures, and protocols for computer 98. Jenkins, R., Globalization, Corporate Social Responsibility and
communications, 2003: p. 27-34. poverty. International Affairs, 2005. 81(3): p. 525-540.
77. Wei, S.J., Corruption in Economic Development: Beneficial 99. Fox, T., Corporate Social Resposibility and Development: In quest
Grease, Minor Annoyance, or Major Obstacle?, in World Bank of an agenda. Development, 2004. 47(4): p. 29-36.
Policy Research Working Paper No. 2048. 1999, World Bank: 100. Pal, J. Computers and the Promise of Development: Aspiration,
Washington DC. Neoliberalism and ‘Technolity’ in India’s ICTD enterprise in
78. Backus, M., E-governance in Developing Countries. IICD Confronting the Challenge of Technology for Development:
Research Brief, 2001. 1. Experiences from the BRICS. 2008. University of Oxford.
79. Hasan, S., Introducing E-Government in Bangladesh: Problems 101. Kuriyan, R., K. Toyama, and I. Ray, Integrating Social
and Prospects. International Social Science Review, 2003. 78(3-4): Development and Financial Sustainability: The Challenges of
p. 111-126. Rural Computer Kiosks in Kerala. Information and
80. Kiran, G.R., E-governance services through Telecentres-Role of Communication Technologies and Development, 2006. ICTD'06.
Human Intermediary and issues of Trust. Information and International Conference on, 2006: p. 121-130.
Communication Technologies and Development, 2006. ICTD'06,
Conference Proceedings, 2006: p. 131-142.
367
Information Communication Technology and

Sustainable Communities in Africa: The Case of
the Niger Delta Region of Nigeria. (Feb. 2009)
Uduak A. Okon
Abstract— Sustainable development is largely seen as I. INTRODUCTION

pertaining primarily to environmental issues and grassroots
social development. This stereotype misses the reality that
sustainable development and the information society are
operationally interconnected. The two phases of the World
D uring the last decades four themes have emerged as a
response to the collective concerns of world citizens: peace,
Summit on the Information Society (WSIS) in Geneva (December
2003) and Tunis (November 2005) provide an excellent freedom, development and environment. As a consequence
opportunity to integrate sustainable development principles and this collective concern and the four themes were manifested in
practices into the institutions and policy frameworks that are a political process towards sustainability labeled “sustainable
shaping the information society. While the World Summit on development”. The concept of sustainability lacks a widely
Sustainable Development (WSSD) and WSIS Phase I brought accepted definition. Sustainable development is a wide
these issues to international attention, Southern voices and concept and has over the years been introduced through many
visions are still notably lacking from the debate different definitions. The term “sustainable development” was
As we move into the age of information it is critically popularized by the World Commission on Environment and
important for us to consider the implications of ICT in
Development (WCED), in its 1987 report entitled, “Our
Sustainable Development and vice-versa. This research
undertaking looks at precisely this intersection with its primary Common future”. This report by the Brundtland commission
focus on sustainable communities. Sustainable Communities (named after it’s chair Dr. Gro Harlem Brundtland) provided
development is an evolving discourse. The locally-owned or the most commonly cited definition of sustainable
adapted knowledge of a community is essential for integrated development. The Commission wrote: “Humanity has the
sustainable development, and is becoming a key priority for ability to make development sustainable – to ensure that it
development practitioners. The popularity of the concept of meets the needs of the present without compromising the
communities is growing, and so are ideas on how to support these ability of future generation to meet their own needs”. This
communities with technologies and how these communities might definition has broad appeal and little specificity, but some
use ICTs to support themselves. These issues are pertinent in
combination of development and environment is found in
assessing the real contribution of ICTs to sustainable
development, and therefore merit a closer examination. The
most attempts to describe it. Since the publication of the
paper presents the findings from the study of 9 communities in report, there has been a mass of literature generated in various
the Niger Delta region of Nigeria. The aim of the study was to fields which has resulted in more specific application of the
explore how ICTs may contribute to the social sustainability of concept, such as sustainable agriculture, sustainable
communities in the Niger Delta Region of Nigeria. The study livelihoods, and sustainable transport. Over the years, the
grounds the understanding of ICT usage among indigenous interest and scope of the sustainable development debate has
communities and consumers in the reality of their everyday lives, grown substantially and this has led to an increasing diversity
in order to promote actions for sustainability. of interpretations. Mostly, sustainable development is
modeled on three pillars used to facilitate the comprehension
Index Terms— Communicative Ecology, Information
of the term: the triangle of environmental (conservation),
Communication Technology, Social Change, Sustainable Communities
economic (growth), and social (equity) dimensions. According
to Selman (1996:31), while such universal aspirations may be
helpful, they are insufficient, at the local level, to identify
local needs, choose meaningful targets, and most important,
Manuscript received February 20, 2009. This work is part of a doctoral “harness the energies of local people and organizations and
research. It is funded by a research Grant provided by Akwa Ibom State
Government, Nigeria. aid development of the local society and economy to change
Uduak Okon is a Doctoral Student at Royal Holloway, University of London in ways which are conducive to sustainability”
and is a member of the ICT4D Collective. (Phone: 01784248786; Mobile: There are newer approaches that argue that the point of
07776474379 ; Email: u.akpan-okon@rhul.ac.uk; okonud@yahoo.com).
departure most be the community. Redclift and Sage (1995)
368
argue that if sustainable development is to mean something, communication ecology with a focus on a (new) definition of
then it must be capable of translating into local action. sustainability. It adds to the ongoing dialogue among
Liemgruber and Imhof (1998) go even further by suggesting academics, development practitioners, business leaders and
that the true scale for sustainability is the local level, where government officials, aimed at pinpointing the ways in which
people interact and communicate and where each individual is ICTs can help make the transition to sustainable development
affected by everybody’s actions. Woodhouse (2000) offers a easier, quicker and economically viable. Its focus is on ICTs
similar argument; he states that the concept of sustainability is as means and tools that enable desired changes, since it is
best understood and evaluated on the basis of a sustainable these changes, not ICTs that lead to collective action and
community. He further states that genuine development sustainable development. The presented research is at the base
enhances the sustainability of the community and this does not of ICT for development in developing Countries. A thorough
necessarily involve economic growth. As the debate on understanding of the interaction ICT with the communities’
sustainability matures it is becoming increasing accepted that communication ecology is of primary importance, before the
development cannot be sustainable unless it builds on cultural deployment of any ICT solution. Thereby it will be of use for
traditions (Nurse, 2004) Traditions, norms and customs are all further research and/or development of ICT as part of a
cultural capital that needs to be preserved and passed to future solution for indigenous communities.
generations. Sustainable development cannot be a product
packaged by the international community and delivered to a
local community; it works best when it draws on existing II. CONCEPTUAL FRAMEWORK
community resources and capacity building efforts (Cooper A. Information Communication Technologies (ICTs)
and Vargas, 2004). It relies heavily on community The definition of ICT for this research uses Hamelink’s
involvement and commitment, since sustainable development definition of ICTs: “Information and Communication
is a way of life. Ideas and projects that are developed locally Technologies (ICTs) encompass all those technologies that
often have the greatest staying power because the community enable the handling of information and facilitate different
develops a sense of ownership. Cooper and Vargas also state forms of communication among human actors, between
that sustainable development is both a bottom-up and top- human being and electronic systems and among electronic
down process. Global commitments made by the nations of systems” (Hamelink, 1997:3) and Duncombe and Heeks
the world are necessary but not sufficient to ensure (1999) simplified definition describing ICTs as an 'electronic
sustainability. Hence, the slogan from earlier Earth Days, means of capturing, processing, storing and disseminating
“Think Globally, act locally”. Localizing global aspirations is information'. This includes the “old” ICTs of radio, television
a key challenge for implementing sustainable development and telephone, and the “new” ICTs of computers and mobile
strategies and even more so for developing countries where technology and the Internet.
most people live in rural and sometimes remote communities. Information and Communication Technologies are rapidly
There is a need to ask local questions and focus on meeting consolidating global communication networks and
local needs. This paper makes a distinction between external international trade with implications for people in developing
physical dimensions of sustainability and the internal socio- countries. Despite this there is a worrying lack of empirical
cultural dimension. The latter dimension is mainly concerned evidence or analysis of the actual experiences and effects of
with individual actions, social structures and the social capital ICTs upon poor people’s economic and social development.
of communities. Contrary to what the proponents that The constraints of existing information systems on poor
prioritise environmental concerns believe, within developing women and men and their intersection with ICTs are also little
countries like Nigeria, pressing problems of social exclusion, understood. Little attempt is made by those promoting ICTs
poverty and unemployment are reducing the attention paid to for development to assess their impact on the cultural identity,
environmental problems. Because of these problems, societies the values, and the state of social equity of the less developed
are less willing to accept the structural changes associated economies. Also lacking are analysis of the social and cultural
with shifts towards a more environmentally sound patterns of factors which determine the effective application and use of
consumption. Thus, for countries like Nigeria, a socio-cultural ICTs by developing countries. Social exclusion in the
perspective is integral to all discussions about susutainable developing world cannot simply be resolved by technology if
development. consideration of the factors that can ensure respect for cultural
The aim of this research is to explore from a socio-cultural values, justice, equity and equality in the distribution of
perspective how ICTs may contribute to and support the wealth - including information - is absent. Rodgers et al,
sustainable development of communities in the Niger Delta (1994) argues that by definition, the nature of these elements
Region of Nigeria. This paper outlines the conceptual imposes certain limitations on how ICTs can be appropriated
framework for the study of ICTs and Sustainable communities for development. Even if the introduction of ICTs is feasible
and presents a narrative of the findings of the empirical over a relatively short period of time as some predict, a
research conducted in Nigeria on the Communicative ecology dialogue between those promoting the technologies and the
of Niger Delta communities. This research contributes to the potential beneficiaries must be the foundation of any
understanding of the interaction of ICT with the communities’
369
development action. In the current information age, the According to the organizers of the World Summit on the
capacity of a society to effectively position itself as a Information Society (WSIS), “We are in the midst of a
consumer and producer of knowledge is crucial to its social revolution, perhaps the greatest that humanity has ever
development. ICTs are increasingly playing a crucial role in experienced. To benefit the world community, the successful
most communities’ capacities to produce, access, adapt and and continued growth of this new dynamic requires global
apply information and thus offer enormous opportunities for discussion and harmonization in appropriate areas.”
facilitating the transfer and acquisition knowledge. They Unfortunately, until recently, few discussions have focused on
present (atleast theoretically) a promising potential to lead harmonizing the visions of the emerging information society
developing countries into the ‘highways’ of development. with the principles and priorities articulated by the United
(Friedman 2006; Castells 2000). Despite the potential role and Nations’ Millennium Development Goals (MDGs) and the
contribution of ICTs to development, there is still a growing World Summit on Sustainable Development (WSSD).
voice in the development field raising serious concerns about Meaningful discussions about national policy coherence
the socio-cultural dimension of their application and use between these processes have been limited by restricted
(Panos 1998; Wade 2002; Gumucio 2001). Proponents of thinking about the information society (IS) and sustainable
ICTs (World Bank 2002, UNDP 2001; Pohjola, 2002; Braga, development (SD). Each has emerged from a different
1998) take an optimistic view and highlight the positive community with a different vocabulary and process for
effects of the Internet and other forms of ICTs to create new determining national priorities. Information society specialists
economic, social and political opportunities for developing have primarily been drawn from the fields of
countries and the poor, but it also highlighted the need to ask telecommunications and economic development. While,
both with regards to individuals and society, especially in sustainable development has been delegated primarily to
third world countries, about the nature and extent of the environmental issues despite the best efforts of its
influence and changes, and about the factors which facilitate practitioners to articulate a holistic vision of integrated
or impede these processes, and more importantly how these economic, social and environmental decision-making.
influences and change may promote actions for sustainability. The lack of interaction between these two policy
It has been found more useful to approach the question in communities is currently serving to reinforce stereotypes of
terms of different predispositions and experiences, and with both fields. Sustainable development is seen as pertaining
different historical, ethnic, linguistic, social, economic or primarily to environmental issues and grassroots social
religious backgrounds in different situations, making different development, while the information society is perceived as
use of ICT media (Avgerou, 2001; Walsham, 1993). Taachi et being more relevant to the economic development potential of
al (2003) further argue that it is important to study the urban elites. These stereotypes miss the reality that sustainable
communication needs, communication patterns and network, development and the information society are operationally
and the increasingly important impact which developments in interconnected. Both terms are increasingly used by civil
communication technology might have on these. O’Farrell society and academics to refer to a desired global future that is
(2001) and Heek’s (1999) share the belief that before one can casting its shadow upon our current time and decisions. For
advocate for the development of ICTs among the poor, it is sustainable development to be effective and efficient, it must
important to understand the existing information systems of harness the institutions and tools of the information society.
the poor, how they interact with more formal information and And for the information society to sustain itself, it must pay
the best way to strengthen them before intervening with new careful attention to the stocks and flows of resources (material
information sources and means of access sources. Questions and human) and energy that underpin it.
such as: who benefits and who loses from the introduction of There is now widespread recognition that sustainable
these technologies; how can ICTs be made useful and development is the responsible way forward and it is
meaningful to the developing countries' poor majority who are enhanced by the recognition that information and
struggling to meet their basic needs; what are the social and communication technologies (ICTs) are key drivers of socio-
cultural opportunities and risks they present; and how can economic change and so has to be brought into focus as part
developing countries meaningfully adopt these technologies of the equation. For example, Radermacher (1998) points out
while lessening their undesirable social and cultural that "Modern information and communication technologies
consequences, are some of the questions that emerge when drive the worldwide economic system and the process of
looking at the potential development impacts of ICTs. These globalisation. In this process we see enormous growth
issues are pertinent in assessing the real contribution of ICTs worldwide with opportunities for overcoming poverty and
to sustainable Communities development, and therefore merit promoting human rights but also with major threats to
a closer examination. sustainability and to social justice".
Only a small amount of literature has attempted to discuss

B. Information Society and Sustainable development: An the information society and sustainable development together.
emerging Convergence Both Felleman (1997) and Pamlin’s (2002) works are good
examples, with each offering important and interesting
370
insights into information policy and tools and their role in development are both value-laden, involving people in the
environmental sustainability. However, more often than not, decision-making and management processes of our society. In
discourse on these two policy fields has a tendency to provide other words the inter dependency between development and
a very narrow cross-section of issues and perspectives focused environment cannot be separated from people’s actions.
primarily on environmental sustainability. To adequately Manning (1990:291) stresses that “the concept underlying
understand the potential for ICTs to enable the achievement of sustainable development is … a human perspective”. Beck
more sustainable development, we must look beyond the (1992:81) argues that environmental problems “are not
direct impacts of the tools themselves to indirect impacts of problems of our surroundings, but – in their origins and
the broader information society that they enable. through their consequences – are thoroughly social problems,
A strong area of convergence is local content—the locally- problems of people. Environmental issues and developmental
owned or adapted knowledge of a community— which is issues are mediated by human beings. According to Giddens’
shown by this paper to be essential for integrated sustainable (1984) theory of structuration, the key to understanding the
development. Technology is crucial to sustainable internal social dimension of sustainability is ‘duality of
development; local information is fundamental to many structure’. His theory argues that individuals and society are
sustainable development activities, such as monitoring and one whole: human action creates the structures of society;
raising awareness; local content is pervasive in many sectors those structures provide the context for the socialization of
of sustainable development. humans, and in turn, the human action which will reflect and
Technology has long been a crucial element in measuring recreate these structures. This concept of ‘duality of structure’
and achieving sustainable development. From satellite which consists of individual actions and social structures is
imagery, environmental information systems and, today, the the lens through which this research explores ICT use and
use of mobile phones to share health information, there has adoption. Attention is placed on the ability of cultures and
been a key dependence on ICTs in the creation, dissemination relationships, enabled by ICTs, to transform socio-economic
and consumption of relevant, local information. But “even patterns. By addressing in more detail how people relate to
after the first United Nations (UN) World Summit on the one another, how shared practices emerge, and how
Information Society (WSIS) in Geneva 2003, the relationship communities evolve, we will be able to understand better
between issues of the global information systems and of when, how, and why such communities use or do not use
sustainable development is not being discussed adequately. It technologies. Insight into the social capital of communities
seems that the interdisciplinary and international research in will also provide better understanding concerning ICT
this field is just beginning (Hilty, Seifert and Treiber 2005). adoption and use by communities than the more traditional
analysis of technological requirements, knowledge
C. Sustainable Communities requirements, and structural aspects of the community. Social
Sustainable development pursues long-term goals and in capital refers to networked ties of goodwill, mutual support,
trying to explain their endurance, attention to the cultural shared language, shared norms, social trust, and a sense of
dimension is essential. The cultural and historical dimensions mutual obligation that people can derive value from. Social
combine to construct the common social, economic and capital then is about value gained from being a member of a
ecological) pillars of interest of sustainable development. network. In general, social capital is often seen as the glue that
Sustainability must become a primary goal if we want to brings and holds communities together (Cohen & Prusak,
ensure the long-term health of communities. 2001). Huysman & Wulf (2004a) go even further by saying
What is lacking in the debate on sustainability is a strong ‘social capital is considered as a necessary “ingredient” that
focus on people. This research does not deal with the all binds communities over time’. There is no universal
encompassing question of sustainable development in general; development model which leads to sustainable communities;
rather it is focusing on the internal, social perspective of instead development is an essential, multidimensional,
sustainability, arguing that it is a deeper explanation of dialectic process that can differ from community to
sustainability issues. It argues that to explore the meaning of community, context to context. Each society and community
social sustainability, as well as its practical implications, is must attempt to define its own strategy for sustainable
crucial to the understanding of sustainability issues. The development. This research seeks to help Niger Delta citizens
central proposition of this research is that the issue of determine what constitutes sustainability in their own
sustainability should be understood as a social problem, a communities, and how they can act on their new knowledge
problem created by, and eventually having its final impact on and how ICTs may facilitate these.
people themselves. Commoner (1993:23) argues that: “When
an environmental issue is pursued to its origins, it reveals an
inescapable truth—that the root cause of the crisis is not to be III. CASE STUDY – THE NIGER DELTA REGION
found in how men interact with nature, but in how they The Niger Delta is the delta of the Niger River in Nigeria
interact with each other…”. The human scale (or social and as defined officially by the Nigerian Government, extends
dimension) is important for the re-conceptualization of over about 70,000 km² and makes up 7.5% of Nigeria’s land
sustainability since the meanings of both environments and
371
mass. Some 20 million people of more than 40 ethnic groups, benefits from the considerable government revenues from oil
speaking some 250 dialects live in the Delta. Their livelihoods and gas sales, as the bulk of revenues have been looted or
are primarily based on fishing and farming. Nigeria has mismanaged by public office holders at all levels of
become Africa's biggest producer of petroleum, including government.
many oil wells in the Oil Rivers. Some 3 million barrels a day The subsiding levels of a culture of public participation in
are extracted in the Niger Delta. The region has accounted for governance and the prevalence of corruption among public
more than 80% of Nigeria's export earnings. Much of the office holders have fuelled fierce competition for political
natural gas extracted in oil wells in the Delta is immediately offices with contenders deploying violent mechanisms to deter
burned, or flared, into the air at a rate of approximately 70 opponents. The use of state apparatus and armed gangs by
million m³ per day. This is equivalent to 40% of African politicians has been on the increase across the country, and
natural gas consumption, and forms the single largest source with it the harassment and abuses of the rights of citizens.
of greenhouse gas emissions on the planet. The environmental The majority of citizens have thus been alienated from the
devastation associated with the industry and the lack of political processes, while those that speak out are sometimes
distribution of oil wealth have been the source and/or key targeted for intimidation and, sometimes, even outright
aggravating factors of numerous environmental movements elimination.
and inter-ethnic conflicts in the region, including recent The increase in militia activities in the Niger Delta is a
guerrilla activity by the Movement for the Emancipation of reflection of disenchantment and local desperation in the
the Niger Delta (MEND). struggle for survival, as well as a quest to seek attention of the
world to their plight in the hands of trans-national oil and gas
corporations and the Nigerian government. Criminal gangs are
also increasingly exploiting the breakdown of communal
order to unleash mayhem on citizens.
However, the increasing global appreciation of the problem
is an opportunity for concerted action. That is why it is
critically important to the region to get on the global
information highway and for ICT media to reflect local
content to enable to people advocate for their rights as
citizens. Community mobilization and collective action is
essential not just to counter looting of revenues allocated to
the different tiers of government, but also to demand for and
defend their democracy.
IV. METHODOLOGICAL APPROACH
The methodological approach is based on Ethnography and

Fig. 1: Map of Nigeria Participatory Research. The research methodology
Map of Nigeria numerically showing states of the Niger incorporated Ethnographic approach into the Participatory
Delta region depicted in red: Research framework. Taachi et al. (2002) have used this kind
1. Abia, 2. Akwa Ibom, 3. Bayelsa, 4. Cross River, 5. Delta, of approach successfully for a research designed to develop a
6. Edo, 7. Imo, 8. Ondo, 9. Rivers transferable methodology for the evaluation of community
multimedia centers. It was developed to focus on actual
practices of use and interactions with technologies in the
A. Community Related Challenges wider context of people’s lives and social and cultural
The Niger Delta remains largely underdeveloped although structures (Tacchi and Slater, 2003) Ethnography is used to
there are areas where expatriates working in oil companies guide the research process and make sense of the complete
like ExxonMobil, Shell, Chevron etc. live, which look like a range of social relationships and processes. The research will
communities in well developed countries. However, amidst involve individuals and groups researching their own socio-
growing exports and increasing revenues to national, state and cultural settings and experiences. They will reflect on their
local governments, poverty levels in the country have been values, shared realities, collective meanings, needs and goals.
increasing while social infrastructures collapse. Dependence A major factor considered in designing the methodology is the
on oil and gas revenues has been a major factor in the inability evolving paradigm in development research. Most notable is
of the state to create an enabling environment for sustainable the ‘things’ versus ‘people’ debate (Korten 1995; Chambers,
development with the result that citizens are disenchanted 2003). Previously research in ICT for development had been
while violence has escalated in the country. The increasing focused on ‘things’ i.e. infrastructure, connectivity, hardware
violence in the Niger Delta complicates a social situation and surveys of ICT use in developing countries etc.
characterized by massive poverty and environmental (Hoffman,1985; Flamm,1987; OECD 1988) More recently
degradation. Corruption has robbed communities of potential
372
there has been a shift in research focus from ‘things’ to development and population to effectively represent the Niger
‘people’. In an evolving paradigm of development there is a Delta as a whole. During phase 1 of the study, 9 communities
new high ground, a paradigm of people as people (Kling, were studied. Three in each state and four focus groups was
2000; Heeks 2001; Madon, 1997). conducted in each state. Each group had similar membership,
The research is people-centered and the methodological and so two of these constituted of women participants (young
design reflected this emphasis. The approach is open, women and older women) and 2 of men participants (young
evolving, participatory, employing diverse methods, men and older men). In total there were therefore be 12 focus
interactions seek to enable local people and it is a bottom-up groups. Three focus groups were conducted in more urban
approach. Central to these methods were participant communities and 9 conducted in rural communities. This was
observation, keeping of field notes, in-depth interviews, focus to enable a comparative analysis of rural versus urban
group discussions and group interviews. The PAR strategy constructions of meanings of ‘community’ and
adopted here aimed to produce knowledge and action directly ‘sustainability’. Participants were drawn from various
through research and to empower people on a deeper level backgrounds ranging from students, professionals and civil
through the process of constructing and using their own servants to artisans and market women. Group size ranged
knowledge. The study is an integrated three stage process of between 6 -12 people. 104 semi structured interviews were
social investigation, education and action designed to support also conducted in the nine communities.
those with less power in their community settings. The field
In phase 2 of the field work, the sampling size was smaller
work for the research is in three key stages that reflects the to allow for a more in-depth and rigorous exercise. The
overall Participative Research Framework (see Table 1). The ethnographic approach meant longer periods of time was spent
findings for the first 2 phases of the research are presented in in each community. The aim of the study was to understand
this paper. the communicative ecologies of the target communities by
investigating the local information flows, social networks and
TABLE 1 structures and how these may affect people’s perception and
RESEARCH FRAMEWORK use of ICTs. It was conducted in Akwa Ibom and Bayelsa
states; four communities were studied, using participant
PHASES Purpose METHODS observation, field notes, in-depth interviews and diaries. Phase
2 spanned four months. The researcher immersed herself in
each community for a month.
Phase 1 Redefining Focus Group
Preliminary ‘Community’ Discussions V. FINDINGS
Study and Interviews
The findings presented in this section are the results from
‘Sustainability’
the first 2 phases of the empirical study.
A. Redefining ‘Community’ and ‘Sustainability’

Phase 2 Understandin Participant Developing a clear picture of what is meant by Community
Ethnographic g the observation sustainability from a group of people who are generally not
Study Communicative Field Notes reflective is not an easy undertaking. The study set out to
Ecology In-depth interviews answer three key questions about Niger Delta Communities
Diaries (1) What kind of community is it? (2)What is to be sustained?
(3)And How will it be sustained?
The definition of community has been contested within
Phase 3 Engaging Focus Group
sociology from Tönnies onwards (Tönnies 1957). For
Participatory communities in Discussions
developing countries with a strong cultural heritage
Action critical dialogue Workshops
‘community’ means something different from the western
Research
construct of community. By understanding what ‘community
means to local people at a local level, it should be possible to
develop sustainable strategies that are meaningful and
germane to people in those communities. Key elements that
are integral to healthy communities like Identity, ownership,
participation, cohesion were given special focus.
A. Sampling The key findings from the preliminary study were that the
The Niger Delta consists of 9 states, and the study was parameters that define communities for the Niger Delta people
conducted in 3 states, Rivers, Bayelsa and Akwa Ibom states. are different from those of western communities. Geography
These 3 states were predetermined and selected based on the and ethnicity played a strong role in defining who belonged to
diversity of their cultures, ethnicity, differing levels of which community, although that was not always the case.
373
There were cases of people who resided in particular An analytical framework for the analysis of the
communities who were considered outsiders and were not communicative ecology with eleven key concepts grouped
involved in community activities because they belonged to a under three categories was developed: These categories are
different ethic group or are not originally from there. In some Community practice; Community Information and
cases people living outside a geographic region have a Communications systems and; ICT Use and Impact.
stronger identity with their communities of origin even though The classification of the concepts under the three categories
they don’t live there. Language is a crucial element in was based on: causal relationships, contextual factors or
belonging to the community, if you had a different language intervening conditions, and actions and consequence of the
you were not considered a community member in the rural previous two. See Fig. 2 below.
communities, Culture, Social Norms and Values, Shared
leadership and Lineage make community bonds much
stronger. Their ‘emic’ definition of a sustainable community
also differed from the western construct of sustainable Concepts Categories
communities. The major reasons for this are that community is
viewed differently and the priorities of the people are based on
their socio-economic situation and their cultural heritage. • Communal life
Also the struggle for survival supersedes all other • Social Networks
development needs of the people. Their construct of a and Connections
Community
sustainable community is: • Structures affecting Practice
• A place where everyone (who shares the same heritage, individual actions
language, values and ethnicity) has an equal opportunity to • Development
participate and contribute. priorities
• A place where citizens shared strong community values,
where their relationships, community leadership structure
and Indigenous customs and culture are preserved for • Access to
future generations. Information
Community
• A community where there is use of ICT with local content • Information sharing
Information
and language, and where people (especially the youths) mechanisms
and
have the skills and capacities to use ICTs effectively. • Constraints to Communication
• One where there is access to information and education. information Access
• Information needs
• A community where there are income generating activities
and employment opportunities.
• A community where there are functioning schools and
• ICT Awareness
health centers.
• Local appropriation
• And people are able to provide for the daily sustenance of of ICT
their families. ICT Use and
• Individual’s Impact
exploitation of ICT
So the sustainable development priorities of these • ICT impact
communities differ significantly from the accepted models for
sustainable communities. These communities also have social
structures that required participation from representative
groups like the youths, women and men associations. The Fig.2. Analytical Framework
traditional leadership structure is very vital in the defining
community laws and future development agendas. There was VI. COMMUNITY PRACTICE
a strong sense of ownership and identity among community
A. Communal life
members which are crucial elements in building healthy
Life in these communities is shaped by three major aspects:
communities.
the conventional face-to-face interaction, villagers’ interest in
With a better understanding of what ‘community’ and
participating in communal issues and the ways and channels
‘sustainability’ meant the people, the next phase of the
for information exchange. A well-built social interweave
research was to study the communicative ecologies of these
where everybody knows one another was observed. This
communities in order to be able to explore how ICTs maybe
network of connections creates the rules of reciprocity,
deployed to meet the sustainable development needs of these
epitomized by membership in community organizations which
communities.
is a platform where people exchange experiences, stories and
every kind of information that is of villagers’ interest.
B. The Communicative Ecology
Communal life is also dictated by the leadership and
374
traditional values of the community. Societal norms are set by careful in their actions to protect this family reputation.
the village head, although most have been handed down by a Another factor that affected individual action was the desire
previous generation. There is a very strong respect for their by the people to ensure community support in times of need
community laws and they believe that obeying these laws is and to avoid punishment and enforcement of community laws.
important to the development of their communities. The There are societal expectations in these communities that
communal life in these communities supports active guide the choices people make; they are expected to enhance
community involvement and mobilization for actions that may the good norms of the society and to serve future generations.
lead to the development of their community. The sense of Membership in community fraternities, associations also has
responsibility and identity the people have is very strong. an influence on how people conduct themselves in this
community. For example there are community laws that
B. Social networks and Connections penalize women for certain offences that do not apply to men.
Communities can be understood as ongoing systems of For example insubordination and adultery. There are also
interactions, a set of social identifications and interactions. certain community development meetings that women cannot
Social networks build and support the development of participate in, like meetings to resolve inter community
community capacity and individual agency critical to social conflicts.
development. The conceptualizaion here, refers to peoples
connections with other people and the community. The idea is D. Development priorities
to focus on how the structure of ties affects individuals and The kind of development people want plays a role in what
their relationships. The shape of a social network helps they would commit to in terms of sustainability. Development
determine a network's usefulness to its individuals. In the that does not meet the people’s needs may not get the required
Niger delta it was observed that the people have very close support. Preserving indigenous knowledge and values is very
connections to family and have strong family ties. The important to the sustainability of these communities. Majority
implication of this is that their actions are greatly influenced of the respondents felt that their local self government and
by their interaction with family members. Outside of the community leadership structure was a positive development
family, the church, community groups and associations that if preserved would benefit future generations because it
provide avenues for social interactions. Types of information had proven effective in running their communities. The
exchanged with their social network are about community observation and interviews revealed that they value their
events, happenings in the community, community traditions and communal norms very highly. As one
infrastructural developments, business, financial challenges respondent put it “our traditions and values are what makes
and politics, personal and family issues. The young men have us unique, that is our identity” These traditions have been
the widest network of connections within the community and passed down to the younger generation by inculcating it in
they are key information carriers within the community. This them when they are young. Community associations also
is because it is acceptable and even expected for them to reinforce these traditions. Social, educational and economic
circulate and interact more. They also act as vigilante for the development was highlighted as the development most desired
community. in their communities. That is the way people interact with
A sense of belonging and the concrete experience of social each other for the advancement of their communities.
networks and the relationships of trust that are involved in Education was also highly desired. In a rural community like
these connections bring significant benefits. However, the Mbiaya Uruan where poverty is high, many children could not
sense of attachment and quality of social networks varies get further than a primary education and because of this they
between different people. It could be argued that for grow up not having many opportunities and many of them
sustainability to take place in a community we should be migrate to the city centers where there are better opportunities.
focusing on enhancing the quality of social networks etc. The youths prioritized human development, capacity building
rather than the creation or strengthening of ‘community’. This and technological advancement as most important for
is the line taken by writers such as (Stacey, 1969). Exploring sustainability. Inconsistent power supply hinders their use of
this concept adds to the context creation of these communities. ICT and unless there was an attempt by the government to
provide consistent power supply, they did not see how they
could maximize ICT for their benefits.
C. Structures affecting individual actions

This concept explores what is responsible for individual VII. COMMUNITY INFORMATION AND
actions in the community, this adds to the causal conditions COMMUNICATIONS SYSTEMS
for people’s action which may prove useful in understanding The concepts under this category are comprised of
how these may facilitate actions for sustainability. contextual, mediating factors that affect access to information
Respondents expressed the desire to uphold the family in the communities. Within communities, people also live and
reputation within the community, so they were constantly
375
participate in what we could call an “information ecosystem.” information. Observations and the interviews revealed that
This does not just refer to the information that a community over 80% of these leaders are illiterate and totally out of touch
presently has available, but all the human and technological with what is happening on the global scene, yet the educated
elements through which people become informed. The ones in the community allow these leaders to preside over
“media,” of course, plays a critical role in the information them. People with mobile phones exchange more information
ecosystem, but they are hardly all of it. Numerous actors, both than those without. Local and state information is received
individual and institutional, contribute to the information from Radio and Television. Radio and town criers are the
ecosystem by creating, storing, disseminating, interpreting, most constant means of receiving information.
and consuming information. Understanding how and why a
community accesses information is critical in exploring the B. Constraints to information Access
role ICTs play in meeting the community’s developmental Gender is one of the factors that hinder access to
goals. information, the widening technological gap between the
sexes was observed and it is reinforcing traditional forms of
A. Information sharing mechanisms power dynamics and hierarchies between men and women. A
Interactions and communication is a vital part of Niger large proportion of women in the region don’t have the
Delta communities. Most communication in the community is capacities to use ICTs. ICT solutions need to mainstream
done through mobile phones and face to face interactions. In gender perspectives into any initiatives to ensure equity.
the rural sites communities share and disseminate information The regions integration and participation in the global
using town criers or announcers and community meetings. information highway is also constrained by factors such as the
One of the key ways of receiving information is through a high cost of access, high Internet access costs, low bandwidth,
town crier, who disseminates community information about poor ICT infrastructure, Inconsistent power supply and often
road repairs, meetings and new laws to community members. unreliable communication facilities.
The information passed by the town crier is by far the most Traditional Leadership can also be a constraint, the
important information of all to the people. Information is also discretion of the leaders affects information access, as stated
received through heads of family. Information from earlier they filter the information coming into the community
neighboring communities is passed on by ‘Okada’ and if it is deemed not suitable, it is withheld from the people.
(commercial motorcycles) riders in the community. Within In oral cultures, the collective memory and importance placed
these communities there are no public transport systems. on the elders to store information creates a strong system for
There are intercity buses that drop people off at the entrance information flow (Slim & Thompson 1993). There is a need
to the communities, but ‘Okada’ riders are the main means of for mediation between the traditional and emerging
transportation within the communities. Because of this they information systems when considering the socio-cultural and
have access to all sorts of information which they pick up economic leap that will be required for societies, accustomed
from the motor parks in town and from the passengers they to receiving information orally from a known and trusted
carry, so they have become key information carriers for their source, to new digital, text based information sources. Lack of
community. See Fig. 4 below. education was another barrier observed, the illiterate ones in
the community where not given certain information because it
is believed it would be of no use to them. For example
information from the Niger Delta Development Commission
that could benefit the whole community is told to only the
leaders and the educated ones. So there are people that are
totally excluded from happenings within the community, the
deployment of ICT could further isolate this socially excluded
groups.
C. Information needs
A number of factors affect how different members of the
community may understand or use information, such as
Okada
gender, economic status, literacy, etc. Each community has
it’s own way of communicating and finding out about what is
going on in their area and outside. The information needs
Fig. 4 Information Flows identified by students and workers were: Latest discovery in
science and technology, Events around the world,
Community leaders filter certain developmental information Entertainment information i.e. music, creative arts, Local and
that comes in and transmit information they consider useful to state news, Health news and information; Social events; Sport
community members. This goes to show the kind of influence news; Business information i.e stock news; Job opportunities
that the leaders have on the community’s access to and Information on community development. Traders and
376
market women are interested in product prices. The elders and desired. Internet use is very limited as stated earlier, because
leadership were more interested in information that has to do cyber cafés are located outside the community and travel costs
with community development and politics. is included with airtime costs. Most people have never used
the internet and can’t use computers but have the perception
VIII. ICT USE AND IMPACT that it is very costly and don’t even try.
The investigations revealed that a combination of old ICTs
ICT Media Use in Anua
(Radio and Television) and new ones like mobile phones,
computers and internet (very limited) were used in the Niger 30
delta region of Nigeria. The impact and use of these ICTs was Television
25
explored in this category to identify the factors that influence
Mobile Phones
the adoption and use of ICTs and identify tangible economic 20 Radio
No. of People
Mobile Phones
and social benefits arising from having access to and using 15
Radio
Internet
ICTs. Non-internet Non-internet computing
10 computing
Internet Television
A. ICT Awareness 5
Computers are viewed as a symbol of modernity. A lot of
0
the older people were quite intimidated by the prospect of 1
using one. People under the age of forty said ICTs are Media Used
important for their individual lives and their communities.
Some young people saw it as a tool that helps learning; some ICT Media Use in Mbiaya Uruan
others view it as a means of getting beneficial information. 25

One respondent said “it can give everybody information about Television
the events happening in different communities and how they 20
Radio
overcame such problems so the community can learn from it” Radio
No. of P eople
15 Mobile Phones
They also thought it can be useful in building capacities and
Mobile Phones Internet
personal growth and can also provide an opportunity for 10 Non-Internet Computing
Non-Internet
wealth creation. Some cited the fact that many people now Computing
Television
earn an income from running calling centers, so could also be 5
Internet
income generating. These calling centers are usually little
0
tents that are constructed by the road side where people make 1
local and national calls from mobile phones for as little as N25 Media Used
(1p) per minute, and international calls for N75 (3p) per
minute. These calling places are all over the place on almost Fig. 3 ICT use
all street corners. Profits are marginal and competition fierce,
but people eke out constant income from offering these C. ICT Impact
services. This concept explores the interaction between Community
Information Systems and ICT Tools. The study showed that
B. Individual’s exploitation of ICT ICT further reinforces the social structures of these
There were no fixed land lines in any of the communities communities. The greatest area where ICT is affecting
studied, so all calls are made with mobile phones. community systems is in strengthening social networks, social
The graphs below show ICT usage in two of the interactions have been greatly enhanced by using mobile
communities studied. Mobile phones and Radio were the most phones, it is getting more people involved in the happenings
affordable and most used. In more rural communities like in the community. ICT use is also strengthening the sense of
Mbiaya Uruan radio is a major source of Information. There identity people have with their communities, there are
were many people interviewed who owned televisions and indigenes that work and live outside of their communities and
some times because of power outage, would get to watch it radion and mobile phones are keeping them connected to
only once a week, and in cases where the outage is due to happenings in their communities, because as revealed from the
faults, they would go for months without watching TV. Radio findings of the preliminary study Niger Delta citizens feel a
has become a primary source for getting information on local, sense of identity with their communities of origin, and are
state, national and international news and events. Radio is committed to the development of their indigenous
considered cheaper to run because it is battery operated, no communities. In this way ICT is helping to strengthen these
power supply needed, there is a huge supply of radios and low bonds.
level batteries imported from china that is readily available in ICT use is also impacting the development priorities of
most corner shops. Phones are also affordable, portable and community members, because it connects them to the rest of
mobile, although not as cheap as radio but definitely highly the world. This connection has positive benefit for the
communities. They are becoming more aware of global issues
377
like HIV/AIDS, environmental degradation and human rights • Preservation of Indigenous cultures
and have mobilized themselves to address some of these • Job creation and skills acquisition
issues. The empirical data provided possible directions for
ICTs have also had an impact on the security of citizens in exploring how ICTs may be employed to contribute to
these communities. The Nigerian Government is very sustainable development for the Niger Delta people. ICT can
ineffective in protecting their citizens from crime. So citizens play a role in facilitating sustainability by:
have been forced to find ways to fill this gap by defending and 1. Strengthening community systems that enable cohesion
protecting their own communities and ICT is playing a crucial and collective action
role in that. Mobile phones are being used effectively by 2. Providing appropriate information on sustainable
vigilantes for security. development practices through:
The following is a summary of areas where ICT is • Public awareness campaigns and social reorientation on
impacting communities in the Niger Delta development issues;
• Community mobilization • Education;
• Community watch • Provide access to information,
• Circulates information faster 3. Electronically documenting indigenous knowledge and
• Connects people together, Fosters good relationships practices for now and future generations.
• Income generation
X. CONCLUSION
The study shows that ICT is being integrated into the social The study provided a better understanding of how the use
fabric of the community, especially mobile phones. of technology affects relationships within the community and
Computers and the Internet are not as yet having any real how individuals use technology to develop their relations with
impact in the communities. There is still a long way to go others in community.
before ICT can begin to meet the information needs of these The study also showed the importance of mobile phones
communities, because they do not have input on content, this and radio in changing the lives of the people. Mobile phones
is designed by others who may not be sensitive to needs of the is by far the most common communication technology to
people. These initial positive development of ICT use shows effect tangible positive change in these communities (market
that there is great potential for ICT, if appropriately deployed and trading information, emergency and security
to meet community information needs. communications, strengthening kinship relations and social
interactions) and is the backbone of ICTs.
At present there is a vast unmet demand for radio
IX. IMPLICATIONS OF ICT IMPACT ON broadcasting with local content in these communities. A
SUSTAINABILITY combination of mobile technology and rural broadcasting will
Incorporating sustainable development at the local level is enable information and communication services reach more
one the biggest challenges facing the movement towards people than any other medium.
sustainability, how to take global principles and make them Overall, the study has provided a better understanding of
concrete locally. Communities which are culturally rooted, the use of technology in community life and presented
locally produced and technologically adapted are being opportunities of how ICTs may be deployed to help these
rapidly eroded. Vast literature frame the sustainability communities towards sustainability.
discourse largely in terms of an environmental agenda pre-
occupied with ‘green’ issues. Literature has largely neglected REFERENCES
the community development as sustainable and livable places [1] Avgerou, C. (2001), The Significance of Context in Information Systems
which adapt their unique cultural identities and specific and Organisational Change. Information Systems Journal 11:43-63.
historical heritage to contemporary needs. Analyzing the [2] Barton, C. & Bear, M. (1999) Information and Communication
Technologies: Are they the key to viable business development services
cultural environment and information ecosystem is a method for micro and small enterprises? Report for USAID as part of the
of understanding the complexity of a given community, its Microenterprises Best Practices Project. March 1999 by Development
culture, its fragility, resilience and it’s reaction to change. Alternatives Inc, MD, USA
[3] Beck, U. (1992) Risk Society: Towards a New Modernity. New Delhi:
The factors relevant to sustainability of communities Sage. (Translated from the German Risikogesellschaft published in 1986
identified from the research are: [4] Bell, S. (2003). Measuring Sustainability: Learning by Doing. London:
• Awareness about sustainable development; Earthscan Publications Ltd.
[5] Braga, C. P. (1998). Inclusion or Exclusion, Information for
• Information availability and thus transparency; Development (InfoDev), The World Bank,
• Public participation in governance; http://www.unesco.org/courier/1998_12/uk/dossier/txt21.htm
• Empowerment of citizens, especially women; [6] Castells, M. (Eds.) (2000) The information Age: Economy, Society and
Culture, Oxford: Blackwell, 3 volumes
• Fostering of cultural diversity; [7] Cohen, D., and Prusak, L. (2001). In good company: How social capital
• Building capacity makes organizations work. Boston: Harvard Business School Press.
[8] Chambers, R. (2003) Whose Reality Counts? Putting the First Last.
• Social cohesion Intermediate Technology Publications, London.
378
[9] Commoner, B. (1993). Population, Development and the Environment: [34] Redclift M. and Sage, C (Eds) (1995) Strategies for Sustainable
Trends and Key Issues in the Developed Countries. In Population, Development – Local Agendas for the South. John Wiley & Sons,
Environment, and Development. New York: United Nations. Chichester
[10] Cooper PJ. and Vargas CM, (2004) Implementing Sustainable [35] Rodgers G. et al. Social Exclusion: Rhetoric, Reality, Responses,
Development: From Global Policy To Local Action. Rowman and International Institute for Labour Studies, Geneva, A contribution to the
Littlefield, Lanham, MD World Summit for Social Development, 1995.
[11] Duncombe R., and Heeks R. (1999) Information, ICTs and Small [36] Selman, P. 1996. Local Sustainability: Managing and Planning
Enterprise: Findings from Botswana. Manchester: University of Ecologically Sound Places. New York: St.Martin’s Press.
Manchester, Institute for Development Policy and Management, [37] Slater, D., Tacchi, J and Lewis, P. 2002. Ethnographic Monitoring and
Working Paper 7.<http://idpm.man.ac.uk/idpm/diwpf7.htm> Evaluation of Community Multimedia Centres: A study of Kothmale
[12] Felleman, J. (1997). Deep Information: The Role of Information Policy Community Radio and Internet Project, Sri Lanka. London: DFID.
in Environmental Sustainability. Greenwich, CT: Ablex Publishing [38] Slim, H. & Thompson, P. (1993) Listening for a Change: Oral History
Corporation and Development. London. Panos Publications.
[13] Flamm K. (1987) Targeting the Computer, New York, the Brookings [39] Stacey, M. 1969. The Myth of Community Studies, British Journal of
Institute Sociology, 29, 134-147
[14] Friedman, T.L. (Eds.) (2006) The World is Flat: The Globalized World [40] Taachi, J. and Slater, D 2003. Modernity under construction:
in the Twenty-First Century, UK: Penguin Comparative ethnographies of internet. Amsterdams Sociologisch
[15] Giddens, A (1984) The Constitution of Society: An Outline of the Theory Tijdschrift 30, No. 1-2 Special Issue Digital contact. pp 205-222
of Structuration. Polity, Cambridge. [41] Tacchi, J , Slater, D., and Lewis, P. 2002. Ethnographic Monitoring and
[16] Gumucio DA. (2001), Making Waves: Stories of Participatory Evaluation of Community Multimedia Centres: A study of Kothmale
Communication for Social Change, Rockefeller Foundation. Community Radio and Internet Project, Sri Lanka. London: DFID.
[17] Hamelink C.J (1997) New Information and communication [42] Tacchi, J, Slater,D and Hearn, G. 2003 Ethnographic Action Research: a
Technologies: Social development and Social change, Discussion paper User’s Handbook (New Delhi:UNESCO,
No.6, Geneva: UNRISD. http://cirac.qut.edu.au/ictpr/downloads/handbook.pdf
[18] Heeks R (1999) Information and Communication Technologies, Poverty [43] Tönnies, F. 1957. [Gemeinschaft und Gellenschaft] Community and
and Development. Development Informatics: Working paper Series. Society. East Lansing, MI, Michigan State University Press.
Paper No 5, June 1995 Institute of Development Policy and [44] UNDP (2001), Making new technologies work for human development,
Management. Published on the www: http://www.man.ac.uk/idpm/idpm Human Development Report 2001, UNDP, Oxford.
[19] Heeks, R. (2001) What Did Giddens and Latour Ever Do For Us? [45] Walsham, G. (1993) The Emergence of Interpretivism in IS Research
Academic Writings on Information Systems and Development', Information Systems Research 6(4): 376-394.
Information Technology in Developing Countries , 0, Vol.11(1) [46] Wade, R. (2002) Bridging the Digital Divide: New Route to
[20] Hilty, L. M., Seifert, E. K., Treiber R. (2005) Information Systems for Development or New Form of Dependency? In Global Governance,
Sustainable Development, Swiss Federal Labs for Materials Testing & Vol.8, No. 4 Oct. Dec.2002
Research, Switzerland; Wuppertal Institute, Germany; Hochschule [47] Woodhouse, P (2000) Environmental Degradation and Sustainability. In
Niederrhein, Germany. Allen, R &Thomas, A (Eds) Poverty and Development – Into the 21st
[21] Hoffman, K. (1985) Microelectronics, International Competition and Century. Oxford University Press, Oxford
Development Strategies: The unavoidable issues- Editor’s introduction. [48] World Commission on Environment and Development. (1987). Our
World Development 13(3): 263-272 Common Future. The Bruntland Commission, UNESCO.
[22] Huysman, M., and Wulf, V. eds. (2004). Social capital and information [49] WSIS (World Summit on the Information Society) Civil Society Plenary
technology. Cambridge, MA: MIT Press. (2003) Civil Society Declaration to the World Summit on the
[23] Kling, R. (2000) Learning about information technologies and social Information Society
change: the contribution of social informatics, The Information Society , <http://www.smsitunis2005.org/plateforme/pdf/civil-society-declaration-
Vol. 16 No.3 en.pdf>.
[24] Korten D.C (1995) When Corporations Rule the World, Kumarian Press.
[25] Liemgruber W. and Imhof G. (1998) Remote Alpine Valleys and the
Problem of Sustainability. In Andersson, L & Blom, T (Eds)
Sustainability and Development - On the Future of Small Society in a
Dynamic Economy. Universit of karlstard research report 98:8
[26] Madon, S. (1997) The Information-Based Global Economy and Socio-
Economic Development: The Case of Bangalore, The Information
Society, 13 (3)
[27] Manning E.W (1990 Conservation strategies: Providing the vision for
sustainable development/ E.W. Manning Environment Canada, [Ottawa,
Ont.] : Alternatives 16, 24–29
[28] Nurse K, 2004, Culture as the Fourth Pillar of Sustainable
Development.
www.fao.org/sard/common/ecg/2700/en/Cultureas4thPillarSD.pdf.
Assessed 15/01/09
[29] O’ Farrell, C. (2001) Information Flows in Rural and Urban
Communities: Access, Processes and People. IRDD, The University of
Reading .
[30] Pamlin, D, ed. (2002). Sustainability at the Speed of Light. Solna,
Sweden: WWF Sweden, [Online] Available:
http://www.panda.org/downloads/general/ict_sustainability.pdf.
[31] Panos (1998), The Internet and Poverty. Panos Media Briefing, No. 28,
Panos Institute, London.
[32] Pohjola, M. (June 2002). The New Economy: Facts, Impacts and
Policies, in Information Economics and Policy.
[33] Radermacher F.J (1998) Intelligenz - Kognition - Bewußtsein:
Systemtheoretische Überlegungen, technische Möglichkeiten,
philosophische Fragen. In: Interdisziplinäre Beiträge zur
Kommunikation und zum Mensch-Technik-Verhältnis (C. Stadelhofer,
ed.), Band 6, S. 146-193, Kleine Verlag GmbH, Bielefeld (1998)
379
Integrating Health Information Systems in

Sierra Leone
Johan Sæbø, Edem Kwame Kossi, Romain Tohouri Golly-Kobrissa, Ola Titlestad, Jørn Braa
health system, suffering from a huge loss of both personnel

This paper presents an ongoing project in Sierra Leone to and infrastructure during the war, is slowly rebuilding the
integrate health information systems at district and national level capacity to improve the service provision across the country.
through a novel approach. Employing solar-powered low-energy This effort is supported by the government and the
computers running Linux, a wide consortium of local and international community through many agencies such as
international actors have tried to counter the severe problems of
UNAIDS, WHO, UNICEF, GTZ etc. and aims at achieving
electricity supply breakdowns and computer viruses. The paper
discusses the experiences from this effort, as well as the
health millennium development goals. The rapid growth of
integration process itself, and the corresponding capacity various health initiatives has created a situation of fragmented
building strategies. The findings so far suggest that alternative information systems, common also in other developing
technologies, namely solar power and open source software, can countries [1, 2]. An initiative to counter this fragmentation by
be fruitful to apply in such infrastructural settings as Sierra integrating and strengthening the HIS together with using
Leone presents. Furthermore, the technical solution to an novel ICT solutions to cope with the extreme infrastructural
intermediary step towards integration shows some promising challenges has been initiated and piloted over the last year.
results. This initiative is supported by the Health Metrics Network
(see http://www.who.int/healthmetrics/) together with other
Index Terms—Sierra Leone, health information systems,
agencies such as the World Bank and UNAIDS.
integration, solar power
In order to address the fragmentation, an integrated district

based data warehouse approach has been followed. As a first
I. INTRODUCTION
step, data reported from the health facilities to the district
using various health program specific reporting formats and
I N this article we describe and discuss a project to develop
an integrated Health Information System (HIS) in Sierra
systems are all being captured in one database application, the
District Health Information Software (DHIS version 2, from
Leone, which has been going on since early 2007. The article now referred to as DHIS). Unified processing and analysis of
focuses on the key challenges facing the project; these previously disparate information flows have then made
1) Building agreement on the need for an integrated it possible to better assess the quality of data and
approach to HIS in order to solve the prevailing fragmented discrepancies between the different data reporting formats. As
situation of multiple vertical reporting systems which have no a result overlaps, gaps and inconsistent definitions of data
coordination and shared data standards – and building a variables between the different reporting formats have been
consortium to carry out the task. identified. As a result of this, a revision of the data collection
2) Handling the extremely poor infrastructure in Sierra tools within an integrated framework is being planned.
Leone by establishing a “new” solar powered computer Furthermore, the district based database application supports
infrastructure. unified data management, data disaggregating (e.g. make it
3) Developing a free and open source database application possible to “follow” the data from the national level down to
which may handle the problem of interoperability with the health facility where it is collected), data analysis,
existing systems and include and manage data from the programmatic reporting, changes in health units organisation,
various health programs and data sources. and local system integration.
4) Capacity development; how to establish a training
scheme and a support structure sufficient for rendering the A number of paper based reporting formats and routines are
new computerised system sustainable. currently in place in Sierra Leone Some of these are also
computerized, such as HIV/AIDS reporting, which consists of
Ravaged by a prolonged civil war, which was declared over paper forms reported from the facilities, and a database
on 18 January 2002, Sierra Leone consistently scores application called Country Response Information System
extremely low on human development indicators. The public (CRIS, from UNAIDS) located in the districts, where the
HIV/AIDS reporting forms are captured. One of the goals is to
.
380
establish interoperability and data exchange between DHIS be more or less integrated, or aligned. Communication and
and CRIS and other software. In order to achieve this interaction between the components of the network may be
integration, many other critical issues such as infrastructure regarded as going through gateways. Integration can then be
that the system will rely on and human capacity need to be perceived as software exchanging data through gateways, but
addressed. The national power supply system is extremely this can also mean integration between paper based and
poor, as many other national infrastructures. The main power computer based components. An integrated HIS allows data to
source of the existing computerized HIS are generators. As the be collected and analysed at one point and to be shared among
government is not able to provide fuel regularly to run the modules or parts of the system. This makes the data
generators, this reliance on diesel contributes to undermining analysis more meaningful as all data is analysed together. It
the system. Human capacity both in terms of data analysis and also entails knowledge integration as well as solving
information use and IT skills is another key issue that has to challenges of political and economical nature [5], [3]. This
be tackled. Many information/IT officers (called M&E process has shown to be non-trivial [6], especially in
officers; Monitoring and evaluation) at district level have developing countries [3]. The fragmented HIS leads to poor
received on-the-job IT training, but because of power quality health information which can be rendered useless.
shortage, computer troubles (viruses, damaged and old Integrated HIS is therefore relevant, but it does not solve all
computers) and absence of follow-up, the training endeavors the problems of HIS in developing countries.
were not capitalized on and they are still unskilled or
uncomfortable in IT. III. METHODOLOGY
The research enveloping the project in Sierra Leone has
The rest of the article will present the case of Sierra Leone been carried out along the lines of Action Research (AR). AR
by focusing on the challenges described above. Some is a form of participative research where the researcher takes
background to HIS and integration, as well as the methods part in the change processes in an organization, actively trying
applied, will be described first. to improve some stated problem [7], [8]. While it was initially
not used much in IS research, and had a lot in common with
II. BACKGROUND: HIS AND INTEGRATION anthropology and social studies, it has over the last two
Fragmentation and poor quality and use of data are major decades increasingly become accepted as a way to generate
problems with health information systems in African very relevant research findings on technology in its human
countries. Integration of information systems is often context [9].
perceived as a technical task involving primarily incompatible
software and infrastructures. While this is an important part of While we acknowledge the critique often raised against
the picture, fragmentation and poor coordination between using AR in the field of IS (or any field, for that matter), we
organizations, and, as in the case of health, between providers also claim that we, through our active participation, gained
of different services, together with political and social aspects knowledge we would not easily come by taking a more
more generally, are as important [3]. This fragmentation leads passive role. Baskerville and Wood-Harper provides an
to gaps and overlaps, and incompatible definitions in the data overview over common critiques of AR. However, they
that is being collected, registered and reported. Experiences conclude that the traps found are not unique for AR as a
from other African countries, as well as from Sierra Leone, specific method, but are likely to trouble a researcher using
show that without shared data standards, data exchange cannot any method in social science [9]:
take place. Integration is therefore first of all about data
standardization. However, while the practical system focus of “A number of problems confront the action researcher such
integration is on standardization, the overall integration will as lack of impartiality, lack of discipline, confusion with
need to involve and enroll as many as possible of the various consulting and its context-bound nature. However, these
actors in the health system in the process. problems confront researchers using alternative methods as
well. The difficulty with action research may be a matter of
Over the last decade the Health Information System degree, and the easy loss of scientific rigour” (ibid, 144)
Programme (HISP, www.hisp.info) has addressed these
problems. The DHIS software was successfully used for The authors have been an active part in the consortium
integrating data and health services in first South Africa, and described in this paper. Through this role, we have been able
later in many other countries [4]. Integration of data and to engage in a close collaboration with users at different
interoperability between information systems as well as levels. Our findings originate in analysis of our work,
increased use of information are key issues in the HISP including software development and customization, database
approach. development, implementation of pilot sites in Sierra Leone,
training of all monitoring and evaluation officers in the
The health information infrastructure may be regarded as country over 6 weeks, on-site training in the pilot districts, and
heterogeneous networks of actors (Latour 1987), which may discussions with district personnel to clarify data flows,
381
reporting, work practices, and the like. The scientific rigour, as interviews and observationary techniques, and carefully
to quote Baskerville and Wood-Harper, has been upheld by recording the findings.
applying qualitative research methods in the daily work, such
IV. PREVIOUS SITUATION

Table 1 summarizes the strengths and weaknesses of the program (only catering one health program) software). This
previous HIS, and Figure 1 depicts the overall fragmented fragmentation is replicated at the national level, with
HIS in Sierra Leone. At facility level, data is captured on various standards for data storage. The result is poor
multiple paper forms and registers. This is sent to the accessibility of data, overlap between the many databases,
districts, where the data is entered into various software and hence less use of data for processing and analysis.
applications (CRIS, Excel spreadsheets, and other single-
Table 1 Strengths and weaknesses of previous HIS

Strengths Weakness
• Clear and understood reporting • Poor IT skills of data management teams at all levels
procedure • Lack of power supply
• Lack of resource to fuel regularly the generators
• CRIS used in all districts
• Lack of IT support and recurrent viruses problems
• Computers in all districts • multiple and not well coordinated tools and formats for data reporting from
• Information awareness and facility to district, with overlaps and gaps, are in use;
computer usage very good in • Multiple computer based tools in use, also with overlaps (same data captured
one of the districts visited; in different systems), and they are not linked
multiple spreadsheets designed • Data analysis and local reporting carried out in districts by re-capturing data in
and used for particular program excel spreadsheets
activities, up-to-date graphs on • Data aggregated to district totals before reporting to national; making quality
program activities displayed on check of data difficult
the wall. • No feedback on reports from national to district;
• Reporting from hospitals very poor (e.g. incomplete, poor coverage);
morbidity and admissions/discharges reported, but not time spent by patients
in hospital (laying days/bed-days), resources utilisation therefore not possible
to deduct
• Anti RetroViral therapy (ARV) patients managed by not-optimal paper record
system; very difficult to provide outcome and cohort indicator reports based
on current system
382
National: Fragmented reporting; gaps & overlaps

Data sources not linked CRIS
Exce
Exce Facility
survey
ICS SU (hospitals- poor)
Other Exce
data
sources
Exce Exce
Data Exce
CRIS
District
Fragmented
Data management SU Data
ICS Exce
Exce
Data
Summary
reports
Compiled Summary report

Summary report monthly reports Data
Facility Mor Hospital AR

ARV patient PMTCT Othe
Multiple IC
idit .
Paper
& registers In & out patients

Records / registers
Fig. 1. The previous Health Information System in. Sierra Leone was characterized by fragmentation and overlap between data collected by the different actors
DHIS. But at the same time, data can be imported from

existing applications into DHIS. So instead of entering
V. PROPOSED SOLUTION overlapping data at district level in many databases, now data
From the current fragmented HIS described above, the main will be entered in one application (DHIS) and then exported in
aim was to provide meaningful and relevant information for appropriate format to other applications. This model brings
decision making and to diminish the workload of staff who is together essential data and indicator sets from all relevant
collecting and reporting the data. The key concepts used to sources disintegrated at the level of the reporting unit.
address this issue are “integration” and sustainability. The Revision, harmonization and standardization of data and
suggested way forward depicted in figure 2 is to use DHIS to indicator sets and tools for registration, collection and
integrate the various data flows and data sources and thereby reporting of data are part of this process and particular
to provide an integrated framework for M&E and data attention must be given to the infrastructure that support the
management. The approach is to design data entry interfaces system as well as capacity building.
that are similar to existing ones in order to enter all data in
383
Figure 2: Dissemination Health &

Census Other
& sharing; data
Demogr.
Vision - surveys data
Reporting, indicators sources
Integrated & M&E, GIS/ Maps
Linked data web-dissemination Gateway:
Sources DHIS CRIS Data extract
Moved to & & integrate Vital
All information Managed by DHIS statistics
available through
“one point” / Facility
surveys Gateway: Other
Data repository Excel Data extract & Epi-Info
Services
ICS
integrate statistics
National level
1. Export data
District level Captured by
Dissemination DHIS
in districts; Other
Gateway: Epi-Info
Reporting, DHIS CRIS Data extract ICS
Services
statistics
indicators & integrate
M&E, GIS/ Maps Data Feedback 2. Import data Data
capture Captured by capture
ICS Epi-info etc
Facility level Feedback reports:

M&E, indicators,
Supervision
District meetings
assessing country HIS (see

http://www.who.int/healthmetrics/). In this project, HMN
VI. INTERVENTIONS plays the role of executive sponsor and project manager.
Health Information System Program – HISP: is a global
A. The consortium : network in HIS development, open source software, ICT for
Once a decision was made to improve the HIS in Sierra Development and research and education. It comprises several
Leone based on the findings of the initial assessment, the actors from a number of countries and has been successfully
Ministry of Health (MoH), the Health Metrics Network implementing sustainable and integrated HIS in developing
(HMN) and HISP had organized several meetings during countries. The software named District Health Information
which many questions regarding project aims were addressed. Software (DHIS) is developed within the HISP network. HISP
In doing so, some key actors were identified and enrolled in started in South Africa 1995 and has been instrumental in
an alliance of complementary and experienced actors to run developing the HIS there. Since 2000 HISP has developed
the project: into a global South-South-North collaborative network with
Ministry of Health of Sierra Leone: is represented at all nodes in South Africa, India, Tanzania, Nigeria, Mali,
levels of the project purposefully to make its participation and Vietnam and Norway (see http://hisp.info).
ownership very strong – as manager, user and beneficiary. Its ICT Development Center – IDC Sarl : is an international
experience in health management in the local context and its IT company based in Mali, specializing in open source and
good understanding of what is needed both as a supplier and eHealth solutions for Africa. It offers eHealth consultancies,
consumer of health information was crucial. eHealth software development such as telemedicine
Health programs: it is useful to look at health programs applications and has a strong experience in HIS and IT project
within the MoH as separate entities, as they are key actors of implementation in developing countries. IDC is also member
the integration process. They have their own HIS, which is the of HISP and has contributed in the design and implementation
main reason for the current fragmentation. Some have in Sierra Leone. The task in Sierra Leone is to provide
computerized HIS and some not. They are autonomous vis-à- regional support and help enabling the local core team assume
vis the MoH. Examples: Mother & Child Health, EPI full ability in running the system.
(immunization), HIV/AIDS, Tuberculosis program Inveneo: is a US IT company specialized in open source
Health Metrics Network –HMN : is a global agency under and low power computing and has deployed low power
the WHO and aims at strengthening developing countries’ infrastructures in many developing countries. Inveneo’s
HIS. They have developed a HIS framework and tools for strategy is to build alliances with local IT companies that can
384
be empowered in order to provide support for the low power competency and responsibility, two pluses indicate substantial
equipment and avoid expensive and problematic traveling competency and involvement, while one plus indicates some
from US. involvement only. This only relates to the roles in Sierra
The attracting characteristics of this consortium lay in the Leone, specifically). This overlapping contributes a lot in the
diversity and the synergy of the actors enrolled. As depicted in common understanding of the problem in its context.
table 2, their expertise cover all fields requested and are
overlapping (Three pluses indicate that this is the core
:
Table 2: Expertise of the consortium members
Actors Politic Policy Health HIS Open Hardware
making source
MoH +++ +++ +++ ++
Health programs + +++ +
HMN +++ +++ + +++
HISP ++ ++ +++ +++ +++
IDC +++ +++ +++ +
Inveneo +++ +++
• The use of Ubuntu Linux OS makes the system virus
free and unlikely to corrupt database files. System
B. Infrastructure start or restart erases any changes in system files.
This reduces significantly the system maintenance
To face the challenging infrastructure context of Sierra
need.
Leone, a decision was made to deploy Inveneo low power
hardware running Linux operating system. In doing so, we use
• The server runs also Ubuntu Linux and hosts the DHIS
one stone to kill two birds: respond durably to the lack of
application. DHIS is then accessed remotely from
power, and counter computer virus.
other computers regardless its operating system.
The Inveneo hardware is an open hardware design (all plan
and software are published through the Inveneo website) that In each site one Inveneo desktop, one server and a mixed –
operates using open sources software. wired and wireless – local area network(LAN) were deployed.
Each site in the project was provided with an Inveneo All equipment deployed are based on low power. The 12v
computing station and an Inveneo hub server. These power is provided by a car battery which can supply the
computers are designed for challenging environments and system for 8 to 12 hours of working when it is charged and
have the following capabilities: can be charged by either solar panel or the existing generator
when this is turned on.
The LAN allows M&E officers to smoothly and seamlessly

access DHIS both from the existing MS Windows computer
and from the new Inveneo desktop. Existing computers can be
used only when the generator is running since they are not low
power based. But the Inveneo desktop and laptops can access
DHIS at anytime when the battery is charged. Thus both M&E
officers as well as the district medical officer are able access
DHIS over a LAN even in times where no external power
Figure 3 Inveneo server source is available. The server comprises two hard drives,
synchronized. In case of a problem, the currently running
• Very low power consumption (total power (now faulty) hard drive can be replaced by the second one as
consumption of 18 watts at peak operation including an emergency solution.
LCD monitor) allowing the units to operate
efficiently on 12 volts DC, with renewable power Inveneo unsuccessfully sought an eligible local partner that
from solar panel, generators, etc. could be trained to be responsible for the in-country support.
Nevertheless, the emerging local core team for the project
• Solid state design (fanless, driveless) eliminates fans,
described here came to constitute this role during the pilot
hard disks, or moving parts of any kind, improves
implementation phase.
durability in hot, humid and/or dusty environments,
making it ideal for use in Sierra Leone.
385
C. Software Development the DHIS software. Key principles of DHIS design strategy
The software application used for the new integrated HIS are; 1) a modular data approach focusing on data item as the
solution is called DHIS, and is an open source software atomic unit rather than the larger data collection forms, 2) dis-
application developed in the global HISP software aggregation of data enabling drill-down to health facilities
development community. Briefly the DHIS can be described from all levels, and 3) to collect and process data across health
as a tool for data collection and visualization of statistical programs on an integrated data warehouse. All these
data, tailored to support integrated health information principles, although arising from many years of
management activities. It is a flexible tool which allows for implementation in various other African and Asian countries,
extensive customization to meet local needs for HIS were highly relevant to the Sierra Leone context.
standardization of collection forms, reports and data analysis. A major problem in the previous solution that had to be
The process of customizing the DHIS to the Sierra Leone addressed in the DHIS was the duplication of data reporting
context has been ongoing since July 2007 and has gone from the health facilities, as the same data items was collected
through many feedback cycles between users and stakeholders in multiple forms. At the same time it was an important
in the field in Sierra Leone, the IDC team responsible for the requirement from the users that the computerised form in
local customization, and the developers in Norway. During the DHIS would look exactly like the paper forms they were
more intense periods the online communication (chat, e-mail familiar with. The software was customised so that all data
lists, issue tracking, source code repository) between Sierra items collected in the various paper forms were defined as
Leone and Norway has been so active that new builds individual elements in the database, but without duplication.
(versions) of the software have been released on a weekly In the DHIS these data elements are the atomic units and the
basis with rapid response to user feedback. Furthermore, the data collection forms are just visual representations of a given
innovative designs of feedback reports in Sierra Leone, collection of such elements, as illustrated in figure 4. While
created in collaboration with HMN, local M&E officers and the user interface for data collection maintained the look of
the IDC/HISP have pushed and inspired the global developers the fragmented paper system, the database behind was
to come up with a better and more generic report solution that organised so that no duplication took place. E.g. if a data
later has benefited users also in other countries. element was shared by two collection forms, then after
While the DHIS software is based on web technologies and registering its value in one form the value would automatically
can be used in networked settings, it also supports stand alone appear in the second form without a need to enter it again.
installations in locations with no internet. This adaptability to Another benefit of such a modular approach was that all data
a varying infrastructure has been especially important in elements are stored in the same integrated data warehouse and
Sierra Leone as there have been local networks in each district available for processing and correlation independent of any
set up around an Inveneo server, but at the same time no collection form. It was also possible to define calculated
network connectivity between most of the districts and the indicators based on these data elements to analyse coverage
national server has been present. DHIS has been installed on a and rates as opposed to the “raw” numbers. E.g. BCG
district Inveneo server enabling multiple desktops in the same coverage (rate of infants given the BCG vaccine against
building to access and use the application simultaneously. tuberculosis) could be calculated by dividing the monthly
This enables sharing of the district data warehouse and access reported data on BCG doses given, by the population estimate
to all data to all program managers, M&E officers and other under 1 year for the same area and thereby provide the M&E
users, as well as a secure server environment where the data manager with information comparable across time and space.
are protected on a virus free Ubuntu Linux system. On a A set of feedback reports were designed to provide districts
monthly basis each district exports its data to an external and chiefdom administrations with a summary report of last
XML file and sends it to the national level either using a dial- month’s data, which combined a few selected key data
up connection or using a flash-drive and physical transport. elements and indicators and used charts to visualise trends
over time, comparisons across areas, as well as an overview of
To address the problems of the previous situation of the the reporting completeness.
HIS, a new computerized system was introduced following
best practices from HMN and HISP, and implemented using
386
Figure 4 Overview of integrated data warehouse
training was analysed, interpreted, and presented, and actions

to improve the situation were discussed. A clear distinction
between the three modules were not drawn, on the contrary,
D. Capacity building
efforts were made to integrate them, for example by using
In each district, two M&E officers are working to collect, epidemiological data in excel spreadsheet training. But on the
prepare, report, and analyse aggregated data. During six whole, the capacity building was centred on the three
weeks in May-June 2008, extensive training was given for all following components:
M&E officers.
1) Using computers, to get the most of DHIS training
The training was given as a three-module course, each 2) DHIS operation, including regular tasks and basic
module taught one week. Given the authors earlier database maintenance
experiences with training similar positions in the use of the 3) HIS and data use, to create understanding and
same software in other countries, we opted for a one-week motivation. Using data from DHIS and analytical and
introductory course to the use of computers in general, presentation tools taught in the first week.
followed by one week of training in the use of DHIS and one
week in use of health information. In the end then, only 1/3 of This approach had a few benefits that justify the relatively
the training was actually in the use of the software in question. long time spent by M&E Officers away from office. First of
Two weeks extra training were seen as necessary to increase all, the basic computer skills taught in the first week were
the level of post-training use. First of all, few M&E officers essential to both understand how to use DHIS, and also to use
actually had any experience with computers at all. So one Microsoft Excel, Powerpoint, and Word, which are common
week was dedicated to lift everyone to the same basic level of to analyse and present data once entered and processed in
computer literacy. Second, sustainability can only be achieved DHIS. Furthermore, it enabled a faster progress during week 2
if the HIS actually serves a need. In a context of poor and 3, which meant that a few, inexperienced M&E officers
information understanding and use, the collection and analysis would not slow down those already familiar with computers.
of data will not be institutionalised. Therefore, the third week Secondly, special topics relevant for the context could be
was dedicated to public health administration using primary examined, with the students, in depth. The training schedule
health care data, and challenges and implications of a was revised daily based on the progress so far, and what the
supporting HIS. Data entered during the week of DHIS students wanted to learn. Motivation and inclusion played a
387
major role in forming the curriculum. For instance, while most center, they closed the books and did the aggregation for the
of the districts are not connected to the internet, we found time district health team a few days earlier. Other facilities just
for a session on browsers, free email hosts, search engines, filled in the form when they had time, and the exact day could
and how to connect with mobile broadband through one of the vary from month to month. Thus some forms could contain
local mobile network providers. Another example is the issue data from 23 days, some for 34 days, etc. The M&E officers
on computer viruses and possible solutions to the problem, being trained knew about these varying practices, although
which were explained in depth by, so to say, popular demand. they did not initially see this as a problem affecting data
inconsistencies.
Thirdly, and quite importantly, the extensive training was
used as an arena for user feedback to the system Lastly, efforts have been made to include the new
implementers. Especially the third week focused on plenary technology applied to the capacity building activities. This has
discussions on topics such as factors influencing the health been done by focusing on strong on-site support by local
status of the community, the validity and use of data forms, partners who have been given extensive training in the
and data quality. In addition, requirements to DHIS and report Inveneo technology.
designs were voiced during the training of the software. The
training was a dialectical process, in which both trainers and VII. DISCUSSION:
students learned from each other and converged on a common Our discussion builds on an ongoing project, and it would
understanding of the challenges ahead and the way forward be too early to draw many conclusions at this time regarding
for the project. A few examples can highlight the mutual sustainability and the long-term implications of the applied
learning experienced. approaches. Nevertheless, the project can already show to
interesting findings regarding the four topics presented in
One form, for filling in among other things stocks of depth earlier; the consortium, infrastructure and software
antigens at the facilities, was fraught with errors. Using the solutions, and capacity building.
validation rules we had set up in DHIS, we could catch these
errors after entering the data into the software. Running these
validation checks, on one occasion only 5 of the 21 M&E A. The consortium
officers found no errors in the form they had just filled in. The As we have seen above, the previous HIS was extremely
errors had two causes; one was wrong calculation of stock fragmented, but the solution proposed was not to solely
balance, another was a wrong interpretation of the data that provide a new system that aimed at integrating everything.
was to be entered. It turned out that the M&E officers, and The focus was not merely on the application but on the whole
most likely also the nurses filling in the forms at the facilities, information infrastructure, which comprises both technical
were not fully aware of what should be recorded; vials or and non-technical components. The diagnosis of the HIS went
doses. For example, BCG vaccines come in vials of 20 doses in depth and revealed that existing infrastructure and
each. If you don’t use all 20 doses at once, you are supposed applications as well as human capacity were weak and needed
to discard the rest, because rapidly they lose potency. So the to be addressed with equal attention.
data elements of Start Balance, Received, and End Balance
should be dividable by 20. Distributed and Discarded, Being mindful of users and collective participation, the
however, could be different, but should add up to vials of 20 strategy was to start with a prototype that would contain data
doses too. Counting the various data elements differently was from several existing data sources such as EPI, facility survey
quite common, however, and several M&E officers did not and population estimates. The prototype was then a useful
understand the correct procedures when the errors showed up concrete example of what and how the proposed integrated
in the software. A quite lively discussion erupted, only to calm HIS could be like. It acted as a boundary object and enabled
down when a representative from the EPI program at national actors participation that gave more input to the project [5]
level clarified correct use. Then, for another data element, [10]. On the one hand, by showing this concrete example,
where the header was “Doses discarded due to”, and the sub- first, major actors such as HMN and MoH got a good
header was “Expiry date”, many forms actually contained the understanding of the potential and benefits of such system.
expiry date of the various antigens in stock, rather than Second, it helped health programs that were reluctant to
amount of doses discarded due to them being too old. This collaborate within the newly proposed framework to realize
revealed both a lack of understanding of the forms at facility that everybody will get more output from the new system by
level, and also a possible ambiguous design of the form itself. gathering all different sources in one data warehouse, while at
the same time pursuing their own trajectory. For health
A similar example comes from the inconsistent closing programs that did not have any computerized HIS, it was a
dates for the data recording books at the facilities. For one great opportunity to move forward. On the other hand, those
form, which was supposed to cover one month, we would find who are feeding the system with data immediately saw that i)
that some facilities reported as early as the 25th day of the at district level, the capability of the system to avoid
month. Since some facilities were far away from the district overlapping data by displaying automatically the value of one
388
overlapping data element already typed for a given facility and boxing” DHIS on the server without screen, keyboard and
a given period will considerably reduce their workload, since mouse, the equipment is kept out of reach of local “IT
overlapping rate among two forms can be as high as 50 %. experts” in districts who used to format hard drives as the
The system’s capability of networking was also attracting for standard solution for any trouble shooting.
them. ii) at the grass root level, health care providers who are
also those who are producing data had understood that the The approach chosen in Sierra Leone has been to go around
proposed solution can trigger a process that will lead to the problem of the current infrastructural problems of no
harmonization of reporting tools which in turn will diminish power supply by establishing a new technical infrastructure
the 17 overlapping forms they have to fill at the end of each based on solar power and low-power computers. Technically,
month. The alignment of interest of actors at all levels [11] this is an innovative and wise solution. However, it will only
made a large alliance of actors that was strong enough to get a work if the consortium succeeds in establishing an effective
momentum in the project and got them very involved and technical support structure, including training and
active. During several months, they all worked together maintenance.
through a process of improvement, feedback, and feed-
forward between developers, implementers, managers and
C. Software solution
users. The heterogeneous and complementary nature of the
consortium was useful in this regard. As far as the integrated data warehouse is concerned, the
prototype strategy helped users at all levels to grasp the
At the early stage of the project, an important attention was system and give more input to its improvement. The DHIS is
given to the infrastructure issue. The team wanted a continuously being improved through these feedback
sustainable and flexible solutions that suit the context of processes involving all the stakeholders; computer users,
Sierra Leone – lack of power supply, no resource to fuel information users, implementers, hardware builders. Users in
generators, viruses to name a few. Then came the idea of the field are aware of the flexibility of the DHIS and see this
bringing Inveno experts with their low power computers that as an opportunity to make the work as easy as possible for
can work with solar panels and car batteries and based on them. For instance, both users and project managers
Linux and open source applications. In the field, IDC experts acknowledged that it is time to improve feedback and
were trained by Inveneo because by then, they were the only reporting system from district M&E to the health programs by
able to have a good understanding of the technical system. shifting from a manually generated report to one “push
They are supposed to continue the training of the core team as button” report generation. This was the highest priority local
well as organizing a local company to provide ongoing requirement for the DHIS from the users. But at the
technical support. The overlapping expertise or the continuous developer’s level, the credo is to always make the globally
expertise domain of the consortium helped keeping in place used software as flexible and generic as possible and keep the
the expertise of one member even when he left. The fact that adaptation to the local context for the customization phase.
IDC has a good background in hardware and open source This mechanism and philosophy improved the software a lot
helped to acquire knowledge from Inveneo and give it back to at the global level as feedback and experiences from the field
the local core team on the basis of learning by doing. The in Sierra Leone were highly relevant to other countries as
HISP-IDC team is engaged in a long term support of the well. The win-win alliance between local needs and DHIS
project and local expertise building. development at global level was enabled by the prototyping
strategy at local level and the need and philosophy of
developing DHIS as a flexible standard; that is, flexible both
B. Infrastructure in terms of use and change.
The new infrastructure ensures data safety, and since its
deployment, users have not experienced loss of data as they For the new software solution to be accepted among the
used to previously. Even when their existing computers are local users it was important that the DHIS was customized so
infected by viruses, the server as well as the Linux desktops that the electronic data collection tools looked exactly like the
still work. So now they have power to run computers and paper forms. The modular data element design enabled
computers to practice what they have learned during the collection forms to be generated simply by combining data
intensive training without spending more money for generator elements and without disturbing the carefully designed
fuel. One month after deployment in four piloting district and integrated data warehouse underneath. The way already
at the MoH, one sever stopped working and the diagnosis has registered values would be automatically filled in the
revealed that the motherboard was off because of overheating, electronic forms when opening a second form using the same
while it was supposed to be designed to overcome heat. The data element was highly appreciated by the health workers
problem was reported to Inveneo who sent a new motherboard that had previously wasted a lot of their time on manually
and it was an opportunity for them to improve the design of filling the exactly same data in various multiple forms.
their computers. All computers were then upgraded and since
then, similar problems have not been reported. By “black- Modularizing data in fine-grained data elements as opposed
389
to the traditional collection forms has several advantages. This effect that the main aim of the training, namely correct use of
metadata model enabled collection forms similar to the paper DHIS, was attained to an acceptable level. In addition, it gave
versions to be generated, which was an important requirement the trainers and students some time to get to know each other,
to get buy-in among the end users, simply by combining data which proved crucial when health information issues were
elements, and without disturbing the carefully designed discussed at a later stage. In this case, the benefits of spending
integrated data warehouse underneath. Furthermore, in an more time with the students became apparent during the third
environment with changing requirements and a range of week.
stakeholders with different needs, the modular design is
flexible to change as standardization takes place at the data The examples from the immunization forms show how the
element level. New collection forms, indicators or report training facilitated peer learning. Discussion amongst the
formats are just new compositions of data elements which can M&E officers established a common understanding on
easily be defined in the user interface without changing the important issues like standards and routines. It also provided
underlying integrated data warehouse. This flexibility has the trainers with an understanding of the situation “on the
been important in the early phases of implementation when ground”, and what could be done to counter any challenges.
the formats of feedback reports and indicator definitions went
through many feedback cycles among the users, the HMN Lastly, since the trainers were also the authors of this paper,
specialists and the developers. the points about discussions and user-feedback highlights the
potential power of action research. Being part of the
The introduction of feedback reports has been an important consortium and taking an active role in the process to improve
step in the DHIS implementation process and a maturing stage the current situation, in this case as trainers for the district
towards a revision of the entire HIS reporting system. The M&E officers, the authors gained knowledge about the issues
integrated data warehouse at district level has provided the raised in the previous paragraphs. As the training unfolded,
local M&E officers access to their data across health and over time the students became more secure on each other
programs, and the tools for processing and analyzing data and the trainers, the discussions flourished and provided an
have helped to illuminate gaps and problems with the existing extremely rich arena for learning about the Sierra Leone
data collection forms. Feedback reports, in stark contrast to health system.
typical data collection forms, have focused on data analysis
and use and not upward reporting, and have given local M&E
officers a quick overview of the situation from month to VIII. CONCLUSION
month. These reports have enabled local monitoring of key The HIS in Sierra Leone, presented in this article, has been
performance indicators of the services provided, and identified developed in an evolutionary participative manner, and it is
gaps in reporting from the chiefdoms and health facilities. In still in the making. Earlier approaches to develop HIS in
stead of presenting data in fragmented and overcomplicated Sierra Leone have been to focus on partial solutions for one
forms, these reports have used just a fraction of the data organization or health program, or, in some occasions, to
collected, extracting only the most relevant data from the implement new reporting formats including several of the
health programs, and used charts and simple tables to health programs. A major problem has been that
visualize the information. fragmentation has rather increased than decreased. In the latter
cases where all-encompassing reporting formats have been
D. Capacity building introduced, various groups of users and health programs have
not been satisfied, and they have continued to use the old
The extensive training provided some benefits, which were
reporting formats in addition to the new ones, and thereby
briefly mentioned above. These were:
worsening the situation.
1. A week-long general computer-course enabled a higher
quality of the HIS-specific training.
2. Time was given for student-led discussions on relevant The particular approach chosen by the ongoing project in
topics, which clarified both for the student body and the Sierra Leone has been different from earlier attempts in that it
trainers various issues of relevance is trying to incorporate all the various requirements for
3. The training was used to get feedback on the software and reporting and use of data in a step-wise approach. By taking
HIS from the students the existing data as a point of departure, users have been able
to join all the various reports being collated at district level
As mentioned earlier, the extensive training was given to within one database framework. In this way they have also for
develop a solid base both for using the software developed, the first time been able to assess the quality of the data they
and for creating the demand that could make it sustainable, by are collecting – through being able to access and analyze the
fostering a culture for using information. While it is too early data – and they have been able to compare the data across the
to make any conclusions regarding sustainability, the fragmented reporting structure. Through this integrated access
preparatory week of general computer knowledge had the to their own data, they have been enabled to identify the
390
problems with the current reporting formats; inconsistent data

definitions between formats, missing data and overlaps. This
learning process is now triggering, and feeding into, a much
needed revision and overhaul of the entire reporting system,
planned for early 2009. A principal difference between the
chosen approach and the previous approaches is the allowed -
and promoted - flexibility to change the system as users and
involved health programs learn concretely about current
shortcomings and potentials for improvements.
Ideally, of course, revisions of reporting formats should

have taken place up-front, as part of the initiation of the
project. Despite the high level profile of the project (Ministry
of Health, HMN/WHO), however, the various health
programs and donor agencies could not agree on shared
revisions and new standards for data reporting. By having
demonstrated how the reported data could be integrated, and
the usefulness of such an approach, the flexible software and
general approach have served as an effective means in
negotiating new data standards, unifying the actors, and
thereby building the consortium pursuing an integrated
approach.
The evolutionary approach followed is relying upon, and

feeding into, the ongoing formal and informal on-the–job
training that is carried out as part of the capacity developing
activities. By basing the training on real data and the use of
the system in practice, the training sessions become a key
participatory input for the improvement of the system, while
at the same time enhancing the learning process.
Introducing radically new technical infrastructure and type

of computers, as the Inveneo technology, represent a
considerable capacity development challenge. The best way to
approach this challenge is to make the technical part of the
HIS integral to the training. Contextualizing the training by
basing it on the problems in the district, including the
hardware parts, is the best way to approach also this part of
the sustainability challenge.
391
1. Braa, J., et al., Developing Health Information

Systems in Developing Countries. The "Flexible
Standards" Strategy. MIS Quarterly, 2007. 31.
2. Sæbø, J., J. Braa, and O. Chandna. A Flexible
Approach to Integrating Health Information Systems:
The case of Data Warehouse as Integrator in
Botswana. in IFIP WG 9.4. 2007. São Paulo.
3. Sahay, S., E. Monteiro, and M. Aanestad,
Configurable politics: trying to integrate health
information systems in developing countries. Special
Issue of JAIS on eInfrastructure, 2007.
4. Braa, J., E. Monteiro, and S. Sahay, Networks of
Action: Sustainable Health Information Systems
Across Developing Countries. MIS Quarterly, 2004.
28(3): p. 337-362.
5. Puri, S.K., Integrating Scientific with Indigenous
Knowledge: Constructing Knowledge Alliances for
Land Management in India. MIS Quarterly, 2007.
31(2): p. 355-379.
6. Hanseth, O., et al., Reflexive Standardization: side
effects and complexity in standard making. MIS
Quarterly, 2006. 31(2): p. 355-379.
7. Checkland, P. and S. Holwell, Action Research: Its
Nature and Validity. Systemic Practice and Action
Research, 1998. 11(1): p. 9- 21.
8. Avison, D., et al., Action Research. Communications
of the ACM, 1999. 42(1): p. 94-97.
9. Baskerville, R.L. and A.T. Wood-Harper, A Critical
Perspective on Action Research as a Method for
Information Systems Research in Qualitative
Research in Information Systems, M.D. Myers and
D. Avison, Editors, Sage Publications. p. 129-145
(2002).
10. Bowker, G.C. and S.L. Star, Sorting Things Out:
Classification and its Consequences. 1999,
Cambridge, MA: MIT Press.
11. Timmermans, S. and M. Berg, Standardization in
Action: Achieving Local Universality through
Medical Protocols. Social Studies of Science, 1997.
27: p. 273-305.
392
Mobile Telephony Access and Usage in Africa

Chabossou, A., Stork, C., Stork, M., Zahonogo. Z.
in developing economies and find that they are playing the

Abstract—This paper uses data from nationally representative same crucial role that fixed telephony played in developed
household survey conducted in 17 African countries to analyse economies in the 1970s and 1980s. Their study covered 38
mobile adoption and usage. developing countries for the period 1996 to 2003 and provides
This paper shows that countries differ in their levels of ICT
evidence of the strong impact of mobile in developing
adoption and usage and also in factors that influence adoption
and usage. Income and education vastly enhances mobile countries.
TABLE I
adoption but gender, age and membership in social networks
ICT DENSITIES IN 2007 (SOURCE: ITU DATABASE 2008)
have little impact. Income is the main explanatory variable for
usage. In terms of mobile expenditure the study also finds Internet Main Mobile
linkages to fixed-line, work and public phone usages. These users per (fixed) cellular
linkages need however be explored in more detail in future. 100 telephone telephone
Mobile expenditure is inelastic with respect to income, ie the inhabitants lines per subscribers
share of mobile expenditure of individual income increases less 100 per 100
than 1% for each 1% increase in income. This indicates that inhabitants inhabitants
people with higher income spend a smaller proportion of their
Sub Saharan 6.47 3.17 26.77
income on mobile expenditure compared to those with less
income. Africa
The study provides tools to identify policy intervention to Low Income 6.04 5.87 24.06
improve ICT take up and usage and define universal service Lower Middle 15.69 15.23 50.99
obligations based on income and monthly usage costs. It help to Income
put a number to what can be expected from lower access and Upper Middle 31.48 21.01 86.18
usage costs in terms of market volume and number of new Income
subscribers. Linking this to other economic data such as national High Income 54.78 43.85 112.42
household income and expenditure surveys and GDP calculation Countries
would allow to forecast the economic and social impact of policy Benin 1.66 1.22 20.98
interventions.
Botswana 4.25 7.28 75.84
Concrete recommendations are being made for policy
interventions and regulatory measures to decrease access and
Burkina Faso 1 0.59 0.70 10.90
usage costs. Cameroon 2 2.23 0.79 24.45
Côte d'Ivoire 3 1.63 1.41 36.60
Index Terms—Economics, Information technology, Mobile Ethiopia 0.35 1.06 1.45
communication, Developing nations Ghana 2.77 1.60 32.39
Kenya 7.99 0.71 30.48
Mozambique 4 0.93 0.33 15.42
I. INTRODUCTION Namibia 4.87 6.66 38.58
T The information, telecommunication and broadcasting Nigeria 6.75 1.07 27.28

sector is increasingly integrated into the day to day Rwanda 5 1.08 0.24 6.53
activities of businesses and lives of households and Senegal 6.62 2.17 33.31
individuals worldwide. Numerous studies have shown that South Africa 8.16 9.56 87.08
ICTs contribute to economic growth, employment and social Tanzania 0.99 0.58 20.40
inclusion. It is therefore important that policymakers monitor
Uganda 6.48 0.53 13.58
progress towards access and usage of ICTs. A paper by Roller
and Waverman (2001) suggests that the spread of modern Zambia 4.19 0.77 22.14
fixed-line networks in OECD countries was responsible for
one third of output growth between 1970 and 1990. The Country level data reveals a global digital divide between
importance of ICTs for economic growth are further supported continents and countries but also within countries
by studies from Jalava & Pohjola (2002), Oliner & Sichel
(2000), Pohjola (2001), Niininen (2001), Sichel (1997), 1
Jorgenson et al (2005), OECD (2003) and UNCTAD (2006). Internet and main fixed line data from 2006
2
Internet and main fixed line data from 2006
Waverman et al (2005) investigate the role of mobile phones 3
Internet and main fixed line data from 2006
4
Main fixed line data from 2006
5
Internet data from 2006
393
(Baliamoune-Lutz, 2003). Table 1 displays the difference levels of access and usage.
between low, lower-middle, upper-middle and high-income TABLE 3
ICT KNOWLEDGE AND USE, AND MOBILE ACCESS ACROSS DISPOSABLE
countries for Internet, fixed-line and mobile density. It also INCOME CATEGORIES
display the same information for countries covered by this
16+ knowing 16+ using the Owning a
studies and the sub-Saharan Africa (SSA) averages. The SSA
what the Internet mobile phone
average for Internet users and fixed-line users is lower than Internet is or active SIM
the low income average. It can also be seen from Table 1 that card
several countries from this 16-country sample are below the
lower top lower top lower top
low income and SSA averages. Countries have a much lower
three dispo three dispo three dispo
ICT density compared to the low-income average. Table 2 and dispo sable dispo sable dispo sable
Table 3 document the digital divide within countries for access sable inco sable inco sable inco
to fixed telephony in urban areas compared to rural ones and inco me inco me inco me
knowledge about and usage of the Internet across disposable me quarti me quarti me quarti
income quartiles. quarti le quarti le quarti le
le le le
TABLE 2 Benin 17% 33% 6% 16% 17% 63%
HOUSEHOLDS WITH WORKING FIXED-LINE PHONES
Botswana 14% 52% 2% 19% 53% 83%
National Major Other Rural Burkina 6% 17% 2% 10% 19% 50%
Urban Urban Faso
Rwanda 0.10% 1.20% 0.00% 0.00% Cameroon 35% 54% 10% 25% 28% 74%
Uganda 0.30% 1.60% 2.00% 0.00% Côte d'Ivoire 12% 31% 4% 14% 33% 63%
Tanzania 0.90% 2.80% 2.50% 0.20% Ethiopia 6% 18% 0% 3% 1% 11%
Mozambique 1.70% 7.30% 1.30% 0.00% Ghana 26% 27% 5% 8% 53% 79%
Cameroon 1.80% 3.50% 3.50% 0.00% Kenya 25% 52% 9% 32% 42% 79%
Kenya 2.30% 11.40% 0.90% 1.40% Mozambique 2% 10% 1% 3% 20% 49%
Ghana 2.60% 6.30% 4.60% 0.30% Namibia 16% 60% 4% 25% 37% 86%
Nigeria* 2.70% 9.70% 5.70% 1.20% Nigeria 35% 52% 10% 22% 74% 93%
Benin 4.60% 15.50% 6.80% 0.60% Rwanda 4% 13% 1% 4% 4% 26%
Burkina Faso 4.70% 18.30% 16.70% 0.30% Senegal 37% 71% 9% 14% 29% 77%
Côte d'Ivoire 4.80% 10.80% 7.10% 1.30% South Africa 42% 75% 7% 38% 54% 84%
Ethiopia 7.60% 46.00% 22.70% 2.90% Tanzania 6% 16% 1% 4% 14% 46%
Botswana 11.00% 7.80% 17.00% 8.70%
Uganda 4% 15% 1% 7% 12% 46%
Senegal 11.70% 21.50% 19.20% 0.60%
Namibia 17.40% 34.10% 29.60% 7.90% Zambia 39% 60% 1% 13% 36% 84%
South Africa 18.20% 38.90% 18.50% 2.10%
Zambia* 2.44% 5.20% 7.55% 0.00% The great potential of ICTs as catalysts of social and
economic development is clearly recognized in continental
discourses of agencies such as NEPAD, economic regional
Bagchi & Udo (2007) are that African countries are not
association strategies and in national policies. ICT
experiencing the benefits from ICT that are being experienced
dissemination and adoption in developing countries facilitate
by OECD nations and lag severely behind even global
the achievement of major development goals in the areas of
averages in ICT adoption in particular in investment-intensive
health, education, governance and others. People in
infrastructure such as fixed-line telephony and broadband
developing countries need easier and cheaper access to ICTs.
Internet access. 6
They need enhanced ICT skills to better employ these
McCormick & Onjala (2007) list the following reasons for
technologies in their homes, schools and jobs. Policies and
the low level of ICT access and usage in Africa:
strategies that have been adopted however have not been
successfully in addressing the question of how African
• weak telecommunications infrastructure countries can catch up with global levels of ICT access and
• generally low level of economic activity usage and how the poor can join the newly developing
• irregular or non-existent electricity supplies; and information societies. For this it is important to understand the
• lack of human resource capacity, lack of skills and brain factors that lead to adoption of ICTs by households and
drain. individuals. It is further of importance to understand what
factors other than income impact on usage once households or
The results from the researchICTafrica.net household
individuals have adopted ICTs.
survey, however, shows that the monthly cost of
telecommunication services is the main reason behind low
II. DATA AND DESCRIPTIVE STATISTICS
The data that is being used for this study stems from
6
See also Sciadas (2003)
394
surveys conducted by Research ICT Africa (RIA) 7 in 17 together in the same housing unit or group of housing units,
African countries during the end of 2007 and the beginning of have common cooking arrangements and share financial
2008 8 . The data is nationally 9 representative on a household resources. Maids, guards and baby-sitters that stay with a
level and for individuals 16 years of age or older. The survey family are part of the household. A head of a household is a
was stratified into metropolitan, other urban and rural areas. person who economically supports or manages the household
Enumerator areas (EAs) were sampled for each stratum using or for reasons of age or respect, is considered as head by
probability proportional to size (pps). The individual to be members of the household or declares himself or herself to be
selected for each household was randomly selected from all head of a household.
household members and visitors that stayed at the home on the
night the household was visited and that were 16 years of age III. STATED AND REVEALED PREFERENCES
or older. The RIA household survey collected information from
The RIA questionnaire was divided in three sections. The mobile users about their monthly expenditure but also from
first part, the household rooster, collected information about non-users about their willingness and ability (WTP) to spend
all household members. The second part collected household on mobile telephony. The first is known as revealed and the
related information. The head of the household or someone latter as stated preferences. WTP measures are widely used to
that manages the household answered part one and two. The provide information to policy makers regarding the economic
third part, the individual section, was answered by a randomly value of non-market, non-pecuniary or environmental goods.
selected individual 16 years of age or older that slept in the For stated preferences, no behavioural changes can be
house the night of the interview and included household observed; the individuals only state that they intend to behave
members and visitors. in that particular fashion (Adamowicz et al, 1994).
TABLE 4
RIA 2007/2008 HOUSEHOLD SURVEY SAMPLE The payment ladder method with exponentially distributed
amounts was used to infer about the willingness and ability to
Major Other Rural Total
Urban Urban pay for mobile services of non-users. The payment card was
Benin 432 336 333 1,101 first used by Mitchell & Carson (1984). The payment ladder is
Botswana 348 241 229 818 a kind of payment card which lists WTP values from low to
Burkina Faso 416 329 332 1,077 high. Enumerators read the values to the respondent, starting
Cameroon 490 347 398 1,235 at the top of the list and moving down. They asked until the
Côte d'Ivoire 502 312 298 1,112 first three values were answered with no. If the respondents
Ethiopia 1,173 631 551 2,355 were almost certain about their willingness and ability to pay a
Ghana 473 324 295 1,092 monetary value then a tick (√) was placed in the space next to
Kenya 472 557 432 1,461 that amount. If the respondents were not sure about an amount
Mozambique 562 312 257 1,131 then it was simply left blank. If the respondents were almost
Namibia 311 294 280 885 certain that an amount was too high then a cross (x) was
Nigeria 895 1,012 844 2,751 placed next to the amount. Three crosses (NOs) were required
Rwanda 415 333 330 1,078 to complete the question to assure that no intransitivities
Senegal 432 312 337 1,081 occurred. The highest amount the respondent was willing and
South Africa 779 465 527 1,771 able to pay and the lowest amount the respondent was not
Tanzania 634 393 463 1,490 willing and able to pay was captured. The difference between
Uganda 436 347 344 1,127 these two values is the range of uncertainty (Bann, 1999).
Zambia 405 212 264 881 Having two points increases the chance of eliciting the
Total 9,175 6,757 6,514 22,446 respondents’ actual willingness to pay. The payment ladder
avoids starting point bias and reduces the number of outliners
(Bateman, 2002). To avoid range and centring biases 10 the
A household constitutes of a person or group of persons,
payment ladder was based on exponential value increment, as
irrespective of whether related or not, who normally live
suggested by Row et al (1996). This approach has three mayor
7
advantages.
Research ICT Africa (RIA) is a network of universities and research First, the scale is consistent with the hypothesis of
institutions from 19 African countries hosted at The EDGE institute,
increasing measurement error with increasing value. For
Johannesburg, South Africa. For more information see
www.researchICTafrica.net. instance a person might value a cup of coffee in a café at
8
These are Benin, Botswana, Burkina Faso, Cameroun, Cote $2,00 plus minus $0,20, but a vacation at $2.000 plus minus
d’Ivoire, Ethiopia, Ghana, Kenya, Mozambique, Namibia, Nigeria, $200. Second, according to Row et al (1996) there exists
Rwanda, Senegal, South Africa, Tanzania, Uganda and Zambia. The evidence that measurement errors in values obtained in
data for Zambia is not included in this analysis yet since the data had contingent valuation studies have a log-normal distribution
not been weighted by the time of submission of this article.
9 and that a logarithmic transformation of the WTP data
The data for Zambia and Nigeria are national extrapolations but not
nationally representative. In Zambia the selection probabilities had to addresses the increasing measurement error in hypothetical
be estimated since enumerator ID information was not unique. In
10
Nigeria the sample was drawn only from 6 of 36 provinces. See Venkatachalam (2004) and Row et al (1996).
395
WTP data. Therefore the exponential payment ladder can be another might just be close to getting one and still saving
seen consistent with the error distribution of WTP values. The money towards it. For both individuals it can only be observed
third argument Row et al (1996) put forward refers to the that they do not have a mobile phone ie Mobile=0. The
concept of “just noticeable” differences and they uses the process leading to that decision is unobserved and referred to a
example of light sources to explain this. A source of light has latent variable. Probit or Logit models tie observable
a brightness B. The differences between two of such sources is (independent variables) to the latent variable though
“just noticeable” if the difference can be detected in 75% of contributions to the probability of the latent variable taking a
the time. When bringing these sources in a sequence arranged value above or below a threshold that would lead to the
in order of increasing brightness B1, B2, …, Bn,so that each
B observable outcome, adoption or not. Many factors are
source is “just noticeable” to the preceding one, than the considered in the economic literature as influencing the latent
relationship to each other is given by Weber’s law: variable, 15 such as socio-economic factors (social networks,
income, prices, household size, education, age and gender)
, (1) and environmental factors (infrastructures, enumerator area,
country’s characteristics).
where k is a positive constant. Hence the “just noticeable” Piccoli et al (2001) established that the education level of
differences are increasing proportionally to the sequence of an individual influences ICTs adoption and usage, through
sources and can also been written by exponential function: influencing an individual’s capability to use technology. The
theory of technology use also points to age as a factor that
(2) influences when and how an individual uses technologies.
Some studies indicate that age is a key factor in the use of
technology, with younger people tending to exhibit higher use
The value of k can be seen as a percentage increase between
levels. Consequently, these individuals tend to use ICTs
adjacent scale values. The survey used 28 values, starting with
technologies more than older users.
a zero (not interested at all) and ending with an empty cell for
Gender and technology studies have found that men and
amounts higher than listed. Equation (2) was used to compute
women adopt and use technology differently (Gefen & Straub,
the 2nd amounts to 27th amount. K was selected so that (1 +
1997; Venkatesh & Morris, 2000). Men’s decisions to use
k)n-1 equals the highest value on the list. 11
technology are more strongly influenced by their perception of
usefulness, while women’s decisions are based more on
IV. MOBILE ADOPTION / ACCESS
perceptions of the technology’s ease of use (Venkatesh &
This section analyses factors that contribute to the Morris, 2000). Further, men and women may view the same
probability of an individual to adopting mobile telephony, ie mode of communication differently. 16
either owning a mobile phone or active SIM card. The
dependant variable is dichotomous and using linear regression Table 5 in the appendix. displays the Probit model
models would not be appropriated. 12 Probit models are being configuration and Table 6 in the appendix details the results
used instead, assuming normal distributed error terms. The for each country. Table 6 provides summary results without
assumption about the error term is arbitrary and its validity reporting coefficients, showing for each of the independent
cannot be tested. 13 A model can only be identified once an variables whether their impact on the probability of mobile
assumption is imposed on the mean and variance of the error adoption is significant and whether it is positive or negative. 17
term. This implies that coefficients cannot be interpreted
directly since they are influenced by the necessary The results suggest that the variables used to analyse mobile
assumptions. Changing the assumptions about the error term phone adoption contribute significantly to explaining the
changes the values of the coefficients uniformly (Long, 2006), probability of adoption. Income increases the probability of
but it does not affect the probability of an event occurring 14 . It mobile phone adoption as expected. Only for Ghana and
affects the spread of the distribution but not the proportion of Rwanda it wasn’t significant. Substituting individual income
the distribution above or below the threshold (Long, 2006). with disposable income increases the pseudo R2 in both cases
The decision to adopt mobile telephony is the result of and let to income being significant with positive coefficient.
several factors which cannot be observed and which will be Disposable income has the advantage over individual income
different for each individual. What can be observed is whether that it takes into account the buying power of dependents that
an individual has a mobile phone or active SIM card do not have own income. It does not generally lead to better
(Mobile=1) or not (Mobile=0). One individual might neither results though. Income being a barrier to adoption has two
be able to afford nor be interested in a mobile phone while consequences. First, people will increasingly adopt mobile
telephony should their income increase. Secondly, they might
11
For example with n = 24 cells which have to be computed and
$1000 the highest value, the equation would be given by 1000 = (1 +
0,35)23 15
See for example : Kwapong (2007) Fabiani al (2005), Bagchi &
12
See for example Maddala (1983) and Gourieroux (1986) Udo (2007), Kabbar & Crump (2006)
13 16
For a more detailed discussion on this assumption see Long (1997). See Gefen & Straub (1997); Ono (2003)
14 17
See Long (2007) pp 49-50 for a proof. The detailed results can be found in the appendix.
396
be adopting it at current income if access (handset) prices and increased traffic in developing countries due to relatively high
usage cost would come down. price elasticities. However, Garbacz & Thompson (2005)
Age influences for some countries positively and for others point out that international calls are less of relevance for poor
negatively the probability of mobile adoption. Age is likely to people in developing countries compared to local and national
impact positively the probability to adopt mobile up to certain calls, where price elasticities are low. They also find that the
level and then decreases. Using age2 instead of age in mobile monthly price elasticity exceeds fixed-line price
subsequent research could get to the bottom of this. 18 elasticity in developing countries by a substantial amount and
The gender variable is mostly insignificant as expected (11 suggest that wireline phones are substitutes in the mobile
countries). Only in Senegal, Tanzania and Burkina Faso does market while mobile phones are not substitutes in the wireline
being a woman decrease the probability of mobile telephony market but complements. This is to some extent contradicted
adoption. In South Africa and Mozambique it even increases by Esselaar & Stork (2005) who find studying usage data for
the probability significantly. nine African countries that mobile telephony is a substitute for
The result from the Probit models confirm the assumptions fixed-line telephony across all income groups.
about education as well as the rural areas. Membership in Clarke & Wallsten (2002) argue that universal service for
social networks contributed positively to the probability of the poor in developing countries is generally ineffective.
mobile adoption in 7 out of 16 countries. For the remaining Modelling income elasticities could establish alternative
countries it was insignificant. approaches to current universal service definitions. Rather
Randomly selected individuals 16 years from each then specifying coverage or geographic spread, universal
household without mobile phone or active SIM card were service obligations (USOs) could be in terms of costs for
asked for their willingness and ability to pay for a handset and monthly usage packages that take into account existing
what they think a handset would cost them. These average income barriers.
figures are being shown in Table 7 in the appendix. The Lee (1999), in estimating access demand for mobile
difference between average willingness to pay and average networks, shows that per capita GDP and the size of the
expected costs confirms income as the main adoption barrier. existing fixed-line network increase the probability of people
Table 7 also indicates the number of new mobile phone users subscribing to mobile telephone services. This implies that the
operators could expect should they offer mobiles for US$20, mobile telephone network is interdependent with the fixed-line
US$15 and US$10 respectively. 20US$ handsets would attract network.
about 3 million new customers in Kenya and Ivory Coast, for Demand, as understood by economists, is defined as the
example. Ethiopia could expect 2,6 million more users if the amount of goods or services people are willing to buy at a
handset price would drop to US$10 and provided that SIM certain price (ceteribus paribus - other things including income
cards would be available. Ethiopia is the country with the equal). In this sense, demand is a concept affected by buying
widest gap between willingness to pay and expected cost and power. A person may have needs but not demand du to
the country with the lowest mobile teledensity (see Table 1). income limitations. With insufficient income, demand can be
The handset price is also known to fluctuate with the zero or low even if the need is urgent. Demand or buying
availability of SIM cards. power for a goods or services arise from the consumers’
preferences for specific goods or services. Thus defining a
V. INCOME ELASTICITY / E-USAGE good or service, and studying how the consumer orders his/her
Perl (1983), Taylor (1994) and Taylor (2002) developed preferences for such a good or service in relation to other
telecommunication demand models that can be addressed available goods and services is crucial.
price and income elasticities for public phones, mobile phones Demand arises from a previous knowledge of the good and
and local, national and international fixed-line call as well as a subjective evaluation of its benefits and costs.
cross price elasticities. 19 Communications services differ from other goods or services
Universal Service Obligations (USO) in this context have since some require a subscription prior to usage such as post
been thoroughly discussed in the literature. 20 Garbacz & paid mobile and fixed line services. Even for prepaid mobile
Thompson (1997, 2002, 2003) find that price elasticities for services a distinction can be made between access (cost of a
landline services in the USA approach zero in recent years and handset) and usage costs. Subscription is necessary to use the
that subsidies therefore have little or no impact on universal Internet as well in most cases (notable exception is the Internet
service. Crandall & Waverman (2000) Eriksson et al (1998) café). In the case of telephony the access service provides
and Hausman et al (1993) support these findings. utility to a customer in terms of granting the ability to make
Wallsten (2001) finds that greatly reduced international and receive calls, even though it is uncertain whether the
settlement rates for telecom traffic between the United States ability will be exercised. This is the so-called option value
and the rest of the world had the effect of reduced prices and (Hee Lee, 2006).
For the purpose of analyzing demand income elasticity for
18
mobile telecommunication services we assume that the
See Tegegne (1999). potential usage basket of a consumer consists of calls (q) and a
19
See for example Kaserman et al (1990) and Parsons (1998)
20 composite good (x) that represents all other goods and
See Gasmi et al (2000), Panzar (2000), Rosston & Wimmer (2000),
Valletti (2000) services. We further assume that the consumer’s decision on
397
individual demand for calls is made by maximizing a utility logarithm of the left hand side and the right hand side:
function within a given budget constraint. In addition to the
factors of calls (q) and a composite good (x) affecting his/her (8)
utility, income (y) can be considered. Moreover, if call
externality is incorporated in the utility of an individual, the
Where and the socio-demographic factors
number of existing subscribers who are making calls to and
receiving calls from the individual affects the demand for set. The transformed demand functions for mobile phone
calls. Thus, after including the size of the network (N) in order services are then expanded by socio-economic factors and
to reflect call externality, the utility function of a customer (U) postulated to be of following form:
is represented in expression (3):
(9)
(3)
m = monthly mobile expenditure in US$, converted using
At the same time, the budget constraint considering access nominal end of 2007 exchange rates (source IMF)
charge and call charge is given by Y= natural logarithm of individual income in US$,
converted using nominal end of 2007 exchange rates (source
, (4) IMF). The income is the sum of any salary or wage, self-
employment income, property income, income from
Where r is the price of access to the telecommunications agricultural produce, pension, transfer income and
system, the price of a call, y is income of the consumer and scholarships.
p the price of the composite good x. For the utility function in A = age in years
(3), the number of calls represents direct benefits of using the PP = dichotomous public phone variable: used public phone
mobile, while the number of subscribers represents indirect in the last three months =1, otherwise 0
benefits through network size. Given the budget constraint in WP = dichotomous work phone variable: access to a phone
(4), and assuming the utility function is strictly quasi-concave at work =1, otherwise 0
and differentiable, using Lagrangian techniques the demand G = dichotomous gender variable, female=1, male=0
functions for calls and the composite good, as shown below in F = dichotomous fixed-line variable: fixed-line access at
expressions (5) and (6), can be derived from the optimal home =1, otherwise 0
solution of the utility function: SN= Dichotomous variable: social network (church groups,
sports clubs etc.) membership= 1, otherwise 0
(5) R = Dichotomous variable: rural= 1, urban (other urban
and major urban) =0
(6) The price variable used in the analysis is represented by the

amount that an individual paid for access and usage of
These functions differ from conventional demand functions mobiles. However since country results are being computed
in two aspects: The budget constraint is y-r, rather than y, individually, there will not be a change for any of the
which reflects the conditionality of having to purchase access responses, the cost of the usage basket is the same for
to telecommunication services, and the demand functions everyone in a country. This lets the price variable become a
depend upon the number of subscribers, which reflects the constant.
access and usage externalities. Perl (1983), Taylor (1994, Table 8 in the appendix presents the results from robust
2002) and Taylor and Kridel (1990) developed regressions round for each country using weighted data.
telecommunication demand models that are now widely used Income is in each case significant and positively related to
in the literature. The model has the following form: mobile expenditure.
The gender variable is insignificant for 10 countries which
, (7) is somewhat surprising given that the average income of
women is lower compared to that of men for all countries but
Where q denotes use of the telephone network, p denotes Mozambique. However income is already taking care off in
the price of access and use, y denotes income, and u denotes a the equation. This means that apart of being at different
random error term. Network externalities as well as now income levels women spend the same on mobile phone as
socio-demographic factors can be included in this model as men. For Botswana, Cameroon, Mozambique and Namibia the
well. 21 Equation (7) can be transformed by taking the natural gender variables was significant and negative, which means
that women spend less. Nigeria had the only positive
21
Haque (2007) suggests ways of testing for the functional form of significant coefficient. 22
the demand function for telecommunication services. This would be
22
beyond the scope of this paper and will be addressed with a separate These finding can be explored further by using mean rank
paper. comparisons for the ratio mobile expenditure / disposable income.
398
Having a working fixed-line phone at home is linked to

higher mobile expenditure in 8 out of 16 countries. This can
be explained by two factors. First, households with fixed-line
phones are generally wealthier which might not directly be
reflected in the individual income of the responded
(dependents). Second, having a fixed-line phone at home
provides more opportunity to talk, such as calling home. These
results confirm the existence of a relationship between mobile
and fixed-line networks. To contribute to the complementarity
and substitutability discussion further the fixed-line
expenditure would need to be investigated for any affect
through the number of mobile phone users in the household as
well. The survey data also can be used to explain usage pattern
in more detail.
Public phone usage has only been significant for 6 out of 16
countries. For Benin, Cameroon and South Africa it has a
negative affect on mobile phone expenditure and for Burkina
Faso, Rwanda and Senegal a positive one. This too needs to be
analyzed in more detail. Public phones might predominantly
be used by lower income groups or might actually substitute
mobile phone expenditure. More country specific research is
required to come to definite conclusions.
Equally for the impact of access to a work phone or private
use. Fore five countries the work phone variable shows as
positive and significant. While having access to a work phone Figure 1: Income elasticity: percentage change in mobile
is a good indicator for income one might have also have expenditure for 1% change in income.
suspected a negative impact on mobile usage expenditure
since the work-phone could be used to substitute for mobile The figure above shows the percentage change in mobile
calls. expenditure for a 1% income change. For all countries the
Age has a significant and negative impact in six countries mobile expenditure is inelastic with regard to income. Nigeria
where that variable is significant. For those countries a youth has the highest income elasticity and Mozambique the lowest.
user’s effect can be established. Since these are national figures based on cross-section data it
The main objective is to estimate demand elasticities. A implies that people with higher incomes spend a smaller
distinction is made between income and price elasticities. proportion of their income on mobile telephony compared to
Income elasticity describes the magnitude and also therefore low income households. Figure 2 confirms this. Income
the responsiveness of a demand change for a good in respect elasticity seems to be linked to relative cost of mobile usage,
to a income change. Graphically expressed, the income penetration and coverage. Those factors need to be explored in
elasticity is –ceteris paribus- the slope of the demand curve, more detail in subsequent studies utilising supply side data and
dependent on the income. qualitative research.
ΔQ /Q
eY =
ΔY /Y
Where: Q = quantity demanded; Y = income and, ey =
income elasticity.
The first derivative of the regression equation for income
yields the coefficient for Y, which is the income elasticity, that
is, the percentage change in mobile expenditure at a 1%
change in income. 23 The estimations results show that income
variable is statically significant and positive for all 17
countries.
23
See Wooldrige (2006) pages 46 and 705, Hej et all (2004) page
296.
399
price elasticity will be relatively low (much less than -1). But,
as prices fall, in a much cheaper service environment, she
argues that people will start using the phone for many non-
essential purposes; ranging from relationship maintenance and
simple conveniences. Added to this, pricing innovations that
enable poorer people to purchase small denomination pre-paid
calling cards to make phone calls it is intuitive that the number
of calls made will increase significantly. Milne (2006) argues
that when this happens, the price elasticity will rapidly rise
towards -1 or even more than -1.
One of the biggest problems in estimating price elasticity of
demand (PED) for telecom services in developing country
situations is the lack of accurate usage data. Unlike in
developed countries where usage is easily obtainable from
monthly bills for post-paid connections, the vast majority in
developing countries, in the case of the current study, between
78% and 99% have no billing records (for prepaid shares see
Table 9). Many use public phones, phones at work or other
people’s phones which equally does not generate records.
Figure 2: Share of mobile expenditure per month in terms of

individual income.
VI. PRICE ELASTICITY / E-USAGE

Price elasticity describes the magnitude and also therefore
the responsiveness of a demand change for a good in respect
to a price change. Graphically expressed, the price elasticity is
–ceteris paribus- the slope of the demand curve, dependent on
the price.
ΔQ /Q
eY =
ΔP / P
Where: Q= quantity demanded; P = price of the good and ep
= price elasticity.
The price elasticity of demand is in almost all cases, except
for inferior goods, negative, if the price increases the demand
decreases. To simplify the notation, demand elasticities are in
general expressed in absolute value and therefore expressed as
positive (Frank & Bernanke, 2003). Figure 3: If calls were cheaper what would you do? (source: RIA
2007/2008 household survey).
Milne (2006) provides a very intuitive argument for the
changing structure of price elasticity in a hypothetical case.
She argues that when phones become available for the first
time in a low-income community and usage charges or tariffs
are relatively high (in terms of income) people will make
limited number of essential calls. In such a scenario, even if
the prices falls by a small amount the number of such calls; for
instance in an emergency or substituting a telephone call for a
bus ride to a city, will not increase by much. In other words
400
could be reached with US$1, US$2 and US$5 Average

Revenue per Users (ARPU) business models.
VII. CONCLUSION
This paper shows that countries differ in their levels of ICT
adoption and usage and also in the factors that influence
adoption and usage. Income and education vastly enhances
mobile adoption but gender, age and membership in social
networks have little impact. Income is the main explanatory
variable for usage. In terms of mobile expenditure the study
also finds linkages to fixed-line, work and public phone
usages. These linkages need however be explored in more
detail in future. Mobile expenditure proofs to be inelastic with
regard to income, ie as income increases mobile expenditure
increases to a lesser extent indicating its importance in
individual budgets.
Key policy interventions would be regulatory measure to
increase access and usage of mobile services by promoting
network investment. The current super profits being enjoyed
by operators across the continent as a result of prices that are
amongst the highest in the world need to be moderated. In
most markets there is not effective competition requiring
regulatory interventions to ensure cost based wholesale and
retail pricing.
Increasing Coverage: Licence fees for infrastructure
investment should cover the administrative and usage costs of
national resources only. The current high cost of licences in
most countries used to generated funds for state coffers
Figure 4: Average reaction to cost of using your phone came
translate in high prices for consumers as operators recover
down by half or going up double (source: RIA 2007/2008
household survey). their licence costs. Countries need to look at incentives to
encourage networks rollout not disincentives.
Another difficulty is that cross section data, like the one that Wider Access: Access prices can be reduced by exempting
is being analysed here, is not suitable to calculate price telecommunication equipment, in particular mobile handsets
elasticities since it only captures data for one snapshot in time. and services from import duties and additional taxes.
Changes in consumer behaviour due to price changes can only Allow more Usage: Policy measures to increase the
be analysed using time series data. Cross section data only competition within the industry are the best mechanism to
allows a rough glimpse at price elaticities by using contingent reduce usage costs. In a competitive environment the operator
valuation methods. Respondents with mobile phone or active can choose to compete on price or on service quality.
SIM card were asked how the would react to price decreases Regulatory measure can improve competition on price by
and increases. The answers to these questions are being creating price transparency.
displayed in Figure 3 and Figure 4. Figure 3 shows that the Price Transparency: Contract and prepaid tariffs vary across
vast majority of respondents would make more calls if prices operators to an extent that a product to product comparison is
would come down. Senegal and Ethiopia are the two countries impossible for consumers. Non- transparent pricing, including
with the highest share of respondents that would use the saved rapidly changing promotions and misleading media
money for something else. campaigns, make it difficult for end-users to make informed
Figure 4 shows how respondents state that they would react decisions. This prevents operators from having to compete on
to a doubling and halving of call charges. What can be seen is prices and leads to higher average prices. Establishing price
the response is asymmetrical, ie respondents would react to transparency on the basis of published monthly usage baskets
price increases stronger than they would do to price decreases. will make the cost to the end-user transparent and force
However, approximating price elasticity from these two operators to compete on price or service quality.
contingent valuations results is strictly speaking not possible. Cost-based Interconnection rates: Above cost
Respondents state their preferences and do not reveal it interconnection charges are often used by dominant operators
(observed behaviour reacting to price changes). to restrict small operator from gaining market share. High
Table presents data on the untapped market, those 16 years interconnection charges make off-net calls expensive and
or older that do not have a mobile or active SIM card at punishes consumer or small operator. Cost based
present. It shows the results for the number of people that interconnection rates (or benchmarked on cost base) increase
the competition between operators and hence lead to lower
401
usage cost. [16] Garbacz, C., & Thompson, H. G., Jr. (2007). Demand for
telecommunication services in developing countries.
Licences for Low ARPU business models: Regulators Telecommunications Policy 31 (2007) 276–289.
should award additional licences to applicants that are able to [17] Gasmi, F., Laffont, J. J., & Sharkey, W. (2000). Competition, universal
offer low ARPU business models, preferably below US$2. service and telecommunications policy in developing countries.
Information Economics and Policy, 12, 221–248.
The correct policy and regulatory measures can boost [18] Gefen, D. & Straub, D. (1997). Gender difference in the perception and
mobile access and usage. Avoiding sector and service specific use of e-mail: an extension to the technology acceptance model. MIS
levies and additional taxes will result in lower recovery costs Quarterly, 21, 389– 400.
and therefore lower prices. Lower prices would allow millions [19] Gourieroux, C. (1986). Econométrie des variables qualitatives,
Economica.
of people currently unable to afford to access mobile services [20] Haque, O., M. (2007). Income Elasticity and Economic Development,
to do so and those currently restricting their usage as a result Springer, ISBN 0-387-24292-9.
of the high cost of services to use services more extensively [21] Hausman, J., Tardiff, T., & Belinfante, A. (1993). The effects of the
breakup of AT&T on telephone penetration in the United States.
and effectively. This is likely to produce greater profits for American Economic Review (American Economic Association Papers
companies overall and Governments will as a consequence and Proceedings), 83, 178–184.
have even more money in their coffers due to taxes on greater [22] Heij, C., de Boer, P., Franses, P., Kloek, T. and van Dijk, H. (2004).
Economteric Methods with Applications in Business and Economics,
operator revenues. Oxford University Press, ISBN 0-19-926801-0.
[23] Jalava, J. and M. Pohjola (2002). Economic growth in New Economy:
VIII. REFERENCES Evidence from advanced economies, Information Economics and Policy,
Vol.14, pp. 189-210
[1] Adamowicz W. & Louviere J. & Williams (1994). Combining Revealed [24] Jorgenson D, M.S.Ho and K .Stiroh (2005). Productivity: Information
and Stated Preference Methods for Valuing Environmental Amenities. Technology and the American Growth Ressurgence. London;
Journal of Environmental Economics and Management, Vol. 26, pp.271- Cambridge: The MIT Press.
292. [25] Kabbar, E.F. and Crump B.J. (2006). The factors that influence
[2] Bagchi, K. and G. Udo (2007). Empirically testing factors that drive ICT Adoption of ICT by recent refugee immigrants to New Zealand,
adoption in Africa and OECD set of Nations, Issues in Information Informing Science Journal, Vol.9, pp.111-120.
Systems, Vol. 8, No.2, pp. 45-52. [26] Kaserman, D. L., Mayo, J. W., & Flynn, J. E. (1990). Cross
[3] Baliamoune-Lutz, M. (2003). An analysis of the determinants and subsidization in telecommunications: Beyond the universal service fairy
effects of ICT diffusion in developing countries, Information tale. Journal of Regulatory Economics, 2, 231–249.
Technology for development, Vol.10, pp.151-169. [27] Kelly, T. (2004). Changing ICT ranking of African Nations. The
[4] Bann C. (1999). A Contingent Valuation of the Mangroves of Benut. Southern African journal of Information and Communication, Issue No.
Forestry, Department, Peninsular Malaysia, Danish Cooperation for 5, 2004.
Environment and Development and the Johor State. [28] Kwapong, O.A.T.F (2007). Problems of policy formulation and
www.soc.uoc.gr/calendar/2000EAERE/papers/PDF/A6-Bann.pdf implementation: The case of ICT use in rural women’s empowerment in
Last Appeal: 10.07.07 Ghana, International Journal of Education and development using
[5] Bateman I., Carson R., Day B., Hanemann M., Hanley N., Hett T., Information and Communication technology, Vol.3, No. 2, pp. 68-88.
Jones-Lee M., Loomes G., Mourato S., Ozdemiroglu E., Pearce D., [29] Long, S. (1997). Regression Models for Categorical and Limited
Sugden R., Swanson J. (2002). Economic Valuation with Stated Dependent Variables, Advanced Quantitative Techniques in the Social
Preference Techniques – A Manual. Edward Elgar, Cheltham, UK, Sience Series, Sage Publications, ISBN 0-8039-7374-8.
ISBN: 1-84064-919-4. [30] Long, S. & Freese, J. (2006). Regression Models for Categorical
[6] Clarke, G., & Wallsten, S. (2002). Universal(ly bad) service: Providing Dependent Variables Using Stata, Stata Press, ISBN1-59718-011-4.
infrastructure services to rural and poor urban consumers. World Bank [31] Maddala, G. S. (1983). Limited-dependent and Qualitative Variables in
Policy Research Working Paper, 2868, 1–54. Econometrics, Cambridge University Press.
[7] Crandall, R. & Waverman, L. (2000). Who pays for universal service? [32] McCormick, D. & J. Onjala (2007). Methodology for Value Chain
When telephone subsidies become transparent. Washington, DC: Analysis in ICT Industry Frameworks for the Study of Africa, Institute
Brookings Institution Press. for Development Studies University of Nairobi
[8] Eriksson, R., Kaserman, D., & Mayo, J. (1998). Targeted and untargeted [33] Milne, C. (2006). “Telecoms demand: measures for improving demand
subsidy schemes: Evidence from postdivestiture efforts to promote in developing countries – a toolkit for action”. Discussion Paper WDR
universal telephone service. Journal of Law and Economics, 41, 477– 0603, Main Report (World Dialogue on Regulation).
502. [34] Mitchell R. & Carson R. (1984). A Contingent Valuation Estimate of
[9] Esselaar, S., and Stork, C. (2005). Mobile Cellular Telephone: Fixed- National Freshwater Benefits – Technical Report to the US
line Substitution in Sub-Saharan Africa, The Southern African Journal Environmental Protection Agency. Resources for Future, Washington
of Information and Communication, Issue 6, pages 64-73, ISSN 1607- D.C.
2235. [35] Niininen, P. (2001). Computers and economic growth in Finland, in
[10] Fabiani, S., F. Schivardi and S. Trento (2005), ICT adoption in Italian Information Technology, Productivity, and Economic Growth:
manufacturing: firm-level evidence, Industrial and Cooperate Change, International Evidence and implications for economic Development,
pp.1-25. Matti Pohjola, ed., oxford University Press.
[11] Frank R. & Bernanke B. (2003), “Principles of Economic”. McGraw- [36] OECD (2003). ICT and Economic growth. Evidence from OECD
Hill, Boston countries, industries and firms. Paris (France).
[12] Garbacz, C., & Thompson, H. G., Jr. (1997). Assessing the impact of [37] Oliner, S.D. and D.E. Sichel (2000). The resurgence of growth in the
FCC lifeline and link-up programs on telephone penetration. Journal of late 1990s: Is Information technology the Story?, Journal of Economic
Regulatory Economics, 11, 67–78. Perspectives, Vol.14, pp. 3-22.
[13] Garbacz, C., & Thompson, H. G., Jr. (2002). Estimating telephone [38] Ono, H. (2003). Gender and the Internet. Social Sciences Quarterly, 84,
demand with state decennial census data from 1970–1990. Journal of 111–121.
Regulatory Economics, 21, 317–329. [39] Orbicom (2003). L’observatoire de la fracture numérique... et au-delà.
[14] Garbacz, C., & Thompson, H. G., Jr. (2003). Estimating telephone Conseil national de la recherche du Canada, Québec (Canada).
demand with state decennial census data from 1970–1990: Update with [40] Panzar, J. C. (2000). A methodology for measuring the costs of universal
2000 data. Journal of Regulatory Economics, 24, 373–378. service obligations. Information Economics and Policy, 12, 211–220.
[15] Garbacz, C., & Thompson, H. G., Jr. (2005). Universal [41] Parsons, S. G. (1998). Cross-subsidization in telecommunications.
telecommunications service: A world perspective. Information Journal of Regulatory Economics, 13, 157–182.
Economics and Policy, 17, 495–512. [42] Perl, L. J. (1983). Residential demand for telephone service. White
Plains, NY: National Economic Research Associates.
402
[43] Piccoli, G., Ahmed, R. & Ives, B. (2001). Web based virtual learning
environments: a research framework and a preliminary assessment of
effectiveness in basic IT skills training. MIS Quarterly, 25, 401–426.
[44] Pohjola, M. (2001). Information technology and economic growth: A
cross-country Analysis, in: Information Technology, Productivity, and
Economic Growth: International Evidence and implications for
economic Development, Matti Pohjola, ed., oxford University Press.
[45] Roller, L. H. & Waverman, L. (2001). Telecommunications
Infrastructure and Economic Development: A simultaneous approach.
The American Economic Review, Vol. 91, No 4, 909-923.
[46] Rosston, G. L., & Wimmer, B. R. (2000). The state of universal service.
Information Economics and Policy, 12, 261–283.
[47] Row R. & Schulze W. & Breffle W. (1996). A Test for Payment Card
Biases. Journal of Environmental Economics and Management, Vol. 31,
pp. 178-185.
[48] Sichel, D. (1997). The Computer revolution- An Economic Perspective,
Washington D.C.: Brookings Institutions Press.
[49] Taylor, L. D. (1994). Telecommunications demand in theory and
practice. Dordrecht: Kluwer Academic Press.
[50] Taylor, L. D. (2002). Customer demand analysis. In M. Cave, S.
Majumdar, & I. Vogelsang (Eds.), Handbook of telecommunications
economics (pp. 97–142). Amsterdam: Elsevier Science.
[51] Taylor, L.D. and Kridel, D.J. (1990). “Residential Demand for Access to
the Telephone Network,” in Telecommunications Demand Modeling,
ed. by A. de Fontenay, M.H. Shugard, and D.S. Sibley, North Holland
Publishing Co., Amsterdam.
[52] Tegegne G.E., (1999). “Willingness to pay for environmental protection:
an application of Contingent Valuation Method (CMV) in Sekota
District, Nothern Ethiopia”. Ethiopian Journal of Agricultural
Economics Vol 3: 123-130
[53] UNCTAD (2006). “Rapport 2006 sur l’Economie de l’information”,
UNCTAD/SDTE/ECB/2006.
[54] Valletti, T. M. (2000). Introduction: Symposium on universal service
obligation and competition. Information Economics and Policy, 12, 205–
210.
[55] Venkatachalam L. (2004). The Contingent Valuation Method - A
Review. Environmental Impact Assessment Review, Vol. 24, pp. 89–
124.
[56] Venkatesh, V. & Morris, M.G. (2000). Why don’t men ever stop to ask
for directions? Gender, social influence, and their role in technology
acceptance and usage behavior. MIS Quarterly, 24, 115–139.
[57] Wallsten, S. (2001). Telecommunications investment and traffic in
developing countries: The effects of international settlement rate
reforms. Journal of Regulatory Economics, 20, 307–323.
[58] Waverman, L., Meschi, M. & Fuss, M. (2005). The impact of telecoms
on economic growth in developing countries. In Africa: The Impact of
Mobile Phones: Moving the Debate Forward, The Vodafone Policy
paper series, No.2.
[59] Wooldrige, J. (2006). Introductory Econometrics A modern Approach,
4th edition, South-Western Cengage Leraning, Canada, ISBN 13 : 978-
0-324-58548-3.
403
IX. APPENDIX
TABLE 5: PROBIT MODEL SPECIFICATION
Variable Variable Type of Variable Expected Comment
Name sign
Dependent Mobile Dichotomous variable: Na
Variable (M) Individual has a mobile or active SIM
card = 1, 0 otherwise
Independent Income Continuous variable: Individual income positive Income is assumed to be
Variables (I) in US$ based on end of year nominal the main barrier to
exchange rates (source IMF) adoption
Gender (G) Dichotomous variable: female = 1, male Not The assumption is that
=0 significant there is no gender
difference when it comes
to mobile adoption
Age (A) In years Negative Technology being a
domain for the younger
generation
Tertiary (T) Dichotomous variable: highest education Positive Education should enable
being tertiary = 1, otherwise 0 individuals to use mobile
Secondary Dichotomous variable: highest education phones, while also being
(S) being secondary = 1, otherwise 0 an indicator for income,
Primary (P) Dichotomous variable: highest education model would drop
being primary, remedial or traditional variables if multi-
=1, otherwise 0; collinearity persists
Vocational Dichotomous variable: highest education
(V) being vocational =1, otherwise 0
Rural (R) Dichotomous variable: rural = 1, urban Negative Mobile network coverage
=0 is less in rural areas
compared to urban (major
urban and other urban)
ones and less adoption is
expected therefore
Social Dichotomous variable: social network Positive Membership in social
Network (as church groups, sports clubs etc.) networks might increase
(SN) membership = 1, otherwise 0 communication need
Constant Captures various factors that either are Negative Captures no education, but
constant for a country such as the price also urban
for mobile or fixed telephony or that
serve as reference for dichotomous
variables such as the educational
variables and rural-urban location.
TABLE 6: NATIONALLY REPRESENTATIVE PROBIT RESULTS FORE EACH COUNTRY USING SAMPLING WEIGHTS (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)
Country Pseudo Prob Positive Coefficients + = significant at 0.1 level, ++ = significant at 0.05 level,
R2 > +++ = significant at 0.01 level
chi2 Negative Coefficients - = significant at 0.1 level, -- = significant at 0.05 level, ---
= significant at 0.01 level
I A G T S P V R SN Constant
Benin 0.396 0.00 +++ ++ No +++ +++ +++ +++ --- no ---
Botswana 0.16 0.00 + No No ++ +++ No Skip -- +++ No
Burkina Faso 0.253 0.00 +++ --- --- +++ +++ +++ Skip --- No No
Cameroon 0.258 0.00 +++ ++ No +++ +++ +++ ++ --- ++ ---
Côte d'Ivoire 0.292 0.00 +++ -- No +++ +++ No +++ --- +++ No
Ethiopia 0.452 0.00 +++ No No +++ ++ No +++ --- No ---
Ghana 0.158 0.00 No - No +++ +++ +++ ++ --- ++ No
404
Kenya 0.258 0.00 +++ No No +++ +++ .++ +++ No ++ ---

Mozambique 0.214 0.00 +++ No +++ +++ +++ +++ No --- No ---
Namibia 0.224 0.00 +++ No No +++ +++ No No --- No No
Nigeria* 0.172 0.00 +++ -- No +++ +++ No + No ++ No
Rwanda 0.244 0.00 No No No Skip +++ +++ Skip --- No ---
Senegal 0.215 0.00 ++ No --- +++ +++ ++ ++ --- No No
South Africa 0.171 0.00 +++ --- +++ +++ +++ No No --- No No
Tanzania 0.172 0.00 +++ + - Skip +++ +++ + --- +++ ---
Uganda 0.357 0.00 +++ + No +++ ++ No +++ --- No ---
Zambia* 0.2894 0.00 +++ No No Skip +++ +++ +++ Ski +++ ---
p
TABLE 7
WILLINGNESS AND ABILITY TO PAY
Average
willingness and Average National
ability to pay for a expected cost of New users at New users at
mobile handset in a mobile 20 US$ for an 15 US$ for New users at 10
US$ handset US$ handset an handset US$ for an handset
Benin 7.45 11.44 124,972 487,176 677,715
Botswana 19.14 27.38 119,014 196,496 228,203
Burkina Faso 8.92 12.84 427,032 1,242,397 1,451,446
Cameroon 15.11 22.16 864,053 1,728,316 1,865,876
Côte d’Ivoire 29.70 30.06 3,057,420 3,539,351 3,914,283
Ethiopia 6.06 64.19 1,436,628 1,637,668 2,644,673
Ghana 14.02 23.15 1,283,271 1,469,652 1,841,837
Kenya 16.98 26.68 2,857,406 4,160,498 5,658,430
Mozambique 4.00 23.2 56,457 79,895 287,147
Namibia 24.64 25.12 162,992 192,395 232,584
Nigeria* 5.65 12.57 356,907 1,004,573 2,527,884
Senegal 19.55 2543.0% 1,336,691 2,169,548 2,301,775
South Africa 17.44 32.41 2,549,812 3,251,782 3,991,768
Tanzania 10.89 17.3 1,422,927 2,102,510 3,272,065
Zambia* 17.42 22.43 682,864 1,061,607 1,598,555
Benin 7.45 11.44 124,972 487,176 677,715
TABLE 8: MOBILE EXPENDITURE - ROBUST REGRESSION RESULTS FOR EACH COUNTRY USING SAMPLING WEIGHTS (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)
Country R2 Prob Only significant coefficients are being displayed: * = significant at 0.1 level, ** = significant
>F at 0.05 level, *** = significant at 0.01 level
Y A G WP PP F R SN
Benin 0.399 0.000 0.35*** 1.07*** -0.27***
Botswana 0.298 0.000 0.46*** -0.26**
Burkina Faso 0.206 0.000 0.33*** 0.28* 0.35***
Cameroon 0.22 0.000 0.27*** -0.32** -0.22* 0.54* -0.4*
Côte d'Ivoire 0.196 0.000 0.36*** -0.02* 0.45** 0.37* 0.34*
Ethiopia 0.353 0.000 0.29*** -0.01** -0.47*** 0.23*
Ghana 0.145 0.000 0.47*** -0.01**
Kenya 0.251 0.000 0.33*** -0.02*** 0.44*** 0.47***
Mozambique 0.189 0.000 0.11** -0.03*** -0.72*** 1.04***
Namibia 0.449 0.000 0.43*** -0.23* 0.37**
Nigeria 0.458 0.000 0.71*** 0.24*** -0.41* 0.24***
Rwanda 0.339 0.000 0.55*** 0.66***
Senegal 0.266 0.000 0.43*** 0.247*
South Africa 0.477 0.000 0.47*** -0.02*** 0.2** -0.2*** 0.3***
Tanzania 0.127 0.000 0.33*** 0.84**
Uganda 0.292 0.000 0.39 *** 0.48** -0.24* 0.43**
Zambia 0.378 0.000 0.52*** -0.01** 0.69***
405
TABLE 9: MOBILE PENETRATION (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)

Monthly Current
average market 16+ with Total
mobile in US$ more than Number of Number of Share of
expenditu million 16+ with mobile phone one SIM duplicated active SIM Prepaid
re in US$ or active SIM card SIM Cards cards users
Benin 8.33 11.38 1,365,851 30.2% 496,917 1,173,454 2,047,486 95.96%
Botswana 10.18 6.67 654,737 59.5% 61,670 129,323 722,390 99.28%
Burkina Faso 5.84 10.77 1,844,701 27.2% 380,945 943,819 2,408,616 96.69%
Cameroon 7.14 21.29 2,979,597 36.5% 240,473 600,756 3,440,472 88.04%
Côte d'Ivoire 12.52 63.13 5,042,524 41.8% 762,295 1,741,585 6,037,870 91.77%
Ethiopia 3.81 5.29 1,387,910 3.2% 8,379 17,282 1,410,159 88.31%
Ghana 10.44 78.23 7,491,378 59.8% 832,341 1,969,676 8,691,409 99.83%
Kenya 10.41 112.11 10,772,696 52.0% 2,796,971 5,932,015 13,984,190 98.89%
Mozambique 6.26 30.47 4,865,758 25.7% 143,404 286,808 5,012,287 98.89%
Namibia 11.41 7.14 625,707 49.3% 39,090 83,807 670,424 89.95%
Nigeria* 10.88 686.54 63,101,014 77.3% 12,265,752 26,381,512 77,954,949 99.33%
Rwanda 6.02 3.13 520,259 9.9% 16,170 32,340 536,429 94.77%
Senegal 11.00 27.54 2,502,300 39.8% 125,251 291,243 2,705,744 99.70%
South Africa 15.88 320.49 20,185,135 62.1% 2,200,647 4,845,907 22,938,052 78.64%
Tanzania 7.44 30.79 4,138,338 21.5% 602,730 1,301,997 5,070,790 96.95%
Uganda 5.75 16.81 2,924,095 20.7% 526,378 1,097,654 3,505,813 97.84%
Zambia* 10.55 25.96 2,459,961 45.5% 110,683 253,279 2,605,368 99.7%
TABLE 10: MOBILE WILLINGNESS AND ABILITY TO PAY (SOURCE: RIA 2007/2008 HOUSEHOLD SURVEY)
Average Monthly
Number of 16+ without a mobile phone or active SIM monthly untapped
16+ without mobile willing and able to spend monthly: WTP in market in
phone or active SIM US$ US$
card 1US$ or more 2 US$ or more 5 US$ or more million
Benin 69.8% 3,162,099 2,797,101 1,402,507 161,217 2.94 8.26
Botswana 40.5% 446,140 311,446 199,511 66,192 4.28 1.47
Burkina
4,929,897 4,371,694 1,875,892 430,952 3.13
Faso 72.8% 13.71
Cameroon 63.5% 5,177,393 3,452,460 1,855,275 550,724 3.75 13.14
Côte d'Ivoire 58.2% 7,033,592 4,485,498 3,645,855 1,677,528 6.86 31.44
Ethiopia 96.8% 42,497,353 10,231,145 3,104,395 74,428 1.53 25.68
Ghana 40.2% 5,036,815 2,849,435 1,953,135 984,279 9.34 38.40
Kenya 48.0% 9,941,748 5,866,299 5,235,785 1,245,083 3.30 25.69
Mozambique 74.3% 14,078,222 1,407,840 1,199,765 376,037 2.96 6.70
Namibia 50.7% 644,056 275,364 247,254 71,171 4.88 1.35
Nigeria* 22.7% 18,541,687 7,989,151 6,234,941 5,128,000 6.09 65.25
Rwanda 90.1% 4,735,492
Senegal 60.2% 3,779,221 3,428,481 1,294,681 502,730 3.28 11.33
South Africa 37.9% 12,331,758 7,604,512 5,551,777 2,209,625 4.34 36.27
Tanzania 78.5% 15,066,652 5,560,959 4,750,935 1,064,087 2.61 21.42
Uganda 79.3% 11,174,801
Zambia* 54% 2,944,732 1,933,833 1,310,576 329,922 3.45 8.2
406
Numeric Paper Forms for NGOs

Gursharan Singh, Leah Findlater, Kentaro Toyama, Scott Helmer, Rikin Gandhi, Ravin
Balakrishnan
Abstract—Non-governmental organizations (NGOs) working possible. In this regard, information technologies frequently
in disadvantaged communities have a variety of data-collection play an important role in the storage, analysis, and display of
and analysis needs, for example, for performing surveys or painstakingly collected data. Microfinance institutions
monitoring programs. Because much of this data collection benefit from back-end databases that store client data [22];
occurs in environments with insufficient IT support and healthcare institutions have need to maintain medical
infrastructure, and among populations not always comfortable
records; and NGOs in general can benefit from longitudinal
with technology, paper forms rather than electronic methods
remain the predominant means for data collection.
data collected over the life of their programs.
We consider the design of machine-readable paper forms for Organizations have also considered the use of information
NGOs. We first examine the unique needs of NGOs that technology for the task of collecting the data itself. Micro-
interact with underprivileged populations through interviews finance accounting information systems [28] and healthcare
with eleven organizations and an in-depth investigation of one information on PDAs [2] are two significant examples. The
NGO’s specific form-filling requirements. These explorations expectation is that these efforts minimize transcription of
led to a focus on numeric forms – forms with questions data from paper forms and allow for more rapid analysis.
requiring responses largely constrained to numbers. There are, however, many who question the value of
We then present an experiment which evaluates how a
electronic means for data collection in poor environments.
variety of formats for numeric data would fare with users from
backgrounds similar to those who might fill out such forms.
One cost-benefit analysis of mobile devices used by
Our goal was to balance the tradeoff between ease-of-use microfinance institutions to interact with their clients
among our intended population and machine readability. suggests that the benefits of electronic technology for data
Combining the results of the experiment with an analysis of collection do not always outweigh the costs [9]. Others point
machine-readability from a technical perspective, we propose out that electronic mechanisms are distrusted by populations
the best numeric input methods for different NGO form filling who are used to physical evidence of transactions [23].
requirements. Paper, on the other hand, is a ubiquitous, low-cost, and
Index Terms— machine-readable forms, paper forms, well-understood medium. Even in the developed world,
input methods, ICT for development paper forms remain widely used for the purposes of
gathering information, despite ongoing advances in digital
I. INTRODUCTION technology. However, while it may be easy to collect data
Monitoring, evaluation, measurement, and self- on paper, transferring that data into a format suitable for
assessment are among several critical tasks for non-profit subsequent computer based storage and analysis remains a
organizations working in global development. Knowing the difficult problem that is currently typically resolved only by
nature of one’s impact, ideally with accompanying tedious manual data transcription.
quantitative information, allows for self-correction, reports In this paper, we first study how paper forms are used for
to sponsors and potential donors, and external influence. data collection in the context of non-profit non-
Most program assessment requires data collection in some governmental organizations (NGOs) that seek to monitor
form as a first step. Data collection can be tedious and their own programs. We performed interviews with eleven
expensive (in labor, time, and financial cost), and it is thus development-focused organizations involved in healthcare,
desirable to extract as much value from the effort as microfinance, education, and agriculture. We then probed
deeper with one of the organizations to better understand
Manuscript received September 22, 2008. Revised February 20, 2009. their data-collection pipeline. This investigation resulted in a
G. Singh was with Microsoft Research India, Bangalore, Karnataka,
India (email: gursharan@alumni.upenn.edu). proposal for data collection that uses paper for the “front-
L. Findlater is with University of British Columbia, Vancouver, BC, end” collection tasks, and subsequently uses a combination
Canada (email: lkf@cs.ubc.ca). of scanner and PC to digitize the data. Rather than attempt to
K. Toyama is with Microsoft Research India, Bangalore, Karnataka,
India (email: kentoy@microsoft.com). handle the myriad of possible paper forms right away, we
S. Helmer is with University of British Columbia, Vancouver, BC, instead focused on an important subset – forms used for
Canada (email: shelmer@cs.ubc.ca). collecting numeric input, which was found to be frequently
R. Gandhi is with Microsoft Research India, Bangalore, Karnataka,
India (email: riking@microsoft.com).
used by NGOs. Although many additional factors besides
R. Balakrishnan is with University of Toronto, Ontario, Canada (email: input type will determine the value of such a system in a real
ravin@dgp.toronto.edu). environment, our study provides a first step towards
building an effective system. We conducted a study of how
407
people with varying educational backgrounds responded to C. Data Collection for the Developing World
form-filling tasks using several different “input methods” In spite of the trend to place digital technology directly in
for recording numbers. Finally we analyzed the literature on front of underserved populations, a few researchers have
machine-readable forms, and gauged how each of the input focused on using paper instead. Parikh et al. [24]
methods would fare for subsequent automatic digitization of systematize paper-based data collection by using paper user
the data. interfaces and automated forms processing for microfinance
institutions. Mackay et al. [20, 21] propose using paper on
II. RELATED WORK top of tablet PC displays. Parikh et al.’s [23] CAM system
We review four categories of research on which our work demonstrates use of a camera-equipped mobile phone to
builds. First, a number of electronic devices and software read bar codes printed on paper forms. Bar codes cause the
tools have been designed to complement paper in phone to issue audio annotations, which instruct the user to
information work. Second, some user studies have compared transfer data on the paper form into the mobile phone.
digital and paper-based methods of data entry. Third, some DeRenzi et al. [6] use a PDA based system to significantly
research has focused on using paper as an interface for increase adherence to medical protocols for pediatric health
systems that interact with underserved populations. And, care. These systems all incur a device cost for the person
finally, existing work on machine-readable forms will be doing the data collection.
reviewed in a separate section. Ellison et al. [7] stress the importance of participatory
monitoring and evaluation for a small scale NGO. Chand et
A. Complementing Paper
al.’s [5] Jadoo, is a paper-only exploration of how structured
The importance of paper as a medium to capture data is content on paper can be made easy to deal with even for
recognized, even in technology-heavy developed-world people with little formal education. Our early investigations
environments: “Rather than pursue the ideal of the agree with this work on the value of paper, and its likely
paperless systems, we should work toward a future in which continued use by NGOs.
paper and electronic document tools work in concert and Our research builds on the above work and focuses on the
organizational processes make optimal use of both” [25]. novel problem of designing paper forms which should, on
And, indeed, many tools have addressed this need. the one hand, allow easy machine readability, and on the
Early research by Johnson et al. [13] showed how paper other hand, be easy to fill in for people with diverse
documents could be stored, indexed, distributed and educational backgrounds.
processed by capturing them as images with scanners.
Wellner’s [29] classic Digital Desk system used a camera III. NGOS AND FORMS
and a projector to record input and project documents.
Over the last several years, we have both formally and
Guimbretière’s PADD [12] attempts to bridge the gap
informally interviewed eleven organizations – five NGOs,
between the paper and the digital world. Built on top of
four microfinance institutions, and two clinics – in India
PADD, Liao et al.’s [17] Papiercraft implemented a gesture
with respect to their data-collection needs and experiences
interface for natural editing. In a lighter-weight approach,
with both digital and paper-based forms. These
the Anoto pen [1] uses paper printed with unique marks and
organizations varied significantly in size and in their use of
a small camera in the pen to synchronize changes on the
forms (from requiring a handful of forms filled in per
paper to a digital version of the document. Arai et al. [3]
month, to hundreds of thousands per year), but all had in
show how a camera-assisted pen can capture words on paper
common that they had data-collection needs and continued
by hyper-linking them to other content. Koike et al. [15]
to utilize paper forms for this purpose, despite isolated trials
used projected Venn diagrams to record input.
with electronic data collection.
Although these systems combine the strengths of paper
Below, we outline what we learned from these
and electronics, their benefits come with additional costs,
organizations, and go into some detail about the data-
which is a concern for NGOs.
collection system for one organization that typifies the NGO
B. Digital versus Paper experience with forms.
Some researchers have compared paper and digital A. Overall NGO Experiences with Forms
systems for data entry, and overall tend slightly in favor of
The need for data collection is widespread among NGOs,
paper. Galliher et al. [8] found that people are more likely to
and means for doing it efficiently and subsequently being
complete paper forms than digital ones. They cite technical
able to analyze the results easily are, in the words of one
difficulties, as well as stolen or lost devices. On the other
NGO, “the need of the hour” [14]. Despite the prevalence of
hand, errors of omission were more common on paper.
paper forms, however, NGOs do not seem completely
Shelby-James et al. [26] disprove an often-made claim that
satisfied with their systems (or lack thereof) for
electronic data capture is more accurate than paper-based
incorporating collected data in their routines.
methods. They found error rates with handheld computers
were over sixty times that for paper-based data entry.
408
The goals of most data collection fall into one of several of data collection. In another case, a bicycle courier was
categories. For surveys, the intent is simply to understand paid 15 cents per village to collect sheaves of paper records
the state of a population as it exists. For monitoring and and bring them back to the NGO office.
evaluation of projects, the objective is accountability, both
in terms of whether NGO staff and beneficiaries are Infrastructure and maintenance: Among the organizations
performing tasks they have committed to, or whether the that had either experience with or expressed interest in
intended outcomes of a project are being accomplished. For information technology (IT) systems for data collection and
baseline surveys, the collected data helps identify various processing, all cited challenges with power and maintenance
parameters and tune efforts before work on a particular issues. Fig. 1 shows a person filling out a form with light
project is started by an NGO, for example, the majority from a flashlight, a frequently encountered situation.
occupation of the population in the area.
Forms can be filled in by different people with different
backgrounds, but for the most part, there seem to be two
classes of people filling in forms: those who are paid by the
NGO to perform data collection and those who are direct
beneficiaries of the NGO’s programs. The former category
consists of people who are almost always literate, and
typically have completed secondary school; some have
advanced degrees, as well. The latter range wider in their
backgrounds, but frequently have not had much formal
education. Anyone asked to fill in a form will likely be
literate, but there are cases where the form-fillers can just
barely sign their name and write numerical digits. Fig. 1. A participant fills out a form during a power outage.
Furthermore, their handwriting may not be very good.
Forms involve recording of numerical data (e.g., age,
price, income, dates, durations), textual data (e.g., name, Training, data accuracy, and consistency: Data-collection
free-text comments), and binary or multiple-choice tasks often require some training, and systems involving
questions. It is worth noting that a significant fraction of the digital devices only add to this need. Training is a time- and
questions asked on forms appear to be reducible to numeric management-intensive activity that NGOs invest heavily in.
information, with multiple-choice answers. The key Ensuring data accuracy and consistency of routine
exceptions are names and free-form comments. Also, in surveys is another headache for many NGOs. Staff, who
some cases, photographs were part of the data collected. themselves are not paid particularly well, may have little
Overall, the repeated concerns expressed by NGOs motivation to perform data-collection tasks with care. In
regarding data collection were: cost, time, training, data many cases, the data-collection staff itself needs to be
accuracy and consistency, storage, and means of data monitored, to ensure that data is being correctly collected
analysis. Among those NGOs who had experimented with according to specification.
electronic systems, these issues were again highlighted, in Even dedicated staff, however, often embed form-filling
addition to difficulties with infrastructure and maintenance. tasks into what can otherwise be busy daily schedules. Thus,
We discuss each of these concerns briefly below. data accuracy and consistency can suffer.
Cost: NGOs are run on tight budgets, and many expressed Data storage and analysis: Almost all of the NGOs we
concerns even about the cost of paper and printing. (Some of spoke with understood the value of digitizing data once it
the NGOs we spoke with explicitly requested a ream of was collected. Many employed data-entry staff whose sole
“white paper” as a meaningful gift in exchange for their job was to take paper forms and convert them into digital
time.) Digital equipment, of course, can be prohibitively format; others outsourced these tasks to transcription
expensive, and even those that can afford the capital costs services; in some organizations volunteers handled this task.
have difficulty with costs of maintaining technology. One These methods are all costly – for some organizations, the
interesting point here is that centralized technology in an costliest part of the data pipeline – or irregular, and
NGO office is easier to justify than devices that must scale depressingly often, we were shown stacks of paper forms
in proportion to the number of respondents (or equivalently, that had yet to be processed; in some cases, they had been
data-collection staff). gathering dust for years.
The cost of staff hours for data collection also adds up, Analysis of data most frequently involved the use of
although these costs match labor costs of the local area, and spreadsheet software as well as tables in word-processing
are typically very low by developed-country standards for software filled in by hand. Two microfinance institutions
hourly wages. Among the NGOs we consulted, data used custom software, which produced fixed reports once
collecting staff were paid as little as US$4 for a 10-hour day the data was entered.
409
Other issues: One thoughtful NGO head mentioned that videos, their interests, suggestions, and so on. The junior
there would be some value in systems that helped staff then summarizes feedback from every session onto a
organizations develop good questionnaires for data form, shown in Fig. 2. These per-session forms are analyzed
collection. He felt that a lot of data collected by NGOs was by the senior staff weekly, and finally aggregated onto a
done without a clear understanding of what information was monthly form. These forms are currently filled in plain text
most useful for evaluation purposes [14]. in Kannada (the local language in the region). The filled
forms are then sent to the regional office. The data is entered
B. One Particular NGO’s Experience
into a database by a human transcriber and is then analyzed
GREEN Foundation is a small NGO based in Bangalore, to spot trends and results in the villages of operation.
India, whose mission is to spread sustainable agricultural We observed that the form-filling sessions in villages tend
practices among rural farmers. With an annual budget of to become interactive “classroom sessions” (Fig. 3) leading
about US$100,000, they have 15 field staff tasked with to vital exchange of information between the NGO staff and
covering an operational area of approximately 100 villages. the participants. Forms filled by the farmers convey what
Recently, they began a project that involves group they want to be taught, and forms filled by on-field staff
sessions with farmers, where facilitators mediate discussion help document their staff progress.
based on video content [11]. As part of the monitoring and Overall, our key findings were that: (1) data collection
evaluation of this project, both facilitators and farmers are and form-filling are important activities for many NGOs; (2)
requested to fill in paper forms. The program has been cost and ease-of-use are major concerns, often preventing
running for over a year, in each of 12 pilot villages, with technology-heavy systems; and (3) digitized data is desired,
typically three sessions per week. The first author visited but digitizing data is the bottleneck for data-collection
these sites several times over a period of two months to efforts. These findings confirm findings from earlier work
better understand how the forms were used. and additionally identify an important problem faced by
Their case is particularly interesting for us, because it many NGOs.
provides information about the two broad classes of form-
fillers: (1) two groups of paid staff of the NGO, all of whom
are literate (senior staff with ten or more years of formal
education, and junior staff, with at least eight years of
education), and (2) smallholder farmers, who earned no
more than $2 a day, and rarely had more than six years of
education; many were all but illiterate.
Fig. 3. The verandah of a house used for meetings and data collection.
IV. PROPOSED DATA COLLECTION PROCESS

A. Proposed System
Based on our findings from NGO interactions, we believe
that the following combination of paper forms and hardware
can solve many of the data-collection challenges outlined
above, while making a good tradeoff among the constraints.
For hardware, one PC, a scanner, and a printer is required.
We then envision that specially designed paper forms will
be printed and used in the field to collect data. Once
completed, the forms are scanned (a scanner with an auto-
feeding mechanism would be ideal), and special software
digitizes the filled-in content. To do this, some software
innovation is required to allow those creating the forms to
easily design customized, machine-readable forms. In
addition, there must be software that can robustly digitize
Fig. 2. A GREEN form filled by the junior staff.
any content that is scanned.
After every video session, the farmers attending the Due to the limitations of technology for digitizing free-
session are asked questions related to the practices shown in form handwriting, the expectation is that any given form
410
will be implemented using multiple-choice or numeric un-coded Bubbles). Coded marking methods, in contrast,
responses as much as possible, because these responses are only provided a single row for all digits (see Table 1a). The
comparably easy to digitize. Free-form text will be given advantage of the un-coded methods is that they allow for a
space within blank rectangles on the form, which can be complete set of numbers: for example, with 4 rows users can
scanned and stored as images (and not converted to enter any number from 1 to 1000. The coded marking
electronic text). reduces the set of numbers that can be input: for example,
The proposed system keeps costs low. No per-staff device “58” and “85” would appear the same when entered as
is required, and all of the equipment is available as mass- coded input and numbers with repeating digits, such as “22”,
market off-the-shelf hardware, which helps to keep costs cannot be input at all. However, the advantage of coded
low and alleviates technical maintenance needs. Since marking is that it greatly reduces the visual complexity of
NGOs that collect data would like to have it in digital form the form and saves physical space, which in turn saves
eventually, the willingness to invest and maintain at least paper. Coded marking could be useful for situations where
one PC is assumed (many will already have a PC). Printers the numeric data is nominal rather than scalar. If the form is
and scanners add costs that are small compared with the PC used to record attendance at village meetings, for example,
itself, and paper is the only additional ongoing cost. The each individual may have an identification code assigned to
system would maintain all of the advantages of paper, them. Since these are nominal values they could be assigned
namely its low cost, a well-understood “interface”, the lack with the goal of entering them as coded input.
of need for power or maintenance, and robustness in the TABLE I
field. Finally, the proposed system addresses the data-entry NUMERIC INPUT METHODS CONSIDERED.
bottleneck by providing an automated means of digitizing
much, if not all, of the data. a. Coded Bubbles
This proposal still leaves us with the following challenge:
on the one hand, existing techniques for machine reading of
hand-marked forms is reliable only when the forms are b. Un-coded Bubbles
designed and filled out in a particular way (e.g., “bubbles”
on standardized test forms) or when digits are neatly written
(e.g., post-office automated digit-reading systems for sorting Coded
c.
mail); on the other hand, the groups that we anticipate will Circles
fill in these forms are less familiar with standardized forms,
and many have borderline penmanship.
This is a non-trivial challenge that requires both d. Un-coded Circles
engineering and interface design. In the remainder of this
paper, we consider how best to handle numeric input on
paper forms with the goal of achieving a reasonable tradeoff Coded
e.
Checkbox
between ease of user comprehension, user accuracy, and
machine readability.
Un-coded
f.
V. NUMERIC INPUT METHODS Checkbox
We considered 10 different methods for entering numbers

on paper forms, shown in Table 1. These were various Coded
g.
Ticks
versions of numeric entry, as commonly found in forms, and
ranging from those that were specifically designed for
h. Un-coded Ticks
machine readability (e.g., bubbles), to those that were
handwritten. Eight were marking-based input methods,
where the user marked the desired number either by filling i. Digit per box
in a bubble, circling the number, drawing a checkmark
inside a box, or ticking the number (Table 1a-h). The
j. Digits in one box
marking-based methods were tested in both coded and un-
coded versions, as described below. We also tested 2
handwritten methods (Table 1i-j): digit per box, where each
VI. TECHNICAL ISSUES WITH MACHINE READABILITY OF
digit was written in a separate box, and digits in one box,
NUMERIC INPUT
where all digits were written in a single box.
The un-coded marking methods provided a row for each Designing paper forms for automatic processing is a
digit, so, for example, a 2-digit number would require two mature field in the developed world, with successful
rows of input (Table 1b shows the number “58” entered in applications in mail sorting using postal codes (e.g., [27]),
academic marking systems, and systems for processing
411
applications (e.g., [10]). Moreover, there are numerous Kannada, an Indian language of interest to the NGOs we
commercial software packages available that require little surveyed, the recognition rate can be less than 90% [16] and
more than a PC and a document scanner. The process itself as mentioned earlier may not be supported by commercial
involves two stages. The first is form registration, which software. It should also be stressed that accuracy rates in
aligns the document so that values can be extracted. The ICR are stated for sanitized data sampled from a completely
second, and more challenging task is extracting the values different distribution of people. It is certain that illiteracy,
from particular locations in the document. These values are poor handwriting, and form degradation will reduce the
extracted using either optical mark recognition (OMR) or accuracy and this adds uncertainty as to whether the
intelligent character recognition (ICR), depending on the technology is applicable in this setting.
input method.
Processing forms in the developing world, however, adds VII. EXPERIMENT
numerous challenges. The cost and ease of deployment may Our goal in designing paper forms is to achieve a
increase since many commercial applications may lack reasonable balance between ease of user input and machine
support for local languages. Another challenge, based on our readability. We thus also conducted a controlled study to test
NGO interviews, is that since forms may be filled out in less user performance, accuracy and preference with respect to
than ideal conditions, they may be wrinkled or dirty, which the 10 different input methods.
can affect both the form registration and value extraction.
Commercial software packages also generally assume that A. Participants
the end users (the individuals filling out the forms) are We recruited 40 participants who ranged in age from 17-
literate and have a certain level of education. These users are 50 years (M = 26.2). Their formal education varied from
expected to have some level of competency in filling out, four years of schooling to undergraduate university level.
which is not always the case in the developing world. All of Most of them spoke at least three of the following
these challenges need to be weighed against the ease of use languages: Hindi, Telugu, Kannada, Sinhala, Tamil,
for different input methods. Bengali, Konkani and English. Every participant could write
The un-coded and coded checkbox and bubble input in at least one language, although many had not held a
methods could be processed using OMR. In general, this writing implement since finishing school. A very basic
method of input is nearly 100% accurate, assuming the user vision test (reading a series of numbers aloud) was used to
has followed the instructions [4]. This rate is reported for screen participants before starting the experiment. The
systems requiring specialized scanners and forms, which is participants came from disparate professional backgrounds:
not applicable for most NGO use; systems likely to be used auto-rickshaw (three-wheeled mini-cab) drivers, farmers,
by NGOs will have less accuracy. In addition, in the restaurant workers, security guards, housekeeping personnel,
developing world, users will likely be less familiar with such cab drivers, army soldiers, and machine operators.
forms, thus resulting in more user errors for OMR Participants’ literacy levels, ages, genders, and occupations
processing (for more detail on errors see Results section, were recorded. Participants were compensated with a small
below). Moreover, if the forms are degraded because they gift for their time.
have not been filled out or stored in antiseptic conditions,
B. Task
less specialized systems may be more likely to mistake
smudges, wrinkles and dirt for marks. The circling and The task was to enter 20 numbers on a paper form, with
ticking input methods we tested would also fall under OMR, five each of the following: 1-digit numbers, 2-digit numbers,
but these types of marks are not generally supported in 4-digit numbers and 8-digit numbers. Numbers to be entered
commercial software. The reason for this is possibly that the were displayed on the forms themselves. A sample form for
recognition accuracy is poor given the variability in the the digit per box input method is shown in Fig. 4. The
location and form of the marks, especially when contrasted numbers included in the tasks were randomly generated with
to checkboxes or bubbles. no digit being duplicated in a number (for the sake of the
The digits in one box and digit per box input methods, on coded input methods) and every participant saw the same set
the other hand, require ICR processing, which is less of numbers (though not in the same order). No participant
accurate. ICR technology leans upon advances in machine entered the same set of numbers for more than one type of
learning that have allowed systems to learn how to input method.
discriminate characters from samples of handwriting, and is C. Experimental Design
still an active area of research [18]. Currently, even the most
A single factor within-participant design was used: each
accurate ICR systems are only capable of 98% accuracy
participant completed the experimental task with 10 forms,
[19], and this is the case for digits written by a more literate
one for each input method. Presentation order was
population than our target users. In the case of forms
counterbalanced using a Latin square design and participants
requiring recognition of characters of partly non-alphabetic
were randomly assigned to an order.
systems, the accuracy is much worse. For example, for
412
Finally, at the end of the study a short interview was

conducted to compare the experience of using different input
methods.
Fig. 5. A participant in a village filling a form on floor – locally known as

the ‘free desk’.
VIII. RESULTS
Participants were divided into two educational groups:
those with up to middle-school education (7 years or less, M
Fig. 4. Sample digit per box form. The numbers above each set of boxes are
the numbers to be entered for the experimental task. = 6.1), and those with more (M = 11.9). A 2x10x10
(educational group x input method x presentation order)
D. Measures repeated measures (RM) ANOVA on the main dependent
We measured speed and accuracy, and asked participants variable of speed showed no significant main or interaction
to provide preference feedback. Speed was measured as the effects of presentation order, so we simplify our analysis by
time to complete the entire task for each input method and only examining the effects of education group and input
was measured manually with a stopwatch. Errors were method (2x10 RM ANOVAs). All pair-wise comparisons
counted after the forms were complete. A mark on a number were protected against Type I error using a Bonferroni
that logically should not be there, or would obviously make adjustment. Where df is not an integer, a Greenhouse-
a machine register the input incorrectly was counted as an Geisser adjustment for non-spherical data has been applied.
error. Subjective feedback was collected for each input One outlier in each of the low and high education groups
method using 5-point Likert scales on difficulty, confusion, were excluded because they were more than two standard
and perceived speed. deviations away from the mean on the dependent variable of
speed. Thus, we report on data from 38 participants (10 in
E. Procedure
the low education block and 28 in the high education block).
Experimental sessions took 45–80 minutes per
participant, as they were all given as much time needed to A. Speed
complete or quit the task. Most of the participants were from Average time to complete the forms with each input
villages in north and south Karnataka and we chose a quiet, method is shown in Fig. 6. Participants in the lower
isolated environment in which to conduct the sessions in education group took 31.8 minutes on average to complete
each of these villages (Fig. 5). The remaining sessions were all forms, while the higher education group took only 21.1
conducted in an office environment, where participants were minutes on average. Note that we ran the RM ANOVA on a
called in. We ensured that there was sufficient lighting in log transform of the speed data since the original speed data
both locations. violated the homogeneity of variance assumption
Participants were first given a background questionnaire (significant Levene’s test).
to collect demographic information. Then, for each of the 10 Both the education level of participants and the input
input methods, participants were given time to examine the method significantly affected the time it took to enter
form and ask questions about the task. Once they were numbers into the forms (main effect of education block: F1,36
comfortable with the nature of the task, they were asked to = 23.3, p < .001, η2 = .393; main effect of input method:
begin. Times were recorded for each of the 4 subtasks, F4.99,180 = 58.8, p < .001, η2 = .620).
namely, 1-digit numbers, 2-digit numbers, 4-digit numbers More interestingly, some input methods were relatively
and 8-digit numbers. After each input method, a better for the higher education group than for the lower
questionnaire was used to collect subjective feedback. education group (interaction effect between input method
413
and education: F4.99,180 = 2.37, p = .041, η2 = .062). To problematic for machine readability, as discussed in Section
understand which input methods were better within each VI. We counted three types of errors: (1) when the number
group of participants, we performed pair-wise comparisons of digits in an entered number was large (four or eight)
and summarize the significant results (p < .05) as follows. participants often found it hard to match the positions of
digits and rows in un-coded versions (Fig. 8a) and marked
two numbers in the same row; (2) ill-formed numbers and
numbers flowing out of boxes were also problematic (Fig.
8b-c); (3) sometimes it was difficult to distinguish which
number had been marked on the control (Fig. 8d).
Fig. 6. Task completion time by education and input method. (N = 38)
1) The handwritten input methods were generally fast, and

this result was clearest with the higher education group.
These two methods were not found to be any different Fig. 7. Average error rate by education and input method. (N = 38)
from each other in terms of speed. For the higher
education group both handwritten methods were faster There was a significant main effect of input method on
than all other methods. The lower education group, on the number of errors (F2.38,85.5 = 3.02, p = .046, η2 = .077),
the other hand, was not as relatively fast with digit per but using a Bonferroni adjustment indicated that no pair-
box: it was not found to be different from several of the wise comparisons were significant. No main or interaction
coded input methods (circles, checkbox and ticks). effects of education group were observed.
2) The bubble input methods were the slowest for the C. Subjective Measures
higher education group, but did not have as large a
Difficulty: Participants generally found all methods easy
negative impact on the lower education group. For the
to use: on a 5-point scale from easy to difficult the average
higher education group, both un-coded and coded
overall rating was 2.0 (SD = 0.8). Some methods were felt to
bubbles were slower than all other input methods with
be more difficult (a 2-way RM ANOVA showed a main
only one exception (coded bubbles and coded checkbox
effect of input method: F5.12,184 = 3.94, p = .002, η2 = .099).
were no different from each other). For the lower
Pair-wise comparisons show that the digits in one box
education group, however, the bubble methods were not
method was less difficult than un-coded bubbles and un-
found to be significantly slower than any of the other 6
coded ticking. No other significant main or interaction
types of marking-based input methods.
effects were found.
3) Coded input methods were also generally faster than
Confusion: Participants generally claimed not to find the
their un-coded counterparts, suggesting that the added
input methods to be confusing: the average rating on a 5-
complexity of having several lines of input instead of
point scale was 4.2 (SD = 1.8), where 5 represented “not at
one increased the difficulty of the task. For all coded
all confusing”. A 2-way RM ANOVA revealed that some
versus un-coded comparisons of ticking, checkboxes
input methods were more confusing than others, but pair-
and circling in both education groups, the coded
wise comparisons showed that the only statistically
versions were faster with one exception (coded ticking
significant comparison was that numbers in a box was less
for the higher education group was not faster than un-
confusing than un-coded ticking (p < .013). Though not
coded ticking).
statistically significant, digits in one box received the best
B. Accuracy rating from participants (M = 4.8, SD = 0.6). No other
Error rates were uniformly low; on average, less than 1 significant main or interaction effects were found.
error was made per form (Fig. 7). Fig. 8 shows examples of Perceived speed: In general, perceived speed matched
entries that were counted as errors and a variety of actual speed. There was a main effect of input method on
additional anomalies (not counted as errors) that could be perceived speed (F6.22,224 = 5.17, p < .001, η2 = .126). Pair-
414
a. Numbers filled in same row in an un-coded

input method b. Ill-formed numbers c. Numbers crossing boundaries
d. Ambiguous mark e. Impatiently filled bubbles f. Digits indexed on the number
g. Darkening on the number in bubble h. Overlapping circles Checkmark extending out of box
j. Double ticks k. Long, overlapping ticks l. Light numbers

Fig. 8. Examples of errors (a-d) and other anomalies.
wise comparisons showed that both the handwritten types One participant commented that she depended on her
(digit per box and digits in one box) were perceived to be husband for filling out textual information since she did not
faster than coded bubbles (p < .05). Digits in one box was know how to write text. However, she was comfortable with
perceived to be the fastest, since participants also felt it was numbers, and felt that if forms only required numeric entry,
faster than two of the un-coded input methods (bubbles and she could do it herself.
checkbox). Participants in the lower education group rated
E. Summary of Results
the tasks as slower than those in the higher education group,
a result that matched actual speed (main effect of education In terms of speed, degree of formal education impacted
block on perceived speed: F1,36 = 7.29, p = .010, η2 = .168). the effectiveness of individual input methods. In particular,
No significant interaction effect was observed between input both un-coded and coded bubbles were the slowest input
method and education group. methods for the higher education group but were no slower
than any other input method for the lower education group.
D. Qualitative Observations We had anticipated that the handwritten methods (digits in
We made a number of observations that, though not one box and digit per box) would be relatively faster for the
directly the focus of the study, illustrate practical issues that higher education participants than the lower education ones,
may be encountered with paper-based forms in this context. but both groups of participants, it turned out, were faster
Several of the less educated participants expressed great with these than with most other input methods. Coded input
enthusiasm with the study. They savored holding a pencil methods were also generally faster than un-coded methods
since they had some formal education but their jobs never and resulted in fewer errors. As expected, higher education
required them to write or read anything. Other participants participants completed the task more quickly than lower
were more apprehensive about holding a pencil because they education participants.
did not know how to write, but they agreed after being Note that since the lower education group had one third as
encouraged by earlier participants. many participants as the higher education group, there was
A few participants without computer backgrounds felt a less statistical power to detect differences among the input
natural inclination to add leading zeros where there were methods; this could explain why some pair-wise
unused “slots”. For example, in a preliminary investigation comparisons were significant for the higher education block
prior to the formal tests, they marked the month of February but not for the lower education block.
in the date field as “02”, filling both the input columns,
instead of only “2” and leaving a column blank. IX. DISCUSSION
Some circumstances of rural life affect the form-filling Combining our understanding of the state of the art for
task in unexpected ways. In one case, we experienced a machine-readability of numeric forms with the user study,
power outage during a preliminary field experiment (again we find that we can make the following recommendations
prior to our formal trials). One of the participants, who was for typical NGO conditions.
over 50 years of age, found it particularly difficult to At the highest level of granularity, if all of speed,
continue, although his vision was fine in bright light. machine accuracy and user accuracy are desirable
characteristics, coded checkboxes seem to be the best fit
415
irrespective of education level of users. If the users are analysis of challenges faced by NGOs around data collection
literate, digits in one box might be a better choice as it was and form filling, (2) a proposal for a semi-automated system
the fastest method for the more highly educated group, with that uses paper as the interface for data collection, and a
low error rate (zero for our participants), and if carefully scanner and PC for digitization, (3) an empirical study with
written these digits have reasonable machine readability 40 participants of how 10 different number-based input
characteristics. methods fare among people with both lower and higher
The relative performance differences between the higher educational backgrounds, and (4) an analysis of how easily
education and lower education groups suggest that there the forms filled using each of these input methods can be
may be a tradeoff between the benefits of structure and automatically digitized.
simplicity for different levels of education. The non- Our main findings are that users tend to prefer and
handwritten methods may be useful for low literacy perform best with techniques that require handwritten
participants since they do not require the same writing numbers. However, these are not the most easily recognized
proficiency as the techniques that require handwritten by a machine. If multiple factors such as speed and accuracy
numerals (digits in one box and digit per box). Our lower of entry, and machine readability are taken into account,
education participants had on average 6 years of schooling. coded checkboxes are likely the best option, although
In our experience from interviewing NGOs, we found that choosing a coded method limits the type of numbers that can
some users of their forms may have even less education, be captured (e.g., numbers with repeating digits are not
which could magnify the relative performance differences easily handled by the coded techniques we considered).
we found based on education. This highlights the need to Another key finding is that the educational background of
consider the educational background of the target user the user can significantly impact performance with different
population. input methods, and as such the demographics of the target
Although coded methods were in general quicker to fill user population should be carefully considered when making
than their un-coded counterparts, and also resulted in fewer decisions as to the type of input method to use.
errors by participants, they are not necessarily suitable for In our study, participants used the same input method
all types of forms. In particular, coded methods cannot across each form. Given the tradeoffs we found with respect
handle numbers with repeating digits. As such, if a form has to coded and un-coded methods, it would be interesting to
some fields that require entry of such numbers, it might be probe further into how people might perform with forms that
best for the sake of consistency to stick to an un-coded have a mixture of input methods, each optimized for the
method throughout the entire form. types of questions being asked on the form. It is unclear if
Interestingly, participants rated handwritten numbers best having more than one input method on a form would be
overall. Both digits in one box and digit per box methods overly confusing to users, particularly those with limited
were reported to be less confusing and easy to fill. In levels of literacy.
contrast, the bubble methods were not particularly favored, Finally, there are many other challenges that will need to
and were the slowest. be addressed in building an effective system to collect and
Taken as a whole, our results indicate that the proposed process data using paper forms as a front-end, and
system of using structured paper forms with subsequent automated input recognition as a back-end. Different types
automated scanning is likely viable. However, the results of data can pose different problems. Our work with only
also caution that there are nuanced tradeoffs that need to be numeric data is a first step in this direction. It would be
made when choosing the type of input method to use. The interesting to actually build a system that handles the
fastest and most preferred techniques (handwritten processing of forms in real use, and to study its performance
numerals) are generally the least accurately recognized by a in the field. We intend to explore this in the near future with
machine, although the digit per box method is arguably a a partner NGO.
reasonable compromise in this regard since the box provides
some structure that eases the task of the recognition REFERENCES
algorithm. 1. Anoto, http://www.anoto.com
2. AED Satellife, http://www.healthnet.org
X. CONCLUSION AND FUTURE WORK 3. Arai, T., Aust D., and Hudson S.E. (1997). PaperLink: a
We have explored the problem space of data collection by technique for hyperlinking from real paper to electronic
NGOs in underserved communities where resources and content. Proc. CHI ’97, p. 327-334.
technological literacy are typically scarce. In particular, we 4. Bergeron, Bryan P. (1998, Aug.). Optical mark
focused on how a particular class of paper forms – those that recognition. Postgraduate Medicine. 104(2).
require only numeric data entry –might be best structured to 5. Chand, A., Dey, A.K. (2006). Jadoo: a paper user
support subsequent automated data transfer to enable interface for users unfamiliar with computers. Proc. CHI
2006. p. 1625-1630.
computerized storage and processing. The main
6. DeRenzi, B., Lesh, N., Parikh, T., Sims, C., Mitchell,
contributions of our work are: (1) an interview-based
M., Maokola, W., Chemba, M., Hamisi, Y.,
416
Schellenberg, D., Borriello, G. (2008). e-IMCI: Remaining Problems. Studies in Computational

Improving Pediatric Health Care in Low-Income Intelligence, SpringerLink.
communities. Proc. CHI 2008. p. 753-762. 19. LeCun, Y., Bengio, Y., Bottou, L.and Haffner, P. (1998).
7. Ellison, M. Firm foundations: Improving NGO data Gradient-Based Learning Applied to Document
collection, http://www.id21.org/id21ext/s7bns1g1.html Recognition. Proc. IEEE. 86(12). p. 2279-2324.
8. Galliher, J.M., Stewart, T.V., Pathak, P.K., Werner, J.J., 20. Mackay, W.E., Pagani D.S., Faber L., Inwood B.,
Dickinson, L.M., Hickner, J.M. (2008). Data Collection Launiainen P., Brenta L., and Pouzol V. (1995). Ariel:
Outcomes Comparing Paper Forms With PDA Forms in augmenting paper engineering drawings. Proc. CHI ‘95.
an Office-Based Patient Survey. Annals of family p. 421-422.
medicine. 6. p. 154-160. 21. Mackay, W.E., Pothier G., Letondal C., Bøegh K., and
9. Gogineni, M., Ratan, A.L. Evaluating the viability of a Sorensen H.E. (2002). The missing link: augmenting
mobile phone-based, SMS/GPRS-enabled, client data biology laboratory notebooks. Proc. UIST 2002. p. 41-
collection channel for urban microfinance. Microsoft 50.
Research India Technical Report. 22. Mifos, http://www.mifos.org.
10. Gravic, 23. Parikh, T.S, Javid, P., Sasikumar K., Dhosh, K.,
http://www.gravic.com/remark/remarkcustomers.html Toyama, K. (2006). Mobile phones and paper
11. Green Foundation, http://www.greenconserve.com documents: evaluating a new approach for capturing
12. Guimbretière, F. (2003). Paper Augmented Digital microfinance data in rural India. Proc. CHI 2006. p. 551-
Documents. Proc. UIST 2003. p. 51-60. 560.
13. Johnson, W., Jellinek, H., Klotz, L., Rao, R., Card, S.K. 24. Parikh, T., Aditya, V.P.S, Vellayutham, M. (2003).
(1993). Bridging the paper and electronic worlds: the Automated Forms Processing and Paper User Interfaces
paper user interface. Proc. CHI ‘93. p. 507-512. for Data Collection from Village Microfinance Groups.
14. Khan, Aqueel. ASKIndia, private communication, 2008. Unpublished draft,
15. Koike, H., Sato Y., Kobayashi Y., Tobita H., and http://people.ischool.berkeley.edu/~parikh/papers/itira20
Kobayashi M. (2000). Interactive textbook and 03.pdf
interactive Venn diagram: natural and intuitive interfaces 25. Sellen, A.J., Harper, R.H.R. (2001). The Myth of
on augmented desk system. Proc. CHI 2000. p. 121-128. paperless office. MIT Press.
16. Kunte, S.R., Samual, R.D.S. (2007). An OCR System for 26. Shelby-James, T.M., Abernethy, A.P., McAlindon, A.,
Printed Kannada Text Using Two - Stage Multi-network Currow, D.C. (2007). Handheld computers for data
Classification Approach Employing Wavelet Features. entry: High tech has its problems too. Trials. 8(5).
Proc. ICCIMA 2007. p. 349-353. http://trialsjournal.com/content/8/1/5
17. Liao, C., Guimbretière, F., Hinckley, Hollan, J. (2008). 27. Siemens,
Papiercraft: A gesture-based command system for http://www.industry.siemens.com/postal-automation/
interactive paper. ACM Transactions on Computer- 28. SKS Microfinance http://www.sksindia.com/techno-
Human Interaction. 14(4). p. 1-27. logy.htm
18. Liu, C., Fujisawa, H. (2008). Classification and Learning 29. Wellner, P. (1993). Interacting with paper on the
Methods for Character Recognition: Advances and DigitalDesk. Communications of the ACM. 36(7). p. 87-
96.
417
Rajnikant’s Laptop: Computers and

Development in Popular Indian Cinema
Joyojeet Pal
parental consent and wisdom would typically be highlighted
in the couple’s struggle.
Abstract: The role of technology in development has seen
an exceptional spike in interest in the past decade in Instead, in Alai Payuthey, the couple turns to a new
academia and industry. India has been a key location in direction for its salvation – technology. Karthik starts a
the study and practice of technology and development computer software company with his friends, eventually
because of the centrality of the technology industry in its winning an outsourced contract from the US that fixes for
growth over the recent two decades. Examining the public good their financial troubles, rubbing in the process a few
discourse on technology in areas where computers are not parental noses in the dirt on the gold-paved streets of south
ubiquitously used or available, we find that cinema is Indian cities. Karthik fails most stereotypes of cinematic
important in constructing how people in Indian villages occupational characterization. He is no idealistic teacher, nor
and even cities perceive computers. Turning back to the upstanding cop, nor charismatic businessman. He is at best a
films themselves, we find a strong aspirational discourse in lovable nerd, traveling on a motorbike to work daily with a
the ways in which computers and technology-users are laptop strapped to his back instead of a holster. In effect, he is
portrayed on Indian popular cinema. These in turn relate the archetype of exactly what he isn’t in the movie – the
to how technology and development have been portrayed perfect candidate for an arranged marriage.
in cinematic discourse over generations, in cinema from
around the world, from the earliest days of silent film. We Before resigning myself to academia, I tried a hand at script-
find from interviews with users of technology in rural writing in India. The process for selling a script typically
India that these images create strong ideas of value and involves going to a movie producer for a ‘reading.’ The
legitimacy derived from using technology. Reflecting on process is somewhat ad-hoc, and depending on one’s
this, we find that study of public discourse has been a familiarity with the producer, it can involve a detailed
critical missing piece of ICTD studies. A study of media description of every element of the screen or quite simply a
imagery such as the one presented here is a small step in very broad outline. In my case, I had to give very specific
that discussion. descriptions of each character, scenario, and outcome. While
my stories were not entirely outlandish, producers would point
out how certain characters needed tweaking, or deleting
I. PROLOGUE
entirely, because the “public will not accept it.” While failing
K ARTHIK is a hormonal rich brat, Sakhti is a

hardworking medical student from a poor family. The
class-crossed coupledecide to marry, against the wishes
to sell scripts, I had a comparatively more successful double-
life as a researcher working with children in villages that had
recently got computers. I had noticed over and over, that
of both sets of parents, and thus the plot thickens in Alai children and adults alike who had never seen computers
Payuthey, a 2000 blockbuster hit from south Indian filmmaker before in their lives had built ideas about technology from
Mani Rathnam. popular cinema. At the end of one particularly disappointing
script reading when all my characters and plotlines were
This kind of face-off between youthful love and parental deemed unsaleable, I saw a Tamil film, Vyapari [5]. The script
opposition to marriage, a veritable obsession of Indian on that one probably sold easier, the plot revolved around a
cinema, is perhaps one of the oldest and most successful protagonist driven by his admiration for Bill Gates which
themes. Back in the old days, Karthik would probably have leads him to use technology to clone himself to make twice as
been reduced to begging office to office for a job wearing a much money.
tie, finding in his useless paper degrees a metaphorical foil for
the oppressive market economy [1] and thereafter turn to a life II. INTRODUCTION
of dubious ethical distinction [2, 3]. Sakhti meanwhile would
In the 1990s, information and communications technology
probably sit home sacrificing square meals and running a
had an extraordinary impact worldwide both in terms of
bare-bones household with a sanctimonious smile [4]. A
number of themes, including the dependence of the man on economic growth and in social change through access. The
the system to earn an honest living, the helplessness of the concurrence of this technological ‘revolution’ with the
woman outside of the home domain, and the importance of changing economic and political conditions in various post-
socialist nations created a renewed interest in the area of
Manuscript received February 20, 2009. technology and development not just within academia [6], but
Joyojeet Pal is a research associate at the Center for Information and within business and management strategy circles [7] and also
Society at the University of Washington. Email: joyojeet@washington.edu
418
within a rapidly expanding international development practice. perspective of gender studies [29, 30]. The theory building
A new wave of scholarly research publishing on ICTD process has also benefited from works that have taken a
(Information and Communications Technology for multidisciplinary approach, looking at ICTs from the
Development) started in the late 1990s, with much work on perspective of one or a few specific existing bodies of social
digital inequality issues [8-11], on technology and sectoral theory including innovation diffusion [31], social inclusion
development and macroeconomic change [12-14], on [32], technology artifacts [33], and stakeholder theory [34].
technology and urban change [15, 16] and a very sizable body Another interesting area of theory work has been that of
of work on the ‘potential’ of technology [17-19]. technology in developing regions, though not necessarily
specifically about projects aimed at development, such work
A look through the history of research in ICTD shows that on transnationalism and technology usage of people from
an inordinately large portion of research and practice in this developing regions [35, 36] and ICT-related research oriented
field has been focused on India. The reasons for this are specifically towards technology use in identity formation and
multifold, but the perceived role of the technology sector in articulation [37-39].
India’s ‘leapfrog’, and the investment of several high-tech
B. Technology and Cinema
companies’ philanthropies as well as international agencies in
the rural technology projects have both played important roles Ideas of technology and society have dated back over a
in this concentration. In the early years, much of the ICTD century to the fanciful silent short film Trip to the Moon [40],
literature in India and the South Asian region in general, and the representation of technology in cinema has straddled
looked more into the role technology had in facilitating a the line between science fiction, and what may be, and
leapfrog either for individuals, or for the economy as a whole; technology and social readjustment. The latter, especially the
consequently, the tone of much work in this period was idea of technology as transformative, came to centerstage in
comparatively more enthusiastic. the 1920s and 30s, around the period and often theme of rapid
industrialization. Scholars have been interested in two
By the mid 2000s, much introspective work looked at the important and fairly consistent themes along technology and
performance of ICTD projects in India and elsewhere of the society in cinema – the first, the dystopian ideas of technology
1990s, and the causes and outcomes of some of the interest in and urban living – either in the present or future as seen in
technology and development [20-23]. While a lot of such Fritz Lang’s futuristic epic Metropolis set in a fractured 21st
work was empirically grounded, most critical discussion in century or Charles Chaplin’s Modern Times [41], set in a
ICTD focused on the supply side – i.e. why certain projects or dehumanized 20th century factory floor. These ideas of
project-types gained centerstage and how they performed, technology as mystical, all pervasive, and potentially harmful
while comparatively a lot less effort went into the demand- have been a consistent theme of cinema and literature
side, i.e. determining why ICTD projects have come to find throughout the decades of vast technological change around
apparent or stated enthusiasm among users populations, and in the world [42, 43]. The second early theme had a more
what ways is this enthusiasm manifested in public discourse. positive view of technology, primarily industrial production,
We begin a discussion in the latter through an examination of and bears its origins in the early Soviet cinema highlighting
representations of technology in popular Indian cinema, technology as a nationalist enterprise, a means of social and
especially in the south of the country, where the technology economic development [44] and later emulated several other
sector has had a comparatively larger economic and social nationalistic cinemas, and this connection of cinema with
impact. modernity has been an important theme of cinema around the
world [45]. Both these themes have been prominent in Indian
III. RELATED WORK cinema, and the two manifestations of these have been in the
struggle between traditionalism and modernism, one seen in
There is little if any specific work within the relatively new
other third world cinema as well [46, 47], and of course the
ICTD field on media representation of computers and its
use of technology in nation-building as the second theme [48].
impacts on technology adoption. There are however a number
So while technology in a more general sense was a consistent
of important areas of related work that are foundationally
useful to this paper. theme of cinema, computers themselves did not appear until
well into the 1950s, and even then they tended to be fanciful
A. ICTD Theory clunky large panels (reasonably similar to what computers
ICTD is still at a fairly young stage for theory-building really looked like!) used in science fiction cinema.
work, and it is here that the multi-disciplinary nature of ICTD
has served as a mixed-bag rather than as a clear advantage. The first films to feature computers were science newsreels
Social theory work in ICTD benefited in its early days from in the immediate postwar period, and in 1951, the first film
Manuel Castells, a sociologist and planning theorist who with a rudimentary computer was When Worlds Collide
formed some of the early foundations of ICTD [24, 25], and featuring a computer in a war to keep earth safe from a
much work in the ‘information society’ studies [26-28]; and renegade asteroid [49]. Several such films followed. The
more recently (and importantly for this paper) from the rising popularity of television in the US spurred the
419
appearance of computers on screen, starting with the 1962 sci- propaganda tool for spreading its message, and began a trend
fi show The Jetsons [50], in scattered episodes of the spy of what is now synonymous with several south Indian
caper serial The Avengers [51], and finally in 1966 with the cinematic cultures, of practically replacing temple idols with
hugely popular sci-fi series Star Trek [52]. This fantasy cinematic stars larger than life, and usually extremely political
archetype is important, since we see this early imagination of [59]. 2 This north-south divide in the manifestation of
computers reprise itself in contemporary Indian cinema. The technology in public discourse exists, especially in looking at
next paradigm featured a cinematic imagination of computers the on-screen use of technology by ‘politically active’ stars
guided by the popular conceptions of artificial intelligence – and examining how that relates to their own political
as typified by the man v/s machine face-off in 2001: A Space motivations. Much work has been done on the cinematic
Odyssey [53]. But the individual user of computers remained culture of India, and its representation of modernity [60-62],
by and large the scientist, as indeed was true for the world at and more recently on television in the post-liberalization India
large, until in the late 1970s and early 1980s, when kids in [63]. Significant commentary on media and public discourse
regular schools started getting access to computers. This has focused on the influence of cinematic and television
triggered a spate of ‘hacker’ films around teenagers and representations on public imagination including on economic
youngsters using computers starting with the Canadian Hide and social decision-making [64-68]. Interestingly, while much
and Seek about a kid who sets off a nuclear emergency by work has focused on the protagonists and their
hacking into a mainframe [54]. But the ubiquitization of characterizations as elements of modernity [69-71], little has
computers on screen, where computers became everyday focused on the actual artifacts themselves, despite as we will
items, happened only around the mid-1990s, with a huge spike find, much continuity among these symbols of modernity.
in technology-related blockbuster cinema made for popular Finally, this article draws at its heart a bountiful literature in
consumption in Hollywood, around the time of the Silicon critical theory and film [72, 73] including in cinematic
Valley economic boom, [55] and the widespread permeation traditions of the developing world [74].
of the Internet. 1 As we discuss the representation of
technology in Indian cinema, we find many of these themes IV. METHODOLOGY
repeated, in Indian cinema – but one striking difference. This paper is primarily a proposition based on readings of
While the computer user as ‘hacker’ is a consistent theme in theoretical work in media studies and ICTD, and an expansive
Hollywood (and used selectively in India, as we shall see), the content analysis of a large number of films. 87 films are cited
computer user as ‘aspirational’ is almost entirely missing. here, of which a subset of 43 films form the core of this
C. Cinema and Public Discourse in India argument. Each of this latter group is an Indian film that has
from the late 1990s or early 2000s some reference or
In exploring why technology in cinema is important to our
characterization relating to technology and society. Given the
understanding of ICTD, we turn briefly to the importance of
limitation of space, I refer primarily to the overall
cinema in public discourse in India. The political importance
characterization and reference to technology rather than a
and development discourse of cinema gained early recognition
analysis of specific details within the script such as dialogue.
among the Bolsheviks following Lenin’s own affirmation of
In addition to the observations from films, excerpts from
cinema as a tool of propaganda [56]. Much Indian cinema
interviews that were part of an earlier interview-based study
under the British was censored and limited to non-nationalistic
on conceptions of technology and computers in rural India are
themes, but the postcolonial period threw open narrative
used here. The excerpts used here are largely those that have
floodgates. Jawaharlal Nehru, the first prime minister,
some mention of cinema, or reflected ideas that relate directly
recognized the importance of cinema in the nation-building
to the themes emergent from our examination of films. These
exercise, commissioning Italian director Roberto Rosselini to
interviews are in fact the motivator for this study, as it
make a film on India in 1950, thereafter supporting films like
emerged in this earlier study that cinematic descriptions of
Mother India and several of Raj Kapoor’s early post-
computers and computer-users were very important in idea
independence films [57]. In this period, a lot of the
formations about the benefits of technology among
screenwriting in the Bombay film industry was done by a
populations that had little or no prior experience of direct use
number of socialist secular writers who were formerly active
of computers.
in the freedom movement, and subsequently incorporated
several nation-building themes into their cinema. But it was
V. ANALYSIS
further south in the Madras film industry that cinema as a
political weapon was actively imagined by politicians and From the early days of Indian cinema, modernity has had a
filmmakers alike [58]. The Dravida Munnetra Kazhagam tenuous track record. Consequently, technology and its
(DMK) as a political party was quick to use films as a representations have often been caught in an awkward
contested space between traditional and modern values, often
1
implying the schism between rural and urban landscapes. One
This is somewhat comparable to the appearance of computers and
computer users in India, which became much more common a decade later
2
when home computers achieved reasonable (though far from ubiquitous) This is further interesting by the fact that DMK was an explicitly atheist
permeation in urban India. party.
420
of the earliest films to straddle this space, Dharti Mata [75], women don’t have jobs. And those that do, typically do so
literally meaning Mother Earth, had two friends, one who because of the failure of some critical male provider. Thus, the
goes abroad to study and returns an urbane engineer home to dead, drunk or incapacitated rural husband or father gives way
find mineral wealth on his ancestral soil, and his buddy, who to the woman who works the field, at the risk of lascivious
stays back and decides to stay true to his rural roots and farm attention of the agrarian landlords or plantation managers [77,
the same earth on which the former wants to drill. The clash 78]. When the female ends up working in the city, if she’s
between the protagonists ends with the village-bound son-of- generally less educated or desperate enough, and will likely
the-soil winning over, an ending that would reprise itself end up as a casual laborer, in which event she is at risk from
repeatedly over the years. the lecherous building contractor [79, 80]. Now these informal
sector jobs tend to go to the ‘mother figures’ in movies, an
Fast forward to the new millennium, and Kandukodain obvious consequence of these positions cutting a harder case
Kandukondain [76], the Indian remake of Sense and for visual glamour. The young heroine in a film tends to work
Sensibility . Here, the protagonist is a woman, and lives in her more often in organized sector.
ancestral village, but when the going gets tough, she leaves
for the city, ends up happily ever after as a software engineer. Here too, many of the same problems exist. There are screen
Indeed Kandukondain may not be the archetype for all films roles where women portray alternatives to the mother figure
of the generation, but as a popular blockbuster, the film such as nurses or doctors [81-83] or caring social workers or
featured several of the key themes that we highlight here on teachers [84-87], reasonably respectable jobs for women, and
the changing view of modernity, especially because of the not macho enough for the male protagonists. The female
reversed gender roles in the film. character who fails to secure any of the above, spills over into
the organized employment sector. The villainy of the rural
A. ICTs and gender empowerment
landlord and the urban building contractor are now reprised in
During field research in rural Karnataka in 2005, at a village the lascivious white-collar bourgeois rogue harassing female
where the local government had just installed computers in a co workers [88-90] and in the rare cases, the daring self-starter
primary school, an interesting insight on computers and the saleswoman [91-93] or a spoilt heiress boss [94, 95]. In most
workplace came from a 21-year old teacher at a village of these cases, a male character, usually the hero, offsets the
computer center. Both her parents were illiterate casual perils of the woman’s tryst with the man’s domain of the
laborers. economy. Of course, the exceptions exist – the woman
embraces the perilous world with vengeance or connivance,
“I want to move out of the village. I am looking for a job the former when she takes on the role of Hindu goddess Durga
with computers because my parents will let me move to as a police woman [96, 97] or avenging angel [98, 99]. What
Udupi or even Bangalore if the work is in computers. For remains degenerates quickly in the caste hierarchy of
any other job, they won’t let me leave the village.” professions from the mildly uncomfortable bar dancer or
Geetha, Computer Teacher, performer [100, 101], to the circumstantial prostitute [102-
Rural Udupi, Karnataka 104], and finally to the campy gangster’s moll [101, 105, 106]
About a year later, an interesting corollary to Geetha’s Kandukondain Kandukondain [76], took a gentle step in a
statement came from a taxi driver interviewee, in the different direction. The female protagonist in the film, a
neighboring state of Tamil Nadu, a father of two girls in their Brahmin girl, Sowmya is reduced to penury after being
20s. hoodwinked out of her ancestral village home by evil
relatives. She proceeds to move from the village to the city,
“Both my daughters work in Chennai in computers. In the where she rectifies her family’s situation by learning to use
early days, we would never let our (referring to the computers and getting a job as a software engineer. The
Thevar caste) women travel to the city to work, but if they actress not only lifts her entire family out of poverty, but
work for computers that is good. There are good facilities emerges as the most valuable technologist in her firm and
with only ladies housing, and many other families from almost flies out to the US to work as an engineer before better
our village have sent their daughters to work in Chennai sense prevails and she marries her wooing admirer. What is
now.” uniquely compelling about the characterization of Sowmya is
Selvaraghavan, Taxi Driver that she remains ‘traditional’ to the end, but straddles the
Coimbatore District, Tamil Nadu modernity space effortlessly. She dresses conservatively,
except for dream sequences with song and dance, serves her
There is already a growing corpus of work on female family dutifully, and acquiesces to an arranged marriage.
technology workers in Indian call centers which probes this When she does start her software job, her workplace does not
issue more closely, [29], and the local movie theatre as a demand flashy hackers, instead, her interactions with her
starting point into some of the changing social conceptions of engineer boss highlight the importance of urban traditional
women’s work. middle-class work ethos. Despite her success, she does not
turn into a slick skirt-wearing executive, and in doing so, kills
If we think back to Indian cinema and think back to the two birds with one stone. She remains the agreeable screen
typical occupations of women on screen … well, reasonable heroine and makes it to poster child for ‘appropriate ICTD’ –
421
a woman who wins her way out of poverty, and manages stay daughters married off. In the 2000s, that role has been taken
culturally rather untainted. over by the US-returned (or not) software engineer. 3
While there had probably never been an Indian film with a The classic ‘catch groom’ of south Indian cinema was
female character playing an engineer save for the oddball Arvind Swamy, who in the blockbuster hit Roja [121] played
automobile mechanic’s daughter, the female software engineer a computer engineer whose arranged marriage is the early plot
and outsourcing center workers hit the stands in a huge way of the film. Over time, two broad strands of software
around this time [107-113], including films with a reversal of engineer-related marriage scenarios have emerged – the first
roles – where the female lead plays an accomplished in which the engineer is the middle-class hero [122-124] and
technologist of some form, and the male lead is portrayed as
the second is the counterpoint – in which parents are shown
professionally subservient [114, 115]. One film stands out as
hankering after a groom who is a software engineer or NRI,
an interesting continuation to Kandukondain. In the Telugu
and the hero in this case is usually a son-of-the-soil type [76,
film Anand [116], the protagonist Roopa takes Sowmya’s
position a step further. Here, she not only supports herself 125, 126], and while the endgame of these films is often an
through a software job, but sheds much of Sowmya’s demure ode to the anti-hero, the focus on the software engineer as
qualities in reinforcing her independent identity and essential to middle class aspiration is nonetheless highlighted.
challenging traditional behavior roles expected from her by What is also interesting is to look at the contrast between the
suitor and in-laws alike. engineer/technocrat protagonist in cinema around industrial
themes with the engineer in a technology film. The factory
An even more interesting addition to the technology and films frequently had a foil – typically industrialist antagonist
gender empowerment idea comes in the film Swagatam [117]. pitted against the engineer [127], and worker, rather is a the
In it, the male protagonist is a demanding customer at an hero [128]. A “good” engineer typically aligns with the
arranged marriage matchmaking bureau. The manager of this worker instead of the industrialist, breaking class barriers [94,
bureau (coincidentally the female lead) has an online 129]. The flatness of the technology sector is reinforced
candidate repository, and tells the male lead to use the through two factors – first, the “proletariat” is missing in the
computer to filter through his requirements. She kindly software company, unlike for instance the range of films about
reminds the hero that a woman’s greatest trauma is being both factories with workers and service sector offices where
rejected at the arranged marriage meeting, and that technology the ‘peon’ symbolized the underclass [130-132] and second,
should be used effectively to circumvent this problem, and the boss himself (not necessarily herself!) being sharing class
thereby empower women. Given that marriage is practically
with the employees [133, 134]. My reason for excluding
an ever-present theme in Indian cinema, it is not surprising
gender is that women and workplace already come loaded
that several other films have used technology as a go-between
with a large literature as emphasized in the previous section,
for arranged or other marriages [113, 118].
thus the grounds for evaluating “flatness” do not exist in the
Although the computer as a cupid connector has been a same vein – women never played engineers or evil
fairly strong theme through most Indian cinema, the female industrialists (heiresses, though) so the terms for historical
software engineer has been less so. Regional cinema such as comparison do not apply. A woman’s professional ascendance
Bhojpuri, Bangla, and Oriya largely cater to rural audiences in the a technology job is in itself shown as having a range of
(or what in trade parlance is referred to as ‘B circuits’), and power dynamics and it would therefore be confounding to
still feature a lot fewer women in professional positions. In analyse the gender depiction based on flatness alone.
Hindi cinema, on the other hand, female characters who used
computers in Hindi films, have tended to be more westernized In some films, the less than desirable character turns to
or upper class, such as the student who does computer science computers as a means of social acceptability and in others
in college as a qualification, before her eventual marriage, like [135, 136], the drive is primarily economic, often explicitly as
Madhuri Dixit in Hum Aapke Hain Koun [119] or the a means of getting jobs in the US [134, 137]. We find among
‘smartypants’ Preity Zinta in Koi Mil Gaya [120], characters male software engineers a distinction between the north and
significantly different than the Sowmyas or Roopas, where the south Indian stars. In Hindi films, the engineer stereotype is
appropriation of technology into the middle class domain has typically applied to the ‘cool youngster’ hacker characters,
been less of a concern. In our discussion of aspiration, we usually teen idols, [138-141], very comparable to the
return to this issue. Hollywood depictions of computer teenager users from the
B. Computers as Aspiration, Engineer as Hero 1980s and 1990s whereas most of the south Indian depictions
There are few windows more insightful into the aspirational are not necessarily hackers who do cool things with
environment in India than the marriage market. Till date, the
highest catches in the dowry market in rural India go to 3
In the last year South Indian actresses, Renuka Menon, Gayathri
holders of government jobs – which offer among other things Raghuram, and Kanika Subramanian have all left the film industry to marry
– stability, an invaluable element in the rain-dependent software engineers in the US. In the past, Swarnamalya, Ravali, Jyothirmayi,
Maheshwari the common stereotype/joke was of actresses leaving the film
agrarian landscape. But films have long portrayed doctors and industry to marry doctors abroad, which is interesting and a potential topic for
‘big officers’ as good catches for families trying to get a larger study on ‘legitimacy’ of actresses.
422
computers, but rather people who have technology-related with honorifics such as ‘Dear Leader’ or ‘Elder Brother,’
jobs. This distinction is doubly interesting since it not only several of them end up in politics, and the opening few weeks
underlines the differences between the viewing cultures of of a film by a mass star can almost certainly expect to run to
popular Hindi and regional cinemas, a difference that has been full houses, not to mention the celebrations and prayers that
articulated in Ashish Rajadhyaksha’s examination of the accompany any release.
‘westernizing trend’ of Hindi cinema [62], an upper class
urbane narrative, in short. In contrast, south Indian cinema, Several stars, especially in the south, have landed in the
while the fantasy sequences such as songs etc. have roughly ‘mass star’ image trap including Mohanlal and Mammootty
the same ‘modernity’ of the Hindi counterparts, the narrative from Kerala Rajkumar in Karnataka, Balakrishna, Nagarjuna,
themes are much closer to the wider middle-class audience and Chiranjeevi in Andhra Pradesh, and Vijaykanth and
appeal. Thus, the persistence of software engineer characters Rajnikanth in Tamil Nadu, icons with possibly the most
is seen not just in the ‘urban’ south Indian cinema (meaning frenzied fan following of all. The use of technology by these
urban in both setting and primarily appeal) but also in the mass stars is particularly interesting, because their audiences
‘mass’ cinema. are often not the middle class aspirants that watch many of the
urban-based films listed above. Interestingly then, we find the
ideas of technology use in such mass star movies somewhat
C. ‘Mass Films’ and the Computer as Benevolent Superstar
simplistic, but in line with the overall theme of what a mass
star does – save the world.
“Computers can be used to fight evil. We can do anything
with a computer”
Typically a mass star is a perfect son-of-the soil, thus almost
Shivraj, 5th grader
always speaking in the vernacular through a film (unlike in
Devanahalli, Karnataka
Hindi films, for instance, where English is often interspersed
with the Hindi dialogue) excelling and promoting local arts
“Computer can save us. When neighbouring country is
attacking, this is known to our scientists by tracking it on and so on. In a sense, the mass stars’ use of computers is
the computers.” comparable to their use of English. Mass stars generally use
Udhaykumar, 5th grader, English only occasionally, laconically, and with much
Coimbatore, Tamil Nadu emphasis, usually when challenged by an English speaking
adversary or intransigent love interest. In the same vein, a
In this section, we turn to the difference between the ‘mass star’ may use a computer to underline just that he can,
aspriational cinema, which we cover in the previous two even if it is entirely irrelevant to the plot. As a son of the soil,
sections, and the ‘mass films’ which not only highlight an he is able not just to play the game by the rules of the
interestingly different view of technology and modernity, but hinterland, but can match up to modernity [142].
also bear uncanny resemblance to some of the Hollywood
characterizations of the past. By ‘mass films’, we refer here to In Sivaji - the Boss 4 [143], by some estimates the top
the films that are primarily intended towards urban lower- grossing India films of all time, Rajnikanth, arguably India’s
class and rural viewers. The term, derogatory as it may seem, most maniacally followed star, plays a software engineer-cum-
is quite commonly used by filmmakers and analysts alike to social worker bent on rectifying the ills of the Tamil world. In
refer to ‘people’s’ cinema, and several major film stars have his quest, Sivaji uses his voice-recognition-enabled laptop to
come to be associated with this movement. While these can control his vigilante operations and outmaneuver his rivals.
also refer to low-budget movies made for rural audiences, a Similarly the use of computers to maintain and control
fairly large share of ‘mass films’ are actually big budget databases of villains has been used effectively by megastars
productions usually featuring a movie star with a significant, Vijaykanth [144], Chiranjeevi [145], and Ajithkumar [146,
often frenzied fan following. Such actors typically associated 147]. The idea of computers as a tool to assist heroes in
with ‘mass’ cinema tend to have reliably larger-than-life cleaning out society has been used to much effect in recent
characterizations and public images [69]. The dialogue in times, mostly in films featuring online websites for
‘mass films’ is typically political and populist, almost ‘submitting un-rectified crimes’ for appropriate appraisal and
propagandist, and can probably be traced back to the cinema corrective action by the hero on the other end of the internet
of M.G.Ramachandran, who never drank or smoked on connection [148, 149]
screen, committed no act of villainy, remained utterly
attractive to women, helpful to every needy person, and The quote from Shivraj at the start of this was not an
entirely unbeatable in any physical or intellectual contest. His isolated one. We were surprised at how often in our interviews
‘mass star’ successors are somewhat less pious in contrast, but in rural India, the same answer was repeated to us over and
fairly comparable in their abilities to commit superhuman acts, over “Computers can do anything” sometimes ranging from
ricochet bullets, and most important of all, save mankind. children’s fantasies to adults with no direct experience with
Fans of ‘mass stars’ frequently refer to their cherished stars
4
Boss here stands for Bachelor of Social Science
423
computers allocating human attributes to the machines, bodies of thought in development studies. Here, I aim to very
“Computers can teach us English” in clear seriousness. It is gently open the windows into the concerns of critical theory
far fetched to ascribe such notions specifically to films, but and discourse analysis.
it’s worth looking briefly at some of the omnipotent deeds of This analysis of computers and cinema in India brings to
computers. Perhaps a compelling analogy would be that light a number of interesting factors about how filmmakers
computers can do anything, just like Rajnikanth. It is not have chosen to portray technology in Indian cinema, and it is
surprising in this light to know that movie star Mammootty particularly interesting to look at where these portrayals are
was selected as the Brand Ambassador for the rural computing comparable to other cinematic traditions, and where, like in
initiative in Kerala, Akshaya. the aspirational characterizations, it is unique and deeply
related to the prevalent discourse of technology being a means
If we look back before the computer-using days of the of social and economic leapfrog. What is also remarkably
mass stars, shades of fantasy and omnipotence among telling is the much greater prevalence of technology-related
computers in India film were akin to western cinema of the themes and characterizations in south Indian films than in
1960s, especially in early Indian science fiction [150, 151]. north Indian cinema, given the disproportionately higher
Computers and their ability for visual magic broke box office concentration of technology-related industries and
records with ‘Miss World’ Aishwarya Rai’s first screen opportunities in the major south Indian metropolises. These
appearance in Jeans [152]. Here, her brother creates a visual cities like Hyderabad, Chennai, and Bangalore saw
double of Aishwarya to convince twin brothers that they were tremendous demographic, landscape, and social
both in love with two different girls, a reasonable exercise in transformations in the late 1990s, the same time that
computers as movie characters spiked up the charts.
balancing technological savvy with the need for romance in
scripts. Jeans and the hugely popular MTV video of Michael Many of the aspirational characters may well draw their
Jackson’s single ‘Black or White’ [153] featuring faces original inspirations from real-life heroes, often engineering
morphed into one another, or simply the glut of Photoshop graduates from the hinterland who moved to the cities, and
engineers in India, also helped create one of the most enduring saw remarkable growths in their own incomes and social
computer tricks in Indian films, still a common feature of statuses, often moving abroad, enjoying opportunities that
‘mass films’ – the ability of a machine to take a scanned would have been much more challenging to aspire to a
picture of a child lost several years ago, and morph out of it a generation back. Most importantly, many of these graduates
perfectly accurate image of the adult version, auto-adjusting were women. What the cinematic analysis shows us is indeed
for sartorial grace and the few extra facial pounds on the likes a reflection of a larger perception of what counted as the
‘India Rising’ metaphor for the urban middle classes and the
of Vijaykanth. [154, 155].
rural aspiring middle classes, which incidentally, largely
comprise the regular film-going population. It is the
CONCLUSIONS
aspirations of these classes that we see portrayed in a
Is there intentionality in these portrayals? My opening the subsection the South Indian cinema of today. It is thus
article with my personal sob story of failing to sell my scripts decidedly interesting to compare the output geared at these
was meant to imply at least partly that there is at least some aspirational markets with the poorer hinterland markets of the
level of production-level engineering of characterizations. I mass films which take on a simplistic form, carrying forth in
would argue that that question impossible to answer given the selective filters the transplant of middle class imagination of
recursive loop on whether society influences media or vice- technology down to the economic and social chain.
versa.
In conclusion, we turn briefly to the iconoclastic
The idea for this research emerged as after confounding proposition of popular Indian cinema being proactively used
outcomes in other research that indicated a mismatch between to impact development. While the producers may indeed have
peoples’ stated interest in technology and their actual use of been sitting by with whips, were screenwriters of the films we
ICTD projects in rural India. People were very excited about see above thinking of issues of technology and development,
computers and the possibility of their own access to them, but and of impacting society, or is what we see here a sheer
unclear on how technology could be practically useful in their reflection of popular psyche. Irrespective of that, as
lives. Such ideas about technology were further seen to not demonstrated, research has already shown that certain types of
just influence researchers’ estimation of what the likely entertainment media experiences in India have had positive
adoption for certain ICTD projects may be, given the apparent social outcomes, especially in gender-related issues like
enthusiasm about technology, but could also influence the female child protection and domestic abuse prevention. This
populations’ own propensities to invest in ICTD projects, could well be dangerous knowledge for the ICTD community,
without a necessarily clear idea of the value of such a significant part of which is comprised of scholar-activists.
technology. For serious scholars of ICTD, closing this Indeed some of us may stand up and ask if cinema can and
phenomenon off as the ‘buzz’ of technology is not adequate, should be proactively be used to impact technology uptake. It
and an examination of the discourse encompassing this buzz is is important that we think of these questions since we are after
necessary. As ICTD matures as a strong body of independent all in development and for many of us, all research implies the
literature, it will be vital to incorporate learnings from existing responsibility of considering real world intervention
424
capability. When we do start thinking in that direction, 19. Nulens, G. and L. Van Audenhove, An Information
perhaps a social scientist in the room can stand up and point Society in Africa?: An Analysis of the Information
out the parallels of such ideas with Leninistic cinematic Society Policy of the World Bank, ITU and ECA.
propaganda, and ask if technology intake ought to be International Communication Gazette, 1999. 61(6):
promoted at all. After all, asking questions is what we do, p. 451.
decisive action is Rajnikanth’s domain. 20. Heeks, R., Information Systems and Developing
Countries: Failure, Success, and Local
Improvisations. The Information Society, 2002.
ACKNOWLEDGEMENTS
18(2): p. 101-112.
21. Keniston, K., Grassroots ICT Projects in India:
Thanks to Eric Brewer for bankrolling my whimsical work, Some Preliminary Hypotheses. ASCI Journal of
and to Divya Ramachandra, Kentaro Toyama and Jenna Management, 2002. 31(1).
Burrell for their interest in this particular research. 22. Madon, S., Evaluating the Developmental Impact of
e-Governance Initiatives: An Exploratory
REFERENCES Framework. EJISDC, 2004. 20(5): p. 1-13.
23. Donner, J., The Use of Mobile Phones by
1. Roy, B., Naukari. 1954, Bimal Roy Productions:
Microentrepreneurs in Kigali, Rwanda: Changes to
India.
Social and Business Networks. Information
2. Chandra, N., Ankush. 1986: India.
Technologies and International Development, 2007.
3. Bapaiah, K., Aaj Ka Daur. 1985: India.
3(2): p. 3-19.
4. Muthuraman, S., Aarilirunthu Arubathu Varai. 1979.
24. Castells, M., The Informational City: Information
5. Chidambaram, S., Vyapari. 2007: India.
Technology, Economic Restructuring, and the
6. Castells, M., Rise of the Network Society. 1996.
Urban-regional Process. 1991: Blackwell Publishers.
7. Negroponte, N., Being Digital. 1995, Cambridge,
25. Castells, M., Information technology, globalization
MA: MIT Press.
and social development. 1999, Geneva: UNRISD.
8. Hoffman, D.L. and T.P. Novak, Bridging the Racial
26. Roche, E.M. and M.J. Blaine, Information
Divide on the Internet. Science, 1998. 280(5362): p.
technology, development and policy. Information
390-91.
Technology, Development and Policy, EM Roche &
9. Compaine, B.M., The Digital Divide: Facing a Crisis
MJ Blaine (eds.). Avebury: Aldershot, UK, 1996.
Or Creating a Myth? 2001: MIT Press.
27. Mansell, R., Knowledge Societies: Information
10. Hendry, J.D., Social inclusion and the information
Technology for Sustainable Development. 1998:
poor. Library Review, 2000. 49(7): p. 331-336.
Oxford Univ Press.
11. Warschauer, M., Reconceptualizing the Digital
28. Webster, F., Theories of the Information Society.
Divide. First Monday, 2002. 7(7): p. 0-262. 2002: Routledge.
12. Singh, N., Information Technology as an Engine of 29. Radhakrishnan, S., Examining the “Global” Indian
Broad-Based Growth in India, in The Future of India Middle Class: Gender and Culture in the Silicon
and Indian Business, P. Banerjee and F. Richter, Valley/Bangalore Circuit. Journal of Intercultural
Editors. 2002, Macmillan: London.
Studies, 2008. 29(1): p. 7-20.
13. Heeks, R., India's software industry. 1996: Sage
30. Kuriyan, R. and K.R. Kitner. Constructing Class
Publications Thousand Oaks, Calif.
Boundaries: Gender and Shared Computing. in
14. Arora, A., et al., The Indian software services
Second International Conference on Information
industry. Research Policy, 2001. 30(8): p. 1267- Technologies and Development, Dec 2007, IEEE
1287. Conference Proceedings. 2007. Bangalore.
15. Madon, S., Information-Based Global Economy and 31. Roman, R., Diffusion of Innovations as a Theoretical
Socioeconomic Development: The Case of Framework for Telecenters. Information
Bangalore. The Information Society, 1997. 13(3): p. Technologies and International Development, 2003.
227-244. 1(2): p. 53-66.
16. Saxenian, A.L. Bangalore: The Silicon Valley of 32. Warschauer, M., Technology And Social Inclusion:
Asia? in Conference on Indian Economic Prospects: Rethinking the Digital Divide. 2004: MIT Press.
Advancing Policy Reform,. 2000. Stanford, 33. Sein, M.K. and G. Harindranath, Conceptualizing the
California: Center for Research on Economic ICT Artifact: Toward Understanding the Role of ICT
Development and Policy Reform. in National Development. The Information Society,
17. Jhunjhunwala, A., Can Telecom and IT be for the
2004. 20(1): p. 15-24.
Disadvantaged? Rural Development, 1998. 17(2): p. 34. Bailur, S., Using Stakeholder Theory to Analyze
321–37. Telecenter Projects. Information Technologies and
18. James, J., Low-cost computing and related ways of
International Development, 2006. 3(3): p. 61-80.
overcoming the global digital divide. Journal of 35. Burrell, J. and K. Anderson, "I have great desires to
Information Science, 2001. 27(6): p. 385-392. look beyond my world:" Trajectories of Information
425
and Communication Technology use among 58. Dickey, S., The Politics of Adulation: Cinema and
Ghanaians. New Media and Society, 2008. 10(2): p. the Production of Politicians in South India. The
203-224. Journal of Asian Studies, 1993. 52(2): p. 340-372.
36. Morton, H., Islanders in Space: Tongans Online. 59. Hardgrave, R.L., Politics and the Film in Tamilnadu:
Small Worlds, Global Lives: Islands and Migration, The Stars and the DMK. Asian Survey, 1973. 13(3):
1999: p. 235–254. p. 288-305.
37. Franklin, M.I., I Define My Own Identity: Pacific 60. Dwyer, R. and D. Patel, Cinema India: The Visual
Articulations of'Race'and'Culture'on the Internet. Culture of Hindi Film. 2002: Rutgers University
Ethnicities, 2003. 3(4): p. 465. Press.
38. Rai, A.S., India On-line: Electronic Bulletin Boards 61. Appadurai, A. and C. Breckenridge, Public
and the Construction of a Diasporic Hindu Identity. modernity in India. Consuming Modernity: Public
Diaspora: A Journal of Transnational Studies, 1995. Culture in a South Asian World, 1995: p. 1–20.
4(1): p. 31-58. 62. Rajadhyaksha, A., The Bollywoodization of the
39. Bastian, M.L., Nationalism in a Virtual Space: Indian cinema: cultural nationalism in a global
Immigrant Nigerians on the Internet. West Africa arena. Inter-Asia Cultural Studies, 2003. 4(1): p. 25-
Review, 1999. 1(1). 39.
40. Méliès, G., Le Voyage dans la Lune. 1902: France. 63. Mankekar, P., Screening Culture, Viewing Politics:
41. Chaplin, C., Modern Times. 1936, United Artists: An Ethnography of Television, Womanhood, and
USA. Nation in Postcolonial India. 1999: Duke University
42. Atkinson, P. Technology and the Vision of Utopia. in Press.
Imagining the Future: Utopia, Dystopia and Science 64. MacPherson, Y., Images and icons: harnessing the
Fiction. 2005. Monash University, Melbourne. power of the media to reduce sex-selective abortion
43. Bendle, M. Zarathustra’s Revenge: The Sordid in India. Gender & Development, 2007. 15(3): p.
Utopia of Contemporary Science Fiction Films. in 413-423.
Imagining the Future: Utopia, Dystopia and Science 65. Fernandes, L., Nationalizingthe global': media
Fiction. 2005. Monash University, Melbourne. images, cultural politics and the middle class in
44. Youngblood, D.J., Movies for the Masses: Popular India. Media, Culture & Society, 2000. 22(5): p. 611.
Cinema and Soviet Society in the 1920s. 1992: 66. Jensen, R.T., E. Oster, and R. National Bureau of
Cambridge University Press. Economic, The Power of TV: Cable Television and
45. Singer, B., Melodrama and Modernity: Early Women's Status in India. 2007: National Bureau of
Sensational Cinema and Its Contexts. 2001, New Economic Research Cambridge, Mass., USA.
York: Columbia University Press. 67. Holbert, R.L., D.V. Shah, and N. Kwak, Political
46. Akudinobi, J., Tradition/Modernity and the Implications of Prime-Time Drama and Sitcom Use:
Discourse of African Cinema. 1995, IRIS. Genres of Representation and Opinions Concerning
47. Lopez, A.M., " Train of Shadows": Early Cinema Women's Rights. The Journal of Communication,
and Modernity in Latin America. Multiculturalism, 2003. 53(1): p. 45-60.
Postcoloniality, and Transnational Media, 2003. 68. Dolnick, S., Keep Movie Heroes Alcohol-Free,
40(1): p. 48-78. Bollywood Urged, in Washington Post. 2008:
48. Mukherjee, R., Hum Hindustani. 1960. Washington, DC. p. C 07.
49. Maté, R., When Worlds Collide. 1951, Paramount: 69. Pandian, M.S.S., The Image Trap: MG
USA. Ramachandran in Film and Politics. 1992: Sage.
50. Zaslove, M. and C. Nichols, The Jetsons. 1962, 70. Jacob, P., From co-star to deity: Popular
Hanna-Barbera Productions: USA. representations of Jayalalitha Jayaram. Women: A
51. Newman, S., The Avengers. 1961, American Cultural Review, 1997. 8(3): p. 327-337.
Broadcasting Company (ABC): USA. 71. Sharma, A., Blood, Sweat and Tears: Amitabh
52. Roddenberry, G., Star Trek. 1966, Paramount: USA. Bachchan, Urban Demi-God. You Tarzan:
53. Kubrick, S., 2001: A Space Odyssey. 1968, MGM / Masculinity, Movies and Men, 1993: p. 167-180.
Warner: USA. 72. Lehmann, C. and L.S. Starks, Spectacular
54. Bonnière, R., Hide and Seek. 1977, Public Shakespeare: Critical Theory and Popular Cinema.
Broadcasting System (PBS): Canada. 2002: Fairleigh Dickinson Univ Press.
55. Faden, E., The Cyberfilm: Hollywood and Computer 73. Bordwell, D., Making Meaning: Inference and
Technology. Strategies, 2001. 14(1). Rhetoric in the Interpretation of Cinema. 1989:
56. Kepley Jr, V., Soviet Cinema and State Control: Harvard University Press.
Lenin's Nationalization Decree Reconsidered. 74. Gabriel, T.H., Towards a Critical Theory of Third
Journal of Film and Video, 1990. 42(2): p. 3-14. World Films. Film Theory: Critical Concepts in
57. Mishra, V., Bollywood Cinema: Temples of Desire. Media and Cultural Studies, 2003.
2002: Routledge. 75. Bose, N., Dharti Mata. 1938, New Theatres: India.
76. Menon, R., Kandukondain Kandukondain. 2000.
426
77. Khan, M., Mother India. 1957: India. 118. Kathir, Kadhalar Dhinam. 1999, A.M. Rathnam:
78. Roy, B., Madhumati. 1958, Bimal Roy Productions: India.
India. 119. Barjatya, S., Hum Aapke Hain Koun? 1994, Rajshri
79. Chopra, Y., Deewar. 1975: India. Productions: India.
80. Krishnamurthy, R., Thee. 1981: India. 120. Roshan, R., Koi...Mil Gaya. 2003, Film Kraft: India.
81. Majumdar, P., Aarti. 1962, Rajshri Productions: 121. Rathnam, M., Roja. 1992, Hansa Pictures: India.
India. 122. Kranthikumar, Kanden Seethaiyai. 2001: India.
82. Sahu, K., Dil Apna Aur Preet Parai. 1960, Mohan 123. Vasanth, Satham Podathey. 2007: India.
Pictures: India. 124. Renjith, Nandanam. 2002: India.
83. Balachander, K., Manathil Uruthi Vendum. 1987: 125. Narayana, K.L., Rakhi. 2006, Sri Durga Arts: India.
India. 126. Selva, V., Youth. 2002, Ayngaran International:
84. Chopra, Y., Daag. 1973: India. India.
85. Padmarajan, P., Koodevide. 1983: India. 127. Chopra, Y., Aadmi aur Insaan. 1969: India.
86. Kapoor, R., Shree 420. 1955, RK Films: India. 128. Chopra, R., Mazdoor. 1983: India.
87. Kapdi, M., Aaitya Bilavar Nagoba. 1979: India. 129. Chopra, Y., Kala Patthar. 1979: India.
88. Prakash Rao, K.S., Secretary. 1976: India. 130. Hingorani, A., Dil Bhi Tera Hum Bhi Tere. 1960.
89. Balachander, K., Avargal. 1977: India. 131. Anthikad, S., Nadodikattu. 1987.
90. Chopra, B.R., Pati Patni Aur Woh. 1978, B.R. Films: 132. Chatterjee, B., Chhoti Si Baat. 1975: India.
India. 133. Menon, R., Kandukondain Kandukondain. 2000:
91. Rawail, H.S., Patanga. 1949, Varma Films: India. India.
92. Neelakantan, P., Raman Thediya Seethai. 1972, 134. Menon, G., Minnale. 2001: India.
AVM Productions. 135. Bhargavan, Thiru Ranga. 2007: India.
93. Paranjpe, S., Chashme Buddoor. 1981, PLA 136. Ravishankar, Varushamellaam Vasantham. 2002:
Productions: India. India.
94. Kanwar, R., Laadla. 1994: India. 137. Rasool, Unnai Paartha Naal Muthal. 2004, Anandi
95. Mirza, A., Raju Ban Gaya Gentleman. 1992, Sippy Arts: India.
Films: India. 138. Singh, A., Raqeeb. 2007: India.
96. Rama Rao, T., Andhaa Kanoon. 1983: India. 139. Ghosh, K., Fida. 2004: India.
97. Dixit, C.P., Fakira. 1976, Eros Entertainment: India. 140. Roshan, R., Krrish. 2006, Film Kraft: India.
98. Bhogal, A., Zakhmi Aurat. 1988, Manta Movies: 141. Anand, S., Bachna Ae Haseeno. 2008, Yash Raj
India. Films: India.
99. Chopra, B.R., Insaf Ka Tarazu. 1980, B.R. Films: 142. Sivamani, Law and Order. 2002, Jyotirmayi Pictures:
India. India.
100. Sippy, R., Shakti. 1983: India. 143. Shankar, S., Sivaji. 2007, AVM Productions: India.
101. Hussain, N., Yaadon Ki Baaraat. 1973, United 144. Murugadoss, A.R., Ramana. 2002, Oscar Films:
Producers: India. India.
102. Balachander, K., Arangetram. 1973: India. 145. Vinayak, V., Tagore. 2003, Oscar Movies: India.
103. Sasi, I.V., Avalude Ravakul. 1980: India. 146. Vasu, P., Paramasivan. 2006: India.
104. Samanta, S., Amar Prem. 1971, Shakti Films: India. 147. Vishnuvardhan, Billa. 2007, Sri Keerthi Creations:
105. Ramachandran, M.G., Ulagam Sutrum Valiban. India.
1973, Emgeeeaar Pictures: India. 148. Shankar, S., Anniyan. 2005, Lakshmi Ganapathi
106. Anand, V., Jewel Thief. 1967, Navketan Films: India. Films: India.
107. Bhadran, Olympian Anthony Adam. 1999: India. 149. Jayaraj, R., 4 The People. 2004: India.
108. Madan, Pellaina Kothalo. 2004, AA Naluguru Films: 150. Parvez, S., Jumbish. 1986, Ghalib Studio: India.
India. 151. Srinivasa Rao, S., Aditya 369. 1991, Sridevi Arts:
109. Selvaraghavan, K., Aadavari Matalaku Ardhalu India.
Verule. 2007, Sri Saideva Productions: India. 152. Shankar, S., Jeans. 1998, Sri Surya Films: India.
110. Yelleti, H., Ankit, Pallavi and Friends. 2008: India. 153. Landis, J., Black or White. 1991, Michael Jackson:
111. Goud, R.K., Jodi No. 1. 2003, RK Movies: India. USA.
112. Shankar, S., Kunjikkoonan. 2002: India. 154. Perarasu, Dharmapuri. 2006, A.M. Rathnam: India.
113. Prabhakar, T., Neethone Vuntanu. 2002, Lakshmi Art p. 150 min.
Pictures: India. 155. Vidhyadharan, Vaitheeswaran. 2008, Annamalai
114. Vamsy, Avunu Valliddaru Ista Paddaru. 2002, Films: India.
Anandi Arts: India.
115. Selvaraghavan, 7G Rainbow Colony. 2005,
Ayngaran: India.
116. Kammula, S., Anand. 2004: India.
117. Dasarath, Swagatam. 2008, Aditya Ram Movies:
India.
427
Regulatory Independence and Wireless Market

Development: A Comparative Analysis of Two
African Nations
Annemijn F. van Gorp & Carleen F. Maitland
Abstract— This study analyzes the nature of regulatory ministries and other government departments is a key feature
independence and its influence on wireless market development of that regulatory quality [5]. Further, empirical studies have
in Tanzania and Botswana. The study finds that the level of found that independence has significant effects on both
regulatory independence is associated with improved market market and regulatory performance [e.g. 4, 5]. However,
conditions. The research has implications for theories of many of these studies are carried out either exclusively in
regulation and market development in low income countries. In
relatively high income countries or in global studies covering
particular the Tanzania case suggests that the independence of
regulation can have secondary effects such as diversity of a broad range of institutional environments. Consequently,
technologies and faster transitions to advanced technologies, they often contain two implicit assumptions that may not
while the reversal of independence in Botswana highlights the apply in low income countries, which in turn raise interesting
need for greater insights into the under-theorized dynamic questions.
nature of regulatory independence. The first assumption is that ministries and other
government departments are able to wield power, from which
Index Terms— regulatory independence, Botswana, Tanzania, independent NRAs are insulated. However, in some low
mobile market development income countries ministries and other government
departments are not well-functioning entities and are unable
to wield power. In these environments, is regulatory
I. INTRODUCTION independence still important and if so how?
T elecommunications infrastructure has long been seen as a

catalyst for economic development [1, 2]. However, its
growth is sometimes hindered by a variety of factors
The second assumption is that gains in independence are
sustained. In low income countries where power shifts and
political upheavals create generally less stable political
including its past as a publicly provided good, regulation, and environments, assumptions about the linear nature of
characteristics inherent to network technologies. As network progress on independence may not apply [6]. Under what
technologies, telecommunication systems require conditions are reversals most likely? What are the
investments that are susceptible to expropriation and implications for market development?
hostage-taking [3, 4], as exemplified in many cases by the Further, whereas earlier research on regulatory quality and
government’s power to specify a variety of conditions market development focused primarily on fixed
including geographic market definition, coverage telecommunications infrastructure, more recent research that
obligations, universal service obligations and even prices. includes mobile market development suggests that it is less
Consequently, the level of investment in, and hence the sensitive to regulatory quality [7, 8]. These findings suggest
widespread availability of telecommunications services, that in low income countries, where mobile markets often
including fixed, mobile and Internet access is likely to be play a more significant role in terms of providing access,
influenced by regulatory quality that reduces uncertainty and regulatory quality may play a much less significant role.
insulates the sector from politically driven actions. Hence, in these markets does regulatory independence
During the past two decades many countries, high and low influence mobile and wireless market development, and if so,
income alike, have sought to improve their regulatory quality how? Also, is the nature of this influence similar to that found
by establishing autonomous national regulatory authorities in fixed networks?
(NRAs). The level of independence of these authorities from In answering these questions this research provides further
nuance to theories of regulation and market development. In
Manuscript received September 22, 2008. This work was supported in particular, it sheds light on the relationship between
part by a grant of the Africana Research Center at the Pennsylvania State regulatory independence and its stability and wireless market
University.
A. F. van Gorp is with the Ted Rogers School of Information Technology
development in low income countries. The analysis employs
Management at Ryerson University, 350 Victoria Street, Toronto, ON a comparative case approach of national regulatory
M4V1N1 Canada (phone: +1-416-979-5000 ext. 2482; e-mail: authorities (NRAs) of two African countries, namely
avangorp@ryerson.ca). Tanzania and Botswana. While being among the poorest in
C.F. Maitland is with the College of Information Sciences & Technology
at the Pennsylvania State University, PA 16802 USA (email:
the world, these nations are recognized as having two of the
cmaitland@ist.psu.edu). better regulatory environments in the region [9]. Data were
428
collected through in-depth interviews and document analysis. component of general regulatory quality (as will be discussed
Comparative case analyses generate insights into the nature further in the following section), research has suggested it has
of independence and stability as well as their implications for a unique contribution to market development [4, 5], although
particular characteristics of the wireless market development. findings are mixed [e.g. 11]. Studies of the relationship
The paper is structured as follows. Section 2 provides between independence and market development have
background on regulatory independence and its effects. employed both qualitative and quantitative methods, with the
Section 3 describes our research method and data collection latter becoming increasingly prevalent along with data
techniques and is followed by Section 4 in which the two availability and relying predominantly on econometric
country cases are presented. Section 5 provides a brief analyses of the formal components of independence.
cross-case analysis and in section 6 discussion and Measures of formal independence vary in their complexity,
conclusions are presented. ranging from simple dummy variables to highly complex
measures [e.g. 4, 5, 11]. The most recent and extensive of
II. REGULATORY INDEPENDENCE AND ITS EFFECTS these is the EURI-I index, which is based on eleven general
While research has found that regulatory independence has measures 1 of formal regulatory independence including: 1.
significant effects on telecommunications market multi-sector, 2. multi-member, 3. funding, 4. reporting, 5.
development, to date it is unclear the extent to which these shared roles, 6. legislative appointment, 7. fixed terms, 8.
findings apply in lower income countries. In particular, given renewable terms, 9. staff, 10. budget, and 11. experience [5].
the differences in the state of institutional development in While many of these variables are self-explanatory, several
some lower income countries and the potential challenges of deserve further discussion.
maintaining institutional gains, it is unclear how The multi-sector measure is based on the presumption that
independence will influence market development if at all. a multi-sector regulator will achieve a higher level of
These questions of influence are particularly salient for the independence from any one ministry, with similar reasoning
wireless market both because it has experienced such rapid for multi-member regulators (1) versus single member (0).
growth in low income countries and because evidence Reporting indicates whether regulators are required to report
suggests the influence of regulatory quality is lower in this only to the Minister (0), to both the Minister and the
sector. These issues are discussed in turn below. legislature (.5) or only the legislature (1). The shared roles
measure indicates whether or not a regulator shares its duties
A. Regulatory Quality with the government (0) or is granted exclusive powers (1).
For years scholars have argued for the necessity of Finally, experience is measured as whether or not the NRA
regulatory independence both for proper functioning of the has been in existence for at least two years.
regulator and to enhance both regulatory and market While the EURI-I index provides a complex measure of
performance [6, 10]. Regulatory performance is the extent to independence it is developed only for the European context.
which regulations fulfill their immediate objective, such as Hence, it is unclear the extent to which their results that
lowering interconnection prices, collecting universal service regulatory independence reduces interconnection rates only
revenues or lowering entry barrier for new entrants. These when the public telecommunications operator is partially
outcomes are expected to create conditions which in turn owned by the state, are generalizable to lower income
improve overall sector performance by, for example, contexts. In addition, a question remains to what extent this
lowering consumer prices, extending networks, and operationalization of regulatory independence relates to
improving service quality. regulatory issues other than interconnection.
Regulatory independence has been characterized by That effects of regulatory quality might be different in
institutional arrangements that foster clarity of roles of the lower income countries are suggested by the research of
regulator, and accountability and transparency in the process Gutierrez [4] and Wallsten [12]. Gutierrez [4] found that
of regulatory decisions [6, 10]. The delegation of authority when separating low versus high income Latin American
and need for independence is driven in part by the desire to countries the influence of regulatory quality on market
establish credibility but also in some cases to insulate the development is indeed different. In particular, regulatory
regulatory policies from future politically-driven changes quality has less of an impact on market development in the
(Gilardi 2007). Further, independence is frequently low income countries. The author proposes this may be result
established at the time the autonomous regulator is of the greater difficulty in implementing a positive regulatory
developed, a frequent driver of which is privatization and environment in low income countries or that their
liberalization of monopoly public telecommunications institutional changes were more recent and therefore have yet
operators (PTOs).To this extent, in a study of PTO ownership to produce changes in the market. Conversely, the variables
in Europe, Bauer [11] found that levels of independence were of competition and privatization have a greater effect in lower
highest for those nations that were slow to privatize their income countries, which may be attributed to the greater
PTO. Thus, he concludes the level of independence may be room for improvement. Similarly, in a study of the impact of
driven in part by the level of state ownership in the PTO.
1
Independence is a multidimensional construct, with both Here we exclude the study-specific measure of interconnect powers,
formal and informal components, each with several which measures the extent to which the regulator has powers over
interconnection. This measure is useful to studies of interconnection but may
dimensions. While it is often conceptualized as one not be appropriate for studies of other regulatory issues.
429
an independent regulator on market performance in Africa regulator, (2) the degree of freedom from political and
and Latin America, Wallsten [12] found the regulatory industry interference, clarity of regulatory functions,
variable on its own is insignificant in explaining teledensity. accountability and transparency, and (3) the legal basis of the
Hence, while research concerned with the implications of creation of the regulatory body.
regulatory quality is becoming more nuanced, with These studies suggest that independence will be influenced
independence, competition, and privatization having distinct by a variety of contextual factors. In particular, factors that
contributions to market development, it is unclear the extent influence the ability of governmental departments to fulfill
to which the direction of these developments are relevant for their role in terms of providing checks and balances will
studying independence is low income countries. The impact independence, positively or negatively. For example,
following section discusses the broader context of as described by Jain [18] and Samarajiva [19] in India and Sri
independence and its implications for regulatory Lanka respectively, the judiciary played an important role in
independence in low income contexts. mediating the relationship between the PTO, ministry and
regulator. Whereas in the Indian case the judiciary challenged
B. The Context of Independence
the position of the regulator, in Sri Lanka it was supported.
While at some level challenges to independence are Further, in both cases the judiciary fulfilled its role in
universal, in different contexts they are likely to vary in their providing checks and balances. However, does this imply that
degree. For instance, while nearly all government in other environments where ministries, judiciaries and other
departments face resource constraints, in low income administrative bodies are unable to wield power, that
countries the lack of resources of the regulator and society in independence is still important?
general raise special issues for independence. Examples
include the inability to pay competitive wages to regulatory C. Changes in Regulatory Independence
staff, resulting in high levels of employee turnover as well as A second assumption implicit in much of the research on
the general scarcity of qualified personnel resulting in the the effect of regulatory quality and independence on market
necessity of hiring staff with potential conflicts of interest development is that gains in quality and independence are
(e.g. are former Ministry or PTO employees). This occurred, maintained. As noted by Stern [6, p. 69], “Even when
for example, in Sri Lanka where the former Managing formally independent regulatory agencies have been set up,
Director of Sri Lanka Telecom (SLT) was named Director as in Argentina or Hungary, the question remains as to how
General of the regulatory authority [13, 14]. far their independence is (a) genuine and (b) sustainable.”
This lack of resources may not only affect the regulator Despite the possibility of reversal little attention has been
itself, but also government departments that serve as partners paid to its implications.
and possibly opponents of the regulator as well. For example, This is not to say, however, that the dynamic nature of
also in Sri Lanka, Balasooriya et al. [13] found that the Fair regulation is not considered. Indeed, several econometric
Trading Commission, which by law was to serve as the studies use time series data that capture changes in regulatory
competition authority for the country, was “almost quality and independence over time. For example, in the
inoperative despite its investigative and quasi-judicial power EURI-I index, which measures independence in the 15
assigned by its Act” (p. 388). original EU member states between 1998 and 2003,
Such circumstances suggest that measures of a regulator’s regulatory institutions declined in two countries, remained
independence that focus solely on the regulator may not constant in four countries, and improved in nine countries [5].
suffice. Earlier studies on the effects of governance or However, because declines are far fewer than improvements
regulatory quality on telecommunications market their implications are not obvious from the broader analysis.
development focused on the broader governmental It is also important to note that even where the changes in
environment. For example, studies examined the effect of the regulatory quality and independence are positive, the effects
credibility of the policy regime [15, 16] through use of the may not be. Gutierrez [4], in a study of 22 Latin American
POLCON index [15], a measure of veto points within a countries during the 1980-1997 period, examines changes in
government that in turn constrain any one political actor from regulatory quality and suggests that effects on market
changing government policy. The studies found that over a development may be nonlinear. In particular, he proposes that
wide range of high and low income countries those with at low levels of regulation, further enhancement of the
lower likelihood of arbitrary policy changes had higher regulatory framework at first increases telecom market
growth in fixed teledensity. performance but then over time its impact is slowly
Examining the relationship between general regulatory diminished.
quality exclusively in the telecommunications realm is the The possibility of nonlinear effects of institutional
work of Gutierrez and Berg [17] and Gutierrez [4]. In these improvements begs the question as to how declines in
two studies, the measure of regulatory quality is developed institutional quality affect market performance. However,
from a dichotomous measure of of the the presence (1) or prior to understanding the effects of declines, more
absence (0) of a regulatory framework, to one in which information is required about their fundamental nature (i.e. in
regulatory development is measured as an index. The index what areas of regulatory quality are declines most common)
includes (1) the separation of operations and regulatory and in what conditions are they most likely to occur.
activities, although not necessarily the existence of a separate
430
D. Regulatory Quality and Wireless Market Development regulatory governance, albeit a general one. Further, both
The dynamic nature of regulatory independence raises a countries have in recent years experienced relatively strong
second issue which is its differential impact on wireless telecommunications regulatory governance, with Botswana
market development. In particular, Andonova [8] finds that being recognized by the ITU [22] and Tanzania by the 2006
whereas an increase in the institutional quality positively ICT Investment Summit [9]. Furthermore, both countries
affects the number of Internet hosts, there is no effect for have recently implemented a converged licensing framework
cellular penetration. Further, the static effects of regulatory as two of the first countries in the continent. The
quality were less for cellular penetration than for fixed line implementation of such licensing frameworks has the
services. potential to significantly affect wireless market development,
Andonova [8] proposes these findings are due to wireless as indeed has been the case in Tanzania [23].
infrastructure being less of a sunk investment, one that can be In addition to these similarities there are differences as
redeployed and thus less prone to expropriation and hostage well. First, the two countries come to the strong regulatory
taking. While this may be true in theory, the rapid pace of positions via different routes, with Botswana having a past
technological change may hinder the redeployment of characterized by political stability and strong governance
wireless infrastructure. Other reasons for the lack of influence throughout the government, as compared to Tanzania that has
of regulatory quality on wireless services are possibly the had a history of political and economic turmoil. Second, most
lower level of regulation, greater prevalence of private recently while Tanzania’s regulator has maintained its
ownership and the high level of demand. independence, Botswana has decreased [9]. This difference
These factors suggest that a closer examination of the enables a comparative analysis of regulatory quality as well
relationship between wireless market development and as its link with wireless market development. These dynamics
regulatory quality is required. Earlier studies of regulatory in regulatory governance in both countries, historically and
quality [e.g. 12, 17, 20] actually used wireless market most recently, facilitate comparisons that can identify their
development as a predictor of fixed line market growth. underlying factors.
Clearly greater understanding of the influence of regulatory Data were collected through 46 face to face interviews
quality on wireless market growth (both fixed and conducted in Tanzania and Botswana during September
mobile/cellular) is required. Also, whereas studies have –November 2006, with managers at incumbent and mobile
focused primarily on market growth in terms of access and/or operators and Internet Service Providers (ISPs), as well as
users as well as efficiency of the sector, the relationship with policy makers from ministries responsible for
between regulation and the wireless market given the less telecommunications and regulators from the national
central role of regulation may require a more nuanced regulatory authorities. Additional data were collected through
understanding. Further, given the rapid growth of wireless policy and document analysis from 2006 until present.
markets worldwide, it may be time to look at higher level Data were analyzed through a combined deductive and
service criteria than merely access. inductive analysis. The goal of the deductive analysis was to
Thus, this research seeks to provide further nuance to characterize the degree of independence according to the
theories of the role of independence in telecommunications EURI-I index as developed by Edwards and Waverman [5]
market development. Research on regulatory independence and changes in policies and market development, while the
in the 80’s and 90’s was primarily normative, given the inductive analysis was used to identify possible underlying
limited experience of regulators. Next, qualitative case study factors influencing these phenomena, with a particular focus
research began to emerge and provide insights into on the dynamics of regulatory independence (i.e.
independence as a contribution to regulatory quality. As sustainability of gains in independence), and the role of, and
experience with independence grew and data became relation between, Minister, judiciary, and other
available econometric studies began to provide systematic administrative bodies with the regulatory authorities.
evidence of its effects. Here, we seek to contribute to research
based on in-depth case studies that can subsequently inform IV. REGULATORY INDEPENDENCE AND WIRELESS MARKET
more systematic analyses. In particular, we seek to provide DEVELOPMENT IN TANZANIA AND BOTSWANA
insights into the nature of independence, how and why it In the following sections the development of regulatory
changes and how it affects wireless market development in bodies and markets for wireless services are discussed. Both
low income countries. These insights may subsequently cases start with an overview of regulatory and wireless
generate greater nuance in both qualitative and econometric market developments as related to the implementation of
studies. converged licensing frameworks, followed by a more detailed
analysis of these countries’ evolving regulatory
III. METHODS independence.
The study takes a comparative, explanatory case study
approach [21]. Tanzania and Botswana have been selected as A. Tanzania Case
cases for their similarities as well as differences. They are
both members of the Southern African Development
Community, which generates a common approach to
431
Regulation and Market Development in Tanzania2 once relegated to particular service categories (e.g. fixed,
In 1993 Tanzania liberalized its telecommunications sector mobile, value added, etc.) could obtain licenses without
by splitting its state owned enterprise, the Tanzania Posts and restriction to the types of services offered. Consequently, all
Telecommunications Corporation, into three separate entities, of Tanzania’s large operators now have a network facility
namely the Tanzania Posts Corporation, the Tanzania license, network service license, and application service
Telecommunications Company Limited (TTCL), and the license, which in some instances are complemented by a
Tanzania Communication Commission (TCC). The latter content services license, alongside frequency spectrum
became responsible for regulating the telecommunications licenses for wireless services provision.
market, whereas Tanzania’s ministry responsible for The new licensing framework resulted in market entry and
communications - currently known as the Ministry of an expansion of the range of (wireless) technologies [23]. By
Infrastructure Development - was responsible for developing May 2006 four new service providers had frequency
higher level policies. Established in 1993, TCC was among assigned, and started rolling out mobile services including 3G
the first 30 autonomous regulatory authorities in the world [5, through CDMA2000 [23]. In addition, one recent (2000)
23]. entrant from the Internet access market has expanded its
Nearly simultaneous to this early liberalization process, in services to broadband and voice services provision, bringing
1993 three mobile operators were licensed [24], with two the total number of mobile providers in Tanzania from 5 to 10
more were added in 2000. These later entrants, Vodacom [23]. Further, the number of internet service providers rose
Tanzania and Celtel, are now market leaders in voice services from 11 to 236. New technologies are reflected in the use of
provision3. The addition of these two new providers spurred CDMA (due to scarcity of GSM frequency bands), as well as
growth: While until 2000 subscriber numbers for mobile upgrades of mobile networks to third generation technologies
telephony remained low, since 2000 the number of telephony (both UMTS and CDMA). Also, most operators indicate an
users has grown exponentially 4 . Nevertheless, adoption interest in the potential for WiMax rollout, which is globally
numbers still remained low as compared to many other still in its infancy7.
countries. Hence, steps to further liberalize the sector and
increase competition were taken. Regulatory Independence in Tanzania: The Roles of the NRA
One of the first steps towards full liberalization of the and Minister
market was the partial privatization of incumbent fixed line The growth of Tanzania’s market has been significantly
provider TTCL in February, 2001. Two companies, one of influenced by regulation, the most recent one being the
which was the mobile operator Celtel International, obtained introduction of a converged licensing framework. As already
35% of the shares, the Government of Tanzania kept 36%5, reported by Van Gorp and Maitland [23], Tanzania’s
and the remaining 29% went to local and international regulatory authority TCRA is characterized by a number of
financial institutions as well as TTCL employees. At the time factors that have enabled it to introduce such competition
of the privatization, TTCL was granted a four year exclusivity stimulating regulations. Its regulatory independence is one of
period for fixed line telephony provision, until 2005. the key features, as will be discussed in more detail next.
Next, as part of Tanzania’s liberalization strategy, regulator The independence of TCRA can be analyzed first by the
TCC merged with the broadcasting and postal regulators into EURI-I indicators suggested by Edwards & Waverman [5].
TCRA, the Tanzania Communication Regulatory Authority, Among these factors, those that suggest the TCRA is
established through the new telecommunications Act in 2003. independent include (1) multi-sector jurisdiction (telecom,
After TTCL’s exclusivity period ended, in early 2005, postal and broadcasting); (2) multi-member board control; (2)
TCRA introduced a range of new regulations to further fixed term appointments of board members; (3) appointments
stimulate competition; the most famous one being the of board members are renewable only once; and (4)
introduction of a converged (technology and service neutral) experience – with 13 years TCRA is one of the older
licensing framework. This meant that all operators as of then regulatory agencies in the world. However, one indicator
were eligible to provide both fixed and mobile services: firms does not necessarily imply independence: while TCRA
receives its funding directly through licensing fees and levies
2
For a more detailed overview of market developments and Tanzania’s
6
liberalization strategies see 23. Van Gorp, A.F. and C.F. Maitland, Quote taken from East African Business Week, “TZ Okays More
Regulatory Innovations in Tanzania: The Role of Administrative Capabilities Cellular Firms”, 22 May 2006. See
and Regulatory Governance. Info: The Journal of Policy, Regulation and http://www.busiweek.com/index.php?option=com_content&task=view&id=
Strategy for Telecommunications, Information and Media, 2009. 11(1): p. 1582&Itemid=39 Last accessed August 11, 2007.
7
64-77. Even though much progress has been made; few challenges have
3
By August 2007 market shares for voice providers are: 1. Vodacom with surfaced. Due to the high demand, regulator TCRA was forced to temporarily
51% market share; 2. Celtel with 26% market share; 3. MIC Tanzania-Tigo stop frequency applications for spectrum that includes the common bands for
with 13% market share; 4. Zantel with 7% market share; and 5. TTCL with GSM, CDMA, and WiMax, and was forced to review the band plan in 2006.
3% market share. See http://www.tcra.go.tz/publications/telecom.html. Last Subsequently, in 2007, plans were initiated to comprehensively survey
accessed August 6, 2007. Tanzania’s ICT infrastructure, with an eye towards determining the extent to
4
For a detailed overview of number of subscribers over time, see also which further opening of the market in the future is desirable, and to
http://www.tcra.go.tz/publications/telecom.html . Last accessed August 6, determine the optimum number of operators in the sector, as well as the
2007. degree of (lack of) competition in different segments of the market. See also
5
See http://www.ttcl.co.tz/about_history.asp Last accessed August 6, http://www.cellular-news.com/story/31356.php?source=newsletter Last
2007 accessed May 30, 2008.
432
by industry, Parliament and Minister play a role in approval July 2007 Sasktel International of Canada started a three year
of TCRA’s budget. contract to take over management control8.
Nevertheless, as per the EURI-I index, there are a number Further, Tanzania’s broader institutional endowment is
of indicators that would actually suggest ‘dependence’. For known for a highly dominant Executive (including President
example, Tanzania’s staff number of 97 is far below that in and Ministers) in national policy making processes. While
Europe, however for African standards is relatively high [9]. Parliament has powers of scrutiny, due to the large majority
Additionally, TCRA primarily reports to the Minister while of one party in Parliament, at the end the Legislature thus
the legislature is hardly involved. Parliament does receive cannot easily hold the Executive accountable [27]. This
TCRA’s budget and annual report from the Minister, but no might explain the limited role of Parliament in
specific approval powers etc. are stated in the TCRA Act of telecommunications matters; besides enacting the TCRA Act
2003. The board members of TCRA are also not appointed by of 2003, Parliament’s single formal role in relation to TCRA
Parliament, but through a committee established by the regards to how it may provide funds to TCRA, and that it
Minister. Further, as per the TCRA Act of 2003, the Minister shall receive TCRA’s annual report from the Minister.
has the power to engage in many regulatory activities. For However, this does not explain the limited role the ministry
example, TCRA is not allowed to award or cancel licenses has played.
with an exclusivity period, universal service obligations or The limited attempts at protecting the incumbent may
any license for a term of more than five years without partially be due to the ministry’s lack of resources. Currently,
consulting the Minister. the Ministry of Infrastructure Development employs roughly
These factors implying certain degrees of dependence vs. 100 people responsible for departments spanning from postal
independence have also changed throughout the years. In to transportation, but as of 2006 only 6 staff are assigned to
favor of independence is first TCC’s evolution from a single telecommunications and ICT matters [23]. Consequently, the
sector regulator into a converged regulator responsible for ministry realistically does not have a strong influence on ICT
postal, telecom and broadcasting. Second, in some areas the and telecommunications policy or regulation. As one
Minister’s powers and influence in the regulatory process manager from the private sector put it, “the regulator is trying
have decreased. Third, during the time that the regulator was to compensate for lack of policy from the Ministry’s side”.
purely in control of telecom, under the name TCC, board Furthermore, the manager mentions that the Minister has an
members could renew their terms an unspecified number of overseeing role, but that in practice TCRA is “independent”.
times, whereas currently more stringent rules apply where Another manager says “frankly speaking, the ministry
board members can only renew their appointment once. doesn’t have enough staff. They should actually give more
However, regardless of developments in the direction of input”. This relative lack of involvement of the ministry in the
more independence for the regulator, the Minister is still telecommunications sector might be further reflected in the
awarded an extensive formal role. Hence, the question lack of engagement of operators with the ministry (as
remains: why is TCRA referred to as an “independent” opposed to the regulator), as is explained by regulatory affairs
regulator? Particularly interesting is the fact that the managers at two mobile operators.
incumbent TTCL remains partially state-owned, which in
many countries has led to government protection and delayed
B. Botswana Case
implementation of liberalization due to incumbent-initiated
court proceedings.
First, in Tanzania, little evidence exists of Regulation and Wireless Market Development in Botswana
incumbent-initiated court proceedings. In interviews with The case of Botswana shows a very different history of
managers at TTCL’s competitors, including mobile operators regulatory quality and wireless market development than
and ISPs, attempts at delaying the implementation of Tanzania. Botswana, due to its population of 1.8 million, has
liberalization mechanisms through court proceedings were a very small market with, besides an incumbent fixed line
not mentioned, and stands in stark contrast with a country like operator, until very recently only two mobile operators.
South Africa [see e.g. 9]. There exist two potential underlying While seemingly a small number of mobile operators, at the
reasons for the lack of judiciary involvement. First, Tanzania time of licensing in 1996 when market liberalization started
is known to have a fairly weak judicial system, which has and the Botswana Telecommunications Authority (BTA) was
restrained market development in many sectors due to the established, many thought that only one mobile operator
lack of (threat of) enforcement, leaving government would be viable due to the small population size. By 2007
unchecked by law [25, 26]. This may have refrained TTCL already 1.2 million people used mobile phones [28]. But
from taking action. nevertheless, it was felt that competition in the Botswana
A second reason for limited action from the side of TTCL telecommunications sector developed “unevenly across
might lie in the problems with management control, which different regions of the country and at different levels” [29, p.
first changed from government to Celtel during TTCL’s 3], as outside of cities and major districts, provision of
partial privatization in 2001. Even though in the years after telecommunications services often remained limited or even
Celtel extended its shareholdership to 65% shares while non-existent. A new service neutral licensing framework was
government remained 35%, problems in management control 8
led to the government taking back management control. In See e.g. http://wirelessfederation.com/news/sasktel-takes-over-at-ttcl/
Last accessed February 18, 2009.
433
developed in hopes of “this imbalance [to] be corrected” [29, generated a significant new impulse to market entry as it has
p. 4]. done in Tanzania. Moreover, even though the PTO market
On June 20, 2006 the Minister of Communications, Science segment shall only in 2009 be considered for further
and Technology announced the intent to further liberalize the liberalization and market entry [31], as explained at the
telecommunications sector through five separate changes to Ministry of Communications, Science & Technology, “The
license conditions, as well as having the incumbent rebalance current regime means there is no intent to get more than three
tariffs and allowing new entrants to tender for service neutral main voice operators”. The future licensing of new national
national licenses [30]. The first five measures were network providers will depend on the Minister, who is in
accomplished by March, 2007 through the introduction of the charge of awarding licenses.
“Service Neutral Licensing Framework in the Era of Since April 2008 BTC has started its mobile operations
Convergence” [31]. Prior to the change in the licensing under the brand name Be Mobile. Managers from industry
framework, the market was categorized into “Fixed, Cellular, explain that indeed more competition is felt due to the
Internet Service Providers (ISP), Satellite and Data etc.” regulatory changes. Perhaps as a result of this, more advanced
Market segments determined as “non-competitive” were services are introduced: Mascom Wireless has launched a 3G
restricted in terms of the number of players within the and 3.5G HSDPA network10, while Orange has introduced
segment. In Botswana these were Fixed and Cellular, whereas the Blackberry in February 2008 on its GSM/GPRS/EDGE
ISPs, data service providers and paging services providers network11. And, even though the new regulations had ISPs
have been determined to be competitive9. Hence, only one remain dependent on infrastructure provision by third parties
fixed and two cellular providers were licensed, while a [9], as of July 2008 value added network services (VANS)
significantly larger number of ISPs and data providers providers (e.g. ISPs) may tender for Fixed Wireless Access
obtained licenses. Spectrum in a number of frequency bands [32].
The new service neutral licensing framework required the
existing fixed and mobile operators (BTC, Mascom Wireless Regulatory Independence in Botswana
and Orange Botswana) to obtain so-called “public Similar to the Tanzanian case, Botswana’s regulator BTA
telecommunications operators” (PTO) licenses. Under this scores well on a number of variables of Edwards and
license, any one of them is eligible to provide both cellular Waverman’s [5] EURI-I index. BTA scores well on (1)
and fixed services. In addition, all PTO licensees have multi-member board control, with 5 non-executive members;
become eligible to operate the international gateway and to (2) fixed term appointments of board members (4 years); and
“self-provide” – a relief to mobile operators that previously (3) experience – with 12 years of experience, Botswana has
were required to use incumbent BTC’s backbone. The three significant regulatory experience. In addition, BTA has both
PTOs however are still not eligible to provide value-added telecommunications and broadcasting departments, and as
Internet services. Hence, BTC continues to use its subsidiary such may be perceived as a multi-sector regulator. However,
Botsnet for Internet services provision [9]. there are separate boards for both sectors.
The introduction of a service neutral license nevertheless There are a number of factors related to independence as
was a surprise to some market players. As explained by identified by Edwards and Waverman (2006) that BTA does
regulatory managers at incumbent BTC, it was expected that not score very well on, however. First, similar to Tanzania,
a third mobile operator was to be licensed instead of a fully the staff number of BTA is low compared to for example
revamped licensing framework to be implemented. BTC European regulatory authorities, at 70 staff in 2006.
applauded this decision, as BTC was “not sure” if it could Nevertheless, for African standards it is very high;
have applied for a mobile license otherwise – as two particularly when taking into account the low population size
managers explain. of Botswana (1.8 million) [9]. A second factor that BTA does
Interestingly however, three regulatory staff at regulatory not score very well on is the lack of involvement of the
authority BTA explain that BTA actually recommended the legislature. Botswana’s Parliament (National Assembly) does
Minister to license a 2nd fixed and 3rd mobile operator after not have a dedicated communications portfolio committee,
stakeholder consultation processes. Nevertheless, the and although the Parliament has passed the
Minister had the authority to either accept or refuse BTA’s Telecommunications Act of 1996 as well as the Amendment
recommendations. Further, as a manager at regulator BTA Act of 2004, Parliament does not have significant formal
explains, the incumbent BTC lobbied the government – powers in relation to BTA. It is only eligible to (1)
concerned that under BTA’s recommended plan BTC could appropriate money for the Authority’s fund; and (2) receive
not have a mobile license. Hence, according to a BTA an annual report and auditing account within 30 days after the
employee, the Minister decided to convert the three major Minister’s reception of both [33]. This also implies that there
operators’ licenses into one service neutral license. are two other factors not in favor of independence: there is no
Thus, while throughout the years the growth in ICT
10
connectivity in Botswana has been greater than expected, the See http://www.cellular-news.com/story/33201.php. Last accessed
introduction of the new licensing framework has not September 20, 2008.
11
See http://www.orange.co.bw/press_room/news_page.php?newsID=11
“Orange and RIM Introduce BlackBerry for the first time in
9
See http://www.bta.org.bw/licensing.html Last accessed March 30, Botswana”, February 14, 2008. Last accessed September 20,
2008. 2008.
434
legislative appointment of board members, and BTA reports budget has been observed immediately: in 2006, BTA
only to the Minister. Third, BTA’s board members may provided to the treasury Pula 10 million, and in 2007 Pula 2.5
renew their appointments an unspecified number of times. million (respectively about USD $1.5 million and USD
Finally, the shared roles of BTA and Minister, and BTA’s $375.000) [28]. Additionally, the government can now take
budget, seemingly are indicative of a limited degree of part of the profits of BTA. A manager at BTA suggests that
independence. The Minister has the power to interfere in a the Minister taking back power is a trend observed in more
number of regulatory areas. First, as per the southern African countries, including Lesotho, South Africa,
Telecommunications Amendment Act of 2004, the Minister and Namibia [9].
has to approve all licensing of fixed line and cellular Indeed, that these developments represent a transfer of
telephone service and may set licensing fees. Additionally, power from the regulator back to the Ministry is confirmed by
the Minister has the “power to make regulations” [33, B101], people in the industry and at the regulator as well. However,
and thus may interfere in many areas of regulation. while theoretically the Minister does have more power,
Thus, as much as six indicators have been identified that reality might be slightly different due to underlying resource
could imply a low level of independence. Further, some issues. As one of the Directors at BTA states, “even though
issues are indicative of the dynamics, including reversal, of the Minister legally has more power […] at present that is not
independence. Especially in its early days, BTA was an issue”. Further, another Director at BTA indicates, “there
perceived as a model regulator, and moreover, governance in is a lot of consultation between the ministry and BTA – BTA
Botswana in general has been of a high standard. BTA’s has a lot to say. Liberalization was initiated by BTA. The
exemplary regulatory governance and independence is for ministry relies a lot on BTA because it is better resourced”.
example reported in a 2001 report by the ITU which states The Director furthermore continues: “The Ministry is really
that “the Botswana experience also offers a number of world under-resourced. Most work is carried out by BTA. A policy
models. Among these are that BTA has achieved a high level direction should come out, which BTA would then have to
of independence as measured by the lack of influence from implement.” This however does not always happen. As the
the government in implementing its mandate. Its virtually Director continues, “a problem […] is that BTA basically
unfettered authority to license operators and self-financing made the national plan. This is not desirable for checks and
operation may also develop as a world model. BTA further balances.”
provides good models of strong legal processes in carrying Nevertheless, regardless of these issues, generally
out its regulatory mandate.” [22]. Additionally, with regard to speaking Botswana is still perceived by many people,
licensing, the ITU (2001) report even states that “BTA is one including those directly involved in Botswana’s private
of the rare regulatory bodies that has been given almost sector, as having a very good regulator.
complete freedom to decide which services are to be licensed,
how many licenses should be granted for each service and V. CROSS CASE ANALYSIS
which operators are to be awarded a license” [22, p. 27]. Not
only the ITU has noticed Botswana’s good governance; The cases of Tanzania and Botswana bring to bear two key
throughout the region regulators refer to Botswana’s strong aspects related to independence: the relation between
governance. For example, a former regulator from South Minister and regulatory authority as well as sustainability of
Africa perceives BTA to be a “model regulator”, which by regulatory independence. First, according to the
industry managers and regulators across the Southern African independence indicators by Edwards & Waverman’s EURI-I
continent is seconded [9]. Botswana’s impressive regulatory index [5], Tanzania currently scores better. The difference
governance is furthermore acknowledged in previous however primarily lies in terms of board members of the
research. McCormick [34], suggests in her article with the regulatory authority of Botswana being renewable an
revealing title “Telecommunications reform in Botswana: a unlimited number of times. Further, in Botswana, the
policy model for African states” that Botswana has been able Minister has significant control over the budget of the
to develop a model of policy and regulatory governance regulatory authority.
known by significant transparency in decision making. However, in both countries there are shared roles between
However, the amendments to the Telecommunications Act the Minister and NRA, and in both countries the Minister
of 2004 have led to BTA’s degree of dependence becoming theoretically has significant room to implement regulations.
point of debate. The Telecommunications (Amendment) Act The case of Tanzania however showed that even though this
of 2004 substitutes a number of sections of the 1996 formally might be the case, it does not necessarily stand in the
Telecommunications Act that give more ‘power’ to the way of independence in day to day operations. Due to limited
Minister. In particular, these changes enable the Minister to 1. resources at the ministry, Tanzania’s regulatory authority has
determine the use of surplus funds that accrue to the enjoyed full freedom to design regulations, and even has
Authority, 2. make regulations, on the recommendation of the played a major role in drafting policies; which typically is a
Board, 3. set licensing fees, and 4. approve all decisions on Ministerial responsibility.
the licensing of fixed line and cellular telephone service [33]. While over the years in Tanzania formal independence has
Thus, a number of decision making powers formerly under increased, in Botswana it has actually decreased. The
the authority of the BTA Board have been transferred back to Minister’s decision to ignore BTA’s recommendation to
the Minister. Further, the Minister’s control over BTA’s tender for a third mobile operator but instead to allow the
435
incumbent fixed line operator to start offering mobile services Indeed, Bauer [11] suggests that complaint records and
illustrates this point. Thus, the cases provide evidence of court decisions may be an adequate measure of one
independence being a dynamic phenomenon. Under what component of independence, as long as they are free of a
conditions are reversals most likely? Comparing Botswana systematic bias. While specific case records may be difficult
and Tanzania one can conclude, although tentatively given to obtain and bias even more difficult to assess, as the above
the limited number of cases, that reversals are more likely in quote suggests, a measure that reflects simply the
countries where the government owns a larger stake in the involvement of the courts may adequately differentiate
incumbent, and thus has incentives to protect the incumbent. different levels of independence.
Finally, the cases provided evidence that suggests that Also, the case data, together with other instances of
regulation has an important role to play in determining the reversals, suggest there are a variety of reasons for these
number of carriers and hence competition, which has been reversals, which may be more or less strategic. The case of
shown to have positive market effects in many countries. Botswana reflects a strategic reversal in that it enabled the
Tanzania’s implementation of its converged licensing ministry to create market conditions favorable to the
framework with an open approach to market entry led to incumbent. However, for example the regulator in Sweden,
market entrance of 4 new operators and additionally an which was recently identified by the EU as needing greater
impetus to innovation (i.e. the offering of more advanced independence due to a court ruling that resulted in reduced
technologies). Botswana’s approach led to significantly less powers for the regulator 12 , reversals can occur for other
impetus, even though competition of course was stimulated reasons as well. This together with the previously mentioned
with there being a third mobile operator now. declines in independence reported in the EURI-I index (see
section 2.3), indicate that independence reversal is an issue
VI. DISCUSSION AND CONCLUSIONS common to both high and low income regions. Hence, a more
nuanced understanding of the dynamic aspects of
This study analyzed the nature of regulatory independence independence, as well as the day to day practices in case of
and its influence on wireless market development in shared roles between Minister and regulatory authority, as
Botswana and Tanzania, and as such, the study has well as the causes and market effects of reversals is required.
implications for theories of regulatory independence in low Further, insight into reversals in regulatory independence
income countries. may also contribute to understanding reversals in the more
First, we find that traditional notions and measures of general realm of regulatory quality [see e.g. 35].
independence may under- or overestimate a regulatory Finally, the research findings reported here and elsewhere
authority’s status, particularly when the broader institutional suggest that regulatory independence has different effects on
context is not taken into account. When other administrative fixed versus wireless market development. Indeed, the lower
bodies, such as the ministry, legislature or judiciary, are levels of regulation in these markets and their rapid growth
weak, independence of an NRA may generate greater across all nations, independent of institutional arrangements,
freedoms than suggested by their legal status. As reflected in has led some researchers to question the role of regulation
the case of Tanzania, to date these freedoms have generated altogether. First, these questions may be more salient in high
positive outcomes for market development. However, this income countries where the diversity of checks and balances
freedom exists with minimal checks and balances, which are is in general greater. Second, if indeed the effects of
typically a requirement for stable political systems. This regulation on traditional measures of market development
finding suggests that independence of the regulator should such as teledensity and efficiency are universal, it may be that
not be assessed solely by characteristics of the regulator itself. new measures of market development are needed. The case of
Measures must further consider: what is the regulator Tanzania suggests that measures such as the diversity of
independent from? technologies and speed of transition to advanced
This may be interpreted as a call to return to studies that technologies, both likely important to consumer satisfaction
examined the broader construct of regulatory quality and not and to enhancing the true benefits of wireless technologies,
just independence. Indeed, some measures of regulatory may be fruitful areas for future research.
quality such as the POLCON index, which measures veto
points in a government, reflect the existence of checks and
balances. However, the checks and balances related to the REFERENCES
independence of the regulator are more subtle. As suggested [1] Saunders, R.J., J.J. Warford, and B. Wellenius, Telecommunications
and economic development. 1983, Baltimore and London: Johns
by Samarajiva [19, p. 81] who described the increasing
Hopkins University Press for the World Bank. 395.
credibility and independence of the regulator in “The [2] Bonciu, F. and H. Williams, The contribution of the mobile industry to
incumbent operator appealed against one of the economic development in Romania. Info, 2006. 8(5): p. 37-55.
interconnection decisions, and for the first time the appeal
went to the courts, not to the political and administrative 12
See “Commission opens three new cases on independence and
authorities as in the past.” In this case the incumbent may effectiveness of telecoms regulators in Latvia, Lithuania and Sweden”,
have had the right to appeal to the courts all along, but never IP/08/1343, September 18, 2008. Retrieved from
http://europa.eu/rapid/pressReleasesAction.do?reference=IP/08/1343&type
needed to. =HTML&aged=0&language=EN&guiLanguage=fr. Last accessed
September 22, 2008.
436
[3] Williamson, O.E., Credible commitments: Using hostages to support Botswana, at the BTA Offices. 13 june 2007. 2007: Gaborone,
exchange. American Economic Review, 1983. 73(4): p. 519-540. Botswana.
[4] Gutierrez, L.H., The Effect of Endogenous Regulation on [30] Venson-Moitoi, P., Press Statement June 20, 2006: Further
Telecommunications Expansion and Efficiency in Latin America. Liberalization of the Telecommunications Sector. 2006: Gaborone,
Journal of Regulatory Economics, 2003. 23(3): p. 257-286. Botswana. Retrieved September 20, 2007 from
[5] Edwards, G. and L. Waverman, The Effects of Public Ownership and http://www.bta.org.bw/Liberalisation/Press%20Statement%20-%20L
Regulatory Independence on Regulatory Outcomes. Journal of iberalisation%203.pdf.
Regulatory Economics, 2006. 29(1): p. 23-67. [31] BTA, Service-Neutral Licensing Framework in the Era of
[6] Stern, J., What makes an independent regulator independent? Convergence. 2007, BTA: Gabarone, Botswana. p. 7.
Business Strategy Review, 1997. 8(2): p. 67-74. [32] BTA, Public Notice Radio Spectrum Licensing Tender No. BTA
[7] Varoudakis, A. and C.M. Rossotto, Regulatory reform and BTA/PT/02/2008.2009: Public Invitation to Tender for Fixed Wireless
performance in telecommunications: Unrealized potential in the Access Spectrum in the Frequency Bands 1785 – 1805 MHz, 2500 –
MENA countries. Telecommunications Policy, 2004. 28: p. 59-78. 2690 MHz and 3400 – 3600 MHz. 2008: Gaborone, Botswana:
[8] Andanova, V., Mobile phones, the Internet and the institutional http://www.bta.org.bw/tenders/2008/Spectrum%20public%20announ
environment Telecommunications Policy, 2006. 30(1): p. 29-45. cement%20(3).pdf.
[9] Van Gorp, A.F., Increasing Regulatory Capacity: The Role of the [33] BotswanaGovernmentGazette, Telecommunications (Amendment)
Region in Shaping National ICT Policy in Southern Africa. PhD Bill, 2004. Bill No. 15 of 2004. 2004: Gaborone, Botswana. p. 3.
Dissertation, the College of Information Sciences & Technology, the [34] McCormick, P.K., Telecommunications Reform in Botswana: A Policy
Pennsylvania State University, University Park, PA, USA. 2008, p. Model for African States. Telecommunications Policy, 2001. 25(6): p.
207. 409-420.
[10] Melody, W.H., On the meaning and importance of 'independence' in [35] Kaufmann, D., A. Kraay, and M. Mastruzzi, Governance Matters VII:
telecom reform. Telecommunications Policy, 1997. 21(3): p. 195-199. Aggregate and Individual Governance Indicators 1996-2007, in
[11] Bauer, J.M., The coexistence of regulation, state ownership and Policy Research Working Paper 4654. 2008, The World Bank
competition in infrastructure industries: Quagmire or Institutional Development Research Group Macroeconomics and Growth Team &
Innovation?, in Quello Center of Telecommunication Management World Bank Institute Global Governance Program: Washington, D.C.
and Law Working Paper Series. 2003, Michigan State University: East
Lansing, Michigan.
[12] Wallsten, S.J., An Econometric Analysis of Telecom Competition,
Privatization, and Regulation in Africa and Latin America. The
Journal of Industrial Economics, 2001. 49(1): p. 1-19.
[13] Balasooriya, A., Q. Alam, and K. Coghill, The effectiveness of the
telecommunications regulatory regime: The case of Sri Lanka
Telecom. Public Administration and Development, 2006. 26: p.
383-393.
[14] Samarajiva, R., The Role of Competition in Institutional Reform of
Telecommunications: Lessons from Sri Lanka. Telecommunications
Policy, 2000. 24(8-9): p. 699-717.
[15] Henisz, W.J. and B.A. Zelner, The institutional environment for
telecommunications investment. Journal of Economics and
Management Strategy, 2001. 10(1): p. 123-147.
[16] Henisz, W.J., The institutional environment for infrastructure
investment. Industrial and Corporate Change, 2002. 11(2): p. 355-389.
[17] Gutierrez, L.H. and S. Berg, Telecommunications Liberalization and
Regulatory Governance: Lessons from Latin America.
Telecommunications Policy, 2000. 24(10-11): p. 865-884.
[18] Jain, R., Interconnection regulation in India: Lessons for developing
countries. Telecommunications Policy, 2006. 30: p. 183-200.
[19] Samarajiva, R., Regulating in an Imperfect World: Building
Independence through Legitimacy. Info, 2001. 3(5): p. 363-368.
[20] Gutierrez, L.H., Regulatory Governance in the Latin American
Telecommunications Sector. Utilities Policy, 2003. 11: p. 225-240.
[21] Yin, R.K., Case Study Research: Design and Methods. Third ed. 2003,
Thousand Oaks, CA: Sage Publications.
[22] ITU, Effective Regulation Case Study: Botswana 2001. 2001: Geneva,
Switzerland. p. 48.
[23] Van Gorp, A.F. and C.F. Maitland, Regulatory Innovations in
Tanzania: The Role of Administrative Capabilities and Regulatory
Governance. Info: The Journal of Policy, Regulation and Strategy for
Telecommunications, Information and Media, 2009. 11(1): p. 64-77.
[24] Moshiro, S., Licensing in the Era of Liberalization and Convergence.
2005, ITU.
[25] Mwaikusa, J.T., The Limits of Judicial Enterprise: Judicial Powers in
the Process of Political Change in Tanzania. Journal of African Law,
1996. 40(2): p. 243-255.
[26] Gloppen, S., The Accountability Function of the Courts in Tanzania
and Zambia. Democratization, 2003. 10(4): p. 112-136.
[27] Lawson, A. and L. Rakner, Understanding Patterns of Accountability
in Tanzania. 2005, Oxford Policy Management, Chr. Michelsen
Institute and Research on Poverty Alleviation (REPOA): Oxford, UK.
[28] BTA, Annual Report 2007. 2007, BTA: Gaborone, Botswana.
[29] Pheko, T.G., Speech by the Chief Executive, Mr T.G. Pheko, at the
Award Ceremony of the Service Neutral License to Mascom Wireless
437
Social Enterprises: A Vocational

Entrepreneurship Framework for Street Youth
Paul Javid12, Kentaro Toyama1, Manna Biswas3
Microsoft Research India1, University of California Berkeley2, Praajak3
Abstract – It is estimated that there are 100 million “street square kilometer [4]. About 4 million live in one of about
children” in the world [1]. Street children typically live 5000 slum communities, and 42% earn between US$132 and
independent of families in makeshift living arrangements, and
$372 dollars per year. These slum communities have an
survive on little but their wits and the camaraderie of small
overall literacy rate of 28.5%, and only 3.9% of males and
gangs.
1.3% of females have graduated from secondary school [5].
To better understand the lives of street children, we conducted Few have access to opportunities, social networks, or
150 days of ethnographic investigations in and around marketable skills that would provide them with a steady
Calcutta, with an emphasis on street children who live in train stream of reasonable income; the majority find irregular
stations. Our hope was to understand their typical challenges, income streams from the informal service industry or small-
and to see whether they could be addressed with information scale industries such as tailoring, carpentry, blacksmithing, or
and communication technology. Among other things, this manufacturing [6]. Indian street children are routinely
study led to the identification of an unreported knowledge gap detained illegally, beaten, tortured and sometimes even killed
among the children. Namely because their world is limited to by police, based on the general population’s mistrust of street
geographical “islands” within a couple-block radius of train children, and the corruption of the police force [7] [8] [9].
stations, they were entirely unaware of the of micro-
In this paper, we describe the three phases of research we
entrepreneurial possibilities beyond their islands.
conducted to provide a technology-based system that provides
In the second half of this paper, we describe a preliminary trial older street children with information about available income-
with an NGO to expose street children to entrepreneurial generating opportunities: First, we conducted an ethnographic
possibilities using mediated video instruction. Such a system exploration of street children in and around Calcutta. Second,
shows promise, if combined with ongoing social support: we tried a couple of short-term interventions to understand the
some children were both quick to understand and try their impact they would have on youths who live in train stations.
hand at new professions, and were to continue their These early investigations led to the novel understanding that
professions for at least two months with ongoing NGO street children have an extremely limited view of the job
support. opportunities available to them, due to a kind of tunnel vision
about the world around them. Third, we then implemented and
observed a preliminary trial of a new system called Social
I. INTRODUCTION
Enterprises, which informs street children of a diverse set of
E stimates suggest that there are approximately 100

million street children or street youth who live and work
on the roads and squares of cities all over the world [1].
micro-entrepreneurship opportunities.
II. AN ETHNOGRAPHY OF CALCUTTA STREET CHILDREN

These terms connote a wide range of children who are either
homeless or who work on the streets but sleep at home In an effort to understand the life of street children, we began
(typically slum community). Many remain in contact with with an eight-month-long ethnographic investigation of street
their families, are male, and work on the streets to augment children in and around Calcutta, beginning October 2007. Our
their household income. Many others have separated from or goal was both to get a background sense for the life of street
were abandoned their families due to war, poverty, family children in this region, as well as to identify opportunities for
alcoholism, or abuse, and other factors [1]. ICT (information and communication technology)
interventions.
India has the largest population of the world’s street children
[2]. At least eighteen million children live or work on the Methodology
streets of urban India, doing work as porters at railway
terminals, informal tour guides, tea sellers, and so forth [3]. The first author moved to Calcutta to live near street children
Calcutta, for example, may have as many as 500,000 street for nine months. He stayed in an apartment near Park Street,
children, who typically speak the language of their state (e.g., which is a busy urban street of food stalls and street youth,
Bengali in Calcutta), as well as a local dialect of their home with whom he had daily interaction.
community. They have little or no formal education, and live On a more formal basis, he spent over 150 days in the field
either in Calcutta or at Malda train station, about 8 hours by
train from Calcutta, spending between 2 and 8 hours a day
in a larger context of urban poverty. Calcutta has a population conducting structured and unstructured interviews with local
of over 14 million people, and a density of 24,760 people per
438
street vendors and micro-entrepreneurs (N=50), street children are employed as beggars. Such beggars typically wear
who lived either on city streets (20) or in train stations (10), “costumes” to disguise themselves as homeless and in need of
and leaders (8 non-profit directors, a dozen non-profit mid- food, and roam the streets with their mothers begging for
level managers, and 2 primary school principals who work money. At times such families might ask for consumable food
with street children) and field staff (6) from a local non- items, but the large majority of food donations is later re-sold
governmental organization (NGO) that specialized in working and liquidates for cash. For such families, begging is not the
with street children. Observations of the routines and main source of income, but such extra income is used by the
behaviors of these groups were also made throughout the family to purchase extra consumer goods such as a radio.
period.
Those in the second group do not have a primary residence,
Among the street children, we cast a wide net, but focused and therefore find shelter on the streets in the city. This group
primarily on subjects who did not claim to possess any formal typically earns less money than those in group 1, through
job skills, expressed interest in wanting a more stable career, “wholesale vending.” They will purchase small food
had received no formal schooling, were considered too old to consumables, such as packets of gum, from wholesaler street
re-enter a traditional public education track, and had found vendors (commonly available in cities like Calcutta), bring the
alternative means to earn a living and survive. food items to high-traffic areas, and resell those items to
passerby clientele.
The NGO, Praajak, has presence in four train stations in West
Bengal, one of which is Malda. Praajak works with For the children at the train stations, a majority of the children
marginalized children and adolescents in west Bengal. Its aim come from families who live in villages within a 12-hour train
is to create a gender- just environment of care and protection journey. They have been living in train stations from
around these children so that they grow up as more anywhere between 3 months and 8 years. All of the train-
“responsible, productive and law-abiding citizens.” Praajak's station street youth are self-employed. They have two typical
programs include education, health and social rehabilitation sources of income: (1) the buying and reselling of various
work. Praajak works mainly with younger boys. goods, such as water bottles or various small food items, at
train stations, and (2) service-related jobs such as sweeping
The interviews focused on general aspects of the lives of street
and cleaning of train stations. Children might board a train
children that are lost in tables of global statistics. Apart from
after it stops at the station, sweep, and ask for recompense
information about their age, gender, and general well-being,
from the passengers. Typically the children earn between
specific questions were asked about life as they knew it: What
US$0.75 and $1.00 per day for such services.
were the children’s relationships to their parents and relatives?
Where do they consider home? Why did they leave home? They do not have a concept of savings. They spend a majority
Who are their friends, and how did they meet? Who are the of their money on food and entertainment, and what is left
people they can’t trust and why? How do they earn money? over, is usually given to other members of their gang (about
How much money do they earn? What strategies do they which more is below). Entertainment might include the
employ to survive on the streets? Do they have career purchase of alcoholic beverages, or watching movies at train-
aspirations? Do they understand what sorts of careers are station “movie theaters”: Some groups will set up TVs and
available? What are their plans for the future? Did they show movies to other youth for a price. The Malda train-
perceive any inhibitors to their plans? station youth frequented such locations approximately twice a
month. While the Malda train station youth had a more stable
Key Findings
sleeping situation through space provided by Praajak, most
Life Overall train station youth are nomadic, moving from one station to
another, spending a few weeks at any one location.
There are two categories of street children (ages 8 to 20) who
Gangs and Hierarchy
live in urban cities like Calcutta: (1) slum dwellers, and (2)
street dwellers [10]. The largest group of street children is Street children as a whole, self-organize into gangs, in order
migrant street children who are first or second generation to increase individual productivity and share livelihood goods.
migrants: either they themselves are migrants, or they have [13] Gangs are territorial, and control particular streets or
migrated with their families [11]. In Calcutta, some of these alleys, preventing those outside the gang from doing business
migrants come from Bangladesh, while a smaller percentage in their geographical area. Gangs also provide their members
comes from surrounding rural village areas. Most of those with lines of trust: i.e., the names and locations of street
who have migrated with their families live in one of 5,000 vendors who will sell and/or purchase items from the gang. It
slum communities in Calcutta (group 1), while those who also is a form of insurance and protection; not only does the
have migrated alone find shelter on the streets (group 2). gang protect its members in terms of helping in dire cases with
While one might assume that children migrating alone is life necessities such as food, but it will train its members to
uncommon, worldwide roughly 48 percent of refugees are avoid the local police, and if caught, how to react
children [1]. These two groups of street children have appropriately to avoid being thrown into jail. The gang
significantly different lifestyles, and sources of income [12]. therefore, is a major broker between any of its members and
Those in the first group typically share some form of family the outside world: providing information about markets and
income from their father and/or mother, while they themselves their vendors, information about police, and information about
439
other children and gangs [14]. While such street gangs are not travel to train stations in areas of India where Diwali
typically contentious, fights between children are celebrations are more concentrated. Train station youth also
commonplace. Initiation into a group requires into a group understand the concept of pricing structures and profit
requires finding a leader and proving your worth through maximization. When they realize that there is a demand for a
sexual favors and financial payments [13]. particular good, say bottles of water, rather than just source
one type of water bottle and sell only that, they understand
It should be noted that the nature of these gangs is different
that if they have different kinds of water bottles they might be
from the street gangs such as those in the United States or
able to build a price structure which increases overall profit.
Brazil, who manage their territory with heavy-duty weapons,
For instance, they might be able to buy larger water bottles for
and whose income sources are typically illegal [15]. In
only a 20% increase in cost, but be able to sell the water bottle
contrast, street children in India live a carefree lifestyle, and
for double the price of the smaller ones.
are not felt to be dangerous or intimidating by surrounding
populations. Street Smarts
The gangs at the train station show further structure and Train station youth also have developed acute survival skills
organization. There are three tiers within the gangs, classified that allow them to continue their lifestyle on the train stations
roughly by age. Leaders, generally between 16 and 21 years [15]. They know the rotations of the police officers, which
old, are responsible for the overall well-being, safety and police officers to avoid at which particular times, which police
material well-being of the group. Below leaders, there are the officers would ask for a bribe, and which ones might be a little
laborers, between 9 and 16 years old, who do much of the more forgiving. They know the times of arrivals and
day-to-day work of the gang and earn approximately $.50 to departures of trains within a six train station radius (stops
$1 per day. While there is no formal payment structure between train stations can be anywhere within 3 – 50 km), and
between the laborers and the leaders, there is a general know which trains will likely be carrying more passengers at
understanding that they need to pay the leaders a percentage which times and days. All of this information is pertinent to
of their extra income that is not used as food, typically less both maximizing profit and keeping as much of it as they can
than $1 per week. Below the laborers, there are peons who (without having to pay bribes to local police officers).
are the youngest children in the group, usually under 10 years Finally, train station youth understand that they are not “liked”
old. While peons will probably bring in an equivalent amount by all – including the management of train stations.
of money into the group as laborers, they are not as self- Whenever management comes to a train station to inspect its
reliant, and typically accompany laborers on cleaning/selling “cleanliness” all of the train station youth vacate its premises
trips when boarding trains. for the day. This partially is to avoid being forcefully thrown
Entrepreneurship out by the local police, but it also is to insure a healthy
relationship between themselves and local police officers.
Street children have an informal understanding of markets and
Technology Usage
pricing. They have a solid understanding of the different areas
of the city where lower, middle, and upper class city dwellers Street children at times use pay telephones to keep in touch
live, which areas of the city to buy the cheapest good and with their clients. Pay telephones in India cost $0.20 per
which areas of the city goods can be resold with the highest minute for a local call. There are two types of clients that are
profit margins [16]. Some items, such as food perishables, are typically contacted through with a telephone (1) helper-
bought from wholesalers in large markets, and resold to clients, and (2) street vendors. Helper-clients refer to those
individual customers on the streets, while other items are people in the city who have built some sort of relationship
bought from wholesalers and are resold to street vendors. with the street child and are involved in a rehabilitation
Such children serve as middlemen in the formal marketplace, program of some sort with them. The first author received
commonly supplying goods to local retailers on a per-need approximately 2 unsolicited calls from street children per
basis. At times, after mutual trust has been built between the week throughout his stay in Calcutta. Such calls either were
street child and the street vendor, the vendor might call upon to ask for money or for moral support. Street children might
him to source certain goods on his behalf, and be paid a small also occasionally call street vendors asking them if they
amount for such services. require any type of good that they could source for them.
Train station youth are entrepreneurs. While one might Rohit, for instance, on occasion would call his eye glasses
readily assume that given the no-skill, no management, no street vendor client to see if he would like him to go to the
savings lifestyle of these youth, that they are not “business market and buy glasses for him. This call was necessary
savvy,” the reality is the contrary. Train station youth are rather than an in person conversation, because the cost for
incredibly good at assessing consumer demand for different Rohit to make the call was less than if he would have taken a
types of products in different locations, and at different times local bus to meet the local street vendor retailer.
of the year. For instance, during Diwali (the festival of Train station children have no interaction whatsoever with
lights), which is celebrated in India in October, youth technology beyond the electronic billboards available at some
understand that the demand for candles and other light-giving train station in India.
instruments will go up, and therefore they go into the business
of buying and reselling candles. At times they might even Microcosmic Perspectives
440
Typical lives of older street youth might look like this: Bablu, history and is intended to help the poor start their own
19, is a migrant to Calcutta, whose family lives in Varanasi (a businesses, and so one question is what would happen if small
13-hour train ride from Calcutta). He passed 6th grade and loans were made available to street children.
then dropped out of school to earn income for the family. Has
In an attempt to understand how such efforts might end up, we
strong ties with his family and has family “emotional”
provided a loan to two youths that allowed them to start two
support, but decided to move to Calcutta in search of better
different businesses: Raja as a shoe shiner, and Stephen as a
job opportunities. Upon arriving in Calcutta, given his lack of
bicycle rickshaw puller (names have been altered). These were
education, he was unable to find formal salaried employment.
businesses which seemed easy enough for someone to pick up
Since Bablu comes from Varanasi which is a major tourist
quickly, and which could be undertaken with capital
destination, he has a smattering of English knowledge, and
investments of less than US$150. To avoid straight monetary
therefore, earns his main source of income as an informal tour
transactions, the youths were provided not with cash, but with
guide in Calcutta. He asks passerby clientele if they would
the tools of their trade as they felt necessary. Raja thus
like further information about a particular area of Calcutta or
received a shoe-shining box, creams for shining, shoe-
if they would like to be guided to a particular destination.
brushes, and Stephen received a cart for his bicycle. A vocal
After services rendered he asks for payment of approximately
agreement was made such that about 20% of revenue would
US$1. Bablu claims to make approximately $5 per week. His
be repaid to the lender at the end of every week. Expected
younger friend, Chotu is 16-years old and a runaway youth,
repayment from the bicycle rickshaw would be about $4 /
whose family occupies a nearby slum in Calcutta. His father is
week, and from the shoe-shining box, $2 / week.
a fruit re-seller; he buys fruit from a wholesaler and resells it
in nearby neighborhoods on a hand-pulled cart. His mother is Raja began his new shoe-shining business within one week, in
a house-worker. Chotu has never attended school, and is a high-traffic location with plenty of middle-class passersby.
addicted to cocaine. His main source of income is through After two weeks in the business, however, Raja reported that
buying and reselling packets of gum. He buys a 12-pack of shoe-shining was not very profitable. On average, he would
gum from nearby street vendors and hawks single packs earn approximately US$1.50 a day – which he acknowledged
around the city. He can make a profit of US$0.10 per packet was approximately double his previous earnings from various
of gum sold, selling an average of 3 packets of gum per day, illicit activities, but he was unsatisfied with the amount of
or $0.30. work needed to make that extra income. He was working
approximately 10 hours per day, every day of the week.
While train station youth might be incredibly perceptive of
Additionally, the local police, who were familiar with Raja as
their own local environment, beyond a mile of the train
a delinquent, pestered him about licenses and so forth (strictly
station, the youth seem to have almost no knowledge
speaking, one is legally required to have a business license,
whatsoever of the business opportunities that might be
but this is enforced only at the pleasure of local police
available to them.
officers, who often extort money from those without licenses),
If particular types of food, for instance, are only of demand and on one occasion, beat his hands to render his shoe-shining
inside the city next to the train station, but not inside of the work more difficult. Within weeks, Raja had sold his shoe-
train station, train station youth would likely have no idea that shining equipment (as the author discovered from the
such opportunities or demand existed. Additionally, their acquaintance to whom Raja had made the sale), and
vocations and acquired skills are strictly limited to what they disappeared.
can directly see and experience at various train stations within
Stephen’s story was similar. Within weeks of being provided
their immediate reach.
the cart, he, too vanished, presumably after liquidation.
There are two key learnings from this experiment: First,
III. ACTION RESEARCH
transitions to more legitimate activities for street youth can be
In addition to the ethnographic explorations, we also challenging. It is difficult for one to make a change in lifestyle
conducted small-scale action research to understand the from one based on gang life and at one’s pleasure, to one that
consequences of two kinds of interventions: micro-loans and requires daily routine and endurance, even if the result is
provision of video-based information. increased income. Second, from the point of view of running
Experiment 1: Loans successful interventions, it will probably help to carefully
screen potential candidates for entrepreneurship and to
Some street youth appear to be eager to start businesses on provide ongoing social support once they begin their
their own, and they actively seek out people who can help businesses.
them with start-up capital. Of course, many of these children
Experiment 2: Video-Based Information
are con artists. Street children, out of necessity, learn how to
turn all kinds of “aid” to their short-term advantage in ways Among information and communication technologies, video is
which may not necessarily be the intent of the benefactor. undoubtedly the richest medium for recording and
Some youth, however, persevere in their requests for help demonstrating activity, at least among those that are available
with legitimate jobs and seem genuinely interested in trying an to the mass market. Street children, who are semi-literate at
alternate lifestyle. Microcredit, of course, has an established best, find it easy to understand. Video captures the reality of
441
the situation; it is very easy for a viewer to grasp the full Information dissemination about local business
context of what is being recorded. In contrast, PCs are opportunities and local demand.
expensive and complex, and audio devices and mobile phones
Social networking tools to connect local youth with
are either feature-poor or portable enough to become easy
mentors and/or entrepreneurs
targets for theft.
Audio-based feedback systems for community
We thus also experimented with producing
building.
vocational training video content and tested approaches to
video screening and mediation. The first video experiment We felt that relatively quick impact could be had by providing
was intended to better understand their receptivity towards information about the kinds of professions that were open to
learning about vocational business skills through mediated- them outside of the small communities they were familiar
video. First, one local entrepreneur in Calcutta was identified, with. This reasoning, in combination, with our positive
an “iron wala,” whose job consisted of sourcing clothing from experiences with video suggested an intervention project we
surrounding neighborhoods, ironing the clothing, and call Social Enterprises.
returning the clothing to its owners. The first author, with
help from a local community member who spoke both English
IV. SOCIAL ENTERPRISES
and Bengali (the language of the iron “wala”) conducted an
interview with the ironing wala. The video focused the skills The long-term goal of Social Enterprises is to provide a
required to run the business (i.e. how to iron, how to source vocational awareness, training, mentoring and monitoring
clients, where to save daily earnings, etc.) The author filmed model to help street children gain self-employment and
his day-to-day operations of the business including: collecting financial independence. In the short term, we focused on
clothes, ironing clothes, collecting payment, and returning providing vocational awareness via mediated video instruction
clothes. In total the entire interview and demonstration lasted for relevant, generically available jobs.
13 minutes. The author then selected a local street child, By “relevant, generically available jobs”, we mean those that
Rahul, of age 18, who had not attended school and whose have the following criteria: a local business that could provide
father was a fruit re-seller in middle class neighborhoods. The at least $1.50 of net income per day, initial capital investment
first author brought Rahul over to his residence, and on a below $200, job-skill requirements that can be learned in less
laptop, mediated the entire video of the local ironing wala, than one week, equipment and material needs that can be met
pausing the video every couple of minutes to insure locally, and unsaturated market demand in surrounding
understanding and knowledge retention. Rahul was then localities and cities.
asked to evaluate the value of the video by answering the
following questions: 1) had you seen an irnoning wala before? In its current form, Social Enterprises includes (1) a video-
2) did this video help you learn the trade of running an ironing content production, (2) mediated instruction for dissemination,
business? 3) if not, would this video help you learn those and (3) a mentoring framework involving a local NGO,
skills? 4) what else would help you learn the skills to learning established entrepreneurs, and train-station youth.
ironing and run an ironing business?
Video-Content Production
There were two main observations based on the feedback
session with Rahul: First, while Rahul had seen ironing wala’s One of the major components of SE involves the production
before, he stated that this video had “good information,” that of video content. Videos produced are informational and
helped him better understand the business. Rahul stated that instructional in nature. All SE videos involve a formal
numeric information such as the number of clients served per interview with a local entrepreneur and an informal
day or the charges for different types of clothing were demonstration of the nature of the business. The first part of
particularly useful. Second, the video was motivating for the production of the video content involves choosing
Rahul. When asked if he would like to start an ironing appropriate entrepreneurs. Once a local entrepreneur is
business, he seemed skeptical, but thought that if he was going located, an NGO field staff, who is trained in the basics of
to, that he would have wanted to see this video first for video recording, sets up a time and day to conduct the
additional information. Key learnings from the feedback interview with the entrepreneur. During the interview the
session include (1) the importance of video in displaying content producer asks the local entrepreneur a series of short-
information about local businesses and (2) information answer questions aimed at giving the viewer a general feel for
contained in the video can be motivating for street children. the tasks required to perform the particular business venture,
and the average revenue and profit of the business. Questions
The severe constraints of street children make it difficult to include: What is your occupation? How long have you had
imagine technology interventions that would be meaningful this occupation? How did you start in this business? How did
for them that don’t devolve into empty or temporary aid. We you learn the skills for this business? How did you find the
believe, however, that several classes of projects are possible: location for your business? How many hours / week do you
Monitoring or check-in systems for lifestyle changes work? What is your average daily income? Etc. During the
(such as a new job). demonstration portion of the video, the local entrepreneur will
demonstrate the activities involved in the business (i.e., if it is
Educational tools to support vocational training. a food business, they will show food preparation). The videos
442
are captured using inexpensive, MiniDV camcorders, and where they legally, and socially would not be disturbed. Once
tripods. Inexpensive wireless microphones are used to the business has been identified, the location for the new
increase audio quality – especially given the loud background venture found, and the materials sourced, the final step before
noise on major streets in Kolkata. Videos range from 2 min. starting the venture requires street youth to learn, first-hand,
to 4 min. in length. through day-to-day observation of a local entrepreneur in the
same business. This process will likely continue for a day to
Video is then edited and stored on a PC. SE has one
three days depending on the complexity of the business. After
technician who has basic computer literacy, and is trained on
these couple of days, the youth is ready to start his venture,
the basics of video post-production. The technician will
and in collaboration with the local NGO, begins their
minimally edit the video, insuring accuracy and clarity of the
operations. During the next couple of months, the NGO will
content before it is store locally. If the content is deemed
monitor the progress of the venture, its revenue and
unusable by the technician, the content is re-recorded. After
profitability, while ensuring that the youth is actively pursuing
the post-production process, the video is stored in a local
the venture on a day-to-day basis, at least 10 hours per day,
database along with title information and additional metadata.
except on Sundays. Monitoring and evaluation by NGO staff
Videos are then given by hand to local NGO partners who
continued at a pace of once a week.
work with SE.
Mediated Instruction
V. PRELIMINARY TRIAL
The principal means of distributing videos from the SE Social Enterprises was implemented by the first author in
database to the train station youth is by physically carrying the conjunction with Praajak over a two-month period. During
DVDs to the local NGO who then hands the DVDs to local this time, our goal was to identify at least three candidates for
field staff. A member of the NGO field staff who is trusted by self-employed entrepreneurship, and to observe (1) how they
the train-station youth is then assigned as a mediator. handled the transition, (2) what benefits and challenges
The mediation sessions are organized by Praajak. Praajak will emerged, and (3) whether the overall system was worthwhile
inform train station youth from surrounding train stations to to continue.
come and attend an event at their site location. Between 5 – Methodology
100 youth have been present at a given mediation session.
Mediation sessions have a dual purpose: (1) to inform train Over a two month time period, two mediation sessions were
station youth of various types of vocations they could get organized by Praajak. They informed train-station youth from
involved in, and (2) to filter likely candidates for the SE Malda and surrounding train stations to come and attend an
system. During a single mediation session *all* of the videos event at their site location. A small projector projected the
are shown (at the moment of the last mediation session, only video on a wall in Praajak’s office where the train station
27 videos had been generated), which is a total of youth were asked to gather. The projector was connected to
approximately 1.5 hours of video. The mediator’s role is to speakers to amplify the sound. Between 5 and 100 youth
help the client audience believe that this in fact is a viable attended each of the sessions. These sessions serve a dual
business opportunity, and that if they desire, one that can be purpose: (1) to inform train station youth of the various
replicated in a nearby locality. Given the short length of the vocations, and (2) to filter likely candidates for further
videos, the mediator generally waits until the end of the video investment and mentorship. During a single mediation session
to have discussion about the video and resolve remaining all of the 27 videos we had recorded were shown, which
questions, evaluate knowledge gained, etc. totaled approximately one and a half hours of video. The
mediators were instructed to provoke discussion, answer
Mentoring Framework questions, and otherwise convince the audience that the
The mentoring framework includes all aspects necessary to vocations displayed were in fact viable business opportunities
assist an SE candidate proceed through the necessary stages of in their own locality. Given the short length of the videos,
exploring, identifying, learning, understanding, implementing, pauses generally occurred between videos. After the video
and continuing a new business venture. was shown the mediator probed the audience for interest in the
business. Some questions asked included “Do you like this
The first step to the mentoring framework involves identifying business?” “Would you be interested in starting this
likely candidates for the Social Enterprises system. These business?” “What types of information were useful in this
decisions, we leave to the NGO, who has the background and video?”
experience to identify those candidates who are sincerely
looking for employment. Such a filter is necessary to avoid Given the large audience, candidates had to be screened
working with and financing clients who do not have the before they would be able to proceed to the next phase of the
emotional or psychological ability to successfully start and SE system. In rough order of importance, candidates were
manage a small start-up enterprise. chosen based on the following criteria: age, trustworthiness,
work ethic, business experience, and vision. Candidates had to
The local NGO would then work with the street youth to be at least fifteen years old; this is due to the fact that younger
acquire the materials to set up the start-up venture, and find a children might still have the possibility to re-enter the formal
specific location within that locality (i.e. a street corner), school environment, as encouraged by Praajak.
443
Ultimately, three youth who indicated interest in the income. While previously train station youth earned between
businesses shown in the video were chosen for the next $.50 and $1 per day, after starting and running their own
phases, which include an introduction to an established business ventures, average daily income increased to
entrepreneur in the chosen lines of business, capital approximately $1.50 per day, an increase of 50% to 100% per
investment by the NGO to start the business, and once-a-week day. However, their overall revenue increased by
monitoring by NGO staff during the two-month course of our approximately 1500%, which meant youth were managing a
trial. All capital costs were recorded. much larger amount of money. At the train stations, a large
percentage (approximately 90%) of daily revenue was profit,
During the monitoring sessions, revenue information from the
because the goods they were procuring were in many cases
new businesses was requested (averages over the week), along
free (such as water bottles left behind by passengers).
with information related to the overall operations of the
However, in their new ventures, average daily revenue was
business. Gathered data included average number of hours
approximately $12. Most of this money however, was used to
worked per day, number of days worked per week, average
recuperate costs, hence only approximately a $1.50 per day of
number of customers who visit and/or purchase from the
revenue was actually profit. It also should be noted that while
business each day. Additionally all information related to
the average daily income for the businesses is greater, the
procuring products for the business was recorded including
income is much more variable. On certain days, for instance,
how much money was spent on which different items for the
a new business might not have any revenue whatsoever (due
business to startup, how often did which different items need
for instance to monsoon weather conditions). Such days can
to be repurchased and how much did they cost.
be extremely frustrating for the youth. Given that their
Results previous earnings were much more predictable (when they
wanted to make money, although little, they likely could find
Overall, things went well for the three youths. There were opportunities at the train station), when low-revenue days
fluctuations in their personal commitment to the business, but occur in their businesses, their first reaction is to liquidate
the regular check-ins by the NGO made a substantial assets and return to their previous income earning profession.
difference. The youths found the new lifestyle a significant
adjustment, and they needed continued encouragement by the The average cost of setting up a business and purchasing its
NGO. They also reported that the videos were a primary first inventory order was $180 dollars. This included various
source of motivation for them to start their businesses. For the necessary one-time purchase items, such as a stand where the
youth, the videos not only demonstrated additional business items would stationed and sold on a street corner that
opportunities, but showcased the lifestyle of an entrepreneur accounted for approximately 30% of the overall cost of the
as one who performs the types of tasks they were already used business.
to conducting in their day-to-day lives. Finally, youth Youth Reactions
reported that the financial assistance from Praajak was another
primary reason why they felt they were willing to take the risk Train station youth remained positive throughout the entire
of abandoning their previous train station lifestyle, including process of learning new business opportunities, setting up
the guaranteed income and their social circle of friends, their own businesses and running their own business.
moving to a new area and running a business. However, youth also expressed equal discomfort with various
aspects of their new lifestyle. Two particular areas of
Value of Video
discomfort for the youth included change in working and time
The advantages of video as a medium were further management lifestyles and change in social scenery. Before
highlighted. Entrepreneurs explained the nature of their beginning their own businesses, train station youth enjoyed
business, many of which are not that difficult to start or the freedom of an “eat what you kill” working methodology:
manage, their income (and the fact that they were able to they worked when they wanted, how they wanted, and with
provide for their families), that it is local (the video visibly whom they wanted. The longer hours they worked, the more
shows the streets of Kolkata and speech in the local language), money they earned. After the transition, Praajak required the
and so forth. In this way, the train-station youth were able to youth to run their businesses 6 days per week and attend to
relate to the local entrepreneur, and gain intuition for whether their businesses at least 10 hours per day. This standardized
or not the vocation was appropriate for them. the lifestyle of the youth, with regular sleeping hours, regular
attendance to the location of their business and attention to
Further indications of the value of video were shown in the customer needs and demands. Additionally, they no longer
overall enthusiasm of the youth to start such businesses after had as much time to enjoy the flippant nature of their social
only having seen a couple-minute video. Over half of the lives on the train station. The work was also more detailed
train station youth expressed their willingness to further oriented: youth had to keep track of their current inventory,
explore the possibility of learning more about and/or perhaps and when necessary, work with Praajak to restock supply.
starting one such business. These two changes were the primary discomforts for the train
Increased Income station youth who participated in the SE system.
On the other hand, youth who were involved felt increased
The greatest incentive for the train station youth to attempt
motivation and began to grow a vision for their future. One
their new businesses was the prospect of increased average
444
youth stated “This business opportunity has changed the way screening, business bootstrapping and ongoing support and
that I am planning for my future.” This youth was referring to evaluation. NGO staff also played a large role in accounting
the increased understanding that he had gained of financial procedures and budging for the business opportunities. Such
planning as a result of running his business for only two interventions need to be replaced with some other form of
weeks. Overall, youth seemed happy to have become selected monitoring and evaluation. Furthermore, train station youth
investees of Praajak, as they knew that they had been filtered, find the social adjustments difficult; moving to day-to-day
and chosen to participate in this pilot project. Furthermore, hourly operations, spending a majority of their time away
the families of the train station youth expressed gratefulness to from their other train station friends, and at times living and
Praajak for having chosen their son for this program, and saw sleeping at new locations. This requires further intervention
the business as a means to long-term sustainable livelihood. on behalf of the NGO. In the future, parents of the children in
Parents, therefore, played a large role in encouraging their collaboration with local village-level mentors could replace
children to push themselves and continuing their business the NGO’s role.
ventures.
One issue is that the profit variability of the businesses can be
System Adjustments discouraging. While on average, income is greater, on bad
days, it is less. For youth who are used to living hand-to-
Through the two-month preliminary trial, various areas of mouth, these instances can be frightening.
potential improvement have been identified for the SE system.
One possibility is with the variety in the video content. Overall the SE system seems to work well but remains at a
Currently all of the videos showcase urban entrepreneurs. small scale, and still requires continuous monitoring and
However, given the heavy support from parents for the encouragement of the newly self-employed youth. It is
program, many of whom live in rural areas, there could be important to note that the train station youth and Praajak have
value in producing video content on village-level a long history of trust. At scale, developing this level of trust
entrepreneurship opportunities. Such content would be used or identifying other organizations with such trust will be
to encouraging youth to return home and pursue village difficult.
businesses in their own home communities.
Through our trial, it became clear that there is little need for VI. RELATED WORK
heavy day-to-day involvement from local entrepreneurs: local Vocational Education for Development
entrepreneurs do not need to take on a mentorship role with
the train station youth in order for them to begin and operate In recent years there has been increased demand in developing
their businesses. Rather, the street youth are savvy enough countries, such as India, for public and private involvement in
that with just basic observation they can pick up the providing skill-based educational programs as a means to
vocational and business skills by watching entrepreneurs in provide the poor with skills to lift themselves out of poverty.
action. The government of India launched the urban Self Employment
Program (USEP) and The Urban Wage Employment Program
However, mentorship, monitoring and evaluation remain (UWEP) in 1997 to assist individual urban poor beneficiaries
necessary to motivate the train station youth through typical for setting up self-employment ventures. The program
start-up pains (variable revenue, change in lifestyle, etc). involves participants in the acquisition of vocational and
Increased parental involvement and encouragement might be a entrepreneurial skills. Additionally, because it is a
viable way to sustain these tasks. government subsidized program, participants only need to pay
Overall Conclusions $50 for six months of training. Approximately 150,000
people in India have been trained through this program [17].
Overall the results of the SE preliminary trial are positive: The Department of Rural Development in India set up the
three youth with no prior experience running or operating a Rural Development and Self-Employment Training
business successfully started, managed, and led their own (RUDSETI) that provides skill training to the rural poor [17].
business venture. Video played a crucial role in informing This program has benefited over 150,000 trainees, of which
youth of the various opportunities available and encouraging approximately 66% have successfully established
them of the likely possibility of higher profitability. Support entrepreneurial ventures, such as repairmen of two wheeler
from Praajak staff in video mediation further enhanced video vehicles, watch and mobile repair, etc [17]. Corporate Social
credibility. Furthermore, ongoing mentoring and monitoring Responsibility initiatives such as that of the Byrraju
from Praajak staff encouraged youth to push through tough Foundation provides training services in areas such as
business times. Motivation levels of the train station youth healthcare, sanitation, and literacy. The foundation currently
were variable throughout the process. While initially there works in 156 villages in Andra Pradesh [17].
was extreme enthusiasm to start and run a business of their
choice, over time, this enthusiasm varied depending largely on The Ambuja Cement Foundation (ACF), which works in over
the profitability of the venture. 700 villages in eight different states in India, trains youth
groups in skills such as bamboo crafts, furniture making, or
Additionally, while the system continues to work well, it is driving. The non-profit sector is also becoming heavily
heavily reliant on intervention from local NGO field staff in involved in vocational training for the urban poor in India in
all stages, including video production, mediation, candidate
445
two keys ways 1) training basic skill and leaving the trainees using local audiences, and spread such knowledge to a larger
to seek wage employment or start their own enterprises or 2) client audience.
“holistic” training which would include other services such as
Other projects involved in ICT for education extension that
assistance in availing credit facilities [17]. The Goodwill
also inspired aspects of SE include the Deccan Development
International Association, for instance, based in Bangalore,
Society’s project based in Hyderabad, India which trains the
trained unemployed youth and school dropouts in various
illiterate to film their own videos, and Video Volunteers
disciplines such as fitting, welding, electric and plumbing.
which partners with Indian non-profits to “empower
Likewise, Gram Vikas provides training to landless unskilled
communities” to address development issues using video
laborers in masonry or stone welding. Most Gram Vikas
[22,23].
clients have managed to double their wages within two years
of returning to the workforce. The Self Employed Women’s ICT for Street Children
Association (SEWA) is another prime example that organizes
women workers for full self-employment by imparting a In countries like India, it is becoming increasingly more
variety of technical livelihood, management, and behavioral popular for young and old, rich and poor alike, to attain IT
skills. 960,000 poor women from five states have benefited related job skills. However many of these opportunities are
from SEWA’s services [18]. left out of the hands of street children. They typically would
not have the financial ability, nor the stability to enter a
While such initiatives are noteworthy, it should also be noted traditionally rigorous IT related training course. There is
that the demand for such services is much larger than such however, one example of a project that does IT training
organizations are able to service, and few, if any, street specifically for street children – in Ecuador, called iConnect
children are easily able to enter into these programs, many of Online. iConnect provides job-related computer and internet
which require formal interactions with government offices. skills to street children who otherwise would not have access
Mediated Video and Education Extension to such educational opportunities [24].
Other initiatives, such as Digital StudyHall and Digital Green

VII. CONCLUSION
have experience extending educational initiatives to larger
audiences using tutored video instruction. In particular, such Based on initial ethnographic studies and experimentation, we
initiatives recognize that while video alone might add little to have presented the Social Enterprises system, which combines
no value in information dissemination, mediated or tutored- video-content production, mediated instruction, and a
video instruction, has shown to add incredible value to the mentoring framework to assist poor street youth move towards
value of information consumption by the poor in India. long-term financial sustainability by founding and running
Tutored video instruction was pioneered by Jim Gibbons in their own local enterprises. In a two month preliminary study,
the 1970’s [19]. Gibbons showed that students that came to involving three local train station youth, we found that the SE
classes where videos were mediated by a teaching assistant system, with an initial capital investment of less than $200
did better than those who just watched the videos passively, dollars can provide youth with the tools necessary to find and
and better, even, than those who attended the live lecture. begin start-up ventures with average profitability up to 100%
Two initiatives in India, Digital StudyHall and Digital Green, higher than previous forms of employment. Local NGO field
have shown promise in the use of mediated video to extend staff play a significant role in the overall implementation,
pedagogy and agricultural instruction, respectively, to the monitoring and evaluation of the progress of such businesses,
rural poor in India. Digital StudyHall digitally records live and are a key ingredient to the success of the system.
classes taught by good teachers in local languages and
according to local curricula, collects them in a distributed However, these results are preliminary, and further study is
database, and distributes them on DVDs to poor rural and necessary to better understand long-term emotional and
slum schools which use TVs and DVD players to playback financial sustainability of street youth in the SE system. To
and mediate the content [20]. Digital Green (DG) disseminates investigate further, additional content demonstrating village-
targeted agricultural information using video and mediated level business opportunities are needed, along with a more
instruction to buttress a system called agriculture extension, localized and sustainable monitoring and evaluation system.
which seeks to spread expert agricultural advice to small and
We look forward to continuing collaboration with Praajak and
marginal farmers [21]. DG records live training sessions
finding increasingly cheap and scalable means of providing
between agriculture extension workers and farmers, collects
vocational tools and skills for the livelihood of millions of
them in a database, and distributes them on DVDs to farming
communities. In villages that receive these DVDs, farmers street children throughout the world.
attend informal training groups where a village facilitator
mediates DG content on a regular basis. ACKNOWLEDGMENT
The authors would like to thank Loreto Day School Sealdah
Digital StudyHall and Digital Green both inspired the initial
and Groupe Development South Asia for initial introductions
steps of the Social Enterprises project. SE was in a large part
and feedback. Special thanks to Praajak and its executive
inspired by efforts made in these two initiatives to capture
local training and educational programs in local languages and
446
team for inviting us to work with them and for ongoing [19] J. F. Gibbons, W. R. Kincheloe, and K. S. Down,
support and collaboration. “Tutored videotape instruction: A new use of electronics
media in education,” Science, 195(3):1139-1146, 1977.
[20] Urvashi Sahni, Rahul Gupta, Glynda Hull, Paul Javid,
REFERENCES
Tanuja Setia,Kentaro Toyama, and Randy Wang. Using
[1] United Nations Childrens Fund (UNICEF) State of the Digital Video in Rural Indian Schools: A Study of Teacher
Worlds Children Report, 2006 (New York: Oxford University Development and Student Achievement, AERA 2008, New
Press, 2006) York, March 25-27, 2008
[2] United Nations Development Programme (UNDP), Human [21] Rikin Gandhi, Rajesh Veeraraghavan, Kentaro Toyama,
Development Report, 1993 (New York: Oxford University Vanaja Ramprasad. Digital Green: Participatory Video for
Press, 1993) Agricultural Extension, ICTD 2007, Bangalore, India.
[3] United Nations Development Programme(UNDP) , Human December 15-16, 2007
Development Report, 1996 (New York: Oxford University [22] Deccan development society website, September 2008
Press, 1996). http://www.ddsindia.com/www/default.asp
[4] Directorate of Census Operations, West Bengal (2003). [23] Video Volunteers website, September 2008
"Table-4: Population, Decadal Growth Rate, Density and http://www.videovolunteers.org/
General Sex Ratio by Residence and Sex, West Bengal/ [24] iConnect website, September 2008.
District/ Sub District, 2001, Census of India http://www.iconnect-online.org/Stories/Story.import4315
[5] United Nations Millennium Project, A Home in the City:
Task force report on improving the lives of slum dwellers,
Executive Summary, Earthscan, London/Sterling, VA, 2005,
p. 10.
[6] World Bank, India: Achievements and Challenges in
Reducing Poverty, 1997
[7] Human Rights Watch, Police Abuse and Killings of Street
Children in India (New York: Human Rights Watch,
1996)
[8] "State of the World's Street Children-Violence"
www.streetchildren.org.uk. Retrieved February 16, 2009.
[9] WHO, World Report on Violence and Health. Geneva:
World Health Organization, 2002.
[10] UNICEF, (2006) State of the World’s Children 2007:
Women and Children. New York: UNICEF.
[11] Natai, Kundu. Kolkata, India: Understanding Slums,
Case Studies for the Global Report on Human Settlements,
2003.
[12] Kidd, S., (2007) “Youth Homelessness and Social
Stigma.” Journal of Youth and Adolescence, 36: 291–299.
[13] Jones, G., Herrera, E. and Thomas de Benítez, S., (2007)
“Tears, Trauma and Suicide: Everyday Violence Among
Street Youth in Puebla, Mexico.” Bulletin of Latin
American Research, 26(4): 462–479.
[14] Droz, Y., (2006) “Street Children And The Work Ethic:
New Policy for an Old Moral, Nairobi (Kenya).” Childhood:
A Global Journal of Child Research, 13(3): 349-363.
[15] Rede Rio Criança, (2007) Criança, Rua e ONG’s: Quem
Faz e o que faz? Mapeamento de Ações das ONG’s Junto às
Crianças e Adolescentes em Situacāo de Rua no Município
do RJ. Rio de Janeiro: Criacāo Gráfica.
[16] Evans, R., (2006) “Negotiating Social Identities: The
Influence of Gender, Age and Ethnicity on Young People’s
‘Street Careers’ in Tanzania.” Children’s Geographies,
4(1): 109.
[17] National Commission for Enterprises in the Unorganized
Sector, India, Skill Formation and Employment Assurance in
the Unorganized Sector, 2008.
[18] Self Employed Women’s Association website, September
2008.
http://www.sewa.org/
447
Speech vs. Touch-tone: Telephony Interfaces

for Information Access by Low Literate Users
Jahanzeb Sherwani1, Sooraj Palijo2, Sarwat Mirza2, Tanveer Ahmed2, Nosheen Ali3, & Roni Rosenfeld1
1
2
Health and Nutrition Development Society
3
Cornell University
surprising that the need for better information access by

Abstract—Information access by low literate users is a difficult CHWs is widely agreed upon: “Providing access to reliable
task. Critical information, such as in the field of healthcare, can health information for health workers in developing countries
often mean the difference between life and death. We have
is potentially the single most cost effective and achievable
developed spoken language interface prototypes aimed at low
literate users, and tested them with community health workers in strategy for sustainable improvement in health care” [3].
Pakistan. We present results showing that 1) in contrast to Over the past three years, we have been researching spoken
previous reports in the literature, well-designed speech interfaces language interfaces for information access by low literate
significantly outperform touch-tone equivalents for both low- community health workers in Pakistan. In the first phase of
literate and literate users, and that 2) literacy significantly our work we have established the importance of telephony-
impacts task success for both modalities.
based information access systems, and pilot-tested one
interface with one group of community health workers [4].
Index Terms—Speech technology, spoken language interfaces,
low literate, illiteracy, information access, community health Since then, we have conducted a number of user studies
workers. testing speech interface prototypes in Urdu and Sindhi, in
different urban and rural sites, with community health workers
of varying literacy.
I. INTRODUCTION In this paper, we present multiple lessons and findings from
L OW literate users face great difficulty in accessing these studies:

information that is often easily available to literate users. • While reinforcement of existing CHW skills is often
This is especially problematic in the developing world, where cited by public health officials as their most
there are many more non literate users, and where the important educational goal, we show that usability
importance of information is often greater. Telephony-based testing should not focus on skills reinforcement
spoken language interfaces offer a unique solution to (Section III).
information access problems for such users, where cell phones • We describe novel improvements to the Poor Man’s
are quickly become ubiquitous, and low literacy renders text- Speech Recognizer, for rapidly achieving accurate
based modalities nonviable. In our research on spoken speech recognition in a new language using a
language interfaces for low-literate users in the developing standard US English speech recognizer (Section
world, we have chosen to focus on a specific domain where V).
information access is especially important: healthcare. • We describe a novel mobile setup (hardware and
Healthcare is a fundamental, yet often under-serviced need software) for carrying out field research in spoken
of citizens in developing countries. These regions have the interfaces, and argue that having such a modifiable
highest maternal mortality and neonatal mortality rates in the system available at the field site is essential for
world, and, not surprisingly, also have the world’s largest rapid iterative development with participatory
unmet need for health service providers. Given the high cost design (Section VI).
of training doctors and nurses, and the low number of medical • We present lessons learnt from three pilot
schools in these parts of the world, many governments have experiments in various urban and rural field sites
begun community health worker (CHW) programs, where (Section VII).
people (usually women) are chosen from their own • We describe a novel method for quickly and
communities, trained in basic health service provision for a effectively teaching novice, technology-shy
few months, and sent back to provide health services in their participants how to use spoken interfaces, and
communities. In some countries, especially in Latin America, show that an effective tutorial is crucial when
their effectiveness is quite high, reducing infant mortality to conducting user studies on such interfaces.
below that of the US [1]. These CHWs vary greatly in literacy (Section VIII).
levels and receive little refresher training [2]. It is not • We present both qualitative and quantitative results
from a comparative user study showing that a well
designed speech interface significantly
Manuscript received August 22, 2008.
448
outperforms a touch-tone equivalent for both low- systems tend to be more constrained, since these are easier to
literate and literate users; and that literacy build, although exceptions do exist, such as Amtrak’s “Julie”
significantly impacts task success for both system which is unusually flexible. Contrasted to the above
interfaces (Section IX). are call routing applications, which are used to direct a caller
• Finally, we discuss the implications of these results to a specific operator, given a few utterances [13]. The major
and situate them in the larger context of research in push for speech interfaces in the developed world has come
ICT4D and SLT4D (Spoken Language from the call center market, and that is what most research has
Technologies for Development) (Section X). focused on. However, since the needs of the populations that
such systems serve are very different, there are entire domains
that are still unexplored (e.g., access to books through speech).
II. RELATED WORK
Thus, there is a need for research in domains relevant to
There have been a number of approaches to GUI design for emerging regions, targeted towards the specific needs and
low-literate users. [5] presents design recommendations for abilities of users in these regions [14, 15, 16].
non-literate users of a proposed PDA-like device, with many The Tamil Market project undertaken by Berkeley’s TIER
recommendations involving speech. However, these group was the first to design, develop and test a spoken
recommendations are not based on empirical evidence from language system with low-literate users in a domain (crop
evaluations with actual semi- or non-literate users – they are information access) relevant to them [17]. Results from a
derived from a literature review of research on Western users. usability study of their speech interface suggest a difference in
[6] focuses on extending access to digital libraries by non- task success rates as well as in task completion times between
literate users, and also gives a short list of recommendations groups of literate and non-literate users, though differences
for such interfaces. However, usability tests reveal that users were not statistically significant. Further, [18] gives a strong
were not able to navigate information effectively, and result in indication that there are differences in skills and abilities
recommendations for keyword search, audio-based help, and between these two user groups, describes the linguistic
limiting the information set to lessen the cognitive load on differences in some detail, and suggests that further research is
users during navigation. [7] describes interface design required to understand the nature of this difference and to
guidelines, and a text-free interface that performed well in a derive principles of dialog design targeted towards such users.
usability test. [8] describes a PDA-based GUI designed for More recently, researchers at the Meraka Institute [19] have
rural community health workers in India. While this may been working on speech and touch-tone interfaces for health
appear to have similarities to our work, their focus is on information services in South Africa. Preliminary results
information entry, while ours is on information access. [9] suggest that touch-tone interfaces may be preferable to speech
describes a system for data entry as well as access to decision interfaces. A study by IBM Research India comparing speech
support by community health workers in India. This is in the and touch-tone interfaces reached a similar conclusion [20].
same domain as our project, and has many similarities to our Taken together, these studies appear to suggest that speech
work. However, our focus is on speech interfaces in this interfaces may not be very useful for low-literate users in the
domain, while their approach is GUI-based. [10] describes the developing world. Based on the results reported in this paper,
iterative & collaborative design process for and evaluation of a we strongly disagree.
GUI targeted to low-literate users for managing community- [21] describes VoicePedia, a purely telephone-based speech
based financial institutions in rural India. While the principles interface for searching, navigating and accessing the entire
of GUI design do not carry across well to speech interface Wikipedia web-site. An evaluation comparing VoicePedia
design, the collaborative design process described has lessons with a GUI-based smartphone equivalent shows comparable
highly relevant to all interface design in such contexts. task success across interface conditions, although the (highly
Speech interface research has resulted in a number of literate) users in the evaluation invariably preferred the GUI
systems in various domains. While the most well known alternative.
speech application is probably desktop dictation, this is just [22] gives an excellent review of the potential contributions
one point on a large multi-dimensional space of potential of CHWs in the developing world.
applications that can be made using speech. These dimensions Finally, [23] describes the difficulties low literate
include: choice of device (e.g., desktop, telephony, respondents face when asked questions that require abstract
smartphone), task (e.g., information access, information
thought.
entry), length of user training (often zero for commercial
applications), vertical domain (e.g., stock prices, news,
III. HEALTH INFORMATION CONTENT
weather), acceptable user input (constrained, open-ended),
interaction style (system initiative, user initiative, mixed Based on our prior ethnographic research, we had initially
initiative) and many others. For instance, Carnegie Mellon identified specific health topics on which to provide
University’s Communicator travel information system [11] information through any automated interface [4]. However,
and MIT’s Jupiter weather information system [12] are two our prior work was focused on urban community health
often-cited examples of speech-based information access workers with a minimum of 8 years of education. Since that
systems usable over the telephone – these are mixed initiative time, we have shifted our focus to low literate, rural
systems that require zero user training, and accept a large community health workers. In collaboration with our partner
range of user inputs, although as in all speech interfaces, NGO in Pakistan, the Health and Nutrition Development
acceptable user input is limited at each step. Most commercial Organization (HANDS), we initially opted to work with
449
reinforcing the material that the health workers were trained Speech Touch-tone
on (maternal and reproductive health), which is what a Hello, I’m Dr Marvi, and I’m here to give you health
deployed system would eventually need to provide. information.
Additionally, this seemed the prudent choice, as it is What would you like to hear For information on Malaria,
preferable to reinforce existing systems and practices than to about? Malaria, Diarrhea, or press 2, for information on
create new ones. The following issues forced us to rethink Hepatitis? Diarrhea, press 3, and for
this approach: information on Hepatitis,
1. For the participant: In a user study, even though press 4.
we clearly stated that “this is not a test of your User says Diarrhea User presses 3
knowledge”, especially when participants are Diarrhea. If this isn’t the Diarrhea. If this isn’t the
tested on information they are supposed to already topic you want, say ‘other topic you want, press 0.
know, they believe that it is a test of their topic’. [Pause] [Pause]
knowledge. In our experience, when participants Let me tell you about Diarrhea. As a Marvi worker, you need
were unable to give answers that they felt they to know that Diarrhea is a dangerous disease that can
should have known from before, they felt potentially be life threatening. You should know about its
embarrassed and uncomfortable. causes, its signs, its treatment, and how to prevent it.
2. For the researcher: It is impossible to tell whether What would you like to learn To learn about the causes of
a response to a question-answer task is being given about: causes, signs, diarrhea, press 2. To learn
based on what the participant found through the treatment, or prevention? about the signs of diarrhea,
system, or from prior knowledge. One way to [Pause] To learn about a press 3. To learn how to treat
cope with this issue is to conduct a pre-test of their different topic, say ‘other diarrhea, press 4. And to
knowledge, but this would further conflict with the topic’. learn how to prevent diarrhea,
previous issue. press 5. [Pause] To learn
3. For both: Reproductive health issues are extremely about a different topic, press
taboo in Pakistani society, and are rarely discussed 0.
in the presence of males. As the primary author (a User says ‘causes’ User presses 2
male) needed to be present during the user studies, The causes of Diarrhea. If The causes of Diarrhea. If
this presented a source of discomfort for user study this is not the topic you want, this is not the topic you want,
participants (e.g., they sometimes leaned in to give say ‘other topic’. [Pause] press 0. [Pause]
a response privately to the female facilitator).
Let me tell you about the causes of Diarrhea… [gives 3 bullet
Based on the above considerations in our pilot tests, we
points on the topic].
have now shifted to working with content that the community
To hear this again, say To hear this again, press 1.
health workers have not been trained on before, without any
‘repeat’. To hear more, say To hear more, press 2.
taboo elements in it.
‘more information’.
User says ‘more information’ User presses 2
IV. TELEPHONY INTERFACES FOR INFORMATION ACCESS
[The system gives 3 more bullets on the topic, and this cycle
continues until there are no more bullets, at which point the
To provide the information identified above, we have built following instructions are given.]
two primary telephony interfaces that we have tested To hear this again, say To hear this again, press 1.
extensively. The first is a purely non-interactive system, ‘repeat’. For a different topic, For a different topic, press 0.
which plays back a specific audio clip from beginning to end. say ‘other topic’.
This was primarily created as a baseline, to assess the
cognitive load on the participants created by the length of the
speech segment. V. IMPROVED “POOR MAN’S SPEECH RECOGNIZER”
The second interface is menu-based. It asks the user to
select a given topic (e.g., malaria, diarrhea, or hepatitis), after For speech recognition, we previously described a “poor
which they are asked to choose from a specific sub-topic (e.g. man’s speech recognizer” [4], using a robust speech
general information, signs, preventative measures, treatment), recognizer trained on US English speech. The basic principle
after which they are given detailed content broken down into of the approach is to map between phonemes in the desired
chunks of three bullet points at a time. The interface was language (Sindhi in our case) and the trained language (US
created in two ‘flavors’: one using touch tone input for English in our case). Thus a word such as ‘wadheek
choosing between the options, and the other using speech maaloomaat’ (transliterated Sindhi for “more information”)
input. Here is a sample call for both flavors, translated from would be given the following US English phonetic
Sindhi: pronunciation: W AH D I K M AA L U M AA DH. In our
initially described approach, the choice of phonemes was left
solely to the discretion of a language expert. We tested this
approach with Microsoft Speech Server (MSS), although the
450
principle would work with any modern speech recognition server physically located in Karachi, accessible over the
system. This approach led to reasonable recognition rates, telephone line connected to a separate telephony server.
although it was not very robust, and prone to error when tested Physically, this consisted of:
in the field. • Windows server running Microsoft Speech Server,
We have improved upon our approach significantly by containing all the logic for the information access
incorporating a novel data-driven method, which we call the interfaces, also running a Voice-over-IP gateway
“Poor Man’s Speech Recognizer++”. The basic idea is to • Linux server running Asterisk/Trixbox for Voice-
enable the developer and/or language expert to quickly over-IP support
generate new pronunciation definitions for words by varying • Uninterrupted Power Supply (UPS) unit as backup
any subset of phonemes in a given word’s pronunciation in case of power failure
definition, and then testing these variants with limited amounts • Monitors, keyboards, mice, routers, and
of data to empirically find the pronunciation definitions that network/power cables
would lead to the highest recognition accuracy. For instance, if While this worked to some extent, it had the following
the developer is unsure of the optimal choice for the last problems:
consonant in the word “maaloomaat”, she could specify a • Any power outage lasting longer than the maximum
wildcard definition of “M AA L U M AA C?”, where the “C?” UPS backup time could potentially bring the
denotes an “any consonant” wildcard. Similarly, if the system down. Running a Windows server for the
developer wants to test the optimal phoneme choice for the speech components, and a Linux server for the
final consonant-and-vowel combination in the word telephony interface meant a high electrical load.
“bachaao”, she may specify “B AX C? V?”, where “V?” • Any modifications to the system could not be made
denotes an “any vowel” wildcard. These pronunciation entries at the field site (often a health center) – they
are automatically expanded to a speech recognition grammar would have to be made in the city, away from the
consisting of all possible pronunciations based on the actual users. This did not facilitate iterative
wildcards. Thus, if there are a total of 20 consonants in the design with short feedback loops, nor did it enable
phonetic dictionary for the source language (in this example, participatory design.
US English), “M AA L U M AA C?” would be transformed • Any software/hardware failure would require trained
into a list of 20 pronunciations, each with a unique final and available personnel at the server site. This
consonant. This speech recognition grammar is then used to was not always possible.
run a re-recognition pass over any sample utterance(s) of the • For extended field research, the above problems
given word, and the best matched pronunciations are then were compounded, and it became very unlikely for
manually chosen by the user to be used as the optimized there not to be a problem
pronunciations in the final system. • The phone line was also prone to temporary
If there are multiple wildcards in the same entry, the blackouts, sometimes for days on end
combinatorial explosion would make it difficult for the speech • It was difficult to physically move the entire
recognizer to work with such a large grammar. For instance, infrastructure to a remote field site, and such a
if the developer was to try the entry “M V? L V? M V? C?”, if move would not solve the power problems, nor the
there are 20 total vowels and 20 total consonants, this would phone problem – in fact, a new phone line would
result in a 20*20*20*20 = 160,000 word grammar, which have had to be provisioned, which could have
might be computationally intractable to run recognition on. In taken months
our experiments with MSS, such large grammars did not
return recognition results even after 10 minutes on one word. Based on the above observations, experiences and
A heuristic to solve this problem is to allow the developer to constraints, we realized the need for a mobile user study setup,
create arbitrary word boundaries, which would reduce the where the actual system would be physically accessible in the
number of combinations in the final grammar. For instance, field, without the power and telephony issues. This led to the
“M V? L V? / M V? C?” (the forward slash denotes an following setup:
arbitrary word boundary) would result in a 20*20 + 20*20 = • Laptop running Windows with Microsoft Speech
800 word grammar, which is much quicker to compute. While Server, along with the Voice-over-IP gateway
the final result may lose some accuracy with the introduction • Linksys SPA3102 device (around the size of a 4-
of an arbitrary word boundary, it is a useful heuristic that port network hub) connected to the laptop through
works significantly faster (less than a few seconds for a one network cable, and connected to a telephone
recognition result with MSS). Using this heuristic, a narrower set through a standard phone cable
set of pronunciations can be derived, which can then be tested • Power for the two devices
without the arbitrary word boundary. Preliminary results
using this improved approach are described in Section VIII. Given the low power requirements for these two devices,
we were able to get much longer backup times using the same
VI. MOBILE USER STUDIES UPS. Further, the portability of the setup meant it was simple
In our initial work, our prototype interface was running on a to take it to any field site. Finally, interoperating with an
451
actual telephone set meant that we maintained the same speech) in a rural health center in Dadu district (part of the
physical interface as before, but removed all the intermediary Sindh province). However, our participants all belonged to
components that were prone to failure. We tested this system migrant communities from Balochistan, and were native
in our final user study, and it worked without a problem. speakers of a minority dialect of Balochi without any written
form. Thus, only those participants who had received at least
VII. PILOT STUDIES some schooling had any knowledge of Sindhi (7 of the 10
participants). The remaining 3 participants did not understand
A. Description Sindhi at all.
We conducted a number of pilot user studies over the past Subjective feedback needs triangulation. When the non-
year, as described below: Sindhi speaking participants were asked if they would prefer a
system in Balochi, none of them replied that they would –
Month Place Avg. Language Sample instead saying that the Sindhi system was fine the way it was.
Education Size This was surprising, as they had not succeeded in any of the
Jan Memon 5-10 years Urdu 10 given tasks. Further probing and questioning showed that
Goth (town) each had a different reason (however valid) for saying this –
Mar Umarkot <5 years Urdu 10 one said it due to peer pressure, thinking that the others would
(rural) “blame” her as the reason why the system was not made in
Jun Dadu (rural) <5 years Sindhi 10 Sindhi. Another participant said that she assumed we were
talking about official Balochi (unintelligible to speakers of
In these studies, we tested the relative effectiveness of their minority dialect), and said she would prefer a system if it
printed text against the baseline speech system as described in were in her Balochi. This reinforces the need to triangulate all
Section III. The system would only play back audio subjective feedback in ICTD research, as the sociocultural
comprising of an Urdu or Sindhi speaker reading out the text complexities inherent in such work are impossible to predict
material verbatim. Users were given an information access and account for in advance.
task (e.g. name any one danger sign during pregnancy), and Speech may be preferable to text, even for a baseline
were then either given the relevant page (e.g. containing a list system. 60% of the participants in the Dadu study said they
of danger signs during pregnancy) or played back the relevant preferred the speech system, while 40% said that both speech
audio clip on the telephone, to answer the question. and text were equal. No participant expressed a preference for
These experiments were meant primarily to validate the text. Based on the previous point regarding triangulation, we
content we had chosen (including the choice of language), as must take this with a grain of salt – however, it is expected
well as to provide a baseline against which further work could that users with limited literacy would prefer a system which
be measured. doesn’t require reading. Also, there was no statistically
significant difference in task success for these conditions in
any of the studies – but it is important to note that the speech
B. Findings system was purposefully poorly designed as it was a baseline
Information presented orally needs to be short. Both low system without any interactivity.
literate and literate users found it hard to hear long passages of Training and working with local facilitators is essential.
text with the purpose of extracting small nuggets of Over the course of these studies, we worked with user study
information. When the length of passages were varied (a few conductors from the city as well as from the locality in which
sentences, to a page, to a pamphlet), the task became the research was conducted. While the local facilitators took
progressively more difficult. more of an effort to train (requiring personalized attention,
Low literate users were less likely to have ever used a instead of assigned readings), they were much more effective
phone. Also, low literate users were more hesitant when in the user study process. Primarily, they were able to
picking up the phone (more likely to ask for permission), and communicate very effectively with participants throughout the
were more likely to hold the phone with the mouthpiece too study, and were able to understand and translate their issues
high or too low. and feedback clearly to the research team. Additionally, they
The national language is not always optimal. Initially, had deep knowledge of the community, the local context, and
our partners had told us that Urdu (the national language of of the specific participants as well – so were able to think of
Pakistan) was a language that “most” of the target users would complications before they happened, and were also able to
be familiar with and that it would be an acceptable choice for provide extra information on past events when needed.
the system. The pilot studies showed that Urdu was not Finally, the linguistic diversity (Sindhi and Balochi) that was
understood at all by 50% of the participants in Umarkot, and required for the Dadu study meant that anyone other than a
66% of those in Dadu. Of the remaining participants, many local community resident would not have been able to
still had difficulty since they were not completely familiar communicate effectively with all participants. Thus, we
with Urdu. strongly recommend training and working with local
The regional language is also not always optimal. Based facilitators for user studies.
on our prior experience, we tested Sindhi content (text and
452
VIII. FORMAL USER STUDY DESIGN • An effective tutorial is essential. Our tutorial did
In September 2008, we conducted a within-subjects user not teach participants how to use either system
study testing the speech and touch-tone flavors of the menu- well. They were not able to complete the second
based system described in Section III. The user study was task (on their own) effectively, and the 5 minute
conducted in Umarkot, Sindh, at a training center for free-form practice was not helpful either. Thus,
community health workers. Participants were recruited their performance on the actual tasks was abysmal,
through HANDS, and came from Umarkot and a nearby town, as they were not able to even navigate through the
Samarro. A day before the actual study began we conducted a system effectively on the given tasks, much less
pre-study pilot with 3 participants answer the questions correctly. It was evident that
we needed a better tutorial.
A. Pre-study Pilot • The tasks were possibly too difficult. Although it
Our initial design was as follows. Participants would be is uncertain whether this was due to the
introduced to the broad goals of the study, and the steps problematic tutorial, participants in the pilot were
involved. Their verbal consent would be requested. Personal not able to succeed in any of the given tasks, being
information would first be collected, including telephone use, especially unprepared for the second and third
educational history, and a short literacy test where the tasks (the moderately difficult and difficult tasks).
participant would read out a standard passage and be • The tasks were possibly too abstract. It is well
subjectively rated by the facilitator. They would then be known that low literate users have difficulty with
verbally introduced to either flavor of the system (touch-tone abstract thinking [23]. Even the task of asking a
or speech), and given a tutorial. After the tutorial, they would question without any context (e.g. naming any
be given three tasks, with increasing complexity, on one symptom of a disease) is an abstract task.
disease. After this they would be introduced and taught the
B. Changes to the Study Design
other flavor of the system, and would then be given three
similar tasks on another disease. At the end of the tasks, they Based on the above observations, we made some
would be given a series of Likert scale1 questions to modifications to the user study design.
subjectively rate the systems on their own and in comparison The tutorial process was increased to three practice tasks
with one another. Finally, the researcher and facilitator would instead of two. The “free-style” 5 minutes were removed.
conduct a short unstructured interview based on the Further, each of the tasks was carried out by the participant,
participants’ experience in the user study. while the facilitator listened in on each dialog, and provided
The tutorial for both flavors of the system consisted of three successively less assistance. Specifically, the facilitator gave
steps. In the first step, the participant would listen in (using explicit instructions on every step for the first task, less help
earphones connected to an audio-tap2) on the facilitator using on the second task, and almost no help (unless the participant
the system to complete a task. The facilitator would was stuck) on the third task.
purposefully make a mistake (choosing the wrong disease) and The tasks themselves were shortened (to make up for the
would then correct it, and successfully complete the task. In lengthened tutorial step) to two instead of three. These two
the second step, the participant would be given a task to were also made easier – with both tasks asking a “name any X
complete, while the facilitator would listen in, giving advice if of disease Y” form question, where X was one of: sign,
the participant had any trouble. In the third and final step, the prevention method, treatment method, cause, and Y was either
participant would be given 5 minutes to use the system as she Malaria or Hepatitis.
pleased. Finally, we thought it may be pertinent to concretize the
The three tasks were roughly equivalent for both systems. tasks by using the Bollywood Method [24]. In the Bollywood
The first task was general: “name any of the signs of disease Method, user study tasks are given a dramatic and exaggerated
X”. The second task was specific: “how many people are back-story to excite the user into believing the urgency of the
affected by disease X every year?” The third task was very problem. We decided to apply this method to only the first of
complex, e.g., “is coughing a sign of Hepatitis?” – note that each pair of tasks. Thus, the tasks were given a back-story
the answer for the third task was always no, meaning that the along the lines of: “Disease X has become prevalent in
user would have to listen through all the signs for the disease, Khatoon’s neighborhood. She is worried about catching the
and would then need to deduce that since they did not hear it, disease and wants to know of any one method to prevent the
it is not a sign. disease. Using the system, find out any one method for
Our findings from this pre-study pilot, covering three prevention of disease X”.
participants, were as follows: After making the above design changes, we conducted the
formal study. We requested Sindhi-speaking participants, and
worked with 9 participants over 3 days, and after two weeks,
1
A standard tool used to elicit subjective feedback from participants. followed these with 11 more participants over 3 more days.
Participants are asked how strongly they agree or disagree with a given
statement, by choosing a number, say 1 through 5, to represent their level of
The order of presentation of the two flavors of the system was
agreement. In our work, we adapted this tool for verbal presentation, and used counterbalanced.
a 3-point scale.
2
Also known as a Telephone Handset Audio Tap, or THAT.
453
IX. RESULTS significant main effect for literacy, F(1,68) = 10.61, p < 0.01,
Of the 20 participants, two were not able to speak Sindhi at with 18 of 32 tasks (56%) successfully completed by low
all, and were unable to complete any of the tasks successfully literate participants, and 35 of 40 tasks (86%) successfully
– their data were removed from the final analysis. completed by literate participants. These results are also
shown in Figure 1.
A. Personal Information Literate participants had a perfect task success rate
Language: Of the remaining 18 participants, it is difficult when using the speech interface. There were no interaction
to classify what language they spoke natively: not only is the effects of literacy and interface. There was a difference of
local language (Thari) very similar to Sindhi, but there is also 25% in task success for both literacy groups between the
significant inconsistency in language and dialog naming. touch-tone interface and the speech interface. Similarly, there
Many participants said they were native speakers of Sindhi, was a difference of 32% in task success between low literate
yet their Sindhi was very different from the Sindhi dialect and literate participants. It is striking to note, however, that
used in the system. The fluidity of local dialects means that it literate participants using the speech interface had a 100% task
is very difficult to tell with a high degree of certainty what success rate (20 of 20 tasks), as shown in Figure 2.
dialect a particular person speaks by simply asking them.
Age: The average age was 23 years (SD = 5.3), with a
maximum of 32 and a minimum of 17.
Years in School & Reading Ability: The average number
of years in school was 6.3 (SD = 3.3), with a minimum of 0
and a maximum of 12. 3 participants were completely unable
to read Sindhi, 5 were able to read with great difficulty, 7 were
able to read with some difficulty, and 3 were able to read
fluently. For the purpose of the analysis, the first two
categories will collectively be referred to as ‘low literate’
participants, while the last two comprise the ‘literate’
participants. Thus, there were 8 low literate participants, and
10 literate ones.
Telephone use: 15 participants had used telephones prior to
the study, with 10 participants reporting using a phone at least Fig. 2: Interaction Plot for Task Success. Literate
once every two days. participants using the speech interface had 100% task success.
B. Quantitative and Qualitative Results
There was no strong consensus on which interface was
Task success in the speech interface was significantly subjectively preferred. 10 users preferred the speech
higher than in the touch-tone interface. There was a interface, while 8 preferred the touch-tone system. A
significant main effect for the interface type, F(1,68) = 6.79, sentiment echoed by a number of participants was “I don’t use
p < 0.05, with 31 of 36 tasks (86%) successfully completed in the phone that often, and I am not used to using numbers – I
the speech condition, and 22 of 36 (61%) in the touch-tone prefer speaking instead”. However, other participants said “I
condition. These results are shown in Figure 1. am afraid I will say the wrong thing”, and that “it is hard to
speak to it, because I say too much”. These participants
understood what they were expected to say, but had a hard
time saying it.
Some participants said that speech might be problematic if
they’re in a crowded area, since the system might end up
hearing the sounds around them and get confused.
The improved tutorial method worked well. All users
were able to complete all of the tutorial steps, even though
some took up to 3 tries on one task to get the correct answer.
The problems they faced in initial practice tasks were
successively corrected over the course of the three practice
tasks, such that by the time they began the actual tasks, they
were much better prepared to use the interfaces than in the
pilot.
Fig. 1: Main Effects Plot for Task Success. There were Low-literate users expressed difficulty understanding
main effects for both interface and literacy on task success. the spoken language output from both interfaces. This was
expressed only in the semi-structured interview at the end,
Task success for literate participants was significantly when asked what main difficulties they faced. P9, for instance,
higher than for low-literate participants. There was a said she understood the facilitator perfectly well, but didn’t
454
understand the majority of what the system said. During her acoustic issues caused by the telephony interface, which may
tasks, it was evident that she wasn’t able to understand the be solved by tuning the parameters of the telephony interface
instructions given to her by either system – as she was waiting device.
without giving any input on certain prompts for up to 20
seconds at a time before hanging up. On further inquiry, it X. DISCUSSION
turned out that while P9 was a native speaker of Sindhi, her
dialect of Sindhi (and in fact, the Umarkot dialect of Sindhi) is A. Task Success vs. Preference for Interface and Literacy
different from the “official” Sindhi that the system’s voice was While earlier studies have suggested an effect of literacy on
recorded in. This includes both the accent as well as the word interface use [17, 18], this study clearly demonstrates that
content – some words are significantly different in the local literacy is a statistically significant determinant of task
dialect. Additionally, the content included some Urdu words, success. Moreover, the speech interface enjoyed a
which completely threw off the low literate participants. significantly higher task success rate than the touch-tone
However, it was difficult to get the participants to explain interface both for low-literate participants and for literate
what they found problematic, as they tended to blame participants. Literate participants were able to solve every
themselves for the problems they faced, rather than blaming single task successfully using the speech interface, suggesting
that lack of literacy constitutes a serious barrier to
the system, or the designers of the system, for creating a
performance of these tasks, irrespective of the interface used.
system that didn’t match her language skills. Finally, it is
One of the surprising, and seemingly contradictory findings
important to note that when asked if her preference would
in the above results is that literate participants reported that
change if the system was made in her language, P9 said that
they had to remember less with the speech interface and
she would prefer the speech interface if both interfaces had
preferred it, yet low literate participants said that the speech
been in her language. This sentiment was shared by other low
interface was harder, and preferred the touch-tone one, even
literate participants for whom the system’s language was
though they performed better with the speech interface on
difficult to understand.
average. This is a known effect with evaluations of speech
Literate users said that the speech system required them
interfaces [16] although with continued use, it is expected that
to remember less. When asked why they preferred the speech
user preferences conform to match task success [25].
system, the literate users responded that with the button
system, they had to remember both the topic they were B. Orality and Literacy
looking for, as well as the number they would need to press to One of the frequently occurring themes in our research is
get it. In some tasks they weren’t sure what the correct label that low literacy involves more than just the inability to read
was (e.g., when hearing the list of options in the task for and write. Low literacy is the result of less schooling, and the
naming a preventative method for Hepatitis, there was an experience of schooling imparts various skills beyond the
initial topic titled “methods of transmission”, with the title mechanics of parsing written text, such as learning how to
“methods of prevention” coming later – the first topic was a learn, learning the process of abstract thinking, learning to
potentially correct choice), and so they would have to trust forms of knowledge other than experience-based
remember two discrete bits of information for any option in knowledge, learning how to answer tests and exams (similar to
the touch-tone case. a user study), and even learning to make sense of other
Speech recognition accuracy was very high. While earlier languages, dialects and accents. We have found Ong’s
experiments with the “poor man’s speech recognizer framework of Orality and Literacy [26] to be a useful lens
approach” had mediocre accuracy (around 50%), with the through which to analyze these issues.
improvements described in Section III, the recognizer’s Ong proposes orality as a way of describing how people
accuracy was 91% for the portion of the data that was think, communicate, and learn in a culture where writing has
transcribed. Specifically, this portion consisted of 150 total not become internalized. Orality theory argues that writing
utterances, of which 133 were in-grammar (i.e., the user said has so fundamentally transformed consciousness in literate
something that the recognizer should have been able to cultures that we (literate researchers) are unable to grasp how
recognize), and 17 were out-of-grammar. For the in-grammar oral cultures and people operate. [27] summarizes Ong’s work
utterances, 121 were correctly recognized, giving an accuracy on orality and discusses its implications for both interface
of 121/133 = 91%. Further, of the 12 errors, only 2 were design and user study design in developing regions. One of the
misrecognitions, while 10 were non-recognitions. Non- key recommendations is the need to fundamentally redesign
recognitions are significantly easier to deal with, as the system any content that was originally created for consumption by
can respond with “I didn’t understand what you said, please literate readers, since oral consumers require content with very
repeat that…” followed by a list of valid options. different organization, presentation and context. Thus, orality
Misrecognitions are harder to recover from, as they result in provides a rich framework to understand why literacy makes a
the system confidently (yet incorrectly) assuming that the user significant impact on task success, suggests ways to improve
said something else, and moves the dialog in the wrong performance by low literate users, and also highlights the
direction (e.g., the user says “Diarrhea”, but the system hears importance of localization.
“Malaria”, and takes the user to information on Malaria).
Finally, of the 10 non-recognition errors, 4 were due to
455
C. Localization in this space (Likert mood knobs, Bollywood Method for task
Localization refers to adaptation of content to local culture, specification [24]), these methods have yet to be rigorously
dialect, terminology and language usage patterns. Although evaluated through multi-site experiments. The need to
crucial, localization is quite tricky. We have found that even develop and improve methods for such research is urgent, and
communicating about languages and dialects is non-trivial: a much work is needed in this direction.
rural participant may self-identify as a “Sindhi” speaker, yet E. Significant Design Decisions
may be unable to understand a “Sindhi” pamphlet recorded by
The system described in this study is the outcome of various
a “Sindhi” speaker from the city. The pamphlet may contain
design decisions made over the course of more than year of
words from other languages (e.g. Urdu), the accent of the city
testing various interface prototypes. Most notably, we
speaker may be unfamiliar, and the dialects of the languages
attempted to optimize each flavor of the interface (touch-tone
may be substantially different.
and speech) as much as possible. Thus, while touch-tone
Low literate participants are less likely to be exposed to
interfaces require a mapping between the specific choice (e.g.
alternative dialects, or to other languages, and find the urban-
diarrhea) and the button to choose that option (e.g. “2”),
Sindhi-accented system’s output more challenging than the
speech interfaces do not have this requirement. This can be
literate participants. Even one unintelligible word can throw
seen when comparing the equivalent prompts from both
off a low-literate listener completely [27]. When participants
flavors:
found the system’s Sindhi difficult to understand, they were
Speech: “What would you like to hear about? Malaria,
hesitant to speak at all after many prompts with the speech
Diarrhea, or Hepatitis?”
interface, though when given the touch-tone interface, they did
Touch-tone: “For information on Malaria, press 2, for
attempt to press buttons – this may be because speech
information on Diarrhea, press 3, and for information on
interfaces require the user to expose their confusion more
Hepatitis, press 4.”
publicly by verbalizing something potentially incorrect, versus
These optimizations meant that both interfaces were
pressing a button, which is less open to scrutiny (and social
optimized in their own right – it would not make sense to
ridicule) than speech.
“cripple” the speech interface by forcing it to match the less
The lesson, then, is that when designing a system for low
natural phrasing of the touch-tone interface, since this is one
literate users, it is crucial to choose both the language content
of the very advantages of speech interfaces that we wished to
and the system speaker (whose voice will speak that content)
test.
based on the local spoken dialect of the target user population.
Similarly, the system was chosen to have a persona (e.g.,
If there are multiple languages and dialects within the group of
“Dr Marvi”), since it is much more natural for a low literate
intended users, the system may need to be designed with
person to interact with a (virtual) person, than with an abstract
multiple language or dialect support if low literate users are
system (e.g. “Health-Line”).
part of the user group. Further, any testing of the system must
Also, the information presented in both flavors was
ensure that low literate users are adequately represented, as
carefully adapted to make the content more conducive for
their experience of any system is qualitatively and
hearing, e.g. simplifying sentence structures, and replacing
quantitatively different from that of literate users, as shown by
difficult words with easier phrases. Additionally, the content
our research. This is substantially different than in the
was designed to give a high-level summary in the beginning,
developed world, where one can often expect uniformity in
and to successively give both greater detail and larger amounts
language in a given region, given the conforming effect of
of content, as the user progressed through the hierarchy. This
schools and of universal access to mass media.
required a substantial redesign of the content by a content
Finally, the choice between speech and touch-tone may be a
expert (a medical doctor who supervises health worker
false dichotomy, as it may be optimal to provide both options,
training), as the existing healthcare material we started with
and let the user choose which option to use based on their
did not follow this paradigm.
current situation (e.g., when in a noisy environment, users may
Finally, we chose a modified form of implicit confirmation
prefer to use touch-tone, but may switch to speech in a quiet
strategy for dialog error-handling instead of explicit
place). This is common practice in the developed world.
confirmation. In our approach, the system repeats what was
D. Literacy and User Study Design recognized, and asks the user to take action only if recognition
It is important to note that user study methodologies have was incorrect (e.g. “Diarrhea. If this isn’t the topic you want,
been developed primarily with Western, literate participants in say ‘other topic’”). Explicit confirmations, on the other hand,
mind. Likert scales require the respondent to read and respond force the user to state whether the recognition was correct or
to the questions. User study instructions are recommended to not (e.g. “Did you say Diarrhea? Please say ‘yes’ or ‘no’.”).
be given uniformly, by reading aloud from a script – which is Explicit confirmation is preferred when recognition accuracy
very foreboding and artificial sounding for a low literate user. is low, but is too tedious and distracting when accuracy is
Finally, the act of asking an abstract question (e.g., name any high, as is the case here. [28].
one sign of Diarrhea) and expecting an answer is also abstract, While not rigorously quantified, we believe that each of
and would be harder for a low literate participant than a these design decisions was significant in making the user
literate, schooled participant. While some work has been done interface more usable by our end-users.
456
F. Speech Recognition Quality semantics that touch-tone interfaces are forced to use (e.g.
While speech recognition accuracy has been a persistent “For information about diarrhea, press 1”), though spoken
problem in our previous work, based on the improvements interaction can avoid this requirement, making the interface
described in this paper, the system’s recognition accuracy more natural. We believe this difference in the interface is
(91%) was comparable to commercial systems deployed in the very significant for low literate and other technologically
West that use robust recognition models trained on the inexperienced users.
language they are used for. We believe that robust speech Next, in their study with 45 participants, the only task that
recognition is a necessary (though far from sufficient) showed a significant benefit of touch-tone over speech was the
condition for the success of a speech system, and great care one that required users to record their voice as the goal of the
needs to be taken to improve speech recognition accuracy interaction. Speech interfaces that combine restrictive
when conducting such research. keyword-based grammars with open-ended “say anything”
recording segments are very difficult for users [29], since it is
G. The Importance of Effective Tutorials not obvious when (or even why) it is not possible to speak in
Through the pre-study pilot, we saw that the initial tutorial sentences in one part of the interaction, but it is required to do
strategy we made was not at all effective. By improving the so in another part.
strategy, we saw large improvements in users’ ability to access Finally, based on our goals (a system for community health
information successfully. With an ineffective tutorial strategy, workers that can be trained), we were able to spend a
both interfaces may have been harder to comprehend for all considerable amount of time training participants in the user
participants, and this might have shifted their reported study on both the touch-tone and speech interfaces. Their
preference towards touch-tone, based on our earlier system was designed and tested for users without any training,
hypothesis. which is why their user study did not involve any training
In this paper, we have proposed human-guided instruction beyond a brief introduction. This difference is noteworthy, as
in which users learn to use the system with a human mentor, even a limited amount of training can make a significant
and have shown that it worked successfully. Compared with difference to the usability of an interface, as we saw during
our prior work using video tutorials, the interactivity and our pilot study.
individually-tailored nature of the cooperative human-guided Thus, when comparing one study with another, it is
tutorial make it a better fit for both low literate and literate important to keep the specifics of the design of the interface,
users. Further work is needed to rigorously prove it as a study and tasks in mind, as well as of the larger goals of the
formal method for speech interface usability research. system involved. Their study is an important and significant
H. Rapid Iterative Development contribution insofar as it warns against the design of speech
interfaces for tasks involving recording a spoken message in
In our most recent study, we used our mobile user study
the context of untrained users. However, this should not be
infrastructure, which enabled rapid development and
extrapolated to mean that touch-tone interfaces are inferior to
modification of the speech system while in the field. This
speech interfaces in the developing world in general. Our
meant that the feedback of local facilitators was used to make
study shows that speech interfaces can be significantly better
both major and minor modifications to the dialog flow of the
than touch-tone interfaces for a different design of the
system. Additionally, it meant that speech recognition tuning
interface, the task, and the user study.
could be done locally and quickly. Finally, it was also
Finally, the study on the OpenPhone interface for HIV
possible to make minor changes after the pilot, as there were
caregivers [19] suggests that users express preference for
some issues that became obvious only when new users started
touch-tone interfaces when privacy is an issue. Privacy was
to use the system. All of this underscores the need for having
never expressed as an important factor by participants in our
a system development setup that enables field-based
study, and it is clear that such issues largely depend on the
modification of the system. We aim to use this method in all
cultural context involved, as well as the specifics of the
our future work.
system’s domain (e.g., HIV vs. neonatal health).
I. Comparison with Similar Research Thus, more work is needed to identify exactly where speech
Our results contradict similar work in the field, most interfaces work well and where they do not.
notably the study by Patel et al. [20] testing speech and touch-
tone interfaces for listening to pre-recorded radio shows and XI. CONCLUSION
recording audio content for a radio talk show. In comparing We draw two main conclusions from this work. The first is
our work with theirs, a number of factors need to be that the ability to perform information access tasks is strongly
considered. hampered by low literacy, and not just because of the inability
First, in our system, the speech-input flavor was more to read. We derived empirical confirmation that literacy is a
conversational (e.g. “What would you like to hear more significant determinant of success in information access tasks.
information about, diarrhea, malaria, or hepatitis?”) as This by itself is not surprising, but our results further suggest
compared to theirs (e.g., “To ask a question, say ‘question’; to that the problems associated with low literacy go far beyond
listen to announcements, say ‘announcements’; to listen to the the inability to read, since they also affect task performance
using the speech interface, where no reading is necessary.
radio program, say ‘radio’”). It is this mapping of keyword to
457
Our second conclusion is that, at least for some target [11] A. Rudnicky, E. Thayer, P. Constantinides, C. Tchou, R. Stern, K.
Lenzo, W. Xu, A. Oh. Creating natural dialogs in the Carnegie Mellon
populations, a well-designed speech interface can significantly
Communicator System, in Proceedings of Eurospeech, 1999
outperform a touch-tone interface for both low-literate and [12] V. Zue, S. Seneff, J. Glass, J. Polifroni, C. Pao, T.J. Hazen, L.
higher literate users. Given the potential utility of Hetherington, 2000 – JUPITER: A Telephone-Based Conversational
information access and the pervasiveness of low literacy Interface for Weather Information, in IEEE Transactions on Speech and
Audio Processing, vol. 8, no. 1, January 2000.
throughout many parts of the world, we hope to see many
[13] A. L. Gorin, B. A. Parker, R. M. Sachs, J. G. Wilson. How May I Help
spoken dialog systems developed, evaluated and deployed in You. Speech Communications, Vol. 23, pp. 113-127, 1997.
the near future. [14] J. Sherwani, R. Rosenfeld. The Case for Speech and Language
Technologies for Developing Regions. In Proc. Human-Computer
Interaction for Community and International Development workshop,
ACKNOWLEDGMENT ACM CHI, Florence, Italy, April 2008.
This project was supported by Microsoft Research through [15] F. Weber, K. Bali, R. Rosenfeld, K. Toyama. Unexplored Directions in
its Digital Inclusion Initiative and by the U.S. Agency for Spoken Language Technology for Development. In Proc. Spoken
International Development (USAID) under the Pakistan-U.S. Language Technology for Development workshop, SLT, Goa, India,
2008.
Science and Technology Cooperation Program. We are
[16] E. Barnard, M. Plauche, M. Davel. The Utility of Spoken Dialog
grateful to both organizations for making this work possible. Systems. Proc. Spoken Language Technology for Development
We would also like to thank all the members of the HANDS workshop, SLT, Goa, India, 2008.
staff that facilitated this research, especially Dr Abdul Aziz, [17] M. Plauche, U. Nallasamy, J. Pal, C. Wooters, and D. Ramachandran.
Speech Recognition for Illiterate Access to Information and Technology.
Bansi Malhi, Laal Jaan, and Dr Anjum Fatma. We are very Proc. International Conference on Information and Communications
grateful to Safieh Shah, Samia, Arbab, and Lata for facilitating Technologies and Development, 2006.
the various user studies. Finally, we would like to thank the [18] E. Brewer, M. Demmer, M. Ho, R.J. Honicky, J. Pal, M. Plauché, and S.
community health workers at Wahi Pandi, Memon Goth, and Surana. The Challenges of Technology Research for Developing
Regions. IEEE Pervasive Computing. Volume 5, Number 2, pp. 15-23,
Umarkot who gave their time and feedback. April-June 2006.
[19] C Kuun, OpenPhone project piloted in Botswana
REFERENCES http://www.csir.co.za/enews/2008_july/ic_05.html, accessed Feb 1st,
2009.
[1] H. M. Kahssay, M. E. Taylor, P. A. Berman. Community Health
Workers: The Way Forward. World Health Organization, 1998. [20] N. Patel, S. Agarwal, N. Rajput, A. Nanavati, P. Dave, T. Parikh, A
[2] S. Hunt. Evaluation of the Prime Minister's Lady Health Worker Comparative Study of Speech and Dialed Input Voice Interfaces in Rural
India. ACM CHI 2009.
Program. Oxford Policy Management Institute.
http://www.opml.co.uk/go.rm?id=380. Accessed Feb 1st, 2009. [21] J Sherwani, Dong Yu, Tim, Paek, Mary Czerwinski, Yun-Cheng Ju,
[3] N. Pakenham-Walsh, C. Priestley, and R. Smith. Meeting the Alex Acero, VoicePedia: Towards Speech-based Access to Unstructured
Information, Interspeech 2007, Antwerp, Belgium.
Information Needs of Health Workers in Developing Countries. British
Medical Journal, 314:90, January 1997. [22] A. Haines, D. Sanders, U. Lehmann, AK Rowe, JE Lawn, S. Jan, DG
[4] J. Sherwani, N. Ali, S. Mirza, A. Fatma, Y. Memon, M. Karim, R. Walker and Z Bhutta. Achieving child survival goals: potential
Tongia, R. Rosenfeld. HealthLine: Speech-based Access to Health contribution of community health workers. The Lancet 369(9579): 2121-
Information by Low-literate Users. In Proc. IEEE/ACM Int'l Conference 2131. 2007
on Information and Communication Technologies and Development, [23] A.R. Luria. Cognitive Development: Its Cultural and Social
Bangalore, India, December 2007. Foundations. Harvard University Press, Cambridge, MA. 1976.
[5] M. Huenerfauth. 2002. Developing Design Recommendations for [24] A. Chavan. 2007. Around the World with 14 Methods.
Computer Interfaces Accessible to Illiterate Users. Thesis. Master of http:// humanfactors.com/downloads/whitepapers.asp#CIwhitepaper.
Accessed on August 22, 2008.
Science (MSc). Department of Computer Science. National University
of Ireland: University College Dublin. [25] A. I. Rudnicky. Mode Preference in a Simple Data-Retrieval Task.
[6] S. Deo, D. Nichols, S. Cunningham, I. Witten, 2004. Digital Library Proceedings of the ARPA Workshop on Human Language Technology.
Access For Illiterate Users. Proc. 2004 International Research San Mateo: Morgan Kaufmann, 1993, 364-369.
Conference on Innovations in Information Technology
[26] W. Ong. Orality and Literacy: The Technologizing of the Word. London:
[7] I. Medhi, A. Sagar, K. Toyama. Text-Free User Interfaces for Illiterate
Routledge, 2002.
and Semi-Literate Users. Proc. International Conference on Information
and Communications Technologies and Development, 2006. [27] J. Sherwani, N. Ali, R. Rosenfeld. Orality-grounded HCID:
[8] S. Grisedale, M. Graves, A. Grunsteidl, 1997. Designing a Graphical Understanding the Oral User. Submitted to the Information Technology
User Interface for Healthcare Workers in Rural India, ACM CHI 1997 and International Development journal.
[9] V. Anantaraman, et al. Handheld computers for rural healthcare, [28] D. Jurafsky, J. H. Martin. Speech and Language Processing. Prentice
experiences in a large scale implementation. In Proceedings of Hall, 2008.
Development By Design, 2002.
[29] J. Sherwani, S. Tomko, R. Rosenfeld. Sublime: A Speech- and
[10] T. Parikh, G. Kaushik, and A. Chavan, Design studies for a financial Language-based Information Management Environment. In Proc. IEEE
management system for micro-credit groups in rural India. Proc. of the Int.l Conference on Acoustics, Speech and Signal Processing, Toulouse,
ACM Conference on Universal Usability, ACM Press (2003). France, May 2006.
458
The Case for SmartTrack

Michael Paik∗ , Ashlesh Sharma∗ , Arthur Meacham∗ , Giulio Quarta† , Philip Smith† , John Trahanas† ,
Brian Levine‡ , Mary Ann Hopkins† , Barbara Rapchak§ , Lakshminarayanan Subramanian∗
∗ CourantInstitute
New York University NY 10012
† School of Medicine
New York University NY

‡ Department of Obstetrics and Gynecology
New York Presbyterian Hospital

Columbia University Medical Center
New York NY 10032
§ Leap of Faith Inc.
23 Brink St, Crystal Lake IL 60014
Abstract—Nearly 40 million people in Africa suffer from Despite the subsidies offered by pharmaceutical companies
HIV/AIDS. African governments and international aid agencies for expensive AIDS drugs, and the enormous effort of African
have been working to combat this epidemic by vigorously promot- governments to promote HAART programs, these initiatives
ing Highly Active Anti-Retroviral Therapy (HAART) programs.
Despite the enormous subsidies offered by governments along have not been adopted on a wide-scale. The fundamental
with free Anti-RetroViral (ARV) drugs supplied by agencies, barriers to large-scale adoption are two-fold. The first problem
the introduction and implementation of HAART programs on is one of lack of accountability in the system, due to the
a large scale has been limited by two fundamental problems: theft of expensive ARV drugs, counterfeiting of drugs, and
(a) lack of adherence to the ARV therapy regimen; (b) lack of corruption. The Global Fund crisis that gripped much of
accountability in drug distribution due to theft, corruption and
counterfeit medication. Africa, with losses amounting to hundreds of millions of
In this paper, we motivate the case for SmartTrack, a telehealth dollars was directly related to the lack of accountability in
project which aims to address these two problems facing HAART the system. In fact, many pharmaceutical companies that
programs. The goal of SmartTrack is to create a highly reliable, provide subsidized AIDS medications are demanding better
secure and ultra low-cost cellphone-based distributed drug in- accountability practices as a prerequisite to their continued
formation system that can be used for tracking the flow and
consumption of ARV drugs in HAART programs. In this paper,
participation.
we assess the potential benefit of SmartTrack using a detailed The second serious problem relates to the lack of patient
needs-assessment study performed in Ghana, using interviews adherence to the medication regimen. A typical HAART pro-
with 516 HIV-positive rural patients in a number of locations gram requires every patient to consume two to five medications
across the country. We find that a system like SmartTrack would per day; in addition, required medication and dosages may
immensely benefit both patients and healthcare providers, and
can ultimately lead to improved patient outcomes and better
change rapidly based on the side effects observed by the
accountability. patient. Lack of medical oversight, or inappropriate use of
Index Terms—SmartTrack, telehealth, drug tracking, drug ARVs not only harms the health of patients but may also
monitoring, patient adherence encourage drug-resistant strains of HIV, posing a substantial
public health risk. Hence, to improve patient outcomes, it is
I. I NTRODUCTION essential for doctors to continuously observe the health status
and the medication consumption regimen of patients. In fact,
According to the World Health Organization, effective it is relatively common for doctors and health-workers to
HIV/AIDS care requires antiretroviral therapy (ART) [1] as physically track patients who have not reported for regular
a treatment method. Without access to antiretroviral therapy, medical visits.
people living with HIV/AIDS cannot attain the fullest possible In this paper, we motivate the case for SmartTrack, a project
physical and mental health and cannot play their fullest role as which aims to address these two fundamental problems with
actors in the fight against the epidemic, because their life ex- HAART programs. The vision of the SmartTrack project is to
pectancy will be too short. While ART is commonplace in de- create a highly reliable, widely available and ultra low-cost
veloped countries, these life saving medications have reached cellphone-based distributed drug information system that can
only a small percentage of the more than 40 million Africans be used for tracking the flow and consumption of ARV drugs
infected with HIV or suffering from AIDS [2]. In order to be in HAART programs, ultimately leading to improved patient
successful in combating this deadly disease, strict adherence outcomes.
to a highly active antiretroviral therapy (HAART) [3] program The motivation to use cellphones in SmartTrack to track the
must be followed. flow of AIDS drugs is in large part driven by the explosive
459
growth in the adoption of low-cost cellphones in the rural II. M OTIVATION AND R ELATED EFFORTS
developing world. For example, around 80% of Rwanda has The SmartTrack project vision is centered on three primary
cellphone coverage [4], [5] and the cost of cellphones is so premises: (a) the need for better accountability in ARV drug
low (less than $50) that the devices have become affordable distribution; (b) the need for improved patient adherence to
by the poor [6]. Cellphones have enormous potential for medication regimen in HAART programs; (c) the utility of
enhancing rural healthcare, given their ability to act as a cellphones as a healthcare platform for monitoring the flow
low-cost computing platform for distributed applications. The and consumption of drugs. In this section, we discuss these
open-source movement in cellphone software [7], [8], [9] three aspects in more detail, along with the corresponding
has also opened up the potential for a wide range of new related efforts.
applications targeting these low-cost devices.
The vision of SmartTrack is to enable patients, health- A. Maintaining Accountability in the Supply Chain
workers and doctors to use cellphones to record, track and The first premise of SmartTrack builds on the position of
transmit information about pharmacotherapy utilization. The the World Health Organization (WHO), which has argued for
idea is to tag every ARV drug bottle with a “smart” tag: an the need for a drug tracking system that can monitor the
RFID or a barcode, that will enable patients to identify drugs flow of medications from the supplier to the patient. The
easily and also enable them to report their drug consumption high cost of ARV drugs makes them an attractive target for
remotely using a cellphone. The drug information will be theft. This is a particular threat in developing countries where
compiled on the device, sent to a central server, and stored corruption can be widespread and persistent. International
at different levels (regional level, district level) to be used HAART programs are not immune to theft, as was illustrated
by the healthcare teams to tailor patient-specific therapies. In by the Global Fund scandal in Uganda in 2006; in that
addition, the device will also be able to remind patient end- incident, an investigation revealed that ”tens of millions” of
users about when to take their specific medications. Thus, dollars in grants intended for AIDS treatment were misspent
SmartTrack will provide the patient with information about by officials, often to finance lavish lifestyles [11]. Similar
his or her specific regimen, and provide the care provider with problems have also been reported in other recipient nations
information about patient compliance to the regimen. [11], and the widespread theft of medicine for resale has been
SmartTrack is based upon eMedonline1 , a technology pre- exposed in a number of instances [12], [13].
viously developed and patented by our partner, Leap of Faith Perhaps more disturbing than outright theft is the growing
Technologies, Inc. [10]. This technology leverages cellphones practice of replacing legitimate medication with counterfeit
and RFID technology to optimize medication compliance, drugs. While it is difficult, if not impossible, to determine the
track medication usage and extend patient care to remote actual scope of this problem, WHO has estimated that in parts
settings. The previous project has been successfully used in the of Africa, Asia and Latin America, over 30% of all drugs sold
US for tracking adverse events and improving compliance with are counterfeit [14]. Consuming counterfeit medication can
oral antineoplastic (anti-cancer) therapy; efficacy is currently lead to drug resistance and death, and many counterfeits have
being demonstrated among cardiovascular disease populations. been found to contain highly toxic substances.
A key requirement for the success of the SmartTrack is In addition to the serious health consequences that can
to first assess needs and user acceptability. In this paper, we result from fake or stolen HAART drugs, the phenomenon
describe our experiences in realizing this first important step could also lead to a loss of confidence in the health system
in the SmartTrack project vision. We recently performed a as a whole. This threatens to undermine support for aid
detailed needs-assessment study, interacting with 516 AIDS programs, as well as discourage patients from seeking out
patients in rural parts of Ghana to determine the potential and adhering to treatment. A supply-chain management system
role and effectiveness of cellphones in improving AIDS care that could track medicines from their acquisition point to the
delivery. To this end, we developed simple cellphone-based end user would be able to detect drug theft quickly, and
healthcare applications which we used as part of our study to would make it easier to discover sources of counterfeit drug
understand user acceptability issues. Our study was conducted substitution within the supply-chain. This added accountability
by medical students who interacted with each patient for over could greatly improve the safety and effectiveness of HAART
30 minutes and used a translator (when language was an issue) programs in developing countries.
to obtain answers to a detailed survey in combination with
B. Tracking Patient Adherence and Symptoms
user interface testing. Based on our detailed patient study
evaluations, we determined that a system like SmartTrack Proper adherence to a HAART regimen is critical to prevent
would immensely benefit both patients, healthworkers and drug resistance and ensure survival [15]. However, HAART
doctors and can ultimately lead to improved patient outcomes treatment requires multiple drugs, taken multiple times per
and better accountability. day, and often results in unpleasant side effects. These dif-
ficulties, combined with the high cost of the drugs and the
1 developed by Leap of Faith Inc., with support from the National Cancer
potential stigma of being identified as HIV+ can act as a
Institute (Contract No.HHSN261200644005C) and the National Institute on deterrent to proper adherence. A number of studies have
Aging (Grant No.2 R44 AG022271-02A1) attempted to measure adherence rates in Sub-Saharan Africa,
460
and they have produced promising numbers, with an average in healthcare environments in Ghana, Uganda, and Kenya.
of 77% of participants reporting proper and consistent use of The project put PDAs into the hands of physicians, medical
the medication [16]. However, these studies rely on unverified students and community volunteers in different settings in
self-reporting, which has been demonstrated to exaggerate order to demonstrate their viability and usefulness, especially
the level of compliance in studies conducted in developed for the collection of health data and dissemination of medical
countries. information.
Tracking adherence is especially difficult in developing Bridges.org, a technology NGO, performed an independent
regions, where lack of medical, transportation, and commu- evaluation of the Satellife PDA trial [28], in which it validated
nication infrastructure can limit contact between patients and the use of handheld computers in healthcare environments in
health-care workers [17]. For example, our study of HAART Africa. Specifically, it found handheld computers to be an
patients in Ghana found that more than 37% of participants appropriate technology for use in the African context, and
went at least three months between clinic visits, that over 94% concluded that they provide an inexpensive alternative to PCs
had never placed a phone call to their clinic, and that the same in terms of computer power per dollar. The technology of
number had never received a phone call from their clinic. This Satellife was found to be simple to use and easily integrated
lack of contact can make it very difficult for health workers into the daily routines of the healthcare professionals.
to keep tabs on patient adherence and progress. Likewise, it Cell-Life [29], [30], a research team at the University
can prevent patients from notifying health workers of critical of Cape Town, has demonstrated the effectiveness of an
symptom information that might require a trip to the clinic or information gathering system using cellphones for improving
a change in treatment. the health outcomes in ART. Their system leverages health
A low-cost method of tracking medication, adherence, and workers to collect real-time information on HIV/AIDS along
symptoms in developing countries could provide great benefit with the spatial infrastructural requirements (person, their
to its users. By preventing fraud, it could save precious finan- environment, and access to basic amenities). Cell-Life has
cial resources for providers, and ensure that medicine gets to been deployed in South Africa and Zambia.
those who most desperately need it. By tracking adherence and
symptoms, it could improve the responsiveness of healthcare III. S MART T RACK
workers to the needs of patients, and ultimately improve and
The SmartTrack project was in part motivated by prior suc-
prolong the lives of those receiving HAART treatment.
cessful research and development by our partner, Leap of Faith
C. Cellphones as a Healthcare Platform Inc., of a cellphone-based telemonitoring system for improving
medication adherence. This system, called eMedonline2 uses
In the past few years, several research and developmental
smartphones equipped with RFID scanners to read smart-tags
efforts around the world have explored [18], [19], [20], [21] the
attached medicine bottles. The phone provides reminders to the
use of cellphones as a potential tool for improving healthcare
patient, and the patient scans each bottle with the phone when
in both the developed and the developing world. For a detailed
taking that particular medication, updating a remote database
overview of globalization and health related issues, please
and providing doctors with detailed information on regimen
refer to [22]. In developed nations, the problem of tracking
adherence. The system validated the integration of cellphones
medication compliance has been addressed successfully with
and radiofrequency identification (RFID) as a therapeutic solu-
smartphone-based telehealth tools [23]. Recent growth in cell-
tion to medication compliance and supply chain management.
phone penetration in developing regions has made mobile tele-
Feasibility and functional tests of the system in a sample
health solutions a real possibility for users in less-developed
of oncology patients demonstrated that drug and compliance-
nations. In Sub-Saharan Africa in 2006, the overall cellphone
related data can be reliably collected, analyzed, and exported
penetration level was estimated at 15%, and it has more than
for use in other clinical monitoring systems. Patient acceptance
doubled in the last two years [4], [5]. Furthermore, certain
and value of the system was very high.
countries, such as South Africa, have attained adoption rates
of over 70% [24]. In our own study, conducted among HAART Building on this prior success, the SmartTrack project vision
recipients in Ghana, we found that cellphone usage was quite aims to develop and deploy a cellphone-based telemedicine
high in the cities and relatively large even in rural areas: and supply-chain sytem that exploits this trend, with the
Nationwide, 54% of participants reported using cellphones, twin goals of tracking the flow of drugs to the field, and of
with more urbanized areas such as the Greater Accra (77%), monitoring the status and adherence of HAART patients. In
Northern (67%) and Volta (57%) regions ranking higher in addition to tracking adherence, we also envision a system that
usage than rural areas, such as the Central region (29%), and acts as a reminder system, that allows users to record symptom
Upper West (26%). information which can be tracked and analyzed by their doctor,
Several groups have examined the use of PDAs and cell and that allows health workers to contact patients. There are,
phone-based tools to augment patient monitoring relative however a number of limitations on the ground that prevent
to ARV therapy, and to extend medical follow-up capabili- 2 developed by Leap of Faith Inc., with support from the National Cancer
ties [17], [25], [26], [27], [18], [19], [20], [21]. One important Institute (Contract No.HHSN261200644005C), and with support from the
project was led by Satellife [25], which tested the use of PDAs National Institute on Aging (Grant No.2 R44 AG022271-02A1)
461
us from simply deploying a modified version of our existing Trusted authority

system or equivalent in rural Africa.
Pharmaceutical /
• The prohibitive cost of smartphones: Smartphones are Subsidized drug
quite simply out of reach for most users in developing
environments, so we must design a system for standard,
low-end cellphones.
• Limited data communication infrastructure: We must rely Health organization National Hospitals
SmartTrack based tracking

on simple Short Message Service (SMS) protocols to
communicate between the local client and the central
database.
• Cost of communication: Every 160-byte message costs
the patient money, so we must minimize the number of local shops
messages and be economical with data.
• Language: In our Ghana study, we found speakers of 11
different languages, many of whom required interpreters Regional Hospitals
to conduct the interviews. Our system would have to
accomodate all of them.
• Illiteracy: A great proportion of potential users are unable
to read or write in any language, requiring us to develop
a simple, non-textual interface. Rural clinics
Later, we describe our detailed needs-assessment user study
based on interactions with HIV positive patients in Ghana that
seeks answers to these specific questions. Based on our user
study assessments, we feel that it is possible to design an Healthcare workers
appropriate and usable SmartTrack system that can be used
by (potentially illiterate) users, health workers and doctors to
improve accountability and patient adherence.
A. SmartTrack Architecture rural patients
Figure 1 illustrates the supply chain flow of drugs in

SmartTrack. Drugs in the supply chain system flow through a Fig. 1. A generic design of tag based supply chain management
system. The Trusted authority provides the Tag along with drug
hierarchy of suppliers, distributors and customers. In Smart- bottles. The users at each point in the supply chain use the cell phone
Track, we require a mechanism to track the flow of drugs at based tag reader to identify and authenticate drugs. Trusted authority
every level in the drug supply chain system. can remotely track items at every node in the supply chain and check
The basic architecture of SmartTrack can be described as for fraud using the cell phone based system.
follows. Every drug bottle is tagged with a “smart” tag (RFID
or barcode) which uniquely identifies a bottle. All the identities
in the bottles should be generated by the supplier of the drugs. server using SMS/MMS or GPRS connectivity. To verify the
Each such identity should be generated from a secret key that authenticity of a tag, an agent has three options: online, offline,
makes it hard to generate new valid identities to primary deal batch authentication. In online authentication, the agent signals
with counterfeiting of bottles. The trusted central authority the tag information to the central server using the cellular
provides the medicines or drugs along with the tag (RFID or network and verifies the authenticity of the tag. In offline au-
barcode) embedded in the bottle for tracking the flow of each thentication, the agent prefetches the list of “admissible” tags
drug. The Tag information of every item or supply is stored from the server into its local-store and verifies each tag with
in the central server. the set of admissible tags. In batch authentication, the agent
Each intermediary point in the supply chain is managed by can assume that the goods are genuine and merely store the tag
an Agent who is equipped with a cellphone and a tag reader information in the cellphone locally. In this scenario, the agent
(barcode or RFID reader) . Upon receipt of any goods, the can collect several tags and perform bulk verification in a lazy
Tag information is captured by the corresponding agent in manner. An important benefit in this delayed batch verification
the supply chain, who stores the Tag information in the cell is that it can significantly reduce the communication cost using
phone. The cellphone acts as the local-store at each supply cellphones and does not require immediate connectivity. In
chain point. summary, this architecture represents a distributed data-store
The central server or Trusted authority uses cellphones of cellphones, where each node maintains a local store of
to remotely track the flow of goods. This type of remote authorized signatures and nodes communicate with each other
authentication is possible by transferring the Tag information using cellular connectivity.
attached to the goods, from a cell phone to or from the central While what we have described is a basic overview of the
462
SmartTrack architecture, we need to address several technical AIDS patients. This system used a voice and text interface
research challenges to make SmartTrack into a low-cost, to prompt users to respond to various questions. In addition,
secure, highly reliable and widely available system. We briefly the system used RFID tag identification which we did not
outline the basic technical ideas that we intend to use in use in our study. We used a HTC P6300 phone for this
SmartTrack to achieve these properties. implementation.
1) Information aggregation: Using an inbuilt network cost Cellphone based system: As a comparison point, we
model, SmartTrack will intelligently aggregate multiple developed an alternative implementation using JavaME [31]
updates from a cell-phone and semantically compress on Nokia 3110 phones that used a text-free pictogram based
(not standard compression) to the smallest number of user interface of specific healthcare symbols to primarily un-
messages possible to reduce transmission costs. derstand whether the end-user population would be conversant
2) Strong identity and secure updates: To provide strong with a pictogram-based interface.
security properties, the system will leverage public-key A. Methodology
and symmetric-key cryptographic operations to develop
an unforgable identity for every cellphone and also The user study for SmartTrack was performed at several
provide the ability to verify the validity of individual sites across Ghana. To perform the study, we developed a
updates using aggregate signatures. survey tool that was administered to cohorts of HIV+ patients
3) Privacy: To provide privacy, the system will develop an in each location. The cohorts were patients affiliated with
anonymized indexing system that cannot be traced back either the West Africa AIDS Foundation (WAAF) [32] or
to individual patients without appropriate information. Korle Bu Teaching Hospital. Participants were referred to us
4) Reliability and Availability: the system increases avail- by these organizations, and we did not attempt to implement a
ability without incurring additional operational cost by randomized study. The ages reported by patients ranged from
maintaining a detailed local data-store within every 17 to 101, and 73.3% of the patients were female, 26.7% male.
cellphone that contains most of the appropriate infor- 77.3% of patients were currently receiving ARV therapy.
mation required by the end-user (health-worker, doctor We interviewed 516 HIV+ patients with a survey consisting
or patient). This local-store is constantly synchronized of 72 questions divided into eight categories:
with the main information store either using aggregate 1) Laboratory values: CD4 count, viral load and WHO HIV
updates or physical synchronization. To improve reliabil- stage (taken from blood tests/patient records).
ity, SmartTrack will intelligently replicate and partition 2) Demographics: Age, housing, occupation, education,
critical information in the data store between the local language skills, access to water and electricity (11
server and the smart-phones. questions).
While these technical issues are certainly important, a 3) Health Care Access: Health practices, frequency of
detailed discussion of them is outside the scope of this paper. medical care, frequency of contact with health worker,
The primary focus of this paper is to make the case for why distance to the hospital (9 questions).
rural users need a system like SmartTrack, and, if provided, 4) Symptoms: HIV-related symptoms, typical response to
will they actually use such a system. The rest of the paper certain symptoms, and histories of other conditions like
will focus on answering these user-centric questions. tuberculosis or malaria (11 questions).
5) Medications: Medications being taken, adherence to
IV. N EEDS -A SSESSMENT U SER S TUDY regimen, acquisition of drugs (10 questions).
When working on any “technology for development” 6) Narrative Text: Difficulties of the disease, difficulties
project, two fundamental questions always arise: the need for of treatment, perceived effectiveness of treatment (9
the technology and user acceptability of the technology. Many questions).
projects in the development space have been successful in the 7) Mobile Phone Usage: Access to mobile phone, usage
beginning phase but have often not been adopted on a wide habits, amounts paid for service and hardware (20 ques-
scale. It is this barrier that we wish to cross in SmartTrack. tions).
Hence, we began with the endeavor of performing a detailed 8) Patient Education: What is the patient’s primary source
needs-assessment study to answer the two questions about of health information (1 question)?
need and user acceptance: (a) Do rural users need a system like The patients gender and use of a translator were not asked
SmartTrack? (b) If provided with a system like SmartTrack, directly, but were recorded. When necessary the aid of a
will they use it? Rather, how would one build SmartTrack to translator was enlisted.
maximize user acceptance. 278 patients were interviewed via healthcare associations
To address these questions, we began with the modest for patients living with HIV, coordinated through WAAF.
goal of designing a detailed survey covering a wide-range of Researchers traveled to five cities in four different regions of
questions. We also designed two simple user-interfaces as part Ghana to interview a total of six cohorts of patients: Volta
of this study: Region (Ho), Central Region (Cape Coast), Northern Region
Smartphone based system: We tailored our smartphone (Tamale), Upper West Region (Wa and Lawra). Blood draws
based telehealth system for cancer patients into a system for to be used for CD4 counts were provided free-of-charge by
463
WAAF in these locations. After having their blood drawn,

the patients were asked to participate in the survey. Before 100
Yes
No
beginning the survey, patients gave verbal consent.
A further 238 outpatients at Korle Bu Hospital Fevers Unit 80
were interviewed. The patients had come to the hospital to

refill medication or to draw blood for CD4 T lymphocyte 60
count. At that time, patients who were willing to participate
percentage
were required to give written consent. The written consent
40
forms were collected by Korle Bu Hospital staff to be included
in the patients chart, and will not be published or disclosed in
this or any other report to maintain patient confidentiality. The 20
HIV+ patient population at Korle Bu Hospital was separated

into groups of those taking ARV therapy, and those that had 0
R
not yet started ARV therapy. Only patients who were currently
rb
ur
an
al
Translator required
taking ARV therapy were interviewed as their experiences are
more relevant to the study.
Adjustments were made to the survey and methods during
Fig. 2. Percentage of patients requiring translators, with urban compared to
the course of the study. The survey itself was edited after the rural
first two administration sessions in order to more effectively
elicit the target information. Several questions were altered,
discarded, or added. The changes have been recorded and will
be taken into account. The overwhelming majority of patients 100
Education percent of patients
were interviewed individually. In certain situations however,
there were time constraints and patients were interviewed in 80
pairs. In these situations each patient answered individually.

60
V. U SER S TUDY
percentage
In this section we present the results of our user study and 40
observations made while interviewing 516 HIV+ patients in

Ghana. 20
A. Literacy & Education

0
Illiteracy is one of the most serious challenges facing the
6t
A-
JS
SS
U
o
ni
-le
h
le
ve
sc
Fo
ve
ve
ho
rs
rm
ity
o
developing world, and is particularly severe in rural areas. We l
Education level
gave participants a number of questions designed to gauge
the level of literacy and education within our population.
Questions we asked included: i) what languages does the
participant speak? ii) in which languages is the participant able Fig. 3. Maximum education level
to read or write? iii) what is the participant’s highest level of
education?
We found that 55.6% of interviewees in urban areas, and
60% in rural areas needed a translator for our interviews. solution might solve the literacy problem, though the voice-
Interviews were conducted in English, which is also the official based UI would not solve the multiple language problem, and
language of Ghana. Additionally, our survey showed that urban would have the additional drawbacks of lack-of-privacy, in
areas tend to have higher literacy rates than rural regions. addition to feasibility and cost problems when using low-cost
Figure 3 shows the distribution of educational attainment phones with expensive network access.
among patients, and Figure 4 shows the percentages of patients To this end, we are exploring pictogram interfaces for
able to read or write in each of the locally-used languages. this application. We developed a demonstration interface for
Notably, at least 35% of patients had never attended school of our Ghanaian study in order to test our ability to convey
any kind, and only 46% reported being able to read English. healthcare-related concepts through graphical symbols. Al-
This high rate of illiteracy, particularly when combined with though extremely preliminary, this demonstration was well-
the lack of a common language among patients, presents a received by users and showed promise for expressing health
fundamental challenge to the deployment and acceptance of concepts. Using simple pictograms, combined with shape or
any telemedicine application. In order for SmartTrack to be color symbols to represent different medications, it should be
effective, it must be operable by a patient with no reading possible to make an effective interface that is usable despite
skills whatsoever. Either a voice-based or a pictogram-based barriers of literacy and language.
464
100 100
Languages known (read or write) percent of patients percentage of ages of patients
80
80
60
percentage
60
percentage
40
40
20
20
0
Ar
En
En
En
En
Ew
Fa
Fr
Tw
on
a
en
ab
nt
gl
gl
gl
gl
i
e
e
is
is
is
is
ch
ic
h,
h,
h,
Ar
Fa
Tw
ab
nt
i
i
ic
20
30
41
51
61
-2
-4
-5
-6
-7
9
3
Languages Age
Fig. 4. Percentage of patients able to read or write, by language Fig. 5. Age distribution of participants
B. Age and Sex To better understand the economic situation of our popula-
A patient’s age is an important factor in determining proper tion, we asked a number of questions pertaining to their basic
medical care, and age statistics can offer some insight into the employment, income, and living conditions. The following are
impact of AIDS on the society at large. Figure 5 shows the age some selected findings:
distribution of the patients, and illustrates the disproportionate • Type of house: 58.3% lived in a compound house, 24.4%
number of people seeking treatment who are in the 30 to 50 in a self-contained house and 16.3% in a rented house.
year-old range, with over 69% of patients falling within that The rest were either homeless or lived in mud huts.
group. For so many people in the prime of their work and • Water source: 75.6% used piped water and 15% used
family lives to be stricken can only have devastating effects on well-water.
a society. The lost productivity of working-aged people slows • Electricity: 26% did not have reliable access to electricity.
the economic development of an already-depressed region, and • Refrigeration: 56% lacked any sort of refrigeration.
the death or debilitation of so many parents will have social • Occupation: The most common job categories reported by
ramifications we have barely begun to understand [33]. patients were petty trading(34.7%), service work(23.1%)
Improving adherence can prolong lives and improve the and agriculture(12.6%).
quality of those lives. In addition to the direct and obvious To live in such a setting can be challenging enough for
benefit to the patient, this can also have the wider benefit of a healthy person, but for an HIV+ person, the difficulty is
allowing the patient to be more productive economically, and multiplied. People with AIDS can experience weakness and
to play a greater role in the lives of family and the community. any number of other symptoms that interfere with their ability
Interestingly, 73% of patients were female. While it is to work and earn money. Of our study participants, 64%
estimated that almost 61% of HIV+ people in Sub-Saharan responded that AIDS has a considerable effect on their ability
Africa are female[34], it is not clear if our higher number to work, and 60% reported missing work as a result of HIV
reflects the actual proportion of women with HIV in Ghana, symptoms.
or if it is a result of our non-random selection process. It is Social stigma for those with HIV has been reported to be a
even conceivable to us that it is a result of the self-selection serious issue in many cultures, particularly in parts of Africa,
process of people seeking HIV treatment or getting diagnosis. but interestingly, 61.5% of our participants said that they did
[34] not have any social problems due to AIDS. If true, this is
wonderful news, but we suspect the response does not tell the
C. Economic and social issues whole story. One possibility is that patients were hesitant to
It is impossible to divorce the problem of medical treatment admit that they had experienced such discrimination. Another
in the developing world from the economic environment in is that they had not experienced discrimination because they
which the patients live. The cost of care is only the most ob- had successfully hidden the fact that they had HIV.
vious element. The patient also must contend with the expense All of these details bring home a key point: Any designer of
of travel to distant clinics, and the cost of missed work during a telehealth system for the developing world must be keenly
time spend traveling. The frequent lack of electricity, clean aware of the tenuous existence its users may lead. Costs must
water, and safe, sanitary conditions makes staying healthy even be kept to a bare minimum, as the pennies charged even
more difficult. to send simple SMS messages may accumulate to become a
465
Disease/symptom % of patients % of doctor visits

serious burden. On the other hand, such systems also provide Tuberculosis 84 -
the opportunity to ease burdens by reducing the amount of Malaria 98.8 -
travel required to see doctors, or by increasing the ease and rashes/lesions 52.7 85
weight loss 60 85
affordability of communication with clinics. Furthermore, by pain while swallowing 26 81.2
helping to manage difficult medication regimens, a system like weakness 60 79.4
SmartTrack can improve the patient’s health and boost his or stiffness,tickling,numbness 41 -
Diarrhea 34 85
her ability to cope in a challenging world. shortness of breath 28.5 74.1
1) Regimen Compliance: Following a strict medication coughing up blood 9.4 84
regimen for HAART treatment is essential for survival. To TABLE I
evaluate the current state of adherence among our population, VARIOUS DISEASE & SYMPTOMS ; PERCENTAGE OF PATIENTS WHO HAVE
we asked about frequency of missed doses, reminder methods THESE DISEASES OR SYMPTOMS AND PERCENTAGE OF PATIENTS WHO
VIST A DOCTOR
for medications, and patient action upon missing a dose. 90%
of patients admitted to frequently missing dosages, a disturbing
figure given the risks associated with drug resistance against
ARVs. 75% of the patients of did not use any kind of external with family and friends. 70% felt that a shared cellphone
reminders, relying on habit alone to take their pills at the would still be useful for a healthcare application.
scheduled time. In the case of a missed dose, 70% reported • Frequency of calls & text messages: 75% make more than
taking their pills when they remember. one call a day. 22% receive text messages daily and 18%
Findings like these boldly underline the need for better send text messages daily.
adherence management. Cellphone-based tracking tools have • Cost: 45% of users spend less than $50 on their cellphone
the potential both to serve as reminder systems, and to alert handsets, and 69% spend less than $20 per month on
doctors to problems with compliance before they result in cellphone services.
tragic consequences. The adoption of mobile phone technology among HIV+ pa-
2) Response to drugs: In addition to adherence questions, tients in Ghana is quite widespread. This presents an outstand-
patients were asked about their perceived response to and tol- ing opportunity to bridge some of the gaps that have prevented
erance of ARV treatment, and their plans for future treatment. adequate healthcare from reaching those who most need it.
• 81.8% said that they felt much better after starting ARV Potential users have shown an interest and a willingness to try
medication. a system like SmartTrack.
• 98% said that the medication was controlling their HIV Patients are already quite reliant and comfortable with their
symptoms. phones, as shown by their frequent voice calls. In addition to
• 96% said that they had plans of continuing medication. the primary voice function, significant minorities of patients
Clearly, patients are happy with the results of treatment, use their phones for music and games, and a large proportion
and participants also expressed interest in a cellphone-based of respondents use them for clocks and alarms. Few people use
reminder system for taking medications, if one were made their phones for SMS texting, and any cost-effective telehealth
available to them. Perhaps with the right tools, patients would system would be based on SMS for the forseeable future.
be better able to manage their ARV treatments and achieve However, the low rate of text-messaging is presumably related
better outcomes. to the lack of literacy among users and language-barrier issues.
D. Mobile phone usage This would not be a problem for an application that used SMS
for data-transport, however, as the user would not be exposed
We asked a number of mobile-telephone-related questions to the underlying messages.
to better understand cellphone usage among the study popula- With a current adoption rate of 54% among our participants
tion and gauge the potential for telehealth applications using in Ghana, and a worldwide growth rate in mobile phone
mobile phones. adoption hitting 25% per year [24], cellphone-based telehealth
• Usage of cellphones: 54% of patients use cellphones and systems promise to become more and more feasible as time
the division of cellphone usage across various regions are goes on.
as follows: Greater Accra (77%), Northern Accra (67%)
and Volta(57%), which are urban areas rank higher than E. Diseases & symptoms
Central region(29%) and Upper West(26%), which are We also collected information on various symptoms, health
rural regions. In other uses of cellphones, only a small practices, secondary diseases, and conditions not related to
percentage of people used cellphones for playing games HIV. This was for the purpose of understanding symptom
and listening to music. 82% of users used cellphones as information that may be communicated by SmartTrack, as-
alarms. sessing the applicability of SmartTrack to conditions beyond
• Would a cellphone reminder be useful? 91% said that it HIV, and getting a sense of the participant’s general health
might be helpful in maintaining their HAART regimens. and medical habits.
• Sharing of phones: About 80% of patients who use cell- Symptom and disease information that we collected is
phones have their own phone and 16% share their phone summarized in Table I. These are symptoms and secondary
466
diseases which may be common to people suffering from it was much more expensive than SMS. Given that the
AIDS, and the percentages of patients who reported having information per patient is extremely limited, we felt that
these conditions at any point. The second column contains the SMS is a much cheaper alternative than voice or data
proportion of patients who traveled to see a doctor as a result services.
of the corresponding condition. These trips can be quite diffi- 3) The system can potentially be extended to other chronic
cult for patients, and might be avoided in certain cases if the illnesses like malaria, TB and Cholera.
doctor can consult with the patient remotely and recommend 4) It is critical to perform a needs-assessment study before
a course of action. Again, a cellphone-based health system performing the actual design. While this may seem
could enable doctors to track patient symptoms remotely and obvious in retrospect, doing a detailed needs assessment
act appropriately, contacting the patient if necessary. study is a tedious and time consuming process that is
often ignored.
VI. U SER R EACTIONS AND L ESSONS L EARNED
When we experimented with two different cellphone plat- VII. C ONCLUSIONS
forms with each user, the reactions were fairly varied across As Bill Easterly wrote in The White Man’s Burden, “2.3
users. Many users were completely new to the concept of trillion dollars of aid over 50 years and we have nothing much
smartphones and had difficulties navigating the screen. This to show for results.” [35] This so poignantly summarizes the
is not to rule out smartphones completely from such an current state of affairs of many philanthropic organizations
environment, but it is well known that rural populations are funding AIDS relief efforts in Africa. While it is indeed true
not accustomed to sudden changes. The cellphone has been that some organizations have made significant progress in
considered a user friendly device due to its similarity with specific areas, this is more of an exception than the norm.
the telephone. While many users found the voice option fairly Our effort is largely centered around technological issues
attractive, our voice interface was based on English which facing the adoption of AIDS relief schemes. We believe that,
was not understandable to over 50% of the users. The users a combination of technological and sociological approaches
we interviewed spoke several languages/dialects that made it are needed to tackle the healthcare(HIV/AIDS) problem in
hard for translators we used in our study to converse with developing regions.
local Ghanaians in rural areas. Another concern that patients From a technical perspective, to make AIDS relief efforts
expressed with voice was that a “talking” phone could disclose more effective, it is essential to improve accountability in the
the HIV status of a patient. health system, and to enhance patient adherence to medication
In contrast, users found pictograms very easy to understand. regimen through constant monitoring. The vision of Smart-
While we used only a limited set of universal healthcare Track is to address these two problems, using a distributed
symbols as pictograms in our study, it remains to be seen as to cellphone-based platform to achieve these goals. Though this
how many different types of questions can be phrased purely project is still in a relative state of infancy, it has undertaken
using pictograms. For example, to represent different drugs, a detailed needs-assessment study using interviews with over
we found that simple color and symbol coding would work 500 AIDS patients in Ghana. Our next step in future work
sufficiently well for several illiterate users. Overall, we felt is the actual deployment of SmartTrack with the aid of the
that a pictogram based user interface holds much promise and West Africa AIDS Foundation and Korle Bu Hospital. We
can also deal with the stigma issues that voice based interfaces plan to use the lessons learned from the needs-assessment
sometimes might present. study, with particular attention to the user interface issues,
The overall response to a system like SmartTrack was to deploy a widely acceptable version within Ghana to non-
very positive. The vast majority of patients (91%) seemed English-speaking and non-literate populations. If successful,
extremely positive and interested in a reminder system that this effort can be replicated in other rural developing parts of
could be placed on their cellphones. Typically, patients have the world.
to travel great distances to reach their treatment facilities;
ACKNOWLEDGMENT
hence, being able to remotely communicate with the doctor
is a huge positive factor for most patients. Social workers, We would like to thank Korlebu Hospital and the West
nurses, community health workers, and doctors seemed very Africa AIDS Foundation for enabling us to do the detailed user
positive and amenable to the system. Many health-workers study and Microsoft Corporation for supporting this project.
also wanted a similar system for other common diseases like R EFERENCES
malaria, TB and Cholera.
[1] “World Health Organization,” http://www.who.int/3by5/en/.
Other important lessons we learned include: [2] “HIV and AIDS in Africa,” http://www.avert.org/aafrica.htm.
1) From a cost perspective, smartphones are much more [3] “Treatment of HIV infection,” http://www.niaid.nih.gov/
expensive than cellphones. In Ghana, users can purchase factsheets/treat-hiv.htm.
unlocked cellphones at extremely low cost, thus we felt [4] “Upwardly mobile in Africa-Business Week,” http://www.
businessweek.com/globalbiz/content/sep2007/gb20070913
the system should be built around people’s cellphones. 705733.htm.
2) From a communication perspective, many rural areas we [5] “Africa Telecom News,” http://whiteafrican.com/2008/08/01/
surveyed did not have GPRS coverage. Even if available, 2007-african-mobile-phone-statistics.
467
[6] “Low cost cellphones,” http://www.cameraphonesplaza.com/ [23] B. Rapchak, “Cell phone technology to enhance medication
new-low-cost-nokia-cell-phones/. adherence and therapeutic outcomes,” in TEPR 2009, Medical
[7] “Open Handset Alliance,” http://www.openhandsetalliance.com. Records Annual Meeting.
[8] “Symbian OS: the open mobile operating system,” http://www. [24] K. Kalba, “The adoption of mobile phones in emerging markets:
symbian.com/. Global diffusion and the rural challenge,” International Journal
[9] “Android - An Open Handset Alliance Project,” http://code. of Communication, vol. 2, Jun. 23 2008. [Online]. Available:
google.com/android/. http://ijoc.org/ojs/index.php/ijoc/article/view/216
[10] “Leap of faith technologies, inc.” [Online]. Available: http: [25] “Satellife: The Global Health Information Network,” http://
//www.leapoffaith.com/ www.healthnet.org.
[11] R. Scheier, “African graft stings donors,” The Christian Science [26] B. DeRenzi, N. Lesh, T. Parikh, C. Sims, W. Maokla,
Monitor, Jun. 1 2006. M. Chemba, Y. Hamisi, D. S. hellenberg, M. Mitchell, and
[12] C. Mangwiro, “Mozambique links health officials to drug G. Borriello, “E-imci: improving pediatric health care in low-
thefts,” Reuters, Aug. 28 2007. income countries,” in CHI 2008. ACM, pp. 753–762.
[13] F. Phiri, “Drug theft poses threat to battle against TB,” Inter [27] I. Medhi, A. Sagar, and K. Toyama, “Text-free user interfaces
Press Service News Agency, Mar. 24 2004. for illiterate and semiliterate users,” in ITID ’07: Information
[14] “Counterfeit medicines fact sheet,” World Health Organization, Technologies and International Development, 2007, pp. 37–50.
Nov. 14 2006. [Online]. Available: http://www.who.int/ [28] “Evaluation of the Satellife PDA Project, 2002: Testing the use
[15] P. Sax, “Patient adherence and response to therapy,” in AIDS of handheld computers for healthcare in Ghana, Uganda, and
Clinical Care August 1, 1999. Kenya,” http://bridges.org.
[16] E. J. Mills, J. B. Nachega, I. Buchan, J. Orbinski, A. Attaran, [29] S. Anand and U. Rivett, “Ict in the management of hiv treat-
S. Singh, B. Rachlis, P. Wu, C. Cooper, L. Thabane, ment: Cell-life: a south african solution,” vol. 6, pp. 56–59, 2004.
K. Wilson, G. H. Guyatt, and D. R. Bangsberg, “Adherence [30] “Cell-Life,” http://www.cell-life.org/.
to Antiretroviral Therapy in Sub-Saharan Africa and North [31] “Java Micro Edition,” http://java.sun.com/javame/index.jsp.
America: A Meta-analysis,” JAMA, vol. 296, no. 6, pp. [32] “West Africa Aids Foundation,” http://www.waafweb.org/.
679–690, 2006. [Online]. Available: http://jama.ama-assn.org/ [33] C. Bell, S. Devarajan, and H. Gersbach, “The long-run
cgi/content/abstract/296/6/679 economic costs of AIDS : Theory and an application to
[17] D. Werner, C. Thurman, and J. Maxwel, Where there is no south africa,” The World Bank, Policy Research Working
doctor? Paper Series 3152, Oct. 2003. [Online]. Available: http:
[18] “OpenMRS,” http://www.openmrs.org. //ideas.repec.org/p/wbk/wbrwps/3152.html
[19] “Partner in Health,” http://www.pih.org. [34] “2007 AIDS epidemic update,” UNAIDS, Tech. Rep., Dec.
[20] “Dimagi,” http://www.dimagi.com. 2007. [Online]. Available: http://data.unaids.org/pub/EPISlides/
[21] “D-Tree,” http://www.d-tree.org. 2007/2007 epiupdate en.pdf
[22] W. Kaplan, “Can the ubiquitous power of mobile phones be [35] W. R. Easterly, The white mans burden : why the Wests efforts
used to improve health outcomes in developing countries?” in to aid the rest have done so much ill and so little good. New
Globalization and Health 2006. York: Penguin Press, 2006.
468
Uses of Mobile Phones in Post-Conflict Liberia

Michael L. Best‡∗ , Edem Wornyo† Thomas N. Smyth∗ and John Etherton‡
∗ School
Interactive Computing
† Schoolof Electrical and Computer Engineering
‡ Sam Nunn School of International Affairs
Georgia Institute of Technology

Atlanta, Georgia 30332–0250
{mikeb, tsmyth3, jetherton}@cc.gatech.edu
edem@gatech.edu
Abstract—Liberia is a country emerging from years of pro-

tracted and devastating civil conflict. Left without any fixed line Fixed Line Teledensity
telephone infrastructure, it relies solely on the mobile phone for 0.5 
Fixed Lines per 100 
telephony. This study investigates the usage of mobile phones 0.4 
Inhabitants 
in this immediate post-conflict setting. In particular, we adopt 0.3 
the uses and gratifications approach to media research, giving 0.2 
focus to both instrumental and intrinsic motivations for use. 0.1 
Mobile phone users in both the capital city of Monrovia and 0 
in various rural areas were surveyed using the Q methodology,
1989 
1995 
1998 
2001 
2007 
1980 
1983 
1986 
1992 
2004 
which identified distinct perspectives within these urban and
rural groups. Participants were then sorted into groups where
Year 
each group contained users with similar perspectives on their
mobile phones. These identified groups included sets of users who
saw their phones as productivity enhancers, means of connectivity (a)
to family and friends, essential business tools, technological
curiosities, and sources of personal security. The idea of a phone Mobile Teledensity
as a stylish object was markedly rejected, especially in rural
areas. We contrast these Q-sort results from Liberia with previous 16 
Subscribers per 100 
work from Kigali, Rwanda, finding differences especially as 12 

Inhabitants 
related to security. 8 
I. I NTRODUCTION 4 
Considerable attention has been given to the role of in- 0 
1989 
1995 
1998 
2001 
2007 
1980 
1983 
1986 
1992 
2004 
formation and communication technologies as tools for de-
velopment within Africa, and increasing levels of excitement
Year 
have concentrated on the use of mobile phones. With some
fanfare the Economist [1] announced that the “real digital
(b)
divide” was in terms of the differential access to mobile
telephones while computers and the Internet were of less use. Fig. 1. Fixed-line and mobile teledensity in Liberia from 1980–2007. Steep
Many writers have disagreed with their pessimistic assessment drops in fixed lines are evident in 1991 and 2003. Introduction of competing
mobile carriers in the mid 2000’s produced a soaring number of mobile
of computers and the Internet (e.g. [2]). Nonetheless, it is subscribers. Note the difference in vertical scale between the two charts.
clear that mobile telephones are playing a substantial and
important role in development within the global south. Indeed,
compelling evidence of the macro and microeconomic effect of
reports that average year-on-year growth rate for mobile phone
mobile phones in low-income countries has been mounting [3].
subscribers in Sub Saharan Africa from 1999-2004 was double
For instance Waverman et al. [4] find that mobile phones offer
what it was in Europe. Indeed Sub Saharan Africa is a
a significant macroeconomic growth dividend and one that is
continent driven by mobile telephony and in 2001 the total
“twice as large in developing countries compared to developed
number of mobile subscribers exceeded the number of fixed
countries”. Microeconomic benefit is also evident. For instance
line subscribers [6]. In 2004, the mobile teledensity across
Jensen [5] shows that mobile phone use by farmers in Southern
all of Africa was 9.1, with the vast majority, 87%, making
India increases productivity, enhances revenues, reduces waste,
use of prepaid cards. Considering only Sub Saharan Africa
and lowers consumer prices.
the mobile teledensity is best approximated at 6.2% [7]. And
A. Mobile phones in Africa while this number describes subscriber penetration it does not
Mobile phone penetration growth rates are today highest give an adequate sense of overall access and usage due to
in Africa compared to all other continents [6]. The ITU widespread sharing of phone subscriptions. Clearly, mobile
469
telephony is the central communication technology for much Furthermore, retrospective empirical scholarship has
of Sub Saharan Africa. demonstrated the critical nature of communication amongst
the people of a nation if there is to be a lasting peace instead
B. Mobile phones in Liberia of an all-too-frequent return to civil conflict [10]. Modern
Liberia, established as a state in 1847 by freed African information and communication technologies can therefore,
slaves from the U.S.A., is situated on the Atlantic coast of on their face, serve as tools in this process of national
West Africa with Sierra Leone, Guinea, and Côte d’Ivoire as reconciliation if they are ably applied to these communication
bordering countries. A relatively small country with approxi- activities. However, to understand what it would mean to
mately 3.3 million inhabitants, it is attempting to right itself “ably” apply modern ICT’s, including mobile telephony,
after decades of civil conflict. to the process of post-conflict development requires at a
Unrest has been a staple within Liberia for more than minimum an adequate understanding of the current uses and
15 years with two major civil wars (1989-1996 and 1999- meanings of mobile telephony in that environment. Such was
2003) in this time period. These years of conflict have seen the motivation for this study.
nearly one-third of the population displaced and taken the
lives of approximately 250,000 people. A peace was brokered II. U SES AND G RATIFICATIONS
and transitional government was established in 2003. A UN In seeking to uncover the everyday, micro-level motivations
peacekeeping mission was positioned to keep this peace, and for mobile phone use among Liberians, this study draws
democratic elections were held in the fall of 2005. This inspiration from the uses and gratifications (U&G) research
resulted in the selection of Africa’s first elected female head tradition. U&G as an approach originated in communications
of state, President Ellen Johnson-Sirleaf. research in the mid-1970’s [11], advancing the view that
An outcome of these years of civil conflict was the complete consumers of mass media make active choices and selectively
destruction of the fixed-line telephone infrastructure. The consume media in order to satisfy specific needs. More tersely,
copper network was wholly destroyed or looted and all but U&G can be said to focus on what people do with media, as
one switch was destroyed [8]. It is clear from Figure 1(a) the opposed to what media does to people [12]. More recently,
steady decline in mainline penetration, starting from a very low the U&G approach has been applied to study adoption and
level to begin with. The precipitous drops evident in 1991 and use of new media technologies, including telephones [13], the
2003 are the outcome of the two major civil wars such that Internet [14], and mobile phones [15]. Also notable about the
by the time of the establishment of peace all mainlines where U&G tradition is attention to a broad range of motivations,
gone. including those which go beyond the purely instrumental or
On the other hand, mobile telephone adoption in Liberia has utilitarian (such as increased productivity or personal safety)
recently been growing at a staggering rate, as shown in Figure to the intrinsic, social, or to quote McClatchey [16], ‘hedonic’
1(b). It has been shown that teledensity phone penetration rates motivations for use.
are likely to over count the number of actual subscribers (due A typical U&G based study proceeds in one of two ways:
to purchased but inactive accounts) and, as already mentioned, either by starting with a hypothetical set of possible uses
significantly undercount the number of actual users (due to and seeking to confirm or deny each one, or in a more
sharing) [7]. In Liberia we estimate the subscribers to users exploratory fashion, starting off with no such initial set. As will
ratio to be as high as one to five. Competition within the be seen, our study walks a line between these two alternatives.
Liberian mobile phone sector is also robust with four active However, it must be noted that U&G in itself is not a method.
operators. Indeed, usage costs are reportedly the lowest in West Indeed, previous studies have employed a variety of different
Africa [34]. All county capitals and most other population methods to investigate uses, including surveys [17], semi-
centers currently receive signal from at least one of the mobile structured interviews [13], and focus groups [18].
providers’ services, and two providers currently offer GPRS A recent investigation of mobile phone usage among mi-
mobile internet services. Operators are actively extending both croentrepreneurs in Kigali, Rwanda [15] also drew upon the
their networks and services. U&G approach. In using the Q-sort methodology (also used in
The striking success of Liberia’s mobile sector, which this study and described in the next section), Donner identified
continues to develop at a feverish pace despite the resource- four archetypal ‘factors’ which speak to predominant uses of
strapped country, is a cause for optimism. mobile phones in Kigali. They were: convenient, intrinsic,
indispensable, and productive. Donner remarked on the diverse
C. Mobile phone usage in post-conflict settings nature of those factors, saying that they ‘’‘suggest numerous
Regrettably, civil conflicts such as those experienced in paths for future research.”
Liberia are not unusual in contemporary times. Indeed, while Our research is intended as an extension of this body of
inter-state wars are increasingly less common, the incidence research on uses of mobile phones to an immediate post-
of civil conflict is on the rise [10]. Thus the study of ICT’s conflict context. To our knowledge, ours is the first study
within countries emerging from civil conflict is an area of of mobile phone uses and gratifications in such an environ-
considerable importance though we note a paucity of work in ment. We believe that this context may give rise to unique
this area [9]. motivations for use, especially given the vibrancy of Liberia’s
470
mobile sector as described above, and the obvious importance

of communication to the task of rebuilding a nation.
III. M ETHODS
A. The Q-Sort Methodology
The Q-sort method was employed to gain insight into the
nature of mobile phone use in post-conflict Liberia. In this
section, the concept of the Q-sort methodology is briefly
explained. However, this paper does not present an in-depth
treatment, as many relevant and well written expositions and
examples of the Q-sort methodology are available elsewhere. Fig. 2. An example Q-Sort, demonstrating the quasi-normal pattern into
The Q-sort methodology, which evolved from factor theory, which statement cards are sorted. In practice, the full statement is printed on
was originally developed by the British physicist William the front of the card, while a reference number is printed on the back. When
the sort is complete, the cards are flipped and the pattern is recorded.
Stephenson for psychological studies [19]. Despite earlier
criticism of the technique in the academic community, the
technique has gained increasing attention and acceptance as
a tool for research in many areas from psychology [20], to to recall that they have enjoyed an additional nine years of
medicine [25], communication [29], social sciences [30], and relative peace. Do those additional years explain some of the
education [32]. inter-state variation we have observed?
In a Q-sort study, a subject is asked to arrange a set of state-
ments (such as those shown in Table 1), pictures, or sounds, B. Protocol
according to some perceptual metric. In most Q-sorts, the
individual is requested to place a statement into one of the slots The study required that participants arrange a set of state-
in a grid akin to Figure 2. This grid is designed to describe ments as listed in Table I according to how these statements
a quasi-normal distribution. Each column along the grid is best describe their use of mobile phones. The statements were
given a relative position along some semantic differential, for printed on flash cards for easy handling. The participants were
instance from strongly disagree to strongly agree. The ordering advised to arrange the statements initially into three piles:
of the statements in each column is irrelevant—only the lateral “Describes me best”, “Neutral”, and “Describes me least”, and
ordering carries meaningful information. Some studies use a then to sort the piles into the appropriate categories in a quasi-
rectangular grid pattern as opposed to a quasi-normal one. normal format similar as shown in Figure 2. In addition to the
We chose the latter as we believe it forces the participant to statements, demographic information was requested from the
think deeply in choosing the strongest points of agreement and respondents. The time taken for the exercise ranged from 30
disagreement. to 60 minutes per participant. The statements were in English.
Data analysis in the Q-methodology establishes groups of
individuals who sort particular traits in common places within C. Participant Selection
the distribution. For example, consider a group of teenage
mobile phone users in Tokyo who place great weight on their Unlike traditional quantitative survey techniques, Q-sorts
connections with friends and the stylish elements of their can be carried out with a relatively small number of par-
phone but who sort customer and work connections as low ticipants from a population space. We sought participants
in importance; this group of people might be detected as a from both Monrovia, Liberia’s capital, and from various rural
“factor” within the Q-sort methodology. Thus, Q-sort is said areas throughout the country. This is in contrast to Donner’s
to be a person-oriented approach as opposed to a trait-oriented study which focused entirely on the capital of Kigali. In total,
tool [33]. 63 participants were interviewed in Monrovia. Participants
Previous work by Jonathan Donner [15] applied the Q- were chosen at random from passers-by on a street corner in
methodology to the study of mobile phone use among mi- downtown Monrovia. Fourteen responses were discarded due
croentrepreneurs in Kigali, Rwanda. The present study is to incompleteness, leaving a total of 49. Outside Monrovia,
intended as an extension of that work, studying the case of a total of 36 participants were selected, also at random from
mobile phone use in Liberia. In order to enable comparisons busy areas. In total, we visited 13 towns and villages in several
between the two studies we have used the same set of state- Liberian counties. The age of the respondents ranged from 19
ments as was used in Rwanda (with only minor modifications). to 62 years. Participants were given a US$5.00 mobile phone
Both Rwanda and Liberia have emerged from recent civil scratch card for their efforts, whether or not they completed
conflict with Rwanda embarking on a path to peace starting the sort.
in 1994 while Liberia saw conflict through till 2003. This Print literacy was a requirement for participation. In ques-
study, therefore, examines how people perceive their mobile tionable cases, prospective participants were asked to read one
phones after only a few years of peace. And when we compare of the statements from a flash card and describe its meaning
those perceptions to Donner’s study from Rwanda it is helpful before they were admitted to the study.
471
TABLE I TABLE II
Q-S ORT S TATEMENTS FACTOR C HARACTERISTICS
Concept Statement F1 F2 F3 F4
Connectivity I use my mobile phone to stay in touch with my cus- Urban # of Defining Participants 9 13 12 4
tomers. % of Variance Explained 10% 13% 11% 7%
I use my mobile phone to stay in touch with my suppliers. Rural # of Defining Participants 13 8 5 6
My phone gives me access to new customers. % of Variance Explained 20% 12% 11% 12%
I use my mobile phone to stay in touch with my family.
My mobile phone helps me come and go without worry-
ing about missing calls.
components analysis to identify initial factors within the data,
I use my mobile phone to stay in touch with my friends.
followed by a varimax rotation to arrive at the final set of
Information My mobile phone helps me find work.
factors.
My mobile phone helps me keep informed about prices
in my business. Each such factor can be thought of as an archetypal per-
Intrinsic Having a mobile phone makes me feel more important. spective; a sorting of the statements that defines one group
Having a mobile phone makes me feel more connected of subjects against the others. Once these factors have been
to the world. identified, a loading score is computed for each partici-
I like customizing my mobile phone with accessories like pant/factor combination, which measures the similarity of that
special sounds and carrying cases
participant’s perspective to the archetypal perspective of that
I enjoy talking to my friends and family on my mobile.*
factor. A participant is said to load on or define a factor if
Having a mobile phone makes me happy.
their loading score for that factor crosses a certain threshold.
My mobile phone is stylish.
As a result of this process, each factor is associated with a set
Productivity My business is easier now that I have a mobile phone.
of participants defining it. In a sense the set of participants
My family is better off because I have a mobile phone.
have now been clustered into a small number of factors (we
My mobile phone saves me time.
find four factors in our study) with each participant assigned
My mobile phone lets me get more done during the day.
to that factor that best represents them.
My mobile phone helps my business save money.
In the final step, the statement rankings for each participant
My mobile phone helps me make more money in a day.
are weighted according to that participant’s loading score for
Security I use my mobile phone for emergency calls.
the factor they are assigned to (therefore if they define the fac-
My mobile phone makes me feel more secure.
tor more closely their ranking will have more weight). Then all
Other Getting a mobile phone changed the way I do business.
I am interested in learning about new features or mobile
of these weighted ranks are combined among the participants
phone models. assigned to each factor such that each factor is then described
I can’t do business without my mobile phone. by a list of single Z-scores, one per statement, along with a
I was among the first of my friends and business asso- p value assessing the value’s statistical significance. Finally,
ciates. these Z-scores are re-projected back into the original space of
I give my mobile phone number to many people. values from -4 to +4 (from “describes me least” to “describes
I share my mobile phone with my family or friends. me best”) as shown in Figure 2. And as an aid to comparison
I keep my mobile phone with me at all times. between the groups each set of Z-scores are projected into the
My mobile phone gives me more control over who I talk space described by the other factors as well.
to, and how/when I talk to them.
PQMETHOD also determines for each a factor a set of
I use my phone more for business more than for social
calls. “distinguishing statements” which differentiate the factor from
* Due to a miscommunication, in the urban study, this statement was the others. These statements are of special importance as they
replaced with “I bought my mobile phone for business.” This change are most representative of the differences between the factors.
was incorporated into the analysis that follows. It is in examining these representative statements that insight
into the meaning of each factor can finally be gained.
Tables III and IV, which we will go on to study below, show
D. Analysis Methods these sets of factors and their most distinguishing statements,
Q-analysis is usually performed using PQMETHOD, a along with those statements’ Z-scores and the -4 to +4 values
software package developed specifically for the task. A typical associated with them.
Q-analysis involves several steps. Initially, a large correlation
matrix is created, describing the similarities between the Q- IV. R ESULTS
sorts of all pairs of participants. We then look for ways to The Q-sort data we obtained from urban and rural par-
reduce the information in this matrix into an interpretable ticipants in Liberia have been analyzed separately. This has
form, a process which is both iterative and partially subjective. allowed us to examine differences in mobile phone use and
There are several routes to this goal, a review of which goes perception between these populations.
beyond the scope of this work. In our analysis we chose a Following the procedures described above, PQMETHOD
procedure similar to Donner [15]; we performed a principal was used to perform our analysis. After the principal com-
472
Commonalities Factor 3: Business

Statement F1 F2 F3 F4 Statement Z F1 F2 F3 F4
I use my mobile phone for emergency calls. 2 1 2 4 Describes Me Best
∗∗I can’t do business without my mobile phone. 1.75 -3 3 4 -2
Factor 1: Productivity ∗My phone gives me access to new customers. 1.49 1 -2 3 -1
Statement Z F1 F2 F3 F4 ∗∗I bought my mobile phone for business. 1.19 -1 -2 3 -1
Describes Me Best Other Distinguishing (Relatively High) Statements
∗∗My mobile phone helps me make more money in 1.57 3 -4 0 -4 ∗∗My mobile phone helps me keep informed about 0.83 1 -2 2 -1
a day. prices in my business.
Other Distinguishing (Relatively High) Statements ∗My mobile phone makes me feel more secure. 0.21 -1 2 1 -1
∗∗I use my mobile phone to stay in touch with my 0.8 2 4 -3 -4 ∗I give my mobile phone number to many people. 0.14 -4 2 0 3
friends. ∗I am interested in learning about new features or -0.15 -3 1 0 -2
∗∗My mobile phone lets me get more done during 0.49 1 -4 -2 3 mobile models.
the day. ∗∗My mobile phone helps me make more money in -0.4 3 -4 0 -4
∗My mobile phone helps me keep informed about 0.19 1 -2 2 -1 a day.
prices in my business. Other Distinguishing (Relatively Low) Statements
∗∗My family is better off because I have a mobile -0.08 0 1 -2 -2 ∗∗My mobile phone lets me get more done during -0.78 1 -4 -2 3
phone. the day.
Other Distinguishing (Relatively Low) Statements ∗My mobile phone saves me time. -0.82 0 1 -2 2
∗∗Having a mobile phone makes me feel more 0.17 0 2 2 3 Describes Me Least
connected to the world. ∗∗I use my mobile phone to stay in touch with my -0.91 4 4 -3 2
∗My mobile phone saves me time. -0.36 0 1 -2 2 family.
Describes Me Least ∗∗Having a mobile phone makes me feel more -1.98 -1 -2 -4 0
∗∗I give my mobile phone number to many people. -1.27 -4 2 0 3 important.
∗∗My mobile phone is stylish. -1.49 -4 -2 -1 1 Distinguishing statements: ∗p < .05, ∗∗p < .01
Distinguishing statements: ∗p < .05, ∗∗p < .01
Factor 4: Security
Factor 2: Connectivity Statement Z F1 F2 F3 F4
Statement Z F1 F2 F3 F4 Describes Me Best
Describes Me Best ∗∗I use my mobile phone for emergency calls. 2.31 2 1 2 4
∗∗I use my mobile phone to stay in touch with my 2.18 2 4 -3 -4 ∗I keep my mobile phone with me at all times. 1.93 0 3 0 4
friends. ∗∗My mobile phone lets me get more done during 1.57 1 -4 -2 3
∗I keep my mobile phone with me at all times. 1.29 0 3 0 4 the day.
Other Distinguishing (Relatively High) Statements ∗I give my mobile phone number to many people. 1.16 -4 2 0 3
∗∗My mobile phone makes me feel more secure. 0.8 -1 2 1 -1 Other Distinguishing (Relatively High) Statements
∗∗Having a mobile phone makes me happy. 0.78 0 2 -1 -3 ∗∗I use my mobile phone to stay in touch with my 1.09 4 4 -3 2
∗I give my mobile phone number to many people. 0.61 -4 2 0 3 family.
∗∗My family is better off because I have a mobile 0.5 0 1 -2 -2 ∗∗My mobile phone helps me come and go without 0.85 -2 -1 -2 2
phone. worrying about missing calls.
∗I like customizing my mobile phone with acces- -0.29 -1 0 -1 -2 ∗I use my phone more for business more than for 0.63 3 -2 3 2
sories like special sounds and carrying cases. social calls.
Other Distinguishing (Relatively Low) Statements ∗∗My mobile phone is stylish. 0.16 -4 -2 -1 1
∗I use my mobile phone for emergency calls. 0.31 2 1 2 4 Other Distinguishing (Relatively Low) Statements
∗∗My mobile phone helps me come and go without -0.3 -2 -1 -2 2 ∗∗I use my mobile phone to stay in touch with my -0.01 2 3 1 0
worrying about missing calls. customers.
∗My mobile phone helps me keep informed about -1.15 1 -2 2 -1 ∗My mobile phone helps me keep informed about -0.51 1 -2 2 -1
prices in my business. prices in my business.
Describes Me Least ∗∗My phone gives me access to new customers. -0.95 1 3 3 -2
∗∗I use my phone more for business more than for -1.33 3 -3 3 2 Distinguishing statements: ∗p < .05, ∗∗p < .01
social calls.
∗∗My mobile phone helps my business save money. -1.43 1 -3 1 1
∗∗My mobile phone lets me get more done during -1.6 1 -4 -2 3
the day.
TABLE III
U RBAN FACTORS
473
Commonalities Factor 3: Intrinsic

Statement F1 F2 F3 F4 Statement Z F1 F2 F3 F4
Having a mobile phone makes me feel more con- 3 3 2 3 Describes Me Best
nected to the world. ∗∗Having a mobile phone makes me happy. 1.4 -2 -2 3 0
I use my mobile phone for emergency calls. 4 3 2 1 Other Distinguishing (Relatively High) Statements
I keep my phone with me at all times. 2 2 2 4 ∗I am interested in learning about new features or 0.74 -1 1 2 0
My mobile phone is stylish -4 -4 -4 -3 mobile models.
∗∗I like customizing my mobile phone with acces- 0.3 -3 -3 0 -2
Factor 1: Business sories like special sounds and carrying cases.
Statement Z F1 F2 F3 F4 Other Distinguishing (Relatively Low) Statements
Describes Me Best ∗∗My business is easier now that I have a mobile -0.79 4 0 -2 2
∗∗My business is easier now that I have a mobile 1.66 4 0 -2 2 phone.
phone. ∗∗My mobile phone helps me find work. -1.03 0 0 -2 3
Other Distinguishing (Relatively High) Statements Describes Me Least
∗∗My mobile phone helps my business save money. 0.77 2 -1 0 0 ∗∗I give my mobile phone number to many people. -1.97 -1 -2 -4 -1
Other Distinguishing (Relatively Low) Statements Distinguishing statements: ∗p < .05, ∗∗p < .01
∗∗I use my mobile phone to stay in touch with my 0.15 0 4 4 2
family. Factor 4: Mixed
∗∗I use my mobile phone to stay in touch with my -0.57 -2 1 0 3 Statement Z F1 F2 F3 F4
friends. Describes Me Best
Describes Me Least ∗∗I keep my mobile phone with me at all times. 1.68 2 2 2 4
∗I was among the first of my friends and business -1.01 -3 0 -3 -4 ∗∗My mobile phone helps me find work. 1.42 0 0 -2 3
associates to get a phone. ∗∗I use my mobile phone to stay in touch with my 1.23 -2 1 0 3
∗∗My mobile phone gives me more control over who -1.53 -3 0 0 -2 friends.
I talk to, and when I talk to them. Other Distinguishing (Relatively High) Statements
∗∗I share my mobile phone with my family or -1.83 -4 1 -2 0 ∗I use my mobile phone to stay in touch with my 1.09 0 4 4 2
friends. family.
Distinguishing statements: ∗p < .05, ∗∗p < .01 ∗∗I enjoy talking to my friends and family on my 1.03 -1 4 -1 2
mobile.
Factor 2: Mixed ∗∗My business is easier now that I have a mobile 0.92 4 0 -2 2
Statement Z F1 F2 F3 F4 phone.
Describes Me Best ∗∗I use my mobile phone to stay in touch with my 0.67 4 -2 4 1
∗∗I enjoy talking to my friends and family on my 2 -1 4 -1 2 customers.
mobile. ∗Having a mobile phone makes me happy. -0.06 -2 -2 3 0
∗∗My mobile phone makes me feel more secure. 1.57 1 3 1 -2 Other Distinguishing (Relatively Low) Statements
∗My mobile phone saves me time. 1.21 0 3 1 -3 ∗∗I use my phone more for business more than for -0.23 3 2 2 0
Other Distinguishing (Relatively High) Statements social calls.
∗I share my mobile phone with my family or friends. 0.34 -4 1 -2 0 ∗My mobile phone helps me come and go without -0.3 -2 -3 -3 -1
∗∗I was among the first of my friends and business -0.07 -3 0 -3 -4 worrying about missing calls.
associates to get a phone. ∗∗My mobile phone makes me feel more secure. -0.85 1 3 1 -2
Other Distinguishing (Relatively Low) Statements Describes Me Least
∗∗My business is easier now that I have a mobile -0.06 4 0 -2 2 ∗∗My mobile phone saves me time. -1.32 0 3 1 -3
phone. ∗My mobile phone lets me get more done during the -1.65 -1 -1 -2 -4
∗My phone gives me access to new customers. -0.11 1 0 1 2 day.
∗∗I use my mobile phone to stay in touch with my -0.68 4 -2 4 1 Distinguishing statements: ∗p < .05, ∗∗p < .01
customers.
TABLE IV
RURAL FACTORS
ponent analysis, several factor rotations were computed and with ranks for the present factor of interest of +3 or +4, or
examined. For both data sets, a set of four factors was found -3 or -4, are placed under the headings ‘Describes Me Best’
to provide the best balance between explanatory power and and ‘Describes Me Least’, respectively. Other distinguishing
succinctness. statements with a high (or low) rank for the present factor
Table II displays the number of participants found to load relative to the other factors are placed under the heading
on each factor, as well as the percentage of initial variance ‘Relatively High’ (or ‘Relatively Low’).
that each explains. Each factor can be taken to represent an
archetypal perspective regarding phone use among Liberians. Below, we review the distinguishing statements for each
The explained variance proportions we obtained are similar to factor in an effort to interpret the nature of the archetypal
those obtained in previous studies. perspectives they represent. We also review commonalities
Tables III and IV show the commonalities across factors for across factors for the urban and rural groups. These common-
both datasets, and the statements distinguishing each factor, alities are statements which are consistently ranked positively
ordered by Z-score. The computed ranks for each factor are or negatively for each of the four factors, and thus indicate
also shown for each statement (F1, F2, F3, and F4). Statements agreement across most participants.
474
A. Urban a day (-2). Furthermore, they do not see the phone as a

Commonalities across factors: The urban group exhibited tool for connectivity with their family (-3), or as a stature
few points of commonality across factors. The only statement booster (-4). Taken together, this set of statements indicates a
which was ranked positively for all four factors was “I use business-minded group of individuals, for whom the phone is
my mobile phone for emergency calls,” which was rated +2, an essential tool. And indeed from our demographic notes,
+1, +2, and +4. No statements received consistently negative this group comprises business people in areas such as the
rankings for all four factors. This lack of consensus could be oil industry, computer technology, money exchange (forex
due to the varied sample of participants and their divergent bureau), and others.
viewpoints. Factor 4: Security: This group is defined by four partic-
Factor 1: Productivity: The group of nine participants ipants, who assign the highest rank of +4 to the statements
defining this factor was comprised mainly of business owners “I use my phone for emergency calls” and “I keep my
with less than 12 employees, except for a computer technology phone with me at all times”. This suggests that the safety
facility owner who had 30 employees. of themselves or their loved ones is of primary concern to
This factor describes responses where people perceive the members of this group. The group also views the mobile as a
statements “I use my mobile phone to stay in touch with my productivity enhancer given the relatively high rank assigned
friends” and “My mobile phone helps me make more money in to the statements “My mobile phone helps me come and go
a day” as rather important where the normalized Z-scores are without worrying about missed calls” (+2) and “My mobile
relatively high (+3 and +2). In contrast the prestige or intrinsic phone lets me get more done during the day” (+3). Further,
statement “My mobile is stylish” was rated quite low. the group ranked the statements: “I use my mobile phone to
It is clear that this particular group does not view the stay in touch with my family” (+2), and “My mobile phone is
mobile phone as an object of style (-4) nor do they give their stylish” (+1) as relatively important, while the statement “My
mobile phone numbers to many other people (-4). Instead, phone gives me access to new customers” (-2) was seen as
affiliates to Factor 1 recognize most significantly the business relatively unimportant.
aspect of mobile telephony, followed by the security utility Thus this group has elements of connectivity oriented users
that the phone provides. This group, therefore, emphasized the as well as gratification since they use phones to stay in touch
importance of the mobile phone as a business tool rather than with family and find their phone stylish. Furthermore, they per-
a fashion item. They also emphasized how the phone allows ceive the phone as offering some productivity enhancements
them to stay in touch with their friends (+2), perhaps as they such as getting more things done and staying in touch with
go about conducting their business. customers. However, what is uniquely distinguishing about this
Overall, this group seems to see the phone as a productivity factor is its emphasis on security with the top two statements
tool. associated with this factor concerned with this issue.
Factor 2: Connectivity: Thirteen participants affiliate
strongly with this factor. This group of respondents has a B. Rural
demographic makeup of sole proprietors and other business Commonalities across factors: One of the most striking
owners with less than three employees, as well as several things about the rural study was that during the initial sort
miscellaneous others. For this group, communication with of the cards into three piles, most participants sorted most
friends is essential (+4), as is constant availability, as they cards into the ‘describes me best’ pile. The second stage of
admit to carrying their phone at all times (+3), and giving the sort then became a difficult exercise in prioritization. Many
their number to many people (+2). The phone also makes participants were visibly torn on which statements to promote
them feel happy (+2), secure (+2), and makes their family to the highest levels and which to leave behind.
better off (+1). On the other hand, this group rated statements Nonetheless, in contrast to the urban data which had few
related to business and productivity rather lowly, saying that pervasive commonalities, several statements emerged as items
they do not use the phone to find out about prices (-2), they of consensus for most participants. All four factors agreed
do not use their phone more for business calls (-3), and they that their mobile phones make them feel more connected to
do not see the phone as enabling them to accomplish more the world. That statement was ranked +3, +3, +2, and +3,
in a day (-4). In sum, it seems that members of this group respectively. Several participants spoke of communicating with
are primarily concerned with being available and in touch family members in other countries in Africa and in the West
with their family and friends, and that they derive feelings as justification for their high rating of this statement.
of pleasure and security from that high level of connectivity. There was also widespread agreement on the phone not
Factor 3: Business: The 12 participants defining this factor being an object of fashion. The statement “My phone is
bought a phone for their business (+3), use the phone to stylish” received rankings of -4, -4, -4, and -3. During the
gain access to new customers (+3) and stay informed about study, many participants openly scoffed at this statement upon
prices (+2), and in general, feel that they can’t do business reading it.
without their phone (+4). On the other hand, perhaps due Finally, while no distinct security or safety factor is iden-
to the centrality of the phone in their business routine, they tified in the rural data, there was nonetheless widespread
do not feel that the phone lets them get any more done in reliance upon the phone for emergency use. The statement “I
475
use my phone for emergency calls” was ranked +4, +3, +2, and I use my mobile phone for emergency calls
+1, while “I keep my phone with me at all times” was ranked F1 F2 F3 F4
+2, +2, +2, and +4. It was clear that many participants felt Urban Liberia 2 1 2 4
strongly about this function of the phone. Several participants Rural Liberia 1 2 3 4
offered compelling stories of using their phone to call for help Urban Rwanda 2 1 -4 -3
during a robbery, to call for medical care for a loved one, or
as a deterrent against sexual violence. My mobile phone is stylish
F1 F2 F3 F4
Factor 1: Business: Thirteen participants defined this factor.
Of the statements distinguishing it from the others, few were Urban Liberia -4 -2 -1 1
positive. Most prevalent among them was the assertion that Rural Liberia -4 -4 -4 -3
business is easier to conduct thanks to the phone, which was Urban Rwanda -3 1 -3 -3
rated +4. Participants in this group also claimed that their
phones helped their business save money. On the other hand, Having a mobile phone makes me feel
members of the group do not share their phone with friends more connected to the world
or family (-4), or especially rely on it for communication with F1 F2 F3 F4
family (0) or friends (-2). Overall, this group views the phone Urban Liberia 0 2 2 3
as a serious tool that has improved their ability to conduct Rural Liberia 3 3 2 3
business, much as in the urban factor of the same name. Urban Rwanda 2 1 3 3
Factor 2: Mixed: This factor describes a variety of personal TABLE V
P OINTS OF COMPARISON BETWEEN RURAL AND URBAN L IBERIAN DATA ,
uses of the phone. Above all, the eight participants defining AND URBAN RWANDAN DATA FROM D ONNER [15].
this factor enjoyed talking on the phone with their family and
friends (+4). However, they rated the statement “My phone
makes me feel more secure” quite high as well (+3), and they
view the phone generally as a time saver (+3). No business-
We discuss them below. Table V presents comparisons of
related statements were rated highly—the phone is not seen
several noteworthy statements.
particularly as making business easier (0), and the phone is not
used to stay in touch with customers (-2). We find this factor Perhaps the most striking and unique result of this study
to be a mix of personal, productivity and security uses with a is the prevalence of security and emergency use reported by
clear bias against business. Unlike the previous factors in this Liberian mobile phone users. Data from Monrovia suggested a
study, this group does not admit to a clear, single description. distinct factor emphasizing security, while rural data revealed
Factor 3: Intrinsic: The five participants defining this factor security as an item of consensus. In both cases, the emphasis
were clearly enthusiastic about intrinsic uses of their phones as on security was considerably stronger than that reported by
technological artifacts. Their most highly rated distinguishing Donner [15] for Rwandan users, as shown in Table V. This
statement was “Having a mobile phone makes me happy” (+3), greater emphasis could be due to Liberia’s much more recent
and they were also uniquely keen to learn about new features civil conflict. From informal discussions with participants, it
or models (+2), and, less strongly, to accessorize their phone was clear that the safety and security of self, of loved ones,
(0). On the other hand, participants in this group were not and of personal property is still a major concern in Liberia.
very interested in business functions of the phone (-2), or in This is by no means surprising. Despite the 15,000-strong
using it to find work (-2). They also reported not giving their UN peacekeeping mission, the country’s police force is still
number out to many people (-4), the reason for which is not under development, and many ex-combatants have turned to
clear. crime as a source of financial support. In such a situation, it
is understandable that a mobile phone is seen as providing
Factor 4: Mixed: Our analysis identified a large number of
security, as it allows the user to call a family member or an
statements with broad meaning as distinguishing this factor.
authority in the event of a crime or transgression. In several
The six participants defining this factor keep their phone with
cases, even police officers themselves spoke of their mobile
them at all times (+4), use it to find work (+3), and stay in
phones as a source of security.
touch with friends (+3). They also find that it makes their
business easier (+2), and they use it to stay in touch with Of all the findings of this study, this emphasis on security
their customers (+1). Thus whereas Factor 2 reveals a variety carries the most implications for possible future technology
of personal uses with a bias against business uses, this factor designs. Current phones, while providing access to centralized
suggests an even more general blend of valued uses. network security services such as 9-1-1, were not designed
for an environment with weak state institutions and a lack of
V. D ISCUSSION centralized security apparatuses. Instead one could imagine a
phone design incorporating a “panic” button feature, which
The results of our Q-sort analysis have highlighted a set of emits a loud noise and flashing light, and automatically
uses and gratifications for mobile phones in Liberia, some of contacts other phones, either in the immediate geographical
which are particularly interesting in light of previous research. area, or on a predefined emergency contact list. This feature
476
could be activated in an attempt to stop a crime in progress, internet cafs, leaving the mobile phone as the only link to the
or to act as a deterrent against potential offenders. outside world.
Another result of interest highlights a difference between Also common to the two studies was the finding of strong
urban and rural users within Liberia. While analysis of urban business-related factors. Our study identified clear business
data revealed four fairly well defined factors, two of the factors for both the urban and rural populations. In addi-
four rural factors admitted to a more blurry description. We tion, several other factors rated productivity-related statements
feel that this may be due to the multifaceted livelihoods highly, such as ‘My mobile phone helps me find work.’ It is
characteristic of rural citizens and/or the fact that the mobile clear that the mobile phone plays a crucial role in the largely
phone is often the only available ICT service in rural areas. We informal Liberian economy.
encountered many rural participants who reported a number of
different occupations. One claimed to work for the Ministry VI. C ONCLUSION
of Immigration in addition to being a farmer. Another taught This study has employed the uses and gratifications ap-
secondary school in addition to selling rubber. Several students proach and the Q-sort methodology to investigate mobile
reported also working in various family businesses. Such users phone usage among urban and rural Liberians. As in previous
are not likely to fit neatly into any one category, such as work, several distinct user groups were identified. Urban users
‘business’ or ‘connectivity’—their use of the phone is more saw their phones variously as productivity enhancers, means
varied. of connectivity to family and friends, essential business tools,
On the other hand, rural users displayed an overwhelming and security providers. A group of business users was also
rejection of the phone as an object of style. As reported, identified among rural users, as was a group of techno-
participants often openly scoffed at the “My phone is stylish” enthusiasts, and two groups which eluded definite description.
statement upon reading it from the card. It seemed that the The multifaceted nature of these groups replicates Donner’s
idea of a phone being stylish was absurd, and to consider principal finding [15] in highlighting the diverse uses and
it so would be shameful. This is in contrast to both the gratifications characteristic of mobile phone users in a low-
urban Liberian and Rwandan data, both of which revealed a income region.
factor in which the same statement received a positive rating, However, the chief difference between these two studies—
as shown in Table V. Rural Liberians universally seem to our identification of the prevalence of security use among
view their phones as serious tools, not fashionable accessories. Liberians—is potentially important in its own right. As stated,
While in central Monrovia there exists a group of relatively we believe that this finding may be related to the nature of
successful business people that exhibit consumerist behaviors, Liberia’s immediate post-conflict environment. Not only does
consumerist populations are mostly absent throughout rural such a finding have implications for future technology designs
Liberia. This is due to poverty and the pervasive lack of a (which we are interested to explore), it also suggests further
formal economy in many of these areas. We suspect that in research into the role of ICTs in the process of stabilizing and
such a context the idea of flaunting or fetishising a phone rebuilding a nation following a civil conflict. The unfortunate
as stylish seems disassociated with local realities. Also due to fact of widespread civil conflict in today’s world makes
limited resources, most participants owned the least expensive, understanding such phenomena even more important.
most basic phone models, and seemed to be aware of the A more general theme that we encountered is the sheer
humbleness of their devices in comparison to the phones indispensability of the phone for most users. In many cases,
available in the city. This awareness is sure to preempt any the mobile phone is their only option for communications other
pretensions of fashionability. than physical travel, which is costly and time consuming. In
With this in mind, the emergence of an intrinsic factor other places where methods of communication are various and
focusing on technological enthusiasms in the rural data be- many, the idea of a single modality being so essential is harder
comes noteworthy. Members of that group expressed interest to fathom. But many participants we spoke to related stories
in learning about newer more advanced phones, despite the of the phone saving them many miles of travel. Businesspeo-
fact that they were likely to be unaffordable. Participants often ple celebrated the time saved in ordering goods from their
spoke about this interest in technology as if it were a civic suppliers over the phone, instead of traveling by costly public
duty - that any good citizen should be up to speed with the transportation, sometimes only to find the supplier out of stock.
latest technology. We suspect that this group is ultimately One participant described a hypothetical situation in which
aspirational ascribing to a vision where technology serves as her daughter had fallen ill and she was without a phone. How
an engine of their personal, and the nation’s development. should she know where to take her, when the only doctor in
One finding from Donner’s study which was mostly repli- the area could be in any of several different towns, each a
cated in our data was the feeling that the phone supports considerable distance away? Add to this the security role they
connectedness to the world, as also shown in Table V. This apparently perform, and it is clear that the phone is a truly
finding was especially prevalent among rural users, which is indispensable item.
not surprising given the fewer options available to rural users A weakness of this work arises from the requirement that
for communicating internationally. Some parts of Liberia, such participants be able to read. This was the unfortunate reality,
as Sinoe County, are without radio stations, newspapers, or since performing a Q-sort requires rapid and repeated visual
477
scanning of the statement cards. Unlike in a traditional survey, [14] J. Eighmey and L. McCord, “Adding value in the information age: Uses
we felt that simply reading the statement to the participant and gratifications of sites on the World Wide Web,” Journal of Business
Research, vol. 41, no. 3, pp. 187–194, 1998.
once would not be sufficient to support the sorting process. Es- [15] J. Donner, “Microentrepreneurs and mobiles: An exploration of the uses
pecially in rural areas, this unfortunately excluded a significant of mobile phones by small business owners in Rwanda,” Information
number of potential participants. In future, we are interested Technology and International Development, vol. 2, no. 1, pp. 1–22, 2004.
[16] S. McClatchey, “The consumption of mobile services by Australian
in experimenting with study designs using iconography and/or university students,” International Journal of Mobile Marketing, vol. 1,
sounds in order to allow non-literate participants to share their no. 1, pp. 1–9, 2006.
view. [17] K. Aoki and E. J. Downes, “An analysis of young people’s use of and
attitudes toward cell phones,” Telematics and Informatics, vol. 20, no. 4,
ACKNOWLEDGEMENTS pp. 349–364, 2003.
[18] F. Williams, H. Dordick, and H. Jesuale, Focus group and questionnaire
We thank Jonathan Donner for kindly providing his data for development for exploring attitudes towards telephone service. Herbert
the purpose of comparison. We also thank the various Liberian Dordick and Associates, 1985.
[19] W. Stephenson, The study of behaviour: Q technique and its methodol-
property owners who provided facilities for this study to take ogy. Chicago: University of Chicago Press, 1953.
place, and the guards at the Carter Center for their help in [20] E. B. Brownlie, “Young adults’ constructions of gender conformity and
pilot testing our study apparatus. Finally we thank the three nonconformity: A Q methodological study,” Feminism & Psychology,
vol. 16, no. 3, pp. 289–306, 2006.
anonymous reviewers for their constructive comments on this [21] D. Shemmings, “‘Quantifying qualitative data: An illustrative example
work. of the use of Q methodology in psychosocial research,” Qualitative
Research in Psychology, vol. 3, no. 2, p. 147, 2006.
R EFERENCES [22] M. Lister and D. Gardner, “Engaging hard to engage clients: A Q
methodological study involving clinical psychologists,” Psychology and
[1] “The real digital divide,” The Economist, 2005, March 12. Psychotherapy-Theory Research and Practice, no. 79, pp. 419–443,
[2] E. J. Wilson, M. L. Best, and D. Kleine, “Moving beyond ‘The real dig- 2006.
ital divide’,” Information Technologies and International Development, [23] P. H. D. Stenner, G. Bianchi, M. Popper, M. Supekova, I. Luksik, and
vol. 2, no. 3, pp. iii–v, 2005. J. Pujol, “Constructions of sexual relationships—a study of the views
[3] M. L. Best and C. Kenny, “ICT’s, Enterprise and Development”. In of young people in Catalunia, England and Slovakia and their health
ICT4D by T. Unwin. Cambridge, UK: Cambridge University Press, implications,” Journal of Health Psychology, vol. 11, no. 5, pp. 669–
2009. 684, 2006.
[4] L. Waverman, M. Meschi, and M. Fuss, The Impact of Telecoms on [24] S. J. Snelling, “Q methodology and the political opportunity of feminist
Economic Growth in Developing Countries. Vodafone, 2005. psychology,” Feminism & Psychology, vol. 14, no. 4, p. 519, 2004.
[5] R. Jensen, “The digital provide: IT, market performance and welfare [25] A. Risdon, C. Eccleston, G. Crombez, and L. McCracken, “How
in the South Indian fisheries sector,” Quarterly Journal of Economics, can we learn to live with pain? A Q methodological analysis of the
2001. diverse understandings of acceptance of chronic pain,” Social Science
[6] V. Gray, The un-wired continent: Africa’s mobile success story. ITU, & Medicine, vol. 56, no. 2, p. 375, 2003.
2006. [26] L. D. Bryant, J. M. Green, and J. Hewison, “Understandings of
[7] J. James and M. Versteeg, “Mobile phones in Africa: How much do we Down’s syndrome: A Q methodological investigation,” Social Science
really know?” Social Indicators Research, no. 84, pp. 117–126, 2007. & Medicine, vol. 63, no. 5, p. 1188, 2006.
[8] M. L. Best, K. Jones, I. Kondo, D. Thakur, E. Wornyo, and C. Yu, [27] R. M. Baker, “Economic rationality and health and lifestyle choices for
“Post-conflict communications: The case of Liberia,” Communications people with diabetes,” Social Science & Medicine, vol. 63, no. 9, p.
of the ACM, vol. 50, no. 10, pp. 33–39, 2007. 2341, 2006.
[9] D. Thakur and M. L. Best, “The telecommunications policy process [28] I. Goldman, “Q methodology as process and context in interpretivism,
in post-conflict developing countries: The case of Liberia,” in The communication, and psychoanalytic psychotherapy research,” The Psy-
36th Research Conference on Communication, Information, and Internet chological Record, vol. 49, no. 4, p. 589, 1999.
Policy (TPRC), 2008. [29] M. W. Kramer and P. M. Pier, “Students’ perceptions of effective and
[10] W. Long and P. Brecke, War and reconciliation: Reason and emotion ineffective communication by college teachers,” The Southern Commu-
in conflict resolution. Cambridge, Mass.: MIT Press, 2003. nication Journal, vol. 65, no. 1, p. 16, 1999.
[11] J. G. Blumler and E. Katz, The uses of mass communications: Current [30] N. J. A. van Exel, G. de Graaf, and W. B. F. Brouwer, “‘Everyone dies,
perspectives on gratifications research. Beverly Hills: Sage Publications, so you might as well have fun!’ Attitudes of Dutch youths about their
1974. health lifestyle,” Social Science & Medicine, vol. 63, no. 10, p. 2628,
[12] W. Chigona, G. Kankwenda, and S. Manjoo, “The uses and gratifications 2006.
of mobile internet among the South African students,” in PICMET 2008: [31] A. C. Borthwick, T. Stirling, A. D. Nauman, and D. L. Cook, “Achieving
Portland International Conference on Management of Engineering & successful school-university collaboration,” Urban Education, vol. 38,
Technology, 2008, pp. 2197–2207. no. 3, p. 330, 2003.
[13] J. W. Dimmick, J. Sikand, and S. J. Patterson, “The gratifications of [32] J. Szente, J. Hoot, and J. Ernest, “Parent/teacher views of develop-
the household telephone: Sociability, instrumentality, and reassurance,” mentally appropriate practices: A hungarian perspective,” International
Communication Research, vol. 21, no. 5, pp. 643–663, 1994. Journal of Early Childhood, vol. 34, no. 1, p. 24, 2002.
[33] A. C. McKennel, “Surveying attitude structures—discussion of princi-
ples and procedures,” Quality & Quantity, vol. 7, no. 2, pp. 203–294,
1974.
[34] R. Southwood. “Liberia: Four mobile companies bring lowest
prices in West Africa” Balancing Act. http://www.balancingact-
africa.com/news/back/balancing-act 352.html#head (accessed February
20, 2009).
478
An Automated Braille Writing Tutor with

Multilingual Exercises and Educational Games
M. Bernardine Dias, M. Freddie Dias, Sarah Belousov, Mohammed Kaleemur Rahman, Saurabh Sanghvi,
Imran Fanaswala, Wael Ghazzawi, Ameer Abdulsalam, Noura El-Moughny1 and S. Raghu Menon2
Abstract—This work advances the state-of-the-art in assistive six buttons placed over the top of the two rows to work as an input
technology for the visually impaired by enhancing an automated area. The stylus is a standard braille stylus that connects to the BT by
tutor designed to teach beginner-level braille writing skills. The its metal tip. Moreover, students can press on one of two control
hardware component of the tutor is a low-cost device that buttons placed on the sides of the BT to perform a variety of mode
provides an intelligent tactile interface that connects to a laptop changes. Further details of the BT are discussed in prior publications
or desktop computer. We demonstrate the many uses of this by Kalra et al. ([1] and [2]). A new design for the packaging the BT
tutor including its multilingual capabilities, interactive exercises, was also explored (Fig. 1.).
and educational games. Preliminary field tests of the tutor in
India, Zambia, and Qatar have demonstrated the usefulness of
this automated tutor, and its potential to enhance braille literacy
in developing regions.
Index Terms— Assistive Technology, Braille, Educational Fig. 1. Prototypes and preliminary field testing of the Braille Writing Tutor
Games, Visually Impaired.
III. MULTILINGUAL EXERCISES AND EDUCATIONAL GAMES

I. INTRODUCTION
We collected ideas from teachers at three partner schools to
B RAILLE is a widely-used language that is the only means of
literacy for blind people. Each braille character is formed using
a subset of six embossed dots placed in a cell of two columns
understand the process for teaching beginners to write braille, and the
use of educational games in the classroom. Based on this research we
and three rows. The positions of the dots are universally numbered implemented a curriculum that includes six staged interactive
from one to six. Despite its significance and the accessibility it exercises for students to practice basic skills needed for learning to
brings, learning to write braille still has a number of barriers. More write braille. These include practicing dot numbers, dot patterns, and
than 90% of the world’s visually-impaired people live in developing the association of dot patterns with letters. This curriculum is now
communities and their literacy rate is estimated to be below 3% [1]. accessible in several languages, including English, Arabic, Chinese,
Poorer areas tend to have both a disproportionately high number of and French. We also implemented some of the educational games
blind people and fewer resources for educating them [1]. Therefore, used in braille-writing classes to improve the fun-factor of the tutor.
the need to improve literacy for the blind in affordable ways is These include an animal game that requires students identify an
paramount. The traditional method of writing braille itself creates animal based on its sound, and then spell its name on the tutor, and
formidable challenges to literacy [1]. In developed countries, braille the popular “hangman” game which helps children practice spelling
is usually embossed with a six-key typewriter known as a Brailler. while trying to guess a word chosen by the computer. We also
These devices are prohibitively expensive in developing countries created two new games that can be played using the BT: a dominos
where braille is almost always written with a slate and stylus and game and a Chinese braille game. The dominos game was inspired
hence children must learn mirror images of letters, get delayed by a popular game with children in Qatar and requires the student to
feedback, and visually impaired teachers have difficulty monitoring match a letter chosen by the computer by picking a different letter
student progress and diagnosing some mistakes. Thus, educational which has the identical dot pattern in the left or right column as the
braille writing technology relevant and accessible to developing
specified letter. Since Chinese braille is based on a phonetic
communities could have significant impact.
representation of the sounds of the language there are 3 categories of
braille characters: Initials (the consonants that are only found at the
II. AUTOMATED BRAILLE WRITING TUTOR beginning of a syllable), Finals (vowels), and Tones (signs that
In response to the observed need for enhancing literacy for the indicate the tone of the word). The Chinese braille game challenges
blind in developing regions, a group at Carnegie Mellon University students to match a “final” to an “initial” chosen by the computer.
developed an automated Braille Writing Tutor (BT) that uses audio We propose to demonstrate all of these capabilities.
feedback to provide guided practice for beginners learning to write
braille [3]. The BT (Fig. 1.) consists of an electronic slate and stylus REFERENCES
which monitors the student’s writing and transmits data in real time
[1] Nidhi Kalra, Tom Lauwers, and M. Bernardine Dias, “A Braille Writing
to a computer to provide immediate audio feedback to the user. The Tutor to Combat Illiteracy in Developing Communities,” in Artificial
latest version of the BT consists of two rows of 16 braille cells and Intelligence in Information Communication Technology for
Development workshop at IJCAI 2007, Hyderabad, Jan 2007.
Manuscript received on January 30, 2009. [2] Nidhi Kalra, Tom Lauwers, D. Dewey, Tom Stepleton, and M. B. Dias,
1
All authors except the last author were affiliated with Carnegie Mellon Iterative Design of A Braille Writing Tutor to Combat Illiteracy, Proc.
University during their involvement with the reported work. Author contact IEEE/ACM International Conference on Information and
information: {mbdias, mfdias, sarahtbw, kaleem, scsanghv, imranf, Communication Technologies and Development (ICTD), Bangalore, Dec
wghazzaw, ameer}@cmu.edu and noura.elmoughny@gmail.com 2007, pp 9-10.
2
S. Raghu Menon is with Microsoft Research India, Bangalore [3] TechBridgeWorld Automated Braille Writing Tutor project website:
(menonsraghu@hotmail.com) http://www.techbridgeworld.org/brailletutor
479
Boosting European Market Access to Malian

Mango Growers
Saskia Harmsen
Abstract—FRUILEMA, a Malian fruit and vegetable economic further inform partners about the supply process and improve
interest group (EIG), is acquiring the tools and resources needed north-south linkages that promote not only trade, but also
to meet the requirements (GLOBALGAP certification) for mutual understanding and collaboration. This platform allows
exporting Malian farm products to international markets. for the better identification of producers, to know their
Starting with the highly marketable mango fruit, an ICT solution collective production forecasts, and to trace the entire chain of
has been developed to enable the identification of producers, to
activities, from the purchasing of agricultural products back to
trace farm plots and produce, and to obtain quantitative and
qualitative information on produce to support analysis and the producer. During the first year of the system being in
decision-making. The information is accessible via a web-enabled operation, FRUILEMA was able to use the platform to
platform and can be updated using Internet and PDAs. The convince European fruit importers to work with them,
information is disseminated to producers, trackers, collectors, resulting in two contracts for the 2008 mango harvesting
importers, and other actors in the mango value chain. season.
The platform was developed during 2007-2008, and
Index Terms—Agriculture, Geographic information systems, involved the identification and translation of all processes in
Information systems, Software the production and "conditioning" of mango products,
software development as well as capacity building of the five
exporting organizations involved. The platform was
I. CASE (CONTEXT, NOVELTY AND SIGNIFICANCE)
developed in collaboration with the five exporting
A GRICULTURAL exporters from Mali face many

requirements and constraints if they wish to sell to the
European market. They need to have exact knowledge about
organisations under FRUILEMA. Since the launch, the
platform is being further developed to allow for the use of
PDAs and other mobile devices to enable direct data
the quantity and quality of farm products, places and periods collection and data uploading while in the field at the farmers’
of production, production inputs used, etc. Meanwhile, site, thus expediting the data collection, management and
international consumers increasingly want to be able to trace dissemination processes.
the products they buy back to the farmer and plot where it
came from. For importers, it is important to be able to assess II. DEMONSTRATION
the quality and reliability of their exporting counterparts. For
The demo will showcase the possibilities offered by the
producers, it is important to know the exact surface of their
platform, by showing the producer, plot and produce
production areas and the requirements of the market. As
information; tracking plot locations and size using
analysis during project formulation has shown, agricultural
Geographical Positioning Systems (GPS); aggregating
producers in Mali rarely have a correct estimate of their
information for the use of export planning; and other
production capacity, often over- or underestimating their real
functionality offered by the platform. If possible, the newly
production capacity. This has led to challenges in developing
developed access modalities (mobile & PDA) will be
realistic production plans, procuring the right amounts of farm
showcased live as well. The different user levels and their
inputs like fertilizers, and being able to deliver the quantities
respective tasks and information provision within the system
of fruits promised to exporters.
will be demonstrated.
To minimize these challenges and optimize the meager
Although the existence of a website strengthened the
resources available, IICD’s Senegalese partner Manobi has
credibility of the exporting partners and even led to concrete
further developed and adapted its existing platform to suit
exporting contracts with companies in the Netherlands and
FRUILEMA’s needs. As an association representing 790
Belgium, developing a database with all the necessary
small producers and five exporter groups throughout the
information does not guarantee a successful market
country, FRUILEMA wants to use all the data gathered to
information system. Many other factors need to be taken into
account such as access to credit, physical infrastructure,
S. Harmsen works for the International Institute for Communication and
Development (IICD), P.O. Box 11586, 2502 AN The Hague, the Netherlands capacities and marketing. The lessons thus far gained will be
(phone: 31-70-311 73 11; fax: 31-70-311 73 22; e-mail: sharmsen@ iicd.org). discussed during the demonstration as well.
480
Creating a Mobile-Phone Based Geographic

Surveillance System for Avian Influenza
Yibo Lin and Claire Heffernan
As such, this demo illustrates a geographic mapping system
Abstract—Highly pathogenic avian influenza (HPAI) is not developed at both the local, regional and global level to aid
only a global thread to human and animal health but also surveillance and remove communication barriers between
disproportionately impacts poor livestock keepers in southern stakeholders. Unique to the system is its ability to RSS global
countries. While billions have been spent on the disease, response
to the epidemic remains fragmented and information channels
outbreak information from the WHO database to rapidly
slow. As such, this demonstration details a geographic mapping inform stakeholders of confirmed outbreaks.
system at the global and local levels to aid information transfer
among policy makers, practitioners and the poor themselves II. DEMONSTRATION
regarding the control of this disease. The demonstration will
During the demonstration, both the regional and global
include hands on case studies in which conference attendees will
be asked to both make decisions and use the tool in a simulated systems will be presented. The regional system illustrates how
outbreak situation. primary data can be gathered by local veterinarians and such
geographic information may be shared by mobile phone.
Index Terms—Avian influenza, Geographic information Conversely, the global system illustrates a five ‘layered’ map
systems, Surveillance system, User-Generated Content representing the five critical factors of an HPAI outbreak:
geographic spread, human mortality, number of poultry culled
and wildlife migration pathways. Use of the system will be
I. INTRODUCTION linked to practitioners in the field. Further, conference goers
can explore the use of the tool from the perspective of the
ACCORDING to WHO [6] ‘Avian influenza is an infectious veterinarian in an outbreak situation. Thus, attendees will be
disease of birds caused by type A strains of the influenza offered a choice of three case studies. In each case, conference
virus’. In recent years, the emergence of highly pathogenic attendees will be asked to use the tool to make decisions in an
avian influenza (HPAI) strain (H5N1) has changed global simulated outbreak situation – and the consequences of their
perceptions regarding poultry, poultry production systems and decisions will be illustrated by an on-line mapping system
the risk of animal diseases being transmitted to man. Indeed, which demonstrates the spread of the disease.
H5N1 is able to be transmitted to humans (presently at low
levels) but has a high fatality rate [5]. REFERENCES
On a practical level, it has been argued that the ability to [1] S. Chang, J. Zhang, X. Liao, X. Zhu, D. Wang, J. Zhu, T. Feng, B. Zhu,
respond to this epidemic remains weak [4]. And that G. F. Gao, J. Wang, H. Yang, J. Yu and J. Wang, "Influenza virus
preparedness for the future pandemic, if the disease does jump database (ivdb): An integrated information resource and analysis
platform for influenza virus research", Nucleic Acids Research, vol.35,
the species barrier, is poor [3]. pp. D376-D380, October 25 2006.
Part of the problem is that poor livestock keepers, who [2] C. Heffernan and F. Misturelli, "The delivery of veterinary services to
globally are the most impacted by the disease, are often the poor: Preliminary findings from kenya", in Report for DFID's
marginalized from services and information delivery [2]. (Department for International Development) Animal Health Programme
(AHP), Livestock Development Group, The University of Reading,
Therefore, present surveillance information are ineffective. Reading, UK, 2000.
Indeed, researchers have already declared that a new [3] M. T. Osterholm, "Preparing for the next pandemic", The New England
system, which directly connects veterinary Journal of Medicine, vol.352, no.18, pp. 1839-1842, May 5, 2005 2005.
[4] I. Stephenson, K. G. Nicholson, J. M. Wood, M. C. Zambon and J. M.
researchers/practitioners to decision-makers is required [1]. Katz, "Confronting the avian influenza threat: Vaccine development for
Nevertheless, direct information links between the poor and a potential pandemic", The Lancet Infectious Diseases, vol.4, no.8, pp.
their service providers is also crucial to the success of any 499-509 2004.
[5] WHO, "Highly pathogenic h5n1 avian influenza outbreaks in poultry
new system. and in humans: Food safety implications", INFOSAN, WHO, Geneva,
Switzerland, 2005.
[6] WHO, "Questions and answers on avian influenza", in A Selection of
Manuscript received February 2, 2009. Frequently Asked Questions on Animals, Food and Water, WHO,
Yibo Lin is a research associate at the Livestock Development Group, Geneva, Switzerland, 2006.
School of Agriculture, Policy and Development, University of Reading, UK
(e-mail: yibo.lin@ reading.ac.uk).
Claire Heffernan is the Director of the Livestock Development Group,
(Tel.: +44 1189 318213, e-mail: c.l.heffernan@reading.ac.uk).
481
Design of a Blood Flow System

A. Osuntogun, S. Thomas, J. Pitman, S. Basavaraju, B. Mulenga and S. Vempala
possible needs to determine how much blood should be

Abstract—Blood is a vital but often scarce resource in distributed to a hospital. To ensure fair and efficient
developing countries. It is crucial that safe blood is made distribution of blood, it is necessary to monitor and understand
available for transfusions in hospitals and clinics to prevent the the patterns of blood collection and requests and to have up-to-
spread of Transfusion-Transmitted Infections (TTI) such as HIV,
Hepatitis and Syphilis. We present a system designed to promote
date data on available blood. The system aims to maximize
hemovigilance in developing countries by monitoring the blood usage while minimizing or eliminating wastage.
collection and usage patterns of blood, predicting collection and
usage for upcoming time periods and finding an allocation III. CONSTRAINTS
assignment for blood distribution that is fair and efficient.
To create a system that is sustainable in a developing country,
Index Terms—Health, Developing Nations, Blood Safety
it is necessary to understand the constraints. An assessment of
the user skills indicate that the system should be user friendly
I. INTRODUCTION since the main users will be blood bank and hospital staff such
as doctors, nurses and biomedical scientists who should not be
B LOOD is a precious resource needed to save lives during

surgeries and other major operations at a hospital. In
many developed countries, there are systems and protocols in
expected to have complex technical skills. It is also important
to consider network availability and power supply for a system
that needs current data to function maximally. Other concerns
place to ensure that the blood is collected safely and only safe for developing countries include funding for the maintenance
and healthy blood is transfused to a patient. In developing and upkeep of the system. These issues were taken into
countries, past research has shown that transfusion of unsafe consideration during the development of the system.
blood units has led to the spread of Transfusion-Transmitted
Infections (TTI) such as HIV, HBV, HCV and Syphilis. The IV. SOLUTION
WHO estimates that the risk of HIV infection through Our work focuses on the design of a blood flow system that
transfusion of unsafe blood is exceptionally high in Africa [1]. tracks a blood unit from collection to utilization. Aggregate
This knowledge has led to the development of protocols in information on a large pool of blood units can then be used to
developing countries to ensure that only safe blood is being determine trends and build probabilistic models for blood
transfused. In addition to this challenge, many developing flow. The major components of this system include monitoring
countries do not have enough safe blood to meet the needs of the collection and usage pattern of blood units, predicting the
the country hence the need for appropriate methods for collection and usage for upcoming time periods and finding a
collection, testing and allocation of a limited blood supply. flow assignment for blood distribution that is fair and efficient.
The system serves as a method for visualization and analysis
II. PROBLEM DESCRIPTION of blood data as well as a decision-informing tool.
To understand and improve the blood supply chain in A live and interactive demonstration of the software system
developing countries, we conducted a series of interviews at showing tracking, prediction and allocation of blood units will
the Zambia National Blood Transfusion Service. Due to the be given in order to facilitate understanding of functionalities
scarcity of blood, units are often moved multiple times of the system.
between collection and transfusion to regions where blood is
needed. Blood allocation is currently being done in an ad hoc REFERENCES
manner in which a blood safety officer uses past and current [1] N. Dhingra, “Making Safe Blood Available in Africa” .World Health
internal knowledge about the amount of blood available and Organization, Blood Transfusion Safety. 2006.
http://www.who.int/bloodsafety/makingsafebloodavailableinafricastate
ment.pdf
A. Osuntogun., S. Thomas and S. Vempala are with the Georgia Institute

of Technology. Atlanta, GA 30328 (e-mail:bola@gatech.edu,
sthomas8@gatech.edu, vempala@cc.gatech.edu).
J. Pitman and S. Basavaraju are with the Centers for Disease Control and
Prevention, Atlanta GA 30333 (e-mail: cgx5@cdc.gov,etu7@cdc.gov).
B. Mulenga is with the Zambian National Blood Transfusion Service,
Lusaka, Zambia (e-mail: bright.mulenga@znbts.gov.zm).
482
DISHA: DISease and Health Awareness for

Children on Multiple Input Devices
Mohit Jain, Aakar Gupta, Navkar Samdaria, Praveen Shekhar and Joyojeet Pal
to both spread the use of a single computer wider across all

Abstract— Much recent work in multiple input use scenarios users as well as increase engagement between the children. We
for children’s learning software has focused either on math or on adopt a hybrid approach of racing and unity models [4], thus
English language learning. The persistence of under-information two teams of children play against each other by connecting to
among children in the developing world on issues of hygiene and
disease prevention remains a massive challenge within the
a server, thus ensuring collaboration among the team mates
scholarly community in public health, especially in the developing and competition among the teams. An Artificial Intelligence
regions that multiple input learning technologies are designed for. backbone tracks the performance of individual children and
DISHA is a collaborative platform for public health information distributes on-screen prompts based on the performance of
for children in low-income regions using multiple mice. The children and the areas they appear to need help. The idea of
system is designed towards collaborative use of screen resources. profiling performance over time used in DISHA also has an
important machine language element in tabulating a child’s
Index Terms— Computer Aided Learning, Multiple Mice,
Developing Regions, Health Education. competencies that can be applied across various learning
subjects for young children, which in turn can provide useful
I. INTRODUCTION cross-domain data on learning skills.
G LOBAL health statistics show a staggering amount of

deaths and disabilities due to preventable diseases
including respiratory infections, diarrhea, malaria and
III. INTERFACE FEATURES
We employ in our design features that have been seen as
tuberculosis that almost exclusively impact the developing effective in past trial applications with education in the
world [1]. The issue of public health education is a relatively developing world. This includes some standard design
unexplored area within the domain of Computer Aided principles of gaming including goal-based progression,
Learning (CAL), despite past evidence of effectiveness of competition, positive reinforcement, personalized scoring and
audio-visual material for hygiene education, as well as the other principles specifically relevant to ICTD such as Same-
currently prevalent use of animation in static learning materials language Subtitling (SLS) and color and shape-coded mouse
for children. With DISHA, we aim to supplement CAL pointers. The interface is designed intuitively for a child’s use,
programs, increasingly prevalent in the developing world assuming minimal availability of teachers.
including many areas with high risk of some of the
aforementioned diseases. DISHA encompasses every aspect IV. DEMONSTRATION
including the symptoms, prevention and cure of these diseases. We demonstrate (a) on a single laptop with four mice the
Using a narrative-interactive loop format of story-telling use of DISHA in a real world, (b) two teams playing against
followed by multiple choice Q&A, DISHA uses MultiPoint each other on two laptops connected to a server.
technology [2, 3] for healthcare learning material. The Q&A
enforces collaboration, competition and turn taking as REFERENCES
machine-induced interactive modules, following a game based [1] Global Disease Statistics
approach. http://www.fightingdiseases.org/main/globalstats.php, Accessed January
30, 2009.
[2] U. S. Pawar, J. Pal and K. Toyama, "Multiple Mice for Computers in
II. IMPLEMENTATION Education in Developing Countries," in Proc. of IEEE/ACM ICTD'06,
Giving each child a mouse of his or her own, DISHA aims 2006, pp. 64-71.
[3] The Microsoft Windows MultiPoint Software Development Kit (SDK).
http://www.microsoft.com/downloads/details.aspx?FamilyID=A137998
B-E8D6-4FFF-B805-2798D2C6E41D&displaylang=en, Accessed
Manuscript received January 30, 2009. January 30, 2009.
Mohit Jain, Aakar Gupta, Navkar Samdaria and Praveen Shekhar are with [4] S. Thota, J. Pal, U. S. Pawar, A. Joshi, M. Jain, S. Teja and S. Anikar,
Dhirubhai Ambani Institute of Information and Communication Technology, "From Pilot to Practice: Creating Multiple-Input Multimedia Content for
Gandhinagar, Gujarat, India (phone: +91 9879031798, +91 9904866639; Real-World Deployment," in Intelligent User Interfaces for Developing
e-mail: {miohit_jain, aakar_gupta, navkar_samdaria, praveen_shekhar} Regions, 2008, Canary Islands, Spain.
@daiict.ac.in).
Joyojeet Pal is with University of Washington, Seattle, USA (phone: 510-
501-8679; e-mail: joyojeet@washington.edu).
483
Freedom Fone: Dial-up Information Service

Bev Clark and Brenda Burrell
cost or free of charge to callers via toll free lines and

Abstract— This demonstration highlights the latest advances of callbacks. Members of the public can also request callbacks
Freedom Fone: Dial-up Information Service. Freedom Fone via SMS.
provides organisations with a simple to use interface to manage The Freedom Fone prototype has been used to pilot the
Interactive Voice Response (IVR) menus and upload pre-
recorded content into an “audio magazine.” Members of the
dissemination of sexual / reproductive health information with
public can then phone in to listen to the latest information, or to adolescents in Zimbabwe, and as an information service at the
leave a message. Freedom Fone prototype software has been Association for Women’s Rights in Development conference
piloted in Zimbabwe and South Africa, with sexual health, in Cape Town in November 2008. We are currently trying to
women’s rights and cholera awareness information. establish a cholera information service in Zimbabwe, in
association with local and international health agencies.
Index Terms—Africa, Information, Mobile communications Elements used by Freedom Fone aren’t new, but their
intuitive convergence makes it innovative and globally
relevant. Global Voices' Ethan Zuckerman said recently "I'm
F reedom Fone is a free open source software tool that can
be used to build and update a dial-up information service
in any language.
surprised there hasn’t been more work done making
interactive voice response systems usable for development
purposes [1]." Freedom Fone provides exactly that.
Its easy to use interface lowers the barriers to using Interactive
Other projects using Interactive Voice Response to make
Voice Response (IVR) for outreach to communities small and
critical development information more available local
large, near and far. Freedom Fone empowers non-technical
communities are VoiKiosks, which “present an alternative
organizations to build automated information services that are
platform to create and host such information kiosks in the
available to the public 24 hours a day, 7 days a week – power
telephony network,”[2] and MobilED, whose “Audio Wiki”
and connectivity allowing.
programme uses a range of software products including
Based on the concept of Interactive Voice Response (IVR),
Asterisk, Kannel and Flite to make audio information
pre-recorded audio files are stored by Freedom Fone in a
available to learners [3].
Content Management System (CMS). This is updated through
a simple to use browser interface. Content is made available to
REFERENCES
telephone callers when they dial phone numbers associated
with the service. Users are presented with an audio menu of [1] Ethan Zucherman. “Delivering Ethiopian teff via Czech taxi?” My
options and using their telephone keypads select channels Heart’s in Accra Blog. October 13, 2007.
carrying information of interest to them. They are able to http://www.ethanzuckerman.com/blog/2007/10/13/delivering-ethiopian-
teff-via-czech-taxi/
listen to a selection multiple times if they so choose. [2] Sheetal Agarwal, Arun Kumar, Amit Anil Nanavati, Nitendra Rajput.
Deployment in any language is possible as key global files for “VoiKiosk: Increasing Reachability of Kiosks in Developing Regions”,
menu prompts can be uploaded through the browser interface WWW 2008 / Poster paper.
to the CMS. [3] Server – MobilED website http://mobiled.uiah.fi/?page_id=98
Individuals can contribute questions, content and feedback
by leaving voice messages via the IVR interface. Freedom
Fone can be operated as a collective, with different groups
managing different channels (IVR menu options) of
information from the same installation.
Currently available as a prototype, Freedom Fone can be
connected to telephony services via the internet using VoIP –
this is the most scalable option; or via telephony cards to
either analog or digital phone lines. Services can be offered at
Manuscript received February 19, 2009.

B. Clark is Creative Director of the Kubatana Trust of Zimbabwe (Phone:
+263-4-776-008; fax: +263-4-746-418; e-mail: info@freedomfone.org).
B. Burrell is Technical Director of the Kubatana Trust of Zimbabwe
(Phone: +263-4-776-008; fax: +263-4-746-418; e-mail:
info@freedomfone.org).
484
FrontlineSMS and Ushahidi – A Demo

Ken Banks and Erik Hersman
Abstract—FrontlineSMS and Ushahidi are both free and open developed in 2005 and updated in 2007 – is being used around
source tools which allow for the collection, dissemination and the world for a wide range of non-profit activities including
visualization of data collected to and from the field. This paper the sending of market prices and other agricultural data to
outlines details of live demonstrations of the tools at the ICTD smallholder rural farmers in Aceh, Cambodia and El Salvador,
2009 conference in Qatar.
the dissemination of news in Iraq, the sending of security
Index Terms—FrontlineSMS, Ushahidi, data, messaging, open alerts to fieldworkers in Afghanistan, for human rights work
source tools, mobile, SMS in places such as Zimbabwe, Pakistan and the Philippines,
and the running of a rural healthcare network for 250,000
people in Malawi. Because the software can be used on a
I. INTRODUCTION single laptop computer without the need for the internet, it has
been widely adopted among the grassroots non-profit
FrontlineSMS and Ushahidi are two fully free and open
community and nominated for several awards.
source tools which leverage information and communication
technologies – of which mobile is a key component – to allow
individuals or groups to collect and/or disseminate FrontlineSMS is currently the focus of a number of academic
information to and from the field. The tools work studies, and the recent UN Foundation/Vodafone report on
independently, but can also work seamlessly together, as they mHealth described it as “one of the largest and most
are currently doing in a deployment in DRC. ambitious mHealth programs in the world”.
Ushahidi, which means “testimony” in Swahili, is a free-to-

access website that was developed to map reports of violence II. PRESENTATION / DEMO
in Kenya after the post-election fallout at the beginning of Representatives from FrontlineSMS and Ushahidi will show
2008. Ushahidi's roots are in the collaboration of Kenyan
how these two tools can be used independently, and in
citizen journalists during a time of crisis. The new Ushahidi
tandem, to change the way organizations large and small can
Engine has been created using the lessons learned from Kenya
to create a platform that allows anyone around the world to set communicate. This isn’t just technology for the developing
up their own way to gather reports by mobile phone, email world, but a way to understand and use technology on a local
and the web - and map them. The core engine is built on the level that fits the everyday needs of ordinary people.
premise that gathering crisis information from the general
public provides new insights into events happening in near Examples of users from various parts of the world will be
real-time. It is being developed by a group of volunteer shown, statistics from past deployments shared, and the
developers and designers, hailing primarily from Africa. So technology will be demonstrated live. In the demonstration, a
far there are representatives from Kenya, South Africa, local tech hub (using a PC running FrontlineSMS) will be set
Uganda, Malawi, Ghana, Netherlands and the US. up and the attendees will see a live display of real time
submissions, response and plotting of this data on a global
A post-mortem study by the Harvard Humanitarian Initiative map immediately available to the world. It will show how
provided qualitative analysis on the added information powerful real-time data collection from disbursed and open
gathering opportunities that this type of platform creates. A communities can be visualized, and then acted upon by others,
report titled, “Digitally Networked Technology in Kenya’s and how two-way group text messaging can be used to help
2007–2008 Post-Election Crisis” from the Berkman Center run and co-ordinate field activities, and deliver information
for Internet and Society also used Ushahidi as an example of services to rural communities.
how mobile phone messaging can be used for good in times of
crisis.
Ken Banks is the Founder of kiwanja.net and developer of
FrontlineSMS is free software that turns a laptop and a mobile FrontlineSMS, and Erik Hersman the Technical Director at
phone or modem into a central communications hub, allowing Ushahidi. Both have extensive experience working in the
users to send and receive text messages with large groups of ICT4D field, and both actively maintain websites and Blogs
people through mobile phones. The software – originally on the application of technology in the developing world.
485
Global Youth Connectivity (GYC) – Using ICT

for Peaceful Recovery and Long-term Change
Annie Bertrand
Abstract—ICT enables peaceful change in communities war, Ziad from Gaza City commented: "The situation is really
affected by conflicts, poverty and injustice. In this demo, we bad in Gaza (…) I am happy about the posts on the blog. It is really
illustrate how young people use multimedia, cell phones and the nice to know that there are people who care about us and support
internet to become global citizens responsible for building secure, us…this is encouraging.” Rebuilding hope is sometimes harder than
productive and just communities at home and around the world. infrastructure but a necessary condition for peaceful recovery.
Global Youth Connectivity allows disaffected youth to express According to a UNDP report on Arab youth, global communication
thoughts and feelings, while also scaling global citizenship systems also provide a gateway for youth to delve into creativity,
education and civic engagement across borders. Thousands of innovation, learning, professionalism, and democratic governance:
Middle-Eastern and American youth have exchanged content
“Their self-confidence is boosted when they expand their practical
and discussed online since 2005. Innovations and partnerships
should enable Mercy Corps to empower millions more living in knowledge and communicate with their peers around the world” 4.
extreme vulnerability.
III. INFORMATION + COMMUNICATION = PROGRESS
Index Terms— Civic engagement, Global Citizenship, Global Many young people in the Middle-East suffer from
Youth Connectivity
repression, isolation and limited opportunities for social or
civic engagement. Mercy Corps and IJMA35 are working on a
grassroots news network with information organized by youth
I. INTRODUCTION
for youth. After a newspaper-based pilot in Iraqi villages, the
he Global Youth Connectivity1 (GYC) program called
T “WhyNot!” was established by Mercy Corps as part of a
relief and development strategy to address psychosocial needs
media outlet will be scaled-up to reach millions of young
Arabs across gender and social groups using internet (e-
newsletter, RSS feed, widgets) and GSM (SMS text and
and reduce the widening knowledge gap in the Middle-East. VOICE news). This digital peer-to-peer education process
Once computer centers are built2, we train young first-time also enables anonymous feedback and asking of questions to
users on digital photos, flip videos and PowerPoint to be addressed. As the content evolves and feeds online
illustrate their culture, voice their concerns, and connect dialogues, young reporters will also be trained on taking
online with others living abroad. Global engagement helps collaborative peaceful actions at the local, regional and
those affected heal their wounds and change from “powerless international levels.
beneficiaries” into active contributors to peace, productivity
and justice at home. The curricula allow diverse youth to IV. DEMONSTRATION
explore cultural differences and find common grounds around During the demo session, we will show how the Global
peaceful change. Such sharing and moderated dialogues also Citizen Corps portal aggregates web, mobile, and other digital
teach American youth about global issues from the media content into a single, youth-friendly and accessible
perspective of those directly affected by them. online resource. Inspiring extracts of youth content such as
video-documentaries, photo-essays, and online dialogues will
II. EXPRESSION + GLOBAL CONNECTION = EMPOWERMENT allow conference participants to witness the role that ICT
During the bombardments in Gaza earlier this year, we used plays in transforming young people into global citizens,
cell phones to give internally displaced youth a chance to capable of and responsible for building a secure, productive
express concerns and be heard. Innovative software developed and just world now and across generations.
by Souktel3 allowed youth to ‘text’ insights to a central
database, which fed into our blog. Reactions from peers REFERENCES
abroad with inspiring words of sympathy had positive [1] Global Youth Connectivity www.globalcitizencorps.org/whynot
[2] “Centers for Excellence” in Iraq and Lebanon were established to
psychosocial effects on their healing process. In the midst of provide access to over 500 computers connected to internet.
[3] Souktel uses SMS text messaging to link young people with jobs and aid
Annie Bertrand is Director of Global Youth Development at Mercy Corps agencies in Palestine and beyond www.souktel.org
supporting youth transformation across 35 countries. She graduated with a [4] Arab Youth Strategising for the MDGs (2006), UNDP, p.37
MBA from Harvard Business School in 2007. (phone: +1 518 282 4783; [5] IJMA3 is a partner of Mercy Corps in implementing ICT projects in the
email: abertrand@nyc.mercycorps.org) Middle-East www.ijma3.org
486
Implementing E-Government Accessible to

Illiterate Citizens
D. Kettani and A. El Mahdi
known as: Electronic Fundamental Etat-Civil System (eFES).

Abstract— Information and Communication Technologies
(ICT) are becoming more and more popular as a tool in the The novelty of eFES system, completed in 2006 as a result of
effective administrations of local and national governments Public-Academia-Partnership, rests on the following:
worldwide. Nevertheless, developing countries still lag behind.
One main challenge is automating the back office associated with
- Completing an ICT-based municipal system that enables
public administrations. Another challenge is designing and the unprecedented shift from cumbersome manual service
implementing ICT systems accessible to illiterate users. This delivery to rapid, more efficient automated service
Demo presents the Fez e-government platform deployed in Fez, delivery in Morocco’s developing country context
Morocco. The intention is to first, explain the workings of the - Succeeding the first-time introduction of “self-service
front office kiosk and how illiterate users request and procure technology” to local public administration.
municipal certificates and second, to outline how the back office
is integrated into the kiosk and other regional administration.
eFES project initiated the introduction of touch-screen kiosk
Index Terms— ICT-based municipal system, illiterate users, to Fez municipal offices so that community members could
touch-screen kiosk, self-service technology have the new possibility of requesting the needed certificates
without full dependency on employee intermediation
(commonly known for abusive and citizen-unfriendly
treatment).
M orocco, a North Africa developing country, has
expressed interest in Information and Communication
Technologies (ICT) since the beginning of the 1990s. More importantly, the innovative qualities of eFES platform
Yet, the development of Morocco’s e-government national lie on initiating the introduction of “self-service technology”
strategy took twelve years (1993-2005). One outcome for such while taking into consideration Morocco’s high illiteracy rate.
a delay was the absence of concrete ICT actions to benefit the Specifically, eFES conducted extensive research to adapt the
daily life of ordinary citizens. e-government touch-screen kiosk to an illiterate user profile.
This involved exploring the different architectures, scenarios
In response to Morocco’s low ICT penetration, action was and technologies for designing ICT-based system adapted to
taken on by the Fez municipality to collaborate with the illiterate users. It required also creating design principles and
ICT4D research laboratory at Al Akhawayn University in conducting usability testing in real life settings with real users
2004. Thanks to research funding granted by Canada’s (i.e. community members).
International Development Research Center (IDRC),
collaboration focused on building an e-Government system Developing ICT solutions that address accessibility issues
for the city of Fez to initiate the introduction of ICT to related to basic and digital illiteracy is a key in ICT growing
Morocco’s local administration. In 2004, the research team use, adoption, and appropriation amongst the masses in
opted for focusing on developing a platform automating the developing countries. eFES significance comes from
delivery of highly and widely requested municipal certificates, developing a methodology and architecture for e-government
application implementation fulfilling the needs in the Arab
World.
Manuscript received January 30, 2009. This work was supported by the
International Development Research Center (IDRC) under Grant 104053. The
project team hereby expresses its appreciation and gratitude to the IDRC for This Demo intends to present Fez e-government system
funding eFez project; without which this research work would have not been (commonly known as eFES); it aims to share the
possible. achievements of introducing ICT to local public
Driss Kettani is an Associate Professor at the School of Science and
Engineering and Head of the ICT4D Research Laboratory at Al Akhawayn administration in Arab developing country context. The eFES
University in Ifrane, Morocco. P.O.Box: 2122, Ifrane 53000, Morocco. Demo would like to show the touch screen kiosk adapted to
(email: d.kettani@aui.ma ) illiterate profile. It intends also to show video footage of kiosk
Asmae El Mahdi is currently with Enhanced Technologies, Al Akhawayn uses in Fez real life settings.
University Technopark; a research team member at the ICT4D Research
Laboratory at Al Akhawayn University in Ifrane, Morocco. Avenue Hassan II,
B.P 2072, Ifrane 53000, Morocco. (corresponding author phone: 00212 35 86
2114; e-mail: a.elmahdi@aui.ma ).
487
Improving Data Quality With Dynamic Forms

Kuang Chen, Harr Chen, Neil Conway, Heather Dolan, Joseph M. Hellerstein, and Tapan S. Parikh
Abstract—Organizations in developing regions want to effi- clinic operations were limited to using paper forms to ensure
ciently collect digital data, but standard data gathering practices timely information access. Only after a labor-intensive delay
from the developed world are often inappropriate. Traditional did the medical researchers enjoy the benefits of digital data
techniques for form design and data quality are expensive and
labour-intensive. We propose a new data-driven approach to form for research and analysis.
design, execution (filling) and quality assurance. We demonstrate We have built a system called U SHER that maximizes data
U SHER, an end-to-end system that automatically generates data quality at entry-time using statistical data modeling, dynamic
entry forms that enforce and maintain data quality constraints interfaces, and collaborative insight. Guided by prior data,
during execution. The system features a probabilistic engine that U SHER learns probabilistic relationships in the data to train a
drives form-user interactions to encourage correct answers.
model, which is then applied to automatically generate forms
I. I NTRODUCTION with the appropriate constraints. U SHER then provides real-
Governments, companies, and individuals routinely make time feedback during the data entry process to dynamically
important decisions based on inaccurate data stored in suppos- guide (or usher) the user toward better data quality.
edly authoritative databases. In healthcare, a simple error may Based on a list of form questions and a sufficient set
have fatal consequences. While data quality can be addressed of answers, U SHER optimizes the form’s question-ordering
at every stage of the data life-cycle, from creation to archival, and layout, mimicking survey design principles. During form
we believe that entry-time is the first and best opportunity to entry, U SHER provides dynamic data-quality feedback to the
improve the quality of manually-entered data. There is much user. When the user enters a value, U SHER automatically
prior work on improving the quality of data that already resides decorates the interface with hints and warnings if the answer
in a database [1]. However, relatively little attention has been is deemed “risky.” Decoration choices are probabilistically
paid to improved techniques for data entry. guided, and include auto-complete, correctness-thermometers,
Survey design [2] has long informed the design of data entry warning/error flags, and other scented widgets [5]. U SHER
forms, applying principles for data encodings, constraints, also invites the user to write and view comments about
and validation rules. For electronic forms, quality assurance form questions or data instances for and by other users.
during entry has centered on the ubiquitous and costly practice Finally, U SHER mimics double-entry by choosing to re-ask
of double-entry [3]. Current standards have failed to take questions with responses likely to be erroneous, based on the
advantage of new technology: pervasive cellular networking probabilistic model.
and low-cost mobile devices allows even remote users to II. D EMONSTRATION
interact with data entry systems that could potentially provide
rich feedback. Our demonstration will show U SHER’s ability to approxi-
For organizations with limited resources, existing standards mate expert form design and double-entry based only on prior
are neither practical nor attainable. In such settings, designing data, both on a PC and a mobile device. Using a real dataset
data collection instruments is too often an ad hoc practice, from a rural health organization, users will be able to 1)
consisting of mapping desired information elements to a set automatically extract training data from a Microsoft Access
of entry widgets (text fields, combo boxes, etc.), guided only database; 2) refine the automatically designed form; and 3)
by the designer’s intuition. According to recent work on data execute the forms with and without smart decorations and
collection in resource-poor settings, lack of expertise and quality assurance.
difficulty of remote data collection are the chief obstacles R EFERENCES
to high data quality [4]. In our previous fieldwork with a [1] J. M. Hellerstein, “Quantitative data cleaning for large databases,” United
well-funded HIV/AIDS treatment program in East Africa, we Nations Economic Commission for Europe, 2008.
found that little thought was given to form design, and a [2] R. M. Graves, F. J. Fowler, M. P. Couper, J. M. Lepkowski, E. Singer,
and R. Tourangeau, Survey Methodology. Wiley-Interscience, 2004.
haphazard double-entry program bottlenecked the data entry [3] S. Day, P. Fayers, and D. Harvey, “Double data entry: what value, what
process to a substantial degree; in fact, the program’s health price?” Controlled Clinical Trials, 1998.
[4] J. V. D. Broeck, M. Mackay, N. Mpontshane, A. K. K. Luabeya,
K. Chen, N. Conway and J. M. Hellerstein are with the Computer M. Chhagan, and M. L. Bennish, “Maintaining data integrity in a rural
Science Division, University of California at Berkeley (email: {kuangc, nrc, clinical trial,” Controlled Clinical Trials, 2007.
hellerstein}@cs.berkeley.edu). [5] W. Willett, “Scented widgets: Improving navigation cues with embedded
H. Chen is with the Computer Science and Artificial Intelligence Labora- visualizations,” IEEE TVCG, 2007.
tory, Massachusetts Institute of Technology (email: harr@csail.mit.edu)
H. Dolan and T. S. Parikh are with the School of Information, University
of California at Berkeley (email: {dolan, parikh}@ischool.berkeley.edu)
Manuscript accepted Feb 10, 2009.
488
IWB4D – Interactive Whiteboards for

Development
John Traxler and Lee Griffiths
including classroom practice and distant, remote or on-site

Abstract — The work presented here describes the concept of mobile teaching. The topics covered will include a general
ubiquitous interactive classrooms by demonstrating emerging, introduction to interactive whiteboards, an introduction to the
low-cost presentation technologies including the Nintendo Wii capability and functions of Nintendo's Wii, the development
and palm-sized projectors. The authors also discuss how they can
be used to promote better learning through ad-hoc digital
of a personal interactive whiteboard, the use multiple input
interaction in traditional classrooms and the field. devices and true mobility, a critique and discussion of
whiteboard disadvantages, the set-up and download of the
Index Terms — active white boards, mobile learning, whole system, and how to enhance teacher mobility in diverse
class teaching settings.
REFERENCES
I. WHITEBOARDS [1] BECTA (2003). What the research says about interactive whiteboards.
T here is considerable literature, for example [1], [2], [3],

[4], [5] & [6], showing the extent to which whole class
teaching in core subjects can be enhanced by interactive
Coventry, England: British Educational Communication and Technology
Agency. Retrieved 10 January2009 from
http://www.becta.org.uk/page_documents/research/wtrs_whiteboards.pd
f
whiteboards. These technologies are however extremely [2] Glover, D., & Miller, D. (2001). Running with technology: the
pedagogic impact of the large-scale introduction of interactive
expensive and thus, untried in a development context.
whiteboards in one secondary school. Journal of Information for Teacher
The Nintendo Wii games console has a unique form of Education 10 (3), 257–276.
interactivity. Since its release programmers have developed [3] Glover, D., & Miller, D. (2002, November 11). The introduction of
example systems which implement futuristic interfaces with interactive whiteboards into schools in the United Kingdom: Leaders,
led, and the management of pedagogic and technological change.
the Wii console’s hand controller or Wiimote which can track International Electronic Journal for Leadership in Learning 6 (24).
moving objects with relative ease. Retrieved January 5, 2009 from
One example of interest to educationalists working in ICTD http://www.ucalgary.ca/~iejll/volume6/glover.html
[4] Glover, D., Miller, D., Averis, D., & Door V. (2005) The interactive
is using the device to implement an interactive whiteboard. whiteboard: a literature survey. Technology Pedagogy and Education 14
Our demonstration will illustrate teacher mobility and student (2), 155–170.
interaction on a very low budget by demonstrating the ad hoc [5] Hall, I., & Higgins, S. (2005). Primary school students’ perceptions of
interactive whiteboards. Journal of Computer Assisted Learning 21 (2),
use of the Wii device to implement interactivity. 102-117. Retrieved February 2, 2007, from EBSCOhost Professional
Participants will see a live demonstration of a personal Development Collection. Hall, I., & Higgins, S. (2005). Primary school
interactive whiteboard using the Nintendo Wii and discuss students’ perceptions of interactive whiteboards. Journal of Computer
Assisted Learning 21 (2), 102-117. Retrieved February 2, 2007, from
teacher mobility and student interaction in a development
EBSCOhost Professional Development Collection.
context using emerging technologies and how to set the [6] Tanner H., Jones S., Kennewell S., & Beauchamp G. (2005) Interactive
system up in their own working environment. whole class teaching and interactive white boards. In Building
Connections: Theory, Research and Practice: Proceedings of the Annual
Conference held at RMIT, Melbourne, 7th-9th July 2005, 720–
727.Pymble, N.S.W.: The Mathematics Education Research Group of
II. THE DEMONSTRATION Australasia. Retrieved January 12, 2009 from
http://www.merga.net.au/documents/RP832005.pdf.
The learning goals of the demonstration are that the
participants will gain the technical skills and competence to
procure and set up an inexpensive interactive whiteboard and
will engage with ideas for improved and innovative teaching
John Traxler is Visiting Scientist, Meraka Institute, Pretoria; Director,

Learning Lab, Shropshire, European Chair ICTD Qatar & Reader in Mobile
Technology for e-Learning, University of Wolverhampton;
john.traxler@wlv.ac.uk, +44 (0)1902 518591
Lee Griffiths is Faculty Learning Technologies Fellow and Lecturer in
Computer Science, University of Salford; L.S.Griffiths@salford.ac.uk, +44
(0)161 2953658
489
Livestock, Learning and Diagnostics: New

Directions in Veterinary Tele-medicine
Jun Yu and Claire Heffernan
Abstract - It has been estimated that 2/3rds of the individuals focus on mastitis was twofold. First, mastitis, historically, is
living on less than $2/day own or keep livestock. However, difficult for farmers both in the North and in the South to
research has illustrated the difficulties that this population face in diagnose. Second, delayed diagnosis has a detrimental impact
obtaining accurate and relevant information regarding animal
health. Therefore, the demonstration offers two tools to support
on milk production and consequently income. As such, the
the animal health needs of the poor in India. First, the creation of disease has a large impact on the livelihoods of poor dairy
a mobile phone based knowledge transfer tool is described and producers, particularly in India [1].
second, a mobile phone-based diagnostic tool to enhance the The reason for choosing a mobile phone application is
capacity of the poor to diagnose diseases is also detailed. The twofold. First, mobile phones are better from both a cost and
demonstration of the tools will support the ‘hands-on’ exposure of battery life perspective [3]. Second, by connecting via
conference goers with problems faced by the poor and the role of
Bluetooth to the Livestock Guru kiosk, the MDS can link to
ICTs in their solution.
learning information provided by the Guru in relation to the
Index Terms - Information Delivery System, Livestock Diseases, livestock disease in question.
Mobile Phone, Poverty Nevertheless, a version of the MDS entirely contained within
a mobile phone was created and will be demonstrated as many
poor livestock keepers may be too remote from a server (The
I. INTRODUCTION Livestock Guru kiosk), in this manner, a farmer may take a
ICTs are critical to addressing the animal healthcare needs of picture of his or her animal and the mobile MDS will provide a
poor livestock keepers in the South [4]. First, such tools may diagnosis. The software was created based upon a combination
be applied as a means of knowledge transfer and secondly, of prototype matching and statistical pattern recognition.
ICTs can help redress historical gaps in animal health service
II. DEMONSTRATION
delivery to the poor. Indeed, research has identified that women
and children, while primarily livestock care-takers are often the In the demonstration, a mobile phone based on Symbian
most excluded from knowledge transfer [5]. Equally, disease S40/60 OS will be utilized to display the Mobile Livestock
diagnosis is particularly problematic for the poor [5]. Guru. Its application among poor farmers in India (including
Therefore, in recent years there has been an increasing interest women and children) will be detailed as part of the display.
in tele-veterinary medicine to address animal healthcare gaps in Additionally, use of the MDS will also be shown. As such, a
southern countries. Mobile diagnostic technologies are a display of symptom pictures will be offered and conference
rapidly emerging field in tele-medicine (both human and goers may utilize a mobile phone to take and send symptom
animal) and are defined by two key characteristics: mobility pictures to a kiosk computer to make a diagnosis of a sick
and remote diagnosis [2]. animal. Alternatively, a mobile phone stand-alone version
In the following demonstration, the authors display three MDS will be also available for conference goers to test.
linked and inter-related tools to address these issues: first, a Further, a video of the tool in use in India will also be provided.
mobile phone version of the Livestock Guru, a multi-media
learning program for poor farmers in the South; second, REFERENCES
multi-media learning games for children and third, a Mobile [1] C. Heffernan, D. Pilling, J. Yu and Y. Lin, “Listening to the Voices of the
Diagnostic System (MDS), which supports the diagnosis of Poor: Enhancing evidence-based decision-making in livestock
development”, Nottingham University Press, Nottingham, United
Mastitis among poor dairy producers in India. The MDS Kingdom, 2008.
enables farmers to take a picture of their sick animal by a [2] M. Papageorges, P. Herbert, “Telemedicine using standard internet
mobile phone and then send this picture to a kiosk, which technologies (telemedicine for less than $1000)”, Clinical Techniques in
Small Animal Practice, vol. 16, no. 2, pp. 102-107, May 2001.
subsequently makes the diagnosis. The reason for the initial [3] L. Naismith, “Literature Review in Mobile Technologies and Learning: A
Report for NESTA Futurelab”. Bristol: NESTA Futurelab, 2004.
[4] S. Warren, “A distributed infrastructure for veterinary telemedicine, in
Manuscript received February 2, 2009. proceeding of the 25th annual International Conference of the IEEE
Jun Yu is a PhD student at the Livestock Development Group, School of EMBS, September 17th-21st, Cancun, Mexico, 2003.
Agriculture, Policy and Development, University of Reading, UK (e-mail: [5] Yu, J. “Improving Development Information flows: The creation of ICTs
aar04jy@ reading.ac.uk). for poverty alleviation in the livestock sector”, Thesis (submitted). The
Claire Heffernan is the director of the Livestock Development Group, University of Reading, Reading, UK.
(Tel.: +44 1189 318213, e-mail: c.l.heffernan@reading.ac.uk).
490
Metamouse: Multiple Mice for Legacy Applications

Kurtis Heimerl (kheimerl)∗ , Divya Ramachandran (divya)∗ , Joyojeet Pal† , Eric Brewer (brewer)∗ and Tapan Parikh‡
∗ Computer Science Division † School of Information ‡ School
of Information
University of California, Berkeley University of Washington University of California, Berkeley
Berkeley, CA 94720-1776 Seattle, WA 98195-2840 Berkeley, CA 94720-1776
(name)@cs.berkeley.edu joyojeet@u.washington.edu parikh@ischool.berkeley.edu
Abstract—Traditional Single Display Groupware (SDG) solu- work. To do this mapping effectively, we have to map mouse
tions have been used to create software for developing regions. movements as well as mouse clicks.
SDG allows for greater utilization of the limited infrastructure
available in such areas. This technology is particularly appropri- A. Movement
ate in classrooms where groups of children often use educational
software on a single computer. However, SDG has faced several To map the many cursors down to one, we use a well known
challenges working with legacy applications. Our technology, SDG technique. We average the mouse locations, and display a
called metamouse, takes a step toward an integrated multi-user metacursor at that position. With this, each user’s actions have
application by allowing users to collaborate within unmodified a small, but noticeable, impact on the metacursor’s location.
legacy educational software. Likewise, if all of the users agree on the position of the cursor,
I. P ROBLEM S TATEMENT the metacursor will be at that same location. This encourages
collaboration, as the users must discuss and agree to place the
Several million children, especially in the developing world,
metacursor where they want it.
never use a computer without a partner sharing the mouse
and keyboard [2]. Single Display Groupware techniques [4] B. Clicking
are particularly relevant to this scenario, allowing for users to We can’t allow the users to click at any time, as the
more effectively share computing resources by using multiple metacursor may be at an inappropriate location. To solve this,
mice. Existing research has shown that these techniques are we only allow clicking when the mice are in close proximity
intuitive, usable, and beneficial in the developing world [3]. to each other. When distant, all of the clicks are ignored. The
Despite these encouraging early findings, two important intuition is that if the mice are close to each other, then the
barriers remain in deploying shared input technologies in real users have agreed on the correct location to make progress,
world settings. First, most legacy software would have to and clicking should proceed. This encourages collaboration;
be significantly modified to effectively use multiple mice, all users must agree on a location before making progress.
but among other issues, many business challenges exist in
doing this. The second barrier, of interest from the HCI III. D EMONSTRATION
perspective, is that of efficiently encouraging coordinated on- We are going to demonstrate the point-and-click metamouse
screen decision-making using multiple mice. Most trials of system running on Microsoft Windows XP. We will run a
multimouse, while emphasizing the importance of collabora- set of educational games from the following vendors using
tion, have been impeded by working within a “fastest-finger- metamouse and four individual mice.
first” race-clicking, or repetitive click modes which require all • The Azim Premji Foundation [1]
users to click on the same link for the function to continue. Our • Disney online
challenge was to work with legacy software, and yet provide • Leapfrog
an alternate way for allowing coordinated clicking to capture
the proven learning gains of collaboration. R EFERENCES
II. S OLUTION [1] Azim Premji Foundation. http://www.azimpremjifoundation.org.
[2] J. Pal, U. S. Pawar, E. A. Brewer, and K. Toyama. The case for multi-user
To solve these problems, we developed the metamouse. design for computer aided learning in developing regions. In WWW ’06:
The metamouse does two things. First, it maps multiple user Proceedings of the 15th international conference on World Wide Web,
pages 781–789, New York, NY, USA, 2006. ACM.
mice and cursors down to one metacursor that interacts with [3] U. S. Pawar, J. Pal, and K. Toyama. Multiple mice for computers
the application. This allows for the use of unmodified legacy in education in developing countries. In IEEE/ACM Intl Conf. on
applications. Secondly, we have to make sure that this mapping Information and Communication Technologies for Development, 2006.
[4] J. Stewart, B. B. Bederson, and A. Druin. Single display groupware:
is intuitive and encourages collaboration among users. a model for co-present collaboration. In CHI ’99: Proceedings of the
We began by trying to create an effective mapping for point- SIGCHI conference on Human factors in computing systems, pages 286–
and-click applications. Tackling other tasks is left as future 293, New York, NY, USA, 1999. ACM.
491
Mobile Phone Job Services:

Linking Developing-Country Youth
with Employers, via SMS
Amber Houssian, Mohammad Kilany, and Jacob Korenblum
Abstract—With its low cost and high penetration rates, mobile people than conventional labor market methods. In its
technology is uniquely poised to address developing-country demonstration session, team members will use mobile phones
labor market inefficiencies, by boosting young job-seekers’ access to show the end-user software experience from job-
to employment information. Simple services like the SMS-based seeker/employer vantage points, illustrating how the
“job matching” application developed by Palestinian-Canadian
application matches youth with employers in five easy steps
NGO Souktel have transformed local labor markets--by giving
young job-seekers real-time information about local jobs, and
connecting them directly with employers. 1) Through short SMS surveys, youth users create
basic “mini-CVs”. These include data on age,
I. CONTEXT: JOB MARKET GROWTH, BUT NO GOOD WAY location, skills, education level, and career
FOR JOB-SEEKERS TO FIND INFORMATION ABOUT WORK interests.
2) At the same time, employers create simple SMS
I N the past decade, developing-country economies have
grown impressively each year—a promising prospect for
local youth, who often make up the bulk of new labour
“job ads” through a similar SMS Q & A process.
3) Both sets of data are sent directly from users’
phones to a central database.
market entrants. However, the reality is much less rosy: While 4) Database matches youth “mini-CVs” with relevant
GDP is growing, young workers are still failing to find “job ads”—pairing identical information from
available jobs in key fields. In emerging markets, one of the youth and employers.
main challenges is basic connectivity: How to give young 5) Users then receive “job-match” messages directly
job-seekers good information about available work, and how to their mobile phones—with details about job
to link them efficiently with employers needing staff. Despite opportunities or potential staff, along with phone
the acute need for labour market information/’matching’ numbers for follow-up interviews. Users can also
mechanisms, few solutions have emerged. Most job-seekers browse for jobs or staff from their phone handsets.
lack web access, government initiatives are minimal, schools
have no resources, and private-sector efforts are limited. III. SIGNIFICANCE: BRIDGING THE DIGITAL DIVIDE;
REDUCING UNEMPLOYMENT; PROMOTING GROWTH.
II. NOVELTY: JOB-SEEKERS AND EMPLOYERS MATCHED BY
MOBILE PHONE. Fast, easy to use, and widely accessible, phone-based job
information services like Souktel offer four significant
Mobile technology, however, is uniquely poised to ‘leapfrog’ advantages over traditional job-find/hiring approaches:
many of these obstacles and help job-seekers in developing
countries find key labor market information. One such • Faster, better access to information: Customized
solution is a new software application developed by engineers matching of youth with employers and “anytime,
at Palestinian NGO Souktel. The application bridges anywhere” access from mobile phone reduces job/staff
information gaps by linking youth with job data through a search time from months to hours.
simple text message-driven process. Piloted in 2006, the
• Gender Sensitive: Virtual (non face-to-face)
technology has now helped over 2,000 youth find work and
communication creates new, safe ways for women in
internships through a process that saves users time and
traditional families to connect with female and male peers
resources, while enabling them to connect with far more
in the labour market.
• Wider Reach – Improved access to remote/rural
Manuscript received January 28, 2009. communities: Mobile phone access option facilitates
Amber Houssian, M.B.A., is the Regional Manager- Middle East for
Souktel, and a past Strategic Planner at leading Canadian marketing firm
service use in areas with no Internet or poor
Cossette (e-mail: ahoussian@souktel.com). communication/transport infrastructure.
Mohammad Kilany, B.Sc., is a co-founder of Souktel and a 2009 Synergos • Ease of use in Conflict/Post-Conflict Zones: No new
Arab World Social Innovation Fellow (e-mail: mkilany@souktel.com). hardware required; no need for regular electricity or
Jacob Korenblum, M.Ed., is a co-founder of Souktel and a past Harvard
University Reynolds Fellow in Social Enterprise (e-mail:
Internet connection. System can be scaled up or down
Jacob@souktel.com). quickly and cheaply, in response to usage patterns and
user demands.
.
492
MultiMath: Numeric Keypads for Math Learning

on Shared Personal Computers
Sunil Garg, Charlotte Robinson, Clint Tseng, Heather Underwood, Richard Anderson, Joyojeet Pal
University of Washington
press, as corresponding to a specific device identifier. Using

Abstract—Past research has shown the benefits of increased this information, we can correlate each key input with a
engagement and collaboration for learning using multiple inputs specific user and react accordingly.
for children sharing computers. Taking the work forth beyond
We employ USB numeric keypads because each unit costs
clicks, we demonstrate the use of inexpensive numeric keypads as
multiple input devices in shared computing scenarios, as many approximately four dollars and can be used with virtually any
have argued are well-suited to the developing world, and show its modern computer, thus allowing for low-cost deployment in
utility in mathematical learning for young children. resource-constrained environments. Moreover, they offer a
high degree of versatility, as they can be used for both
I. INTRODUCTION numeric and directional input, and their keys can be re-
W HILE a significant amount of work exists in the use of

gaming in education, there is little literature on the
value of multi-player gaming environments in developing
mapped and labeled for specialized applications. Previous
applications have used soft keyboards on the display which
depend on mouse input (Moraveji, Kim et al. 2008) – we
world contexts. The work with multiple mice (Pawar, Pal et al. demonstrate that hardware input devices offer several
2006) has succeeded in showing some gains with multiple- advantages, such as by allowing faster input and by saving
choice environments for young children constrained to sharing valuable screen real estate for other purposes.
computers, and subsequent work has shown the impacts of For display purposes, we split the screen area into portions
split screens (Moed, Otto et al. 2009) in increasing the reserved for each participant, allowing for individualized
collaboration and engagement for learning. MultiMath is a attention. Not only does this enable each player to complete
multi-player educational game for basic arithmetic skills. problems presented to them at their own pace, but this also
Utilizing input via multiple numeric keypads and a split- allows us to introduce adaptive questioning in order to best
screen display, the application allows for multiple individuals address the skill level of each individual.
to participate at their own speed and skill level while
simultaneously introducing competition between players. The
split screen concept enables the application to provide each III. DEMONSTRATION
player with personalized curriculum and also allows for the In this demonstration we will present an initial version of
use of adaptive questioning. the MultiMath application, including multiple keypad input,
split screen display, and a simple problem set. Using the
implementation described above, the application will comprise
II. TECHNICAL IMPLEMENTATION of a game in which four players can individually complete
The MultiMath game is implemented as a desktop application randomly generated arithmetic questions and receive visual
using the Microsoft .NET framework. While Windows does feedback of their progress and ranking within the game,
support multiple mice and keyboards, it does not make any through display elements such as a progress bar. At the
distinction between devices by default, as all mice control the demonstration, we also plan to display results from
same cursor and all keyboards provide a combined stream of preliminary field tests with children during our iterative
input. However, Windows offers a RawInput API which can design process.
be used to simultaneously interface with multiple human
interface devices by identifying each event, such as a key
REFERENCES
Sunil Garg, Charlotte Robinson, Clint Tseng, Heather Underwood, and [1] Moed, A., O. Otto, et al., "Reducing Dominance Behavior in Multiple-
Richard Anderson are with the Department of Computer Science and Mouse Learning Activities," 8th Int. Conf. on Computer Supported
Engineering at the University of Washington, Seattle, WA 98195 USA. Collaborative Learning (CSCL 2009), Submitted, Rhodes, Greece.
(email: {skgarg, solstinc, cxlt, hmu2, anderson}@cs.washington.edu) [2] Moraveji, N., T. Kim, et al., "Mischief: Supporting Remote Teaching in
Joyojeet Pal is with the Center for Information and Society at the Developing Regions," Proc. 26th Annu. SIGCHI Conf. on Human
University of Washington, Seattle, WA 98195 USA. (email: Factors in Computing Systems (CHI 2008), Florence, Italy.
joyojeet@u.washington.edu) [3] Pawar, U., J. Pal, et al., "Multiple Mice for Computers in Education in
Developing Countries," 2006.
493
A New Generation of
Open Source Data Collection Tools
Yaw Anokwa, Carl Hartung, Adam Lerer, Brian DeRenzi, Gaetano Borriello
Abstract—Organizations in low income regions need tools JavaRosa and AndroidRosa. We will also demonstrate Server-
for collecting and reporting on data. Existing paper-based ap- Rosa, software which processes and hosts data submitted by
proaches are often slow and incomplete when compared to data mobile phones.
collection tools on mobile devices. In this demonstration, we
present the authors’ Android-based system in addition to broader To demonstrate JavaRosa, we will use a few examples of
work by the OpenRosa Consortium – a community dedicated phones which support Java 2 Platform Micro Edition (J2ME).
to building open source, standards-based tools for mobile data The Nokia 3110c and 6085 both sell for around $100 and are
collection, aggregation, analysis, and reporting. widely available in low income regions. The Nokia N95 is
Index Terms—data collection, mobile phones, ICT available for around $400 and is one of the most advanced
smartphones available. We will demonstrate form filling and
submission functionality designed for community health care
I. I NTRODUCTION workers on these phones.
variety of tools are urgently needed to address the lack AndroidRosa will be demonstrated on the T-Mobile G1
A of data in low-income countries. This data is required for
decision making and research in many sectors. Due to recent
phone running Android. Android is an open source operat-
ing system developed by Google that is more feature-rich
technological advancements, there is great excitement for and easier to develop for than J2ME. AndroidRosa uses the
using mobile devices to address current gaps in information. JavaRosa core code and leverages Android’s unique abilities.
Phones and PDAs have proven to be dramatically faster and We will demonstrate form filling which builds on the G1’s
more complete than traditional methods of pen and paper, and input devices (touchscreen and keyboard) and sensors (SMS,
can be more accurate and less expensive as well[5], [6]. location and images).
The OpenRosa Consortium[1] is a group working together To collect the completed form data sent from the phones,
to foster open source, standards-based tools for mobile data we will demonstrate ServerRosa. This server component is
collection, aggregation, analysis, and reporting. Participants in- written in Java and can run on local servers as well as in
clude AED-Satellife, Cell-Life, CIDRZ, D-Tree International, the cloud. Phones can submit completed form data to the
DataDyne, Dimagi, Google, MRC-SA, University of Bergen, server, which can export submitted data and generate reports
Makerere University, and the University of Washington. The if necessary. The server can also summarize and aggregate
group has active developers in Bangladesh, Kenya, India, submission data and send it to a separate Android application,
Norway, Pakistan, South Africa, Tanzania, Uganda, and the which supervisors can use to monitor data collection.
United States primarily working on JavaRosa – an open source
framework for data collection on Java-enabled phones. III. C ONCLUSION
By developing open source solutions and conforming to The OpenRosa Consortium is close to consolidating the
standards based on the W3C XForms specification, OpenRosa necessary functionality for the first public release of JavaRosa,
projects can interoperate their components. So while many AndroidRosa and ServerRosa. Together, we are working to-
of the member organizations are currently using JavaRosa wards solutions to allow organizations with minimal technical
for form collection and management systems, others are capacity to collect data using a variety of mobile phones.
building complex solutions like clinical trial software[4], data
reporting on servers[3], and even community health worker ACKNOWLEDGMENT
management[2]. The authors thank all members of OpenRosa, especially
Neal Lesh, Drew Roos and Clayton Sims.
II. D EMONSTRATION
R EFERENCES
In this demonstration, we will present some of the data
collection and management tools under development by the [1] OpenRosa, http://openrosa.org
[2] CommCare, http://commcare.cs.washington.edu
authors for the OpenRosa Consortium. We will show form [3] Gather, http://blog.gatherdata.org
filling, submission and monitoring on mobile phones using [4] OMEVAC, http://epihandy.org/index.php/OMEVAC
[5] K. Shimira et al., The use of personal digital assistants for data entry at
Yaw, Carl, Brian, and Gaetano are affiliated with the University of Wash- the point of collection in a large household survey in southern Tanzania,
ington and Adam is affiliated with the Massachusetts Institute of Technology. Emerging Themes in Epidemiology, 2007.
This work was done while Yaw, Carl, Adam and Gaetano were interns at [6] T. Parikh, et al., Mobile Phones and Paper Documents: Evaluating a New
Google. All authors can be reached at yanokwa@cs.washington.edu. Approach for Capturing Microfinance Data in Rural India, CHI, 2006.
494
RuralScope: An Information System for

Tracking Rural Disbursements
1
Sai Gopal Thota, Rabin Patra, Murali Medisetty, Sivananda Reddy, Vivek Mungala, Joyojeet Pal
Abstract—Tracking and ensuring disbursements make it to their II. IMPLEMENTATION

intended recipients is among the problems development
professionals are most concerned about. To examine the potential RuralScope is an information system that helps make the audit
monitoring role playable by technology, we present RuralScope, process more efficient and less risky for the auditor by
an information repository designed to handle the information arranging the data and making it available on disconnected
needs of a newly started welfare scheme offering daily wage devices, replacing cumbersome muster rolls and dramatically
payments to over 31 million families in rural India. Since the reducing the amount of time an auditor needs to spend on the
problem of monitoring development extends beyond project ground. On the front end, RuralScope offers a single gateway
specific needs to a general sense of awareness, RuralScope works into all the information on outlays and disbursements for
on two levels. First, for project auditors, RuralScope acts as an NREG allowing researchers the ability to identify trends using
easily searchable, disconnected database of disbursements that graphs and maps and annotate visualizations. Finally, by
helps speed up the audit process. Second, for a more general allowing researchers to plug in relevant demographic,
audience, an organized and annotatable front end for casual
agricultural produce and related data from other sources,
viewers, policy-makers, and researchers alike to keep track of
documents and data related to the project RuralScope.
RuralScope makes it easier for researchers and casual users
alike to contextually understand the disbursements. The
I. INTRODUCTION descriptions for all visualizations and annotations created by
users as stored as structured XML, and we use here data from
M UCH recent interest in development has focused on the

need for robust systems that can enable efficient
the state of Andhra Pradesh for 2007 disbursements
representing 28000 villages and over 6 million workers.
governance of public funding in various parts of the world.
Increasing transparency within the functioning of III. DEMONSTRATION
governments and public finances is seen as an important goal
for development (Qureshi 2004 [1], Moody-Stuart 1997 [2]) We will demonstrate use cases for casual online users with
Gaps in adequate information, legal loopholes in the limited understanding of the NREG system or database
implementation of such schemes have been cited as reasons through queries to the RuralScope database that can create
for weakened monitoring of corruption several developing visualizations. For sophisticated users familiar with the NREG
nations (Levin and Satarov 2000[3]). The National Rural or activist/monitors, we demonstrate a working model for
Employment Guarantee Act (NREGA) was passed by the audit queries to trace by individual job card holder or family
Government of India in 2005 to address the problems of ID exactly how much a family has been paid over a certain
seasonal employment in Rural India. The scheme broadly time period. To emphasize the value addition in the audit
offers 100 days of employment at minimum wage to every process that RuralScope brings, we plan to contrast the new
rural household through projects that create local interface with the currently prevalent muster roll data outputs.
infrastructure, often roads or schools. The scheme officially This feature is specifically targeted at making social audits
already reaches 31 million households, making it the largest quicker. Finally, we would like to demonstrate the
welfare administration scheme ever in the world. But within disconnected version of RuralScope designed for a mobile
the first few years of its progress, many questions have arisen computing device, on which data can be collected or viewed
on disbursement of the funds and the need for better from local storage as needed, and thereafter synchronized in
transparency. Two kinds of problems have been highlighted: areas of network connectivity. This last feature has much
first the misrepresentation of welfare payments and siphoning relevance to several prevalent ICTD project domains
of funds. A second, and more fundamental issue faced by rural including delay tolerant networking and OpenMRS
development issues, has been the limited coverage of the
scheme and its impacts by media. media. The only effective REFERENCES
way of tracking disbursements in NREG has been through [1] Qureshi, Z. (2004). Millenium Development Goals and Monterrey
social audits, perilous processes in which activists have been Consensus: from vision to action. ABCDE Conference of the World
known to get murdered during the time they spend in actual Bank, Monterrey, Mexico.
audits as they make a lot of enemies during the process of [2] Moody-Stuart, G. (1997). Grand corruption: how business bribes
damage developing countries.
recovering the misused funds through these audits.
[3] Levin, M. and G. Satarov (2000). "Corruption and institutions in
Russia." 16(1): 113-132.
1
Sai Gopal Thota(thota_gopal@daiict.ac.in), Rabin Patra(rkpatra@berkeley.edu) , Murali Medisetty(murali_krishna_2006@daiict.ac.in), Sivananda
Reddy(sivananda_reddy@daiict.ac.in), Vivek Munagala(pavan_v_munagala@daiict.ac.in), Joyojeet Pal(joyojeet@washington.edu)
1
495
T-Cube Web Interface in Support of

Real-Time Bio-surveillance Program
Artur Dubrawski, Maheshkumar Sabhnani, Michael Knight, Michael Baysek, Daniel Neill,
Saswati Ray, Anna Michalska and Nuwan Waidyanatha
pilot project currently under development in Sri Lanka and

Abstract—T-Cube Web Interface is a generic tool to visualize Tamil Nadu state of India. It will consist of information
and manipulate large scale multivariate time series datasets. The gathering system based on mobile handheld devices and
interface allows the user to execute complex queries quickly and wireless networking, and of a specialized IT system designed
to run various types of statistical tests on the loaded data. We
show its utility in an important application scenario: real-time
for automated surveillance of the incoming data and for
bio-surveillance system designed to support rapid detection and interactive navigation through data and extracted information.
mitigation of bio-medical threats in developing countries. In this demonstration, we present a prototype application of
the T-Cube Web Interface in monitoring of public health data
Index Terms—bio-surveillance, event detection, interactive in Sri Lanka and India.
analytics, data cubes. T-Cube Web Interface [1] is a generic tool for interactive
analysis, visualization and manipulation of large scale
I nfectious diseases such as tuberculosis or avian influenza
pose a persistent threat to human population all over the
world. Their mitigation and containment is especially difficult
multidimensional temporal and spatio-temporal datasets
commonly encountered in public health domain. It allows the
users to execute complex queries quickly and to perform
in the least developed countries, where existing surveillance various types of statistical tests against the loaded data. The
systems cannot gather and process the necessary information underlying data representation technology, the T-Cube [2], is
in time for public health authorities to effectively respond to an in-memory data structure designed to improve the response
rapidly developing bio-events. For example, the system time to ad-hoc time series queries against large datasets. T-
currently in use in Sri Lanka is paper-based and relies on the Cube has been tested on synthetic and real-world datasets
postal service to communicate community health reports to the containing millions of records and hundreds of dimensions.
central office for analysis, which may take even ten days. Results show that its response time can be orders of
Communication latency is exacerbated by limitations of magnitude shorter than that of the state-of-the-art commercial
analytic resources (in terms of their quantity as well as database tools. The attainable speedups have been shown to
training) available to process the available information. We qualitatively change the way the users interact with their data,
believe that that affordable, existing information and leading to better situational awareness, comprehension, and
communication technologies can play the key role in interpretability of data and results of analyses. It has already
collecting, communicating and analyzing the relevant data, been successfully used in equipment maintenance, logistics
enabling reliable and timely identification of emerging threats and food safety domains. We intend to leverage the benefits of
and, therefore, effective containment of their consequences. T-Cube technology in RTPB.
The International Development Research Centre of Canada Demonstrated capabilities of the tool include geospatio-
supports the Real-Time Bio-surveillance Program (RTBP), a temporal visualization of syndromic data, navigation through
different levels of data aggregation, and selected statistical
Manuscript received January 30, 2009. This work was supported in part by analyses including spatial scan for rapid detection of emerging
the International Development Research Centre of Canada (p;roject number
105130), the U.S. Department of Agriculture (award 1040770), Centers of outbreaks of diseases. Presentation emphasizes user-perceived
Disease Control and Prevention (award R01-PH000028), National Science utilities due to computational efficiency of the analytic
Foundation (under grant IIS-0325581) and by the U.S. Department of Defense algorithms and the underlying data representation technique.
(AFOSR award FA8650-05-C-7264).
A. Dubrawski, M. Sabhnani, M. Knight, M. Baysek, D. Neill, and S. Ray
are with the Auton Lab, Carnegie Mellon University, Pittsburgh, PA 15213 REFERENCES
USA (Corresponding author: A. Dubrawski, phone: 412-268-6233; fax: 412- [1] S. Ray, A. Michalska, M. Sabhnani, A. Dubrawski, M. Baysek, L. Chen,
268-6233; e-mail: awd@cs.cmu.edu) (Other authors email: J. Ostlund, T-Cube Web Interface: A Tool for Immediate Visualization,
sabhnani@cs.cmu.edu, mikek@cs.cmu.edu, mjbaysek@cs.cmu.edu, Interactive Manipulation and Analysis of Large Sets of Multivariate
neill@cs.cmu.edu, sray@cs.cmu.edu). Time Series, 2008 AMIA Annual Symposium, Washington, DC, 2008.
A. Michalska is with the Institute of Computer Science, University of [2] M. Sabhnani , A. Moore, and A. Dubrawski, Rapid processing of ad-hoc
Warsaw, ul. Banacha 2, 02-097 Warszawa, Poland (e-mail: queries against large sets of time series. Advances in Disease
michalsk@mimuw.edu.pl). Surveillance 2, 2007.
N. Waidyanatha is with LIRNEasia, 12 Balcombe Place, Colombo 08, Sri
Lanka (e-mail: waidyanatha@lirne.net).
496
Web Search over Low Bandwidth

Jay Chen, Lakshminarayanan Subramanian, and Jinyang Li
developed a system that explicitly exposes the slow network

Abstract— Web search and browsing have been streamlined connection to the user while providing features to make
for a comfortable experience when the network connection is fast. progress despite it.
Existing systems, however, are not optimized for scenarios where
connectivity is poor, as is the case for many users in developing
regions where fast connections are expensive, rare, or
unavailable. We examined the challenges faced by users behind II. DEMONSTRATION
one of these connections in a previous study, and found that the In this demonstration we show a prototype of our web
existing web interface is incapable of providing a good experience search system over a low bandwidth connection. Our system is
when the connection is extremely slow. In this demonstration we
present a prototype implementation of a web browsing system
comprised of two proxy servers: one local proxy that is behind
that incorporates what we learned. Our system helps the user’s the low bandwidth link, and one remote proxy that is well
search offline as much as possible, and in a single search query connected to the Internet. Requests are made at the local proxy
when information from the Internet is required. and sent to the remote proxy for processing and eventual
return of optimized results. This architecture in itself is not
Index Terms—Low Bandwidth, Search, Internet, Intermittent new, and similar to the TEK project [5]. What is novel about
Network
our system is that it has several unique features designed with
asynchrony as a primary consideration. Due to lack of space
we list only the search process along with a few of the most
I. INTRODUCTION
visible application level features to be demonstrated here.
S low, expensive, or non-existent Internet access is a fact of

life for many computer users in the developing world
where the physical infrastructure has yet to catch up to the
The web search interface provides local search and query
construction assistance. These two features enable users to
avoid using the network if the result may be satisfied locally; if
increasing demand. In these regions connectivity is often query cannot be satisfied locally then the search query has
intermittent due to power cuts [1], and bandwidth is generally already been iteratively refined to be as specific as possible.
an expensive and rare commodity because none of the The query construction interface has fields for users to
traditional wire-line connectivity solutions (fiber, broadband explicitly specify the type of response expected by the user.
and dial-up) are economically viable for rural regions with low The remote proxy uses this additional information to
user densities [2]. compress, filter, and sometimes prefetch results based on the
Existing work on bringing the Internet to the developing user’s preferences. Once a request is issued to the remote
world generally focuses on connectivity issues [3, 4], but even proxy, the local proxy also provides an estimated time until
after Internet connectivity is established via any of these request completion. Users are free to queue up requests and
physical channels, the link is slow. Systems typically attempt return after the queries are satisfied. Finally, after results are
to mask the poor connection from the application using some returned, they are integrated into the local cache and available
combination of traditional methods such as compression, to the user for browsing.
caching, filtering, and prefetching [5, 6]. However, when the
connection is extremely slow (hours or days for mechanical REFERENCES
backhauls) the iterative process of web search [7] becomes [1] S. Surana, R. Patra, S. Nedevschi, M. Ramos, L. Subramanian, and E.
impractical due to the long round trip time. The existing Brewer, “Beyond Pilots: Keeping Rural Wireless Networks Alive,”
literature is unclear on how to further improve the web NSDI, 2008.
[2] S. Mubaraq, J. Hwang, D. Filippini, R. Moazzami, L. Subramanian, and
browsing interactivity when the process is nearly or completely
T. Du, “Economic analysis of networking technologies for rural
asynchronous. Our work identifies this problem as pervasive developing regions,” Workshop on Internet Economics, 2005.
and unsolved in developing countries, and as a result we [3] K. Fall, “A Delay Tolerant Network Architecture for Challenged
Internets,” Computer Communication Review, 2003.
[4] R. Patra, S. Nedevschi, S. Surana, A. Sheth, L. Subramanian, and E.
Manuscript received January 28, 2009. Brewer, “WiLDNet: Design and Implementation of High Performance
J. Chen is with New York University, New York, NY 10003 USA (e-mail: WiFi Based Long Distance Networks,” NSDI, 2007.
jchen@cs.nyu.edu). [5] W. Thies et al., “Searching the World Wide Web in Low-Connectivity
L. Subramanian is with New York University, New York, NY 10003 USA Communities,” WWW, 2002.
(e-mail: lakshmi@cs.nyu.edu). [6] Loband, http://www.loband.org.
J. Li is with New York University, New York, NY 10003 USA (e-mail: [7] T. Joachims, L. Granka, Bing Pan, and G. Gay, “Accurately Interpreting
jinyang@cs.nyu.edu). Clickthrough Data as Implicit Feedback,” In ACM SIGIR, 2005.
497
Author Index
Abdulsalam, Ameer 478 Crow, Ben 212

Agarwal, Sheetal K. 56 Dasgupta, Tirthankar 201
Ahmed, Tanveer 447 Davis, James 212
Ali, Nosheen 447 DeRenzi, Brian 493
Anderson, Richard 492 Dias, M. Frederick 129, 478
Anderson, Ruth 233 Dias, M. Bernadine 129, 478
Anokwa, Yaw 493 Dolan, Heather 487
Aoki, Paul 35, 118 Donner, Jonathan 17
Asuntogun, A. 280 El Mahdi, A. 486
Aziz, Sherif M. 288 El-Moughny, Noura 478
Balakrishnan, Ravin 337, 406 Escobari, Marcela 17
Banks, Ken 484 Etherton, John 468
Barnard, Etienne 95 Fanaswala, Imran 478
Basavaraju, S. 481 Findlater, Leah 406
Basu, Anupam 201 Gandhi, Rikin 337, 406
Baysek, Michael 495 Garg, Sunil 492
Belousov, Sarah M. 129, 478 Ghadially, Rehanan 310
Bertrand, Anne 485 Ghazzawi, Wael 478
Best, Michael L. 192, 468 Gigler. Björn-Sören 267
Bhatnagar, Subhash C. 183 Golly-Kobrissa, Romain Tohouri 379
Biswas, Manna 437 Gong, Haijun 129
Blagsvedt, Sean 150 Griffiths, Lee 488
Borriello, Gaetano 233, 493 Grover, Aditi Sharma 95
Braa, Jørn 379 Gunawardane, Prabath 212
Brewer, Eric 490 Gupta, Aakar 482
Brunette, Waylon 233 Harmen, Saskia 479
Brunskill, Emma 74 Hartung, Carl 493
Bubrawski, Artur 495 Hefferman, Claire 480, 489
Burrell, Brenda 483 Heimerl, Kurtis 490
Bussell, Jennifer 173 Hellerstein, Joseph M. 487
Canny, John 139 Helmer, Scott 407
Chabossou, A. 392 Hersman, Erik 484
Chakraboty, Dipanjan 85 Ho, Melissa R. 35, 118
Chauhan, Himanshu 85 Hollow, David 27
Chen, Harr 487 Hopkins, Mary Ann 458
Chen, Jay 495 Houssain, Amber 491
Chen, Kuang 487 Hussain, Faheem 252
Chen, R. 300 Hutchinson, Kelly 163
Chib, Arul 328 Kim, Matthew 139
Chu, Gerry 337 Jain, Mohit 482
Clark, Bev 483 Jain, Shirley 139
Clayton, Peter 320 Javid, Paul 437
Coles, J. 300 Kettani, D. 486
Conway, Neil 487 Khan, Farida 310
498
Kilany, Mohammad 491 Pawar, Udai Singh 150

Kleine, Dorothea 108 Pitman, J. 280, 481
Knight, Michael 495 Planché, Madelaine 95
Kolko, Beth E. 46, 192, 233 Poon, Anthony 233
Komathi, A.L.E. 328 Putnam, Cynthia 46
Korenblum, Jacob 491 Quarta, Giulio 458
Kossi, Edem Kwame 379 Rahman, Mohammed Kaleemur 478
Kumar, Arun 56, 85, 139 Rajput, Nitendra 56
Kuun, Christiaan 95 Rao, F. R. 300
Kuriyan, Renee 66 Ramachandran, Divya 490
Lee, J. 300 Rapchak, Barbara 458
Lerer, Adam 493 Ratan, Aishwarya Lakshmi 150
Levine, Brian 118, 458 Ratra, Rabin 494
Li, Jinyang 496 Ray, Isha 66
Lin, Yibo 480 Ray, Saswati 495
Lodha, Suresh 212 Reddy, Sivananda 494
Lustig, Caitlin 233 Robinson, Charlotte 492
Luk, Rowena 118 Rosenfeld, Roni 447
Maitland, Carleen F. 427 Sabhnani, Maheshkumar 495
Masperi, Paola 27 Sæbo, Johan 379
Mathur, Akhil 139 Sahyraoui, Sofiane M. 348
Meacham, Arthur 458 Samdaria, Navkar 482
Medisetty, Murali 494 Sanghvi, Saurabh 478
Meirer, Patrick 3 Satpathy, Sambit 150, 337
Menon, S. Raghu 337, 478 Sharma, Ashlesh 458
Michalska, Anna 495 Shekhar, Praveen 482
Middleton, Erin 212 Sherwani, Jahanzeb 447
Mills-Tettey, G. Ayorki 129 Singh, Gusharan 406
Mirza, Sarwat 447 Singh, Nupur 183
Molla, Alemayehu 163 Smyth, Thomas N. 468
Mostow, Jack 129 Sweet, Tracy Morrison 129
Mulenga, B. 280, 481 Stodden, Victoria 3
Mungala, Vivek 494 Stork, C 392
Nanavati, Amit Anil 56 Stork, M. 392
Nedevschi, Sergui 357 Subramaniam, Lakshminarayanan 222, 458, 496
Neil, Daniel 495 Subramaniam, Thanuja 150
Nyella, Edwin 243 Terzoli, Alfredo 320
Okon, Uduak A. 367 Thakur, Dhanaraj 192
Osuntogun, A. 481 Thies, William 74
Owusu, Emmanuel K. 35 Thinyane, Mamello 320
Paik, Michael 222, 458 Thomas, S. 280, 481
Pal, Joyojeet 357, 414, 482, 490, 492,494 Thota, Sai Gopal 494
Palijo, Sooraj 447 Titlestad, Ola 379
Parikh, Tapan S. 487, 490 Tongia, Rahul 252
Patnaik, Somani 74 Toyama, Kentaro 150, 337, 406, 427
Patra, Rabin 357 Trahanas, Philip 458
499
Traxler, John 488

Tseng, Clint 492
Underwood, Heather 492
van Gorp, Annemijin 427
Vempala, S. 280, 481
Waidyanatha, Nuwan 495
Wornyo, Edem 468
Yu, Jun 489
Zaharia, Matei 118
Zahonogo, Z. 392
Zia, Lilian 150

ICTD2009 Proceedings

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

ICTD2009 Proceedings

Enviado por

Direitos autorais:

Formatos disponíveis

3rd International Conference on

Information and Communication Technologies and Development

A Global Empirical Evaluation of New Communication Technology 3

A Review of the Research on Mobile Use by Micro and Small 17

An Evaluation of the Use of ICT within Primary Education in Malawi 27

Claim Mobile: Engaging Conflicting Stakeholder Requirements in 35

Computer Games in the Developing World: The Value of Non- 46

Content Creation and Dissemination by-and-for Users in Rural Areas 56

E for Express: “Seeing” the Indian State through ICTD 66

Evaluating the Accuracy of Data collection on Mobile Phones: A Study 74

FOLKSOMAPS – Towards Community Driven Intelligent Maps for 85

HIV Health Information Access using Spoken Dialogue Systems: 95

ICT4What? – Using the Choice Framework to Operationalize the 108

Improving Child Literacy in Africa: Experiments with an Automated 129

Improving Literacy in Rural India: Cellphone Games in an After-School 139

Kelsa+: Digital Literacy for Low-Income Office Workers 150

Mapping the Dynamics of Social Enterprises and ICTD in Cambodia 163

Results from a Study of Impact of E-government Projects in India 183

The Contribution of User-Based Subsidies to the Impact and Sustainability 192

Poster Papers 201

A Speech Enabled Indian Language Text-to-Braille Transliteration 201

Analyzing Statistical Relationships between Global Indicators through 212

ATMosphere: A System for ATM Microdeposit Services in Rural Contexts 222

Challenges in Health Information Systems Integration: Zanzibar 243

Cross Technology Comparison for Information Services in Rural 252

Decentralization, Clientelism and Popular Participation - Is There a 267

Design and Deployment of a Blood Safety Monitoring Tool 280

Dimensions of IT Literacy in an Arab Region Study in Barkha(Oman) 288

Emergency Communication and System Design: The Case of Indian 300

Empowering Muslim Youth through Computer Education, Access, Use: 310

eServices Provisioning in a Community Development Context through a 320

Extending the Technology-Community-Management Model to Disaster 328

Featherweight Multimedia for Information Dissemination 337

Information Communication Techology and Sustainable Communities in 367

Integrating Health Information Systems in Sierra Leone 379

Mobile Telephony Access and Usage in Africa 392

Numeric Paper Forms for NGOs 406

Rajnikant’s Laptop: Computers and Development in Popular Indian 417

Regulatory Independence and Wireless Market Development: A 427

Social Enterprises: A Vocational Entrepreneurship Framework for 437

Speech vs. Touch-Tone: Telephony Interfaces for Information Access 447

The Case for SmartTrack 458

Uses of Mobile Phones in Post-Conflict Liberia 468

An Automated Braille Writing Tutor with Multilingual Exercises 478

Boosting European Market Access to Malian Mango Growers 479

Creating a Mobile-Phone Based Geographic Surveillance System 480

Design of a Blood Flow System 481

DISHA: DISease and Health Awareness for Children on Multiple 482

Freedom Fone: Dial-up Information Service 483

FrontlineSMS and Ushahidi – A Demo 484

Implementing E-Government Accessible to Illiterate Citizens 486

Improving Data Quality with Dynamic Forms 487

IWB4D – Interactive Whiteboards for Development 488

Metamouse: Multiple Mice for Legacy Applications 490

Mobile Phone Job Services: Linking Developing-country Youth with 491

MultiMath: Numeric Keypads for Math Learning on Shared Personal 492

A New Generation of Open Source Data Collection Tools 493

RuralScope: An Information System for Tracking Rural Disbursements 494

T-Cube Web Interface in support of Real-Time Bio-surveillance Program 495

Web Search over Low Bandwidth 496

Author Index 497

Dr. Hessa Sultan Al-Jaber Dr. Kentaro Toyama

Dr. M. Bernardine Dias

i = ρ ∗ i−1 A. Modeling Global ICT Penetration as a Function of Demo-