Você está na página 1de 2

Doctoral Consortium ICER ’18, August 13–15, 2018, Espoo, Finland

Towards an Instructional Design of Complex Learning in


Introductory Programming Courses
Rodrigo Duran
Aalto University
Aalto, Finland, 00076
rodrigo.duran@aalto.fi

ABSTRACT 2 CONTEXT AND MOTIVATION


To provide effective learning strategies to complex content, edu- To develop successful learning interventions, instructional design-
cational psychology produced frameworks that holistically inte- ers must consider student’s learning trajectory. Ideally, tasks should
grate skills at a controlled pace achieved by sequencing tasks by its not overwhelm students but at the same time challenge them, as
complexity. However, to apply such frameworks to introductory proposed by Vygotsky’s Zone of Proximal Development. To assist
programming courses, there is a dearth of established methodolo- teachers in designing instruction to complex content, research in ed-
gies for assessing how difficult such tasks are from the student’s ucational psychology has produced frameworks such as the 4C/ID
cognitive perspective. My goals are twofold: in order to design [11]. At the core of this framework is the required skill to sequence
an instructional intervention that applies and adapt frameworks tasks based on its complexity, therefore promoting effective schema
aimed to complex learning, I first establish and empirically validate construction and automation. However, there is a gap in research
a model to evaluate the complexity of programs from a cognitive since current methods described by Computing Education Research
perspective, providing metrics to quantify different aspects of pro- (CER) do not formally define or provide enough detail in how each
gram’s complexity. I want to explore how to refine and expand this program’s components impact program’s complexity from a cogni-
model analysis, what kind of evidence can be extracted from its tive perspective. How to compare program’s complexity to another?
empirical validation and how to impact instructional design with What makes a program complex?
the results of my work.
3 BACKGROUND & RELATED WORK
CCS CONCEPTS While Bloom’s and SOLO taxonomies were extensively used by
• Social and professional topics → Computer science educa- CER to classify activities [5, 6], this work aims to complement such
tion; Model curricula; Student assessment; approaches by characterizing the content of an activity, not the ac-
tivity itself or the connectedness of its outputs. Whether discussing
how programming structures can have different complexities (“Is
KEYWORDS
an assignment statement easier to read and understand than a print
Model of Hierarchical Complexity; Cognitive Load Theory; Program statement?” [9]), or building learning trajectories to CSE [8], under-
Cognitive Complexity; Complexity; Plan-Composition Strategies standing the complexity of programs by evaluating its structures is
a significant part of the analytical work.
ACM Reference Format:
Rodrigo Duran. 2018. Towards an Instructional Design of Complex Learning The Cognitive Load Theory (CLT) [1] aims to improve the in-
in Introductory Programming Courses. In ICER ’18: 2018 International Com- structional design by investigating how learning is constrained by a
puting Education Research Conference, August 13–15, 2018, Espoo, Finland. limited working memory (WM) where a small number of schemas,
ACM, New York, NY, USA, 2 pages. https://doi.org/10.1145/3230977.3231007 chunks of related elements, are processed. Element interactivity
(EI) determines the degree of interconnectedness between elements
1 PROGRAM CONTEXT in the WM, which is the main component of Intrinsic Load (IL), the
I am in my 3rd year of doctoral studies in Computer Science Educa- amount of simultaneously necessary elements required for learn-
tion (CSE) at Aalto University under the supervision of Prof. Lauri ing, characterizing complexity in CLT. As experience grows, we
Malmi. I recently first-authored one paper outlining the theoretical construct hierarchies of increasingly complex higher-level schemas
framework of my thesis [4]. As I intend to defend my dissertation that encompass many low-lever schemas, which in turn decreases
in 2020, I am working towards publishing papers on my model’s IL by reducing the number of active elements in the WM. Soloway
empirical validation and evaluation of my pedagogical intervention. [10] introduced a plans (schema) and goals tree in the programming
context representing abstractions of strategies to solve a problem
that students later realize by composing a program with code.
Permission to make digital or hard copies of part or all of this work for personal or The Model of Hierarchical Complexity (MHC) [3] is a Neo-
classroom use is granted without fee provided that copies are not made or distributed Piagetian framework applied to analyze the complexity of tasks. It
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for third-party components of this work must be honored. defines stages of development based on a hierarchy of actions orga-
For all other uses, contact the owner/author(s). nized by coordinating rules. The axioms of MHC establishes that a
ICER ’18, August 13–15, 2018, Espoo, Finland more complex action coordinates in a non-arbitrary way at least
© 2018 Copyright held by the owner/author(s).
ACM ISBN 978-1-4503-5628-2/18/08. two lower-level actions, thus defining complexity as a recursion
https://doi.org/10.1145/3230977.3231007 over constituent elements of a task.

262
Doctoral Consortium ICER ’18, August 13–15, 2018, Espoo, Finland

4 STATEMENT OF THE THESIS/PROBLEM instrument to later collect data in a larger cohort (N=600) of CS1
My work aims to inform the instructional design of a complex students from the same university. I expect that this data can val-
subject (programming) to novices (introductory courses) adapting idate the model and provide me enough insight and evidence to
theoretical frameworks from educational psychology specifically design the final pedagogical intervention.
intended for complex tasks. To be able to achieve my overall goal, Since my model is not yet supported by empirical evidence, at
first I aim to develop and validate a model of cognitive complexity the doctoral consortium I would like to discuss its limitations and
that enables an a priori fine-grained analysis of programs. This possible refinements to improve it. I hope to receive feedback on
model extends the previous analysis of plans by providing metrics my empirical validation plan and discuss how to make it more
to evaluate different facets of complexity, including the trajectory of reliable, what kind of additional data could be collected and how to
schema (plan) acquisition, and an evaluation of how plans interact- present concrete evidence of the model validation. Since the model
ing using different code composition strategies impact complexity. and data analysis itself is a necessary step to design a pedagogical
intervention aimed to teach a complex subject (programming), I
5 RESEARCH GOALS & METHODS hope to receive feedback on future applications of such intervention,
how to evaluate it and future extensions of the model.
Conceptualize a model to evaluate computer programs from
a cognitive perspective: The Cognitive Complexity of Computer 7 EXPECTED CONTRIBUTIONS
Programs (CCCP) framework is a theoretical model for reasoning
about the complexity of computer programs and generating metrics This research will improve the conceptualization and evaluation of
that summarize aspects of complexity [4]. The CCCP analyzes program’s cognitive complexity providing a more detailed analysis
the schema construction applying the MHC axioms to create a based on the content of programs and the cognitive effort demanded
hierarchical representation of the plans in a program, defining to comprehend them. The final result of my work, a pedagogical
the plan depth metric. The plan interactivity metric, an adaptation intervention aimed for complex learning of programming may
of CLT’s element interactivity, estimates which plans need to be influence practitioners, teachers, and examiners. My model can
kept in the WM simultaneously while the programmer mentally influence stakeholders in several CER areas, such as: assessment,
manipulates the (higher-level) plans of the program using a given creating learning trajectories to CS1 courses, personalized exams,
code composition strategy. design of “more learnable” programming languages based on the
Validate the framework with empirical results: I will col- complexity of its structures, plan-composition strategies and its
lect data from CS1 and K12 students to evaluate their program impact on learning and performance, and theoretical impacts of
comprehension using tracing and code completion exams with interventions that deal with complexity of programs.
different levels of complexity. The data consists of performance
measurements, subjective ratings of clarity and difficulty[7] and
REFERENCES
[1] Hwan-Hee Choi, Jeroen JG Van Merriënboer, and Fred Paas. 2014. Effects of the
identifiable concepts to evaluate the program. A pre-test will be physical environment on cognitive load and learning: towards a new model of
administered to select subjects with no previous knowledge, and a cognitive load. Educational Psychology Review 26, 2 (2014), 225–244.
post-test will summarize a rank of program complexity. The exams [2] Michael Lamport Commons, Eric Andrew Goodheart, Alexander Pekker,
Theo Linda Dawson, Karen Draney, and Kathryn Marie Adams. 2008. Using
will be administered along with the regular coursework in order Rasch scaled stage scores to validate orders of hierarchical complexity of balance
to ensure that they are measuring schema construction for each beam task sequences. Journal of Applied Measurement 9, 2 (2008), 182.
[3] Michael Lamport Commons, Edward James Trudeau, Sharon Anne Stein, Fran-
content. Performance outcomes will be analyzed by an extended cis Asbury Richards, and Sharon R Krause. 1998. Hierarchical complexity of tasks
Rasch model [2] to rank the exam items. I will investigate a correla- shows the existence of developmental stages. Developmental Review 18, 3 (1998),
tion between the predicted complexity of programs with student’s 237–278.
[4] Rodrigo Duran, Juha Sorva, and Sofia Leite. In review. Towards an analysis of
performance and triangulate it with existing methods [7] using program complexity from a cognitive perspective. In review (In review).
quantitative and qualitative data. [5] Cruz Izu, Amali Weerasinghe, and Cheryl Pope. 2016. A study of code design
Create a pedagogical intervention for CS1 informed by an skills in novice programmers using the SOLO taxonomy. In Proceedings of the 2016
ACM Conference on International Computing Education Research. ACM, 251–259.
instruction design aimed at complex learning: My goal is to [6] Susana Masapanta-Carrión and J. Ángel Velázquez-Iturbide. 2018. A systematic
adapt the 4C/ID framework to create a pedagogical intervention review of the use of Bloom’s Taxonomy in computer science education. In Pro-
ceedings of the 49h ACM Technical Symposium on Computer Science Education
designing learning tasks and performance assessments. I will se- (SIGCSE ’18). ACM, 441–446.
quence learning tasks to create tasks classes (same knowledge con- [7] Briana B Morrison, Brian Dorn, and Mark Guzdial. 2014. Measuring cognitive
tent, varying complexity), design supportive information, mental load in introductory CS: adaptation of an instrument. In Proceedings of the tenth
annual conference on International computing education research. ACM, 131–138.
models representations, part-task practices and cognitive rules. I [8] Kathryn M. Rich, Carla Strickland, T. Andrew Binkowski, Cheryl Moran, and
will also perform analysis of cognitive strategies and pre-requisite Diana Franklin. 2017. K-8 Learning Trajectories Derived from Research Literature:
knowledge aimed at introductory programming courses. Sequence, Repetition, Conditionals. In Proceedings of the 2017 ACM Conference
on International Computing Education Research (ICER ’17). ACM, New York, NY,
USA, 182–190. https://doi.org/10.1145/3105726.3106166
6 DISSERTATION STATUS [9] Simon, Mike Lopez, Ken Sutton, and Tony Clear. 2009. Surely we must learn
to read before we learn to write!. In Conferences in Research and Practice in
I have conceptualized the first iteration of the CCCP model using Information Technology Series (ACE ’09), Margaret Hamilton and Tony Clear
a series of case studies of short programs as a proof of concept to (Eds.), Vol. 95. Australian Computer Society, 165–170.
[10] Elliot Soloway. 1986. Learning to program = Learning to construct mechanisms
present the plan depth and plan interactivity metrics. To validate and explanations. Commun. ACM 29, 9 (1986), 850–858.
this model I will pilot the first data collection with a smaller cohort [11] Jeroen JG Van Merriënboer and Paul A Kirschner. 2017. Ten steps to complex
learning: A systematic approach to four-component instructional design. Routledge.
(N=200) of CS1 students, refine the model and the data collection

263

Você também pode gostar