AND
PERFORMANCE
DIPLOMA DI LICENZA
Scuola Superiore SantAnna
CANDIDATO
NICOL DE BENETTI
RELATORE
GUALTIERO FANTONI
1
STATE OF ART................................................................................................................... 3
INTRODUCTION.................................................................................................................9
EXPERIMENT................................................................................................................... 12
AFFORDANCE EXISTENCE .............................................................................................. 18
IMPACT OF AFFORDANCE............................................................................................... 26
LIGHT AND LIFE BATTERY 31
DESIGNER AND MARKET................................................................................................. 44
CONCLUSION ..................................................................................................................50
SURVEY 51
STATE OF ART
The author that coined the term affordance was James J. Gibson, a perceptual psychologist. Gibson
defined affordance as:
The affordances of the environment are what it offers the animals, what it provides or furnishes,
either for good or ill. It imply the complementary of the animal and the environment.
This concept of complementarity can be applied in design, and in particular for analyzing the
interaction between animal (user) and a product in a particular environment. Gibson added also:
As an affordance. For a species of animal, however, they have to be measure relative to the animal.
They are not unique for that animal. They are not just abstract physical properties. So an affordance
cannot be measure in physics An affordance is neither an objective property nor a subjective
property; or it is both if you like. Affordance are properties taken with reference to the observer.
In his book, The Ecological Approach to Visual Perception, Gibson try to study how animals perceive
environment.
The psychologist Donald A. Norman applied for the first time the concept of affordance to design. In his
book The Design of Everyday Things, Norman shows how some objects should afford and should not
afford and he gives some examples of affordances of ordinary objects:
A chair affords (is for) support and therefore, affords sitting, A glass is for seeing through, and for
breaking, Knobs are for turning, Slot are for inserting things into but he also adds Wood is
normally used for solidity, opacity, support or carving.
However, the meaning of the term affordance for Norman is a bit different compare to the Gibsons
one caused by the different purpose of their research. For Gibson affordance are offerings or action
3
possibilities in the environment in relation to the action capabilities of an actor, while for Norman
affordances are perceived properties that suggest how to use the product and they may or may not
actually exist. This different definition imply many differences. The main one is that for Gibson an
affordance has a binary existence, or it exist or it not exist and therefore it is independent from the
animal past experience while for Norman, since affordance depends on the perception of the animal,
it is dependent on the experience, knowledge, or culture of the animal or user. For example, both
ecologist and designer focus on the affordance sitability of a chair in a visible environment. However,
if this chair is moved to a room so dark that a person entering it cannot perceive the existence of the
chair, for ecologist the affordance sitability is still useful as long as it exists (it can support the weight
of the person without any change with the environment); while according to Norman, in this situation
the actor cannot perceive the sitting on the chair, and hence sitability is useless for designers
unless the person accidentally touches the chair or turns on the light and perceives the chair.
In order to explain this ambiguity Gaver tried to classify affordance in different type.
Similar to Gibson, Gaver consider affordance independent from the perceptual information since he
put these two variable on the two different axis. However Gaver agree with Norman that only the
affordance that can be perceive are useful. He introduced also the concept of false affordance that
means affordance perceived by the user but not thought by the designer.
Gero and Kennengiesser proposed the functionbehaviorstructure (FBS) model shown in Figure 2.8,
simulating the view of designers:
4
Figure 2: Geros functionbehaviorstructure (FBS) framework (Gero and Kannengiesser, the left is the 8step version published
in 1990, and the right is the 20step version published in 2002); Xe means the expected X (X=F, B or S)
This model integrates the cognition of users, their perceptions and the environment into the three
levels of the world moving from the specific to the universal, suggesting that affordances are
Structur (Gero and Kennengiesser, 2010). The FBS model shows
the steps in designing a product as continuous processes, comparing the designers expectations with
the practical operations of users and the behaviors and functions of the structure. Gero and
Kennengiesser introduced for the first time a new actor in affordance world, the designer. Based on
the FBS model Cascini and Fantoni developed a new framework in which they separate the different
entities from the Geros circles making the process more clear. In particular Cascini and Fantoni
focused they attention on the bias between their expectation and the products practical use,
resulting in the misuse, alternative use and failed use.
Figure 3: Schematic representations of links between the entities of the proposed extension of the FBS framework and
relations with the situated model (Cascini et al., 2010)
A great contribution to the development of affordances has been brought by Maier and Fadel. They
tried to extend the application of affordance to the whole design world and not just only to some
things as done by Norman.
They created two different categories: ArifactUser Affordances and ArtifactArtifact Affordance.
The first one refers to the interaction between product and user. For them the ArtifactUser
Affordance is a characteristic neither of the object not of the human. They also distinguished between
the set of all the possible interactions and the subset that are affordance. As Gibson said typical
physical properties such as mass, texture, color, etc. are not, in and of themselves affordance. But to
the extent, and only to the extent, that these properties or combinations of these properties, are
useful in some way to the user than an affordance exists. Starting from these considerations they
coined a new definition of affordance:
As the set of interaction between artifact and user in which properties of the artifact are or may be
perceive by the user as potential uses. The artifact is the said to afford those uses to the user.
The affordance type of interaction is first perceptual; it may also require cognition. The perception of
affordance indeed, most of the time requires some kind of sensory information. For example through
the visual perception a person can understand if a chair afford sittability, through the tactile
perception instead she can say if a floor or in general an object afford sustenability etc. It is clear how
every sense could be involve in perceiving affordances.
The ArtifactArtifact Affordances instead, are defined by the authors as:
The set of interactions between two artifact in which some properties of one artifact interact in
some useful way with properties of the other artifact.
Maier and Fadel define also some important properties of the affordances:
Polarity: affordance can be either positive or negative depending upon whether the potential
behavior has beneficial or harmful consequences.
Quality: affordance can be varying quality depending upon how well the subsystem support
potential behavior.
Form dependence: affordance depends on the physical structure of artifacts differently from
functions.
Given these characteristic they introduce a new design approach, the AffordanceBased Design. In
their view the design process can be thought as the specification of an artifact that possesses certain
desired affordances, and does not possesses certain undesired affordances.
INTRODUCTION
Maier and Fadel define the affordance as the set of interaction between user and product in which
properties of the artifact are or may be perceived by the user as potential use. In this work we
want to add to this definition an improvement related to the concept of performance. Our
purpose is to analyze how the user perceives the product performances, and how this perception
impacts the use and the customer satisfaction. During the interaction, the user figures out how
the object works, and more precisely how to reach his desired goal through the interaction with
the product. Focusing on this part it can be said that there are two different moments: the first
one where the user figures out how to interact and the second one where he tries to create
expectations of the interaction results. For example, seeing a chair the user understand how he
can interact with it, that means where he has to sit, where he has to put his hands or his back . In
a second moment he figures out how many kilos the chair can support. If the kilos are less than his
weight, he will interact differently, or he will not interact. In this case he could prefer a less
comfortable object than the chair. It is important to underline that as even if the chair can resist to
double the weight of the person but it seems weak the user could prefer to sit on a stair or a table
than on the chair itself. This example shows how the expectation of the performance can have a
huge impact on the user choice and on the customer purchasing decision. From another point of
view, suppose that the chair can afford sittability and that the user can expect that it can support
his weight but actually it cannot. In this case we will have a failure, the chair will be broken and the
user will suffer some damages. Another thing to consider is that if the user doesnt think that the
chair can support all his weight he could decide to use his legs to support a part of his weight.
With this example we want to highlight three interesting areas where the perception of the
performance impacts the user. First, the interaction between product and user could be different
according to his or her perception of the performance. Second, there are many risks related to a
potential misunderstanding of the real performance. Third, the impact on the customer
purchasing decision and the use or not use decision.
has to accelerate a new car for the first time the force that he will use to push the accelerator
depends on the expected acceleration of the car and to the control that the accelerator has on the
engine. Another example could be the use of a knife. When the user see a new knife, he tries to
figure out the force that he has to put to cut a piece of bread or a watermelon. He decides to use
one hand or two hands according to the force the he thinks is necessary to use. Also, the position
of the second hand according on that value and so the interaction will be different and maybe it
can generate some failure or risk that the designer did not expect. From the first example we
highlight how it is not only the final performance that impact, but also the expectation on the sub
components performances while with the second example we want to show how the interaction
itself depends on a users expectations.
POTENTIAL RISK
Maier and Fadel introduced the concept of false affordance as .. Adding to the definition of
affordance, there is, in our view, a new case that is worth analyzing. If the expected performance
is really different from the real one the user interacts with the product according on his
expectations and not on the reality. This means that the final performance obtained will be really
different from the desire goal. In some cases this might be really harmful for the user and for the
product itself. Consider the previous example of the acceleration of the car. If the user uses too
much force he will obtain a great acceleration that could cause a crash with the environment
around him. So we want to introduce the concept of extreme use.
USE COMPREHENSION
CORRECT
UNCORRECT
CORRECT USE
MISUSED
CORRECT
EXTREME USE
MULTIPLE EFFECTS
UNCORRECT
PERFORMANCE
COMPREHENSION
We want to focus on the extremeuse since the misuse has already discussed in (Cascini, Fantoni.
Maier and Fadel). The multiple effects situation is the sum of the effects of misuse and extreme10
use while the correct use analysis is the desired goal. We believe that an extremeuse could be
really harmful, particularly for some types of products such as automotive products (cars,
motorbikes.) knifes and more generally all the products that can create a damage to the user, to
other things or to themselves when they are over or under used. Marketing messages often
increase the gap between the real performance and the expected performance. The
communication that the marketer usually expresses about the product performance is not
oriented always to the users correct perception. Rather it is oriented to the perception that
maximizes the probability of purchase. This fact makes extremeuse highly probable compared to
misuse, since the marketing communication always shows the correct use of the product and tries
to reduce the gap between perception and reality though images, tutorials, and through what is
called the educational process of the customer.
11
EXPERIMENT
In order to investigate this phenomena an experiment has been conducted. We showed four
flashlights to the participants of the experiment from Clemson University that could be potential
users or customers. We have asked them to score the flashlights and their performances before
interaction and after interaction and how much they would be willing to pay. We have chosen the
participants randomly between students inside the campus
The performances investigated are the light that they produce, the last of their batteries and their
weight. For each performance we have two different products that have the same level of
performance (same light, same weight and same last of the batteries). Therefore we can say that
the ratio between the two performances is one. What we want to analyze is the ratio or the
difference between the judgments that the participants score. We want to see if:
1. There are some regularity in the judgments that means a common perception between the
people
2. The judgments change when the participants can interact with the products
3. There is some relationship between the second judgments and the first ones (we consider
for each feature analyzed only the product that have equal performance).
We attach our survey at the end of this document (page 51).
12
PERCEPTIBILITY LEVEL
EXAMPLE
Class 1
Low
Protection, last.
Class 2
Medium
Acceleration
Class N
High
13
In any case we want to try to identify and study performances belonging to different classes for
the product that we will choose according to our expectations.
PRODUCT IDENTIFICATION
In order to set the experiment we search an object with the following characteristics:

Known performances. This means that performances are easy to define and measure.
Easy to test. This means that our product should be known by our participants and that
they can easily test it in the place where the experiment takes place. It is also important
that it could be test in a small amount of time in order to increase our efficiency.
Easy to find and to change. This is necessary since our time is limited and we need to
conclude the experiment as soon as possible. Therefore we have to consider the possibility
that the object will broke during the experiment.
Easy to handle: in order to perform the experiment in a very crowded place and to allow us
to change easily the location.
PERFORMANCE DEFINITION
We developed our experiments setting in the following passage.
First, we decided the performances to test. As said above we wanted to choose performances that
we expect they belong to different classes of perceptibility. Performances chosen are:

Light
Weight
Batteries life
Second, we collected all possible data regarding the flashlights into the market. To do that we took
data from sites like HomeDePot and Amazon. The three parameters that we considered were:

Weight
14
Batteries life
What we were looking for was the possibilities to have different flashlights with completely
different look (in order to maximize the potential affordance for a person) but with the same
performances. Unfortunately, there were no so many flashlights with the same performances for
more than one characteristic.
To solve this problem we decided to use a group of four flashlights in which there were at least
two of them with the same level of performance for each characteristic.
The flashlight we decided to use are the followings:
LIGHT (lumen)
BATTERY LIFE
WEIGHT (lb)
125
1.235
PRODUCT B (2)
10
0.13
PRODUCT C (3)
35
0.275
PRODUCT D (4)
25
0.275
15
Figura 4: Torchlights
We can observe as product A and D have the same light, product B and C have the same battery
life while product C and D have the same weight.
EXPERIMENT CONDUCTION
The experiment was conducted in different locations inside the Clemson Universitys campus
during night time. Environment conditions were the same for all the experiments. Our attention
was focused on the following parameter:
1. Light (flashlights should be clearly observable, but at the same their lights should be easily
evaluable)
2. Surface around (same distance and same level of brightness)
16
The place where the flashlights were shown was a bench close to a street lamp during the night
time. All street lamps produced the same light with the same color and the flashlight position
respect them were approximately the same.
17
AFFORDANCE EXISTENCE
Our first goal is to determine if there is a common perception between people regarding the
perception of the performance. In order to understand it we start with a consideration. If there is
no common perception the averages of each product scores should not be statistically different
each other otherwise this means that the people have a common preference. Therefore, in case
we consider standardize data for each person we should not have an average statistically different
from zero.
Starting with this consideration our hypothesis is that there is no common perception between
people and so the difference between the averages and zero is not significant.
To test this hypothesis we develop the analysis in two parts:
1 First, we calculate averages and variance for each product and for each characteristic
2 Second, we obtain pvalues. (we consider a good pvalue if it is smaller than 0.05)
The first step of our analysis is to calculate the average and the standard deviation for each
product and for each category before the interaction.
The results are illustrated in the following tables (first table consider the scores given while the
second one is based on the same data after standardization, the third one instead shows the
standard deviations referred to the standardize variables):
AVERAGE SCORE
PRICE
LIGHT
BATTERY
WEIGHT
Product 1
8.4
6.5
6.6
6.8
Product 2
4.0
4.4
4.9
2.0
Product 3
7.9
6.5
6.3
4.9
Product 4
6.4
6.0
6.1
4.9
PRICE
LIGHT
BATTERY
WEIGHT
Product 1
0.49
0.29
0.24
0.96
Product 2
0.89
0.75
0.50
1.27
Product 3
0.51
0.40
0.24
0.20
Product 4
0.11
0.06
0.02
0.11
LIGHT
BATTERY
WEIGHT
Product 1
0.09
0.10
0.11
0.06
Product 2
0.07
0.09
0.10
0.04
Product 3
0.08
0.08
0.07
0.07
Product 4
0.06
0.07
0.08
0.05
Table 6: Standard deviation of the average of standardize score for each product
PRICE
0,60
0,40
0,20
0,00
0,20
0,40
0,60
0,80
1,00
19
LIGHT
0,60
0,40
0,20
0,00
0,20
0,40
0,60
0,80
1,00
BATTERY
0,30
0,20
0,10
0,00
0,10
0,20
0,30
0,40
0,50
0,60
20
WEIGHT
1,50
1,00
0,50
0,00
1
0,50
1,00
1,50
FIRST ANALYSIS
In order to evaluate if the differences are significance we consider the distribution of the average.
We want to verify if 0, that according to our hypothesis should be the real value, is too far or not
from our average.
Figure 10: Gaussian distribution. The black area are pvalue of our test
Since we are considering a group of 112 individuals and in particular their averages, we suppose a
Gaussian distribution according to the Central Limit Theorem.
In the table below we show standard deviations (relative to the averages) of each variable,
averages, ratio between these two measures and in the end the significance of this difference for
each characteristic.
PRICE
21
STANDARD
AVERAGE
RATIO
DEVIATION
PVALUE
AVERAGESTANDARD DEV
Product 1
0.09
0.49
5.40
<0.0001
Product 2
0.07
0.89
12.65
<0.0001
Product 3
0.08
0.51
6.50
<0.0001
Product 4
0.06
0.11
1.68
0.093
Table 7: Standard deviation, average and ratio between average and standard deviation of normalized score for each product
and pvalue for price
LIGHT
STANDARD
AVERAGE
RATIO
DEVIATION
PVALUE
AVERAGESTANDARD DEV
Product 1
0.10
0.29
2.97
0.003
Product 2
0.09
0.75
8.64
<0.0001
Product 3
0.08
0.40
5.08
<0.0001
Product 4
0.07
0.06
0.90
0.368
Table 8: Standard deviation, average and ratio between average and standard deviation of normalized score for each product
and pvalue for light
22
BATTERY
STANDARD
AVERAGE
RATIO
DEVIATION
PVALUE
AVERAGESTANDARD DEV
Product 1
0.11
0.24
2.26
0.024
Product 2
0.10
0.50
4.80
<0.0001
Product 3
0.07
0.24
3.53
<0.0001
Product 4
0.08
0.02
0.32
0.749
Table 9: Standard deviation, average and ratio between average and standard deviation of normalized score for each product
and pvalue for battery life
WEIGHT
STANDARD
AVERAGE
RATIO
DEVIATION
PVALUE
AVERAGESTANDARD DEV
Product 1
0.06
0.96
16.77
<0.0001
Product 2
0.04
1.27
31.47
<0.0001
Product 3
0.07
0.20
2.94
0.003
Product 4
0.05
0.11
1.98
0.048
Table 10: Standard deviation, average and ratio between average and standard deviation of normalized score for each product
and pvalue for weight
As we can see, averages are significantly different from zero. This means that there is a common
perception of the performances and so people base their perception on some common factors.
SECOND ANALYSIS
23
We conduct an analysis of variance in order to confirm our results. Before to do that we have to
be sure that the distributions of our variables are Gaussian since this is a hypothesis for applying
the analysis of variance.
We conduct a ShapiroWilk test. The results show that our variable are not Gaussian
LIGHT
PVALUE (SHAPIRO TEST)
PRODUCT 1
0.00003377
PRODUCT 2
0.0002234
PRODUCT 3
0.002414
PRODUCT 4
0.002899
BATTERY
PVALUE (SHAPIRO TEST)
PRODUCT 1
0.000027
PRODUCT 2
0.000037
PRODUCT 3
0.000159
PRODUCT 4
0.001619
24
WEIGHT
PVALUE (SHAPIRO TEST)
PRODUCT 1
0.0006467
PRODUCT 2
0.00000017
PRODUCT 3
0.00087
PRODUCT 4
0.008456
The null hypothesis in the Shapiro test is that the population is normally distributed. Since all our
pvalue are smaller than 0.05 we have to refuse this hypothesis. Therefore the distribution of our
variables are not Gaussian and so we cannot apply the analysis of variance.
To further investigation we have done to standardize data in order to verify if in this case we could
have a Gaussian distribution. The results are the same and the distributions are not Gaussian.
25
IMPACT OF AFFORDANCE
In order to understand if the affordance has an impact on the final scores, we analyze the changes
in preferences. For this analysis we use the scores of products with same performances. We
consider the first gap (before interaction) as affordance and the second gap as final perception. In
other words we consider the two products with the same performance for each characteristic and
we calculate the difference of the scores before the interaction as an indicator of the affordance
and instead the difference of the scores after interaction as indicator of final perception. For
example we consider light. The first person gave the following scores:
SCORE BEFORE INTERACTION (LIGHT)
PRODUCT A
PRODUCT B
PRODUCT C
PRODUCT D
Since for light the two products with the same performance are product 1 and product 4 we
consider:
Affordance= 76 =1
Final performance = 86=2
We want to underline that we calculate always score product 1 minus score product 2. It means
that the sign of the result indicate the preference. If sign is positive it means that the person
prefers product 1 otherwise it means that she prefers product 4. Since we are interested to
understand if there is an impact and not to quantify this impact, as we should consider more
variables and it is not easy to determine, we use only two possible value, +1 and 1. We indicate
with +1 in case the product preferred is the first one while we indicate with 1 if the product
preferred is the second one. In our example we obtain:
Affordance= +1
26
Final Performance= +1
We consider a 0 significant the case in which the affordance is different from zero while the final
perception is equal to zero. We do not use cases in which affordance is equal to zero since we are
interested in understanding if affordance has a positive impact so we cannot consider cases in
which affordance is not present or not understandable.
HYPOTHESIS
Our hypothesis is that affordance has no impact on the final perception.
To investigate it, we have conducted two different analysis:
FIRST ANALYSIS
CONSIDERATION 1
If the performances of the products are the same and there is no impact of the affordance on the
final perception (previous hypothesis), the probability a person thinks that product B has an higher
performance than product C, after the interaction, is equal to the probability that she thinks that
product C has an higher performance than product B. (we dont consider the cases in which there
are no preferences, in other words cases where both the product obtain the same score).
DEFINITION
We call change of preference the situation in which a person believes that product A has a higher
performance than product B before the interaction and she believes that product B has a higher
performance than product A after the interaction. A change of presence is also if a person believes
that product B has a higher performance than product A before the interaction and she believes
that product A has a higher performance than product B after the interaction. (In other words
when affordance and final perception have different signs).
We call confirmation of preference the situation in which a person believe that the same product
(A or B) has a higher performance than the other one, both before and after the interaction. (In
other words when affordance and final perception have the same sign).
CONSIDERATION 2
27
We can infer from CONSIDERATION 1 and from the original hypothesis (affordance has no impact
on the final perception) that change of preference has the same probability to happen than
confirmation of preference, if the two products have the same real performance.
In the next step of our analysis we determine the number of change and confirmation of
preference. We consider only the two products with the same performance for each characteristic
in order to use CONSIDERATION 2 in our analysis. We conduct the following passages:
a) For each characteristic we calculate affordance and final performance for each
person
b) We calculate the product of them
c) We indicate with 1 if the product of the differences is positive (+1*+1 or 1*1), that
means person does not change her idea (confirmation of preference), 1 if the
product is negative (+1*1, or 1*+1), that means person changes her idea (change
of preference) and 0 if in at least one of the judgment is the same for the two
products.
RESULTS
The results of our first part of the analysis are shown in the table below. The same data are
reported also in the graph below.
Number of Confirm
Number of change
Number of 0
of preference
of preference
significant
Light
48
38
16
battery
48
29
18
weight
33
22
11
Total
129
89
45
Table 15: Number of confirm of preference and number of change of preference for each characteristic
28
60
50
40
30
20
10
0
LIGHT
BATTERY
CONFIRMATION OF PREFERENCE
WEIGHT
CHANGE OF PREFERENCE
As we can see the number of confirmations is higher than the number of changes of preference
for every characteristic.
We want to know, now, if this difference is statistical significant. Therefore our new hypothesis is:
affordance has not a positive impact on the final perception.
Considering that our variable can assume only two states (confirm or change) we use a binomial
distribution and we want to calculate the pvalue of our hypothesis. Since all the results show that
confirmations are more than changes of preference we consider pvalue equal to the probability
to obtain a number of confirmation higher than the one we have obtained. To use binomial
distribution we have to identify the number of the population. We place this number equal to the
sum of confirmation and change of preference. We conduct the same analysis in which we
consider the 0 significant. We have split them half in confirm of preference and half in number of
change of preference. This means that we consider half of the cases as positive impacts but too
small to be perceived and half of the times negative impacts but too small to be perceived.
29
Number of
Number of
Number of
PVALUE
PVALUE
confirmations
changes of
0 significant
considering
of preference
preference
light
48
38
16
0.138
0.118
battery
48
29
18
0.02
0.011
weight
33
22
11
0.068
0.05
Total
129
89
45
0.0047
0.0027
0 significant
Table 16: Number of confirmations of preference and number of changes of preference and pvalue for each characteristic
SECOND TEST
To further investigations of the significance of this measure we conduct a chisquare test. Thus,
for each characteristic we consider the number of confirmation of preference and change of
preference. Given the hypothesis we have done, we consider the average between the number of
confirmations and the number of changes one (Average) and we calculate the :
Average = (Number of confirmations + Number of changes)/2
= ((Number of confirmations) Average)2/ Average + ((Number of changes) Average)2/
Average
We consider the distribution with 1 degree of freedom in order to determine our pvalues.
As done in the previous analysis we consider both the cases with significant 0 and without
significant zero.
The results are illustrated in the table below:
30
Number of
Number of
Number of 0
PVALUE
Pvalue
confirm of
change of
significant
considering 0
preference
preference
light
48
38
16
0.32
0.2809
battery
48
29
18
0.05
0.0304
weight
33
22
11
0.17
0.138
total
129
89
45
0.011
0.0067
significant
Table 17: Number of confirm of preference and number of change of preference and pvalue for each characteristic
The results show as in general the affordance has a positive impact on the final perception but
with different level of impact. In our case it seems that affordance has a greater impact in which
characteristic that cannot be easily perceived by the customer.
31
We call NEGATIVE CORRELATION the cases in which light and life of the battery have an
INTERACTION CHANGE with opposite sign, for the same person.
CONSIDERATION 1
We can infer from the hypothesis that, since there are no significant relation between the two
variables, the probability to have a POSITIVE CORRELATION is equal to the probability to have a
NEGATIVE CORRELATION and so equal to 0.5 for each person.
In the next step of our analysis we determine the number of positive and negative correlation. We
conduct the following passages:
a)
For each characteristic and for each product we calculate the INTERACTION CHANGE for
each person
b)
c)
We indicate with 1 if the product of the differences is positive, that means person changes
in the same way the two scores (POSITIVE CORRELATION), 1 if the product is negative, that means
person changes in the opposite way the two scores (NEGATIVE CORRELATION) and 0 if in at least
one of the judgment is the same for the two product.
For example we consider again the first person. His scores are:
PRODUCT A
PRODUCT B
PRODUCT C
PRODUCT D
33
LIFE)
(BATTERY LIFE)
PRODUCT A
10
PRODUCT B
PRODUCT C
PRODUCT D
Table 19: first person of the sample scores of the battery life
PRODUCT A
1
+3
PRODUCT B
1
PRODUCT C
3
2
PRODUCT D
2
Table 20: first person of the sample score difference of the light and the battery life
We dont consider the zero again because they dont give us information related to what we are
looking for. Therefore we can conclude that it is present a NEGATIVE CORRELATION for product A
and a POSITIVE CORRELATION for product C. We consider a 0 significant the case in which the
difference of light is different from zero while the different in the battery life is equal to zero. Since
we want to know if change in perception of light has a positive impact on the change in battery
light perception we do not consider cases in which light perception does not change. So in our
example product D is a 0 significant while product B is not.
RESULTS
34
The results of our first part of the analysis are shown in the table below. The same data are
reported also in the graph below.
ZERO
SIGNIFICANT
POSITIVE CORRELATION
NEGATIVE CORRELATION
PRODUCT 1
40
29
26
PRODUCT 2
46
30
24
PRODUCT 3
58
26
18
PRODUCT 4
53
24
19
Table 21: Number of positive correlation and number of negative correlation for each product
58
53
46
40
29
PRODUCT 1
30
26
PRODUCT 2
POSITIVE CORRELATION
PRODUCT 3
24
PRODUCT 4
NEGATIVE CORRELATION
As we can see the number of positive correlation are always higher than the number of negative
correlation.
We want to determine now if this difference is statistical significant, so we need to understand if it
is true that change of lightscore has a positive impact on life of the batteryscore.
35
Considering that our variable can assume only two states (positive or negative) we use a binomial
distribution and we want to calculate the pvalue of our hypothesis. Since all the results show that
positive correlation are more than negative correlation, we consider the value of the pvalue equal
to the probability to obtain a number of positive correlation higher or equal compare to the one
we have obtain. To use binomial distribution we have to identify the number of the population.
We place this number equal to the sum of positive and negative correlation. We conduct the same
analysis including the number of 0 significant. In this case we have split half and half the number
of zero in the two categories. This means that we consider half of the cases as positive impacts but
too small to be perceived and half of the times negative impacts but too small to be perceived.
The table below show the results of our analysis:
POSITIVE
NEGATIVE
ZERO
CORRELATION
CORRELATION
SIGNIFICANT
PVALUE
PVALUE
considering 0
significant
PRODUCT 1
40
29
26
0.11
0.074
PRODUCT 2
46
30
24
0.045
0.025
PRODUCT 3
58
26
18
0.0005
0.0001
PRODUCT 4
53
24
19
0.002
0.0003
TOTAL
197
109
37
0.0000022
0.0000003
Table 22: Number of positive correlation and number of negative correlation and pvalue for each product
36
SECOND TEST
To further investigations the significance of this measure we conduct a chisquare test. So for
each characteristic we consider the number of confirmation of preference and change of
preference. With the hypothesis done we consider the average between the number of positive
and the number of negative correlation (Average) and we calculate :
Average = (Number of positive + Number of negative)/2
= ((Number of positive) Average)2/ Average + ((Number of negative) Average)2/ Average
We consider the distribution with one degree of freedom in order to determine our pvalues.
As done in the previous analysis we consider both the cases with significant 0 and without
significant zero.
NEGATIVE
CORRELATION
CORRELATION
ZERO
SIGNIFICANT
PVALUE
PVALUE
considering 0
significant
PRODUCT 1
40
29
26
0.26
0.1854
PRODUCT 2
46
30
24
0.11
0.066
PRODUCT 3
58
26
18
0.0015
0.0005
PRODUCT 4
53
24
19
0.003
0.001
TOTAL
197
109
37
0.000002
0.0000005
Table 23: Number of positive correlation and number of negative correlation and pvalue for each product
As the results show, the significance for product three, product four and all the 4 products
combined is impressively high (pvalue lower than 1 over one thousand), while for the second
product the significance is between 0.1 and 0.05 and for the first it is not good. In light of this
37
results we can conclude that for our torchlights the change in the perception of the light has a
positive impact on the battery perception.
We can conclude that there is a positive relation between these two variables.
RESULTS
PRODUCT 1
INCRESE LIGHT
DECREASE LIGHT
INCREASE BATTERY
18
14
DECREASE BATTERY
15
22
38
22
18
15
INCRESE LIGHT
INCREASE BATTERY
14
DECREASE LIGHT
DECREASE BATTERY
Figure 13: Increase or decrease of battery in two different scenarios relative of the increase or the decrease of light for product 1
PRODUCT 2
INCRESE LIGHT
DECREASE LIGHT
INCREASE BATTERY
36
DECREASE BATTERY
21
10
36
21
INCRESE LIGHT
INCREASE BATTERY
10
DECREASE LIGHT
DECREASE BATTERY
Figure 14: Increase or decrease of battery in two different scenarios relative of the increase or the decrease of light for product 2
39
PRODUCT 3
INCRESE LIGHT
DECREASE LIGHT
INCREASE BATTERY
15
17
DECREASE BATTERY
43
43
17
15
9
INCRESE LIGHT
INCREASE BATTERY
DECREASE LIGHT
DECREASE BATTERY
Figure 15: Increase or decrease of battery in two different scenarios relative of the increase or the decrease of light for product 3
PRODUCT 4
INCRESE LIGHT
DECREASE LIGHT
INCREASE BATTERY
30
DECREASE BATTERY
18
23
40
30
23
18
6
INCRESE LIGHT
DECREASE LIGHT
INCREASE BATTERY
DECREASE BATTERY
Figure 16: Increase or decrease of battery in two different scenarios relative of the increase or the decrease of light for product 4
Starting from the same considerations we do similar analysis using, this time, normalized data.
POSITIVE CORRELATION
NEGATIVE CORRELATION
PRODUCT 1
61
45
PRODUCT 2
64
43
PRODUCT 3
64
43
PRODUCT 4
70
36
Table 28: Number of positive correlation and negative correlation for standardize data
41
70
64
61
45
64
43
43
36
PRODUCT 1
PRODUCT 2
POSITIVE CORRELATION
PRODUCT 3
PRODUCT 4
NEGATIVE CORRELATION
As we can see the number of positive correlation are always higher than the number of negative
correlation.
We want to determine now if this difference is statistical significant, so we need to understand if it
is true that change of lightscore has a positive impact on life of the batteryscore.
In order to verify this we want to understand if our original hypothesis is true or not. We use the
same method used for the last analysis but with new data.
The table below show the results of our analysis using the binomial distribution in the first one and
the chisquare test in the second one:
POSITIVE
NEGATIVE
PVALUE
CORRELATION
CORRELATION
PRODUCT 1
61
45
0.0491
PRODUCT 2
64
43
0.0165
PRODUCT 3
64
43
0.0165
PRODUCT 4
70
36
0.0003
TOTAL
259
167
0.000003
Table 29: Number of positive correlation and negative correlation for standardize data and pvalue of binomial test
42
POSITIVE
NEGATIVE
XSQUARED
PVALUE
CORRELATION
CORRELATION
PRODUCT 1
61
45
2.42
0.12
PRODUCT 2
64
43
4.12
0.042
PRODUCT 3
64
43
4.12
0.042
PRODUCT 4
70
36
10.9
0.001
TOTAL
259
167
19.9
0.000008
Table 30: Number of positive correlation and negative correlation for standardize data and pvalue of chisquare test
43
Our goal is to evaluate if there is a significant difference between the averages of each products
score between the two groups.
We divide the process in the following step

ANALYSIS
We divide our sample and we calculate averages and variances for each category.
The results are shown in the table:
PRICE
LIGHT
BATTERY WEIGHT
Product 1
5.7
5.8
6.1
6.9
Product 2
4.3
4.8
5.4
2.1
Product 3
7.0
7.1
6.9
5.4
Product 4
5.8
6.4
7.0
5.2
44
For the following analysis we use normalized data since we are interested in the relative
preference:
PRICE
LIGHT
BATTERY WEIGHT
Product 1
0.08
0.11
0.14
0.69
Product 2
0.31
0.65
0.48
1.07
Product 3
0.22
0.54
0.24
0.31
Product 4
0.01
0.22
0.38
0.07
The standard deviations of the averages are shown in the table below:
PRICE
LIGHT
BATTERY WEIGHT
Product 1
0.09
0.19
0.20
0.14
Product 2
0.07
0.15
0.20
0.08
Product 3
0.06
0.13
0.11
0.15
Product 4
0.08
0.14
0.14
0.10
Table 33: Standard deviation of the average for engineers group (standardize data)
COMPARISON
In order to better highlight the differences between the two groups we create some graphs which
show the averages of the whole group (blue) and the ones of the engineers group.
45
LIGHT
0,8
0,6
0,4
0,2
0
0,2
0,4
0,6
0,8
1
ALL
ENG
Figure 18: Comparison between averages of the products between engineers and others (standardize data)
Looking this graph it is absolutely evident how engineers have a different perception of the
product 1. While people associate dimensions of the object to the real performance engineers
seems to think differently and based their evaluation on different parameters like for example the
material of the product.
BATTERY
0,6
0,4
0,2
0
0,2
0,4
0,6
ALL
ENG
Figure 19: Comparison between averages of the products between engineers and others (standardize data)
It is clear again how product 1 and product 4 are perceived completely differently between the
two groups.
46
WEIGHT
1,5
1
0,5
0
0,5
1
1,5
ALL
ENG
Figure 20: Comparison between averages of the products between engineers and others (standardize data)
In this case it seems that the phenomena we highlights is still present but in a lighter measure.
STATISTICAL ANALYSIS
In order to evaluate if the differences are significant we conduct an analysis similar to the first
analysis we have done to investigate the existence of the affordance.
We calculate the differences of the averages and variance ratio and we since we have a group of
24 individuals we suppose a Gaussian distribution of the average according to the central limit
theorem.
In the table below we show the standard deviations of each variable, the difference between the
averages (whole group and engineers group), ratio between these two measures and in the end
the significance of this difference for each characteristic.
47
LIGHT
STANDARD
DIFFERENCE
RATIO
DEVIATION
BETWEEN
AVERAGE
AVERAGES
STANDARD DEV
PVALUE
Product 1
0.19
0.40
2.12
0.0354
Product 2
0.15
0.10
0.66
0.5092
Product 3
0.13
0.14
1.07
0.2846
Product 4
0.14
0.16
1.08
0.2800
PVALUE
BATTERY
STANDARD
DIFFERENCE
RATIO
DEVIATION
BETWEEN
AVERAGE
AVERAGES
STANDARD DEV
Product 1
0.20
0.37
1.87
0.0608
Product 2
0.20
0.02
0.08
0.9362
Product 3
0.11
0.01
0.07
0.9442
Product 4
0.14
0.35
2.50
0.0124
48
WEIGHT
STANDARD
DIFFERENCE
RATIO
DEVIATION
BETWEEN
AVERAGE
AVERAGES
STANDARD DEV
PVALUE
Product 1
0.135
0.269
1.99
0.0466
Product 2
0.079
0.199
2.54
0.011
Product 3
0.147
0.107
0.72
0.4716
Product 4
0.096
0.037
0.38
0.7040
The results show that for product 1 the differences are always significant that means engineers
have a different perception of the performances of this product compare to the whole group.
49
CONCLUSION
The analysis done show that affordance has an impact on the customerfinalperception of the
product. Therefore it is important to study them in order to maximize the customer satisfaction. It
is necessary also consider the moment in which the client has to judge, her level of knowledge and
experience that can reduce the importance of the affordance. It is also worth to be mentioned
that impact of affordance seems differ according to the perceptibility of the performance.
Our study also highlights there are some characteristics that drive other in the customer
perception and they do not always respect the physic relationship. This is really important in order
to understand the customer perception and the relationship between each characteristic in this
world. The study shows also that sometime the expectation of the designer could be different
from the reality. This fact is also strengthen by our investigation on designers and market, where
we prove how some perceptions of the designers could be really different from the customer
perceptions.
Further investigation will be conduct in order to better understand if exist common path in the
perception of the affordance and if we can group people according to their expectation in terms of
affordance.
A possible evolution to this work could be conducted in order to better understand the level of the
perceptibility of the performance and at the same time to try to quantify the impact of each
affordance on final perception.
50
Age
Sex
Major
A
B
C
D
How much light do they produce?
A
B
C
D
How much do their batteries work?
A
B
C
D
51
A
B
C
D
NOW YOU CAN INTERACT WITH THEM FOR 5 SECONDS EACH ONE AND ONLY ONE
PER TIME.
52
A
B
C
D
How much do their batteries work?
A
B
C
D
What is their weight?
A
B
C
D
53
A
B
C
D
How much will you pay for those flashlight?
A
B
C
D
54