Content uploaded by Hideyoshi Yanagisawa
Author content
All content in this area was uploaded by Hideyoshi Yanagisawa on Aug 15, 2016
Content may be subject to copyright.
1 Copyright © 2008 by ASME
Proceedings of DETC’08
ASME 2008 Design Engineering Technical Conferences and
Computers and Information in Engineering Conference
DETC/CIE 2008
August 3-6, 2008, Brooklyn, New York, USA
DETC2008-49279
FACTORS AFFECTING VIEWPOINT SHIFTS WHEN EVALUATING SHAPE AESTHETICS
TOWARDS EXTRACTING CUSTOMER’S LATENT NEEDS OF EMOTIONAL QUALITY
Hideyoshi Yanagisawa
The University of Tokyo
School of Engineering
Department of Engineering Synthesis
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan
hide@design.t.u-tokyo.ac.jp
Tamotsu Murakami
The University of Tokyo
School of Engineering
Department of Engineering Synthesis
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan
murakami@ design.t.u-tokyo.ac.jp
ABSTRACT
There are three main issues when trying to capture the
customer's need for a product's emotional quality such as its
aesthetics. The first is that customers have difficulty
externalizing their emotional needs even if they have a clear
mental image of those needs. The second is that people have
different sensitivities when perceiving emotional qualities. The
third is that customers have a latent sensitivity of which they
are unaware. Evoking such latent sensitivity is effective when
extracting the customer’s potential needs. Latent sensitivity
may be evoked by shifting a fixed viewpoint for evaluating an
emotional quality [1]. In this paper, we focus on the third issue,
which has not been dealt with in conventional studies. The
authors address the question of how to provide information that
can shift the customer’s fixed viewpoint and evoke his/her
latent sensitivities on a product’s emotional quality. To
determine what factors are involved in such information, we
conduct an experiment in which the subjects exchange and
mutually evaluate their shape solutions for an emotional image
and the associated viewpoints. Because people have different
sensitivities, customers have different viewpoints and images
toward an emotional design concept as expressed by a
subjective word. We assume that different viewpoints and
images may contain information that can evoke the latent
sensitivity of a customer. To help the subjects to externalize
their images for a given emotional concept, which is the first
issue, we developed an interactive shape generation system in
which the customer as non-designer can easily shape his/her
image. The system generates design samples, which the user
synthesizes using genetic operation. From the experiment, we
observed different types of subjects and different patterns of
effective viewpoints that can shift one’s fixed viewpoint.
Keywords: Emotional quality, latent sensitivity, shape
aesthetics, shifting viewpoints, exchanging viewpoints.
INTRODUCTION
A product's emotional quality, such as its aesthetics, has
become an important factor for increasing the value of mature
products. Such emotional qualities regarding the customer’s
need are difficult to determine. Without a clear understanding
of those needs, the designer has to rely on his/her assumptions
that may be different from the customer’s wants.
There are three main issues when trying to grasp the
customer's needs on emotional quality. Firstly, emotional
images or feelings are tacit. People often have difficulty
externalizing their internal feelings. Although a customer can
explain his/her image of some requirement using a subjective
word, such as "stylish", he/she usually has difficulty describing
it clearly. Secondly, our sensitivities towards an emotional
quality differ from person to person. If the designer's definition
of "stylish" and the customer's are different, the designer will
misinterpret the customer's needs. Such diversity in sensitivities
among customers makes it more difficult to grasp their needs.
Thirdly, the customer has a latent sensitivity of which he/she is
unaware. Evoking latent sensitivities can enrich a customer’s
sensitivity towards a product’s emotional quality. In other
words, the customer’s definition of an emotional concept word,
such as “stylish,” and his/her viewpoint for evaluating a
product’s emotional quality are enriched when his/her latent
sensitivities are evoked.
Most previous studies address the first issue. For example,
Kansei engineering studies ways to formulate the relationship
between emotional words, which are often adjectives, and
measurable design parameters to describe emotional quality.
2 Copyright © 2008 by ASME
This approach statically captures the current state of the
customer's sensitivity for a target emotional quality. However,
the customer's sensitivity is dynamically changing.
Rather than trying to capture such dynamically changing
and diverse sensitivities, the authors aim to find a way to
induce changes in the customer's sensitivity by shifting his/her
fixed viewpoint and evoking his/her latent sensitivity.
Capturing the customer’s needs based on evoked latent
sensitivities means capturing potential needs. Unexpected
viewpoints and images on an emotional concept help customers
to shift their fixed viewpoints. The second issue given above is
that people have different viewpoints, so that exchanging
different viewpoints and images of a concept among the
customers may shift their fixed viewpoints and evoke their
latent sensitivities. In our previous work, we confirmed that
exchanging different viewpoints among customers is effective
for evoking their latent sensitivities and externalizing their
emotional needs based on enriched sensitivities [1]. However,
we did not analyze the factors associated with others’
viewpoints that work to evoke one’s latent sensitivities.
The purpose of this work is to find such factors related to
reference information externalized by others that can shift one’s
fixed viewpoint. We conduct an experiment to seek such
factors, where multiple subjects each externalize an image and
viewpoint against a given emotional design concept and then
they mutually evaluate those images and viewpoints. We regard
shape models or viewpoints made by others that were different
from those created by a subject, but later accepted by him/her
as those that shifted his/her fixed viewpoint. We analyze the
patterns of such effective information. To help the subject to
externalize his/her image as a shape, we developed a shape
generation system in which a subject who is not a designer can
easily generate a shape that fits his/her emotional image.
RELATED WORKS
To understand the customer’s emotional needs, several
methods have been developed in the field of kansei engineering
to quantify the human sensitivity towards a product’s emotional
quality[2][3]. Kansei is originally a Japanese word that refers to
the human sensitivity of a sensory organ where sensations or
perceptions take place in response to stimuli from the external
world (e.g., a product). Kansei includes evoked senses, feelings,
emotions and impressions. The most common approach is to
compose a formalized relation between subjective words that
represent the feeling or impression towards a product and
measurable design attribute parameters extracted from that
product. To obtain the emotional responses towards a product,
several sensory tests have been developed, such as the semantic
differential method[4] and pair-wise comparison[5]. The
relations are formulated using multi-regression analysis, fuzzy
reasoning[6], neural networks[7] and so on. This approach is
suitable for composing an average model of the customer’s
sensitivity towards an existing product or prototype. Thus, it
can only be applied for evaluating existing design samples.
Moreover, it does not deal with changes in human sensitivity.
There is another approach, based on generating design
samples, that helps the customer to externalize his/her
emotional image towards a product. An example of this
approach is the shape design support of the product’s
appearance. To help customers who are non-designers to
generate a design sample that fits his/her unclear emotional
needs, several computer support systems have been developed.
Interactive evolutional computation (IEC) is one such
computational technique. In an IEC-based support system, the
user’s evaluation is regarded as the fitting function of an
Evolutional Computation (EC), such as a genetic algorithm
(GA), where the design parameters are coded as chromosomes
[8]. There are several applications of IEC in design support
systems [9-11]. In this scheme, analysis and synthesis interact
with each other in short cycles. It is known that the user's
kansei model changes after each interaction. Because the user’s
preference and sensitivity change with the influence of design
samples, this scheme is suitable for supporting personal design.
The IEC-based approach only supports the externalization
of the user’s conscious images for product shapes. Yet,
customers also have latent sensitivities of which they are not
aware. For example, customers do not expect a novel design
before it appears, although they potentially have a latent
sensitivity to perceive its goodness. A novel design that evokes
the customer’s latent sensitivity is often more preferable than a
design that they had expected.
The authors proposed a shape generation system that helps
the user to externalize his/her unconscious preference by
showing multiple shape features to which he/she may pay
attention[12-14]. Those features, which we called favored
features, are estimated using reducts in rough set theory[15].
The reduct derives all possible combinations of design
parameters that distinguish the preferred designs from other,
unpreferred designs, as the favored features. Favored features
include shape features that the user prefers but is not
consciously aware of. We observed that such unaware features
stimulate the user and evoke his/her latent sensitivity. In the
proposed system, the user generates shapes by changing
selected favored features based on an interactive scheme. This
approach helps the user to become aware of his/her own
unconscious favored features and to see his/her preferences
from a different point of view. In other words, this approach
supports inner reflection.
However, it does not produce totally unexpected solutions
or viewpoints that are beyond his/her fixed sensitivity, whether
conscious or subconscious. It is assumed that a design solution
generated with an enriched sensitivity improves its value. The
experience of shifting their own viewpoints also supplies a
value for customers. The authors developed a shape generation
system that allows a group of users to exchange design
solutions and viewpoints on a target design concept. Using this
system, we verified the effectiveness of exchanging solutions
and viewpoints in terms of evoking latent sensitivity[1].
In this paper, we examine those factors contained in
effective information externalized by others that induce a
customer to shift his/her viewpoint on an emotional quality.
SHAPE GENERATION SYSTEM FOR EXTERNALIZING
EMOTIONAL IMAGE OF SHAPE AESTHETICS
System overview
A customer who is not a designer will usually have
difficulty describing his/her image of a product’s shape
3 Copyright © 2008 by ASME
aesthetics. To extract such an image from a customer, we
developed a computer support system that helps a customer to
externalize his/her image on shape aesthetics.
Figure 1 shows the user’s view of the system. The system
works in a virtual 3D space consisting of a generation area,
where design samples generated by the system are displayed,
and an operational area, where the user saves design samples
used for generating new design samples. Figure 2 shows the
flow of the system. Firstly, the system generates initial design
samples and displays them in the generation area. Each initial
model consists of parameters selected using an orthogonal array.
The user selects one or more samples that he/she intends to use
for further shape generation and moves them to the operational
area. If the user cannot find any suitable sample in the
generation area, the system generates other samples. To
generate new design samples, the user selects design samples
from the operational area, which we call parent models, and
executes a generation command. The system prepares six
different generation commands: “random”, “partially random”,
“crossover”, “local search” and “morphing”. These are
explained in detail below in the “Shape generation operator”
section. The system then generates new design samples using
the parent models based on the selected command. The user
repeats the above process until he/she obtains a solution that
fits his/her emotional image.
Figure 1 User’s view of shape generation system for
externalizing emotional image
Figure 2 Flow of shape generation system for externalizing
emotional image
Handling interface of 3D shape model
To manipulate the 3D shape model, we introduce an
interface system as shown in figure 3. A magnetic sensor that
detects 6DOF positions and angles is embedded to the bottom
of a plastic tube. The positions and angles are synchronized
with the handling operator in figure 1. The handling operator
enables the user to grab, move and rotate a shape model.
Figure 4 shows how to select, grab and release a shape model
using the operator. To select a shape model, the user moves the
magnetic sensor to area 1 in figure 4. To grab a shape model,
the user moves the sensor to area 2. To release a shape model,
the user moves the sensor to area 3.
The interface system enables the user to view the shape
model from various angles to evaluate its aesthetics. The action,
such as rotating the object using the hand, and view angle affect
the appearance of the virtual object.
Figure 3 Picture of object handling device
User selects parent models
and selects a generation operator
System generates initial samples
(using orthogonal array)
System generates samples
using selected operator and parent models
Fit emotional image?
End
Yes
No
Generation area
operational area
Handling operator
Magnetic sensor
4 Copyright © 2008 by ASME
Figure 4 Object handling using a magnetic sensor
Shape generation operators
In the shape generation system, the user selects one of the
following six generation operators.
Orthogonal search: The system sets parameters using an
orthogonal array. The orthogonal array, which is widely used in
the Design of Experiment (DOE), is a mathematically derived
table in which each column corresponds to an individual
variable that has a small set of discrete values. Let < f1, f2, ….,
fn > (fi∈R) be n design parameters, normalized to [0 1]. Let
[fimin, fimax] be the search area in which the middle point is fimid.
Let Lm(3p) = [ljk] ( ljk∈{1, 2, 3}, m≧ms, p≧n) be an orthogonal
array, where ms is the number of design samples generated at
one time. Allocate a part of the array to the search area, fimin, fimid,
and fimax. A submatrix of the array is allocated to the design
parameters generated at one time. The sub-matrix shifted by ms
rows is assigned to the next generation. The user uses this
method to search broadly for design samples, such as in initial
searching.
Weighted orthogonal search: This operator is the same as
“orthogonal search” except that the search area is changed
based on the parent models, as follows:
2/)(
),max(
),min(
max
21max
21min
midmid fff
ppf
ppf
=
=
=
(1)
where p1 and p2 are parameters of the parent models.
Crossover: This operation is the same as the uniform crossover
operation in genetic algorithm[17]. The new design sample
inherits parameters that are randomly selected from the parent
models.
Local search: This operator sets the parameter for generating
design samples as follows:
jjj
jjij
ppw
n
wpf
21
1)1
1
1
(
=
+=
(2)
where fij denotes the jth parameter of the ith generated sample, p1j
and p2j are the jth parameters of the parent models and n is the
number of generated design samples. This operation generates
design samples around parent p1 based on a vector formed from
p1 and p2.
Morphing: This operation generates design samples whose
design parameters are gradually changed between two parent
models. The jth parameter of the ith generated sample, fij, is
calculated as follows:
jjj
jjij
ppw
n
wpf
21
21
1
=
+=
(3)
where fij denotes the jth parameter of the ith generated sample, p1j
and p2j are the jth parameters of the parent models and n is the
number of generated design samples.
EXPERIMENT
Overview of experiment
The purpose of the experiment is to identify what types of
information externalized by others do people refer to when
shifting their fixed viewpoints on an emotional quality
regarding shape aesthetics, and in what manner. Shifting one’s
own viewpoint on an emotional quality has the effect of
evoking one’s latent sensitivities. The customer is likely to
meet his/her latent wants better based on an evoked latent
sensitivity rather than normally expected needs[18].
The experiment consists of three phases, as shown in figure
5. In the 1st phase, we present the design concept “stylish
shape” to eight subjects. The subjects externalize their
emotional images of the design concept as shape models, which
we call shape solutions. In the 2nd phase, we ask them to state
their viewpoints on the design concept for each shape solution.
To find those factors associated with effective information that
are accepted by people when they shift their previously fixed
viewpoints, we ask the subjects to evaluate the shape solutions
created by others, as well as their viewpoints, in the 3rd phase.
The experimental procedure is described in the following
sections.
(a) Select
Area 2
Area 1
Object boundary area
Operator
Magnetic
sensor
Area 2
Area 1
(b) Grab
(c) Release
Area 3
5 Copyright © 2008 by ASME
Figure 5 Flow of experiment (Si: ith subject, Mij: jth shape
solution of ith subject, Vij : viewpoint of Mij, Et : evaluation of
shape solution without viewpoint, Et’: evaluation of shape
solution with viewpoint, Ve: viewpoints given by others, Ev:
evaluation of Ve and Ev’: evaluation of Vij )
Parametric shape model
In the experiment, we use a cylindrical shape model
formed by a revolving spline as the parametric model example.
The spline has five nodes. The model consists of eight
parameters as follows.
rmin: minimum of radius
h1~h3: height between nodes
x1~x4: distance between nodes and rmin
Figure 6 Parametric shape model used in the experiment
We selected a simple, abstract shape to avoid the influence
of factors unrelated to shape aesthetics, such as product brand
image, etc.
1st phase: Externalization of emotional image as
shape solution
Eight subjects (Si: i=1,2,…,8) externalized their emotional
images using the developed shape generation system. To
control the emotional design concept, we presented the
emotional term “stylish shape” to the subjects. Each subject
created three different shape solutions (Mij : i=1, 2, …, 8, j=1, 2,
3) that fit his/her image of “stylish”. Externalized shape
solutions enable us to see the variety of images for the target
concept expressed by a subjective word.
2nd phase: Extraction of viewpoints for shape
solutions
In the 2nd phase, we asked the subjects to state their
viewpoints on the emotional design concept for each shape
solution (Vij : i=1, 2, …, 8, j=1, 2, 3). We asked for the
viewpoints using three types of expression: “adjective”,
“association” and “shape feature”.
“Adjective” is a viewpoint expressed in adjectives such as
“slim”, “sharp”, etc. Such a viewpoint offers a context for the
concept.
“Association” is an analogical expression that explains
how an externalized shape solution fits “stylish”. For example,
the subject may state that “This shape is stylish because it looks
like a cocktail glass.” In this example, his image of “stylish” is
related to a “cocktail glass”.
“Shape feature” is the shape feature to which the subject
paid attention. We asked the subjects to state the “feature” and
its associated “parameter”. For example, a subject may pay
attention to a “constricted part” (feature) and choose a
constriction which is “very narrow” (parameter).
3rd phase: Mutual evaluation of shape solutions and
viewpoints among subjects
The subjects mutually evaluated shape solutions and their
viewpoints. Firstly, we showed each of the subjects only the
shape solution externalized by another subject, and asked three
questions about the solution, as follows:
Q1: “Do you think that this shape solution fits the design
concept (stylish)?”(concept appropriateness),
Q2: ”Did you expect this type of shape solution as fitting
for the design concept when you were creating a shape solution
in the 1st phase?”(expectedness) and
Q3: “Do you think that this shape solution is informative if
you are to make a new shape solution based on a different
viewpoint?”(informativeness).
The set of answers to the above questions appears as Et in
figure 5. We also asked the subject to state his/her viewpoint on
the shape solution presented, in the same manner as in the 2nd
phase (Ve in figure 5). We then asked three questions on each
viewpoint he/she gave:
Q4:“Do you think that this viewpoint fits the design
concept?”(appropriateness),
Q5:“Did you expect this viewpoint when you were
creating a shape solution in the 1st phase?”(expectedness) and
S1
M11
M12
M13
V11
V12
V13
S2
M21
M22
M23
V21
V22
V23
S1
S2
Sn
Mn1
Mn2
Mn3
Vn1
Vn2
Vn3
Sn
Ve
Et
Et’
Ev
Ev’
Ve
Et
Et’
Ev
Ev’
Ve
Et
Et’
Ev
Ev’
:
:
:
:
1st exp
2nd exp
3rd exp
rmin
x1
x2
x3
h1
h2
h2
x4
6 Copyright © 2008 by ASME
Q6: “Do you think that this viewpoint is informative if
you are to make a new shape solution based on a different
viewpoint?”(informativeness).
The set of answers to these questions appears as Ev in
figure 5.
Next, we showed the subject the viewpoints given by the
original shape creator, obtained in the 2nd phase, to determine
their effectiveness. The subject evaluated each viewpoint by
rating it from Q4 to Q6 (Ev’ in figure 5). The subject then
evaluated the shape solutions once more using Q1 to Q3 by
taking into account the presented viewpoints (Et’ in figure 5).
Each subject repeated the above evaluation for each of the
shape solutions made by all other subjects.
RESULTS AND DISCUSSIONS
Variation of shapes and their similarity
We obtained several shape solutions based on the design
concept "Stylish" in the 1st phase of the experiment. We divided
them broadly into five types based on the cross-section shape,
as shown in figure 7. These types are characterized by the
pattern of convex and concave curves. For example, (a) in
figure 7 consists of one concave curve and (d) consists of a
convex and a concave curve. We use such curve patterns to
determine similarity between shapes. In the following
discussion, we regard any shape that is of the same type as
one’s own shape solution as a predictable shape.
Figure 7 Shape types of obtained solutions
Evaluation of solutions and their personal variation
In the 3rd phase of the experiment, the subjects evaluated
the others’ shape solutions in terms of their concept
appropriateness. Figure 8 shows the average number of other
subjects who evaluated a subject’s shape solutions as
appropriate for the design concept. The maximum number is
seven because the total number of subjects is eight. From the
results, all subjects created shape solutions that were evaluated
as appropriate for the concept by two or more other subjects. In
most subjects, presenting the associated viewpoint increased
the average number but the difference was not statistically
significant. The average number varies among subjects. For
example, subjects #6, #7 and #8 created shape solutions that
were evaluated as appropriate by most subjects (5-6.5). In
contrast, only a few agreed with subjects #2 and #3 in terms of
concept appropriateness for their shape solution.
0
1
2
3
4
5
6
7
1 2 3 4 5 6 7 8
Target subject no.
Average number of other subjects
Befor e sho wing vi ewpo int s
After showing viewpoint s
Figure 8 Average number of other subjects who evaluated
target subject’s shape solution as appropriate for “stylish”
Figure 9 shows the average number of other subjects who
evaluated a subject’s shape solution as unexpected. On average,
three to four subjects stated they had not expected a target
subject’s solutions. The variation among target subjects was not
statistically significant.
Figure 10 shows the average number of other subjects who
evaluated a subject’s solution as informative. This number
varies among subjects, and roughly follows the distribution
pattern for concept appropriateness, shown in figure 8. In other
words, a subject whose shape solutions were evaluated as
appropriate for the given design concept provides informative
solutions for other subjects, and vice versa.
In the following discussion, we call such a subject who
provides appropriate and informative solutions a solution
leader and the opposite one a solution follower. A solution
follower provides unexpected solutions, as shown in figure 9,
but most of the other subjects did not accept them as being
appropriate or informative. A solution leader provides fewer
unexpected solutions, but they are accepted by most of the
subjects including the solution followers.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
1 2 3 4 5 6 7 8
Target subject no.
Average number of other subjects
Before showing viewpoint
After showing viewpoint
Figure 9 Average number of other subjects who evaluated
target subject’s shape solution as ”unexpected”
(a)
(b)
(c)
(d)
(e)
7 Copyright © 2008 by ASME
0
1
2
3
4
5
6
1 2 3 4 5 6 7 8
Target subject no.
Average number of other subjects
Before showing viewpoint
After showing viewpoint
Figure 10 Average number of other subjects who evaluated
target subject’s shape solution as ”informative”
Effect of viewpoint as reference information
From figure 10, the number of other subjects who
evaluated a subject’s solution as informative increased
markedly in all target subjects after presenting the viewpoints.
Figure 11 shows a comparison of the total number of solutions
evaluated as informative before and after the viewpoints were
presented. The difference between them is significant (p<0.01).
Showing the viewpoints helps the subjects to find
appropriateness for the concept as well as informative parts in
the shape solution, even when they had not done so earlier
when only the shape solutions had been presented without the
viewpoints. The informative part found in a shape solution has
the effect of shifting a fixed viewpoint because a subject
evaluated a shape solution as informative when it includes a
part that can be used to create a new solution based on a
different viewpoint.
0
10
20
30
40
50
60
70
80
Befor e sho wing
viewpoint
After showing
viewpoint
Cum ulative n umb er o f su bjects who evalu ate
solution s as in formativ e
Unexpect ed and informative
App ropriat e and in formativ e
App ropriat e, unexpected and
informative
Figure 11 Comparison of number of solutions evaluated as
informative before reading viewpoint and after reading
viewpoint
Figure 11 shows that most of the informative shape
solutions were appropriate (grey and black portions) both
before and after presentation of the viewpoints. About half of
them are evaluated as unexpected (black portion), and the rest
expected (grey portion). Both increase after viewpoint
presentation at a similar rate. But those shape solutions that
were unexpected and informative (but not appropriate)
increased drastically after viewpoint presentation (white
portion). This result shows that presenting the subjects with the
extracted viewpoints along with the shape solutions is effective
in terms of providing information that can shift their fixed
viewpoints.
Next, we examine the content of the effective viewpoints.
In the 2nd phase of the experiment, we extracted three types of
viewpoints: "adjective", "association" and "shape feature".
Figure 12 shows the standardized numbers of effective
viewpoints given by each subject. We define a viewpoint that
was evaluated as appropriate for the design concept and
informative for making new shape solutions based on a
different viewpoint as an effective viewpoint. Because the
number of extracted viewpoints differed among subjects, we
standardized the number of effective viewpoints based on the
number of extracted viewpoints for each subject. The
standardized number of effective viewpoints varies among
subjects, as shown in figure 12.
Figure 13 shows the number of others’ viewpoints that
each subject evaluated as effective. A comparison of figures 12
and 13 shows that a subject who offered the others more
effective viewpoints tends to evaluate fewer viewpoints of
other’s as effective, and vice versa. We found a high inverse
correlation coefficient (r = -0.82, p<0.01) between the
standardized number of a subject's effective viewpoints and the
number of others’ viewpoints evaluated by the same subject as
effective. In the following discussion, we define a subject
whose viewpoints were evaluated as effective by many other
subjects, such as subjects #7 and 8, as a viewpoint leader. We
define a subject who has viewpoints evaluated as effective by
few other subjects, such as subjects #1, #2 and #5, as a
viewpoint follower. Viewpoint followers refer to the viewpoints
of viewpoint leaders. Viewpoint leaders tend not to refer to the
others’ viewpoints.
A comparison of figures 10 and 12 shows that a solution
leader tends to be a viewpoint leader. Subjects # 7 and #8 are
two examples. They also have more effective viewpoints
expressed by adjectives (“impression”) than the other subjects.
Subject # 6 gave the most association-based effective
viewpoints among all subjects. As exceptional cases, subjects
#3 and #4, who are solution followers, have relatively many
effective viewpoints expressed by association. Their viewpoints
are evaluated as effective by the other solution followers, but
not by the solution leaders, as can be deduced by comparing
figures 12 and 13.
8 Copyright © 2008 by ASME
0
1
2
3
4
5
6
1 2 3 4 5 6 7 8
Subject
Standardized number of effective
viewpoint
Features
Impression
Association
Figure 12 Standardized numbers of effective viewpoints for
each subject
0
5
10
15
20
25
30
35
40
1 2 3 4 5 6 7 8
Subject
Total of viewpoints that a subject
evaluated as efective
Feat ures
Impression
Association
Figure 13 Number of viewpoints that each subject evaluated as
effective (i.e., appropriate and informative)
Patterns of effective viewpoints
(1) Effective viewpoints expressed by adjectives
We confirmed that a solution leader also tends to have
effective viewpoints. By considering their combination with the
shape solution, we see three patterns for effective viewpoints.
The first is an unpredicted viewpoint on a predicted shape
solution. The second is a predicted viewpoint for an
unpredicted shape solution. The third is an unpredicted
viewpoint for an unpredicted shape solution. We regard a
viewpoint given by another subject that is the same as one’s
own viewpoint as a predictable viewpoint. We call all other
viewpoints as unpredictable viewpoints. We regard a shape
solution that has the same pattern of concavity and convexity as
one’s own shape solution as a predictable shape solution and
all others as unpredictable shape solutions. The three patterns
above contain unpredicted information in different ways, which
may affect a subject to shift his/her own fixed viewpoint.
Table 1 shows a percentage breakdown of these patterns
among effective viewpoints expressed by adjectives. No subject
evaluated a predictable viewpoint for a predictable shape
solution as effective. Unpredictable viewpoints, for both
predictable and unpredictable shape solutions, account for the
great majority. Predictable viewpoints for unpredictable shape
solutions are few.
Table 1 Percentage of patterns of effective viewpoints
expressed by adjectives
Shape solution
Viewpoint
Predictable
unpredictable
Predictable
0.0%
5.4%
Unpredictable
52.7%
41.9%
Figure 14 shows an example of an unpredictable viewpoint
for a predictable shape that was evaluated as effective. Subject
#8 created a shape solution of the same type as subject #7. (i.e.,
predictable shape) The adjective-based viewpoints given by
subject #8 for his own shape are "casual" and "joyful".
However, he evaluated “unisex,” given by subject #7 to
describe his own shape, as being effective although it was
unexpected.
Figure 15 shows another example of an effective
unpredictable viewpoint for a predictable shape. Subject #7
made a similar type of shape as the one made by subject #1.
Subject #7 evaluated subject #1's viewpoints "formal" and
"adult", which were unexpected, as effective.
Figure 14 Example of unpredictable viewpoint of predictable
shape solution
Figure 15 Example of unpredictable viewpoint of predictable
shape solution
Even when two subjects externalize similar shapes to fit a
given emotional design concept, the viewpoints to capture the
concept may be different. Exchanging different viewpoints
expressed by adjectives as a context for a design concept is
effective for shifting fixed viewpoints on an emotional keyword.
(2) Effective viewpoints expressed by association
Table 2 shows the percentage breakdown of patterns of
effective viewpoints expressed by association. The result is
similar to that of effective viewpoints expressed by adjectives
(table 1). Unpredictable viewpoints for predictable and
unpredictable shape solutions make up the great majority.
casual
joyful
(a) subject #8
unisex
simple
(b) subject #7
Formal
Adult
(b) subject #1
Novel
(a) subject #7
9 Copyright © 2008 by ASME
Table 2 Percentage of patterns of effective associational
viewpoints
Shape solution
Viewpoint
predictable
unpredictable
predictable
2.0%
2.0%
Unpredictable
44.0%
52.0%
Table 3 shows the effective viewpoints expressed by
association whose frequency is high among all effective
viewpoints found through mutual evaluation in the 3rd phase of
the experiment (column (a)). We can divide the viewpoints into
two groups. One is a viewpoint which occurs frequently among
those initially given by the subjects in the 2nd phase (column
(b)), such as “glass” or “base”. These are easily associated with
because they represent physical prototypes of the given
parametric shape models. The other type of viewpoint is that
which does not occur often among all associational viewpoints,
such as “dress” and “warhead” (see figures 16 and 17). In other
words, not many subjects use words like “warhead” to associate
with their created shapes.
Table 3 Effective viewpoints expressed by association and their
frequency
Effective association
Freq. in
effective
association
(a)
Freq. in all
association
(b)
(a) / (b)
Dress
9
2
4.5
Warhead
9
3
3.0
Flower
3
1
3.0
Flagrant
3
1
3.0
Calamus leaf
3
1
3.0
Female
8
3
2.7
Glass
7
8
0.88
Base
6
8
0.75
We divide the frequency among effective viewpoints,
found through mutual evaluation (column (a)), by that among
all associational viewpoints initially given by the subjects
(column (b)), which results are shown in the rightmost column
of table 3. This value represents the average number of subjects
who evaluate that viewpoint as effective. For example, the
associational viewpoint “dress” is used for two different shape
solutions. An average of 4.5 subjects (per solution) evaluated
the viewpoint as effective. A high value means that the
viewpoint is effective for many subjects.
Viewpoints that are generally difficult to make associations,
such as “dress”, scored high in this value. On the other hand,
“glass” and “base”, which are ideas generally easily associated
with, have values of less than 1.0. This means that on average
such viewpoints were evaluated as effective by less than one
subject.
Furthermore, we observed that viewpoints that were more
descriptive, such as "cocktail glass", tend to be effective (see
figure 18).
Figure 16 Examples of shape solutions whose associational
viewpoints are "like a female dress"
Figure 17 Examples of shape solutions whose associational
viewpoints are "like a warhead"
Figure 18 Examples of shape solutions with associational
viewpoint "cocktail glass"
(3) Effective viewpoints of shape features
We divide the viewpoints describing shape features into
three groups: 1) both the feature and the relevant parameter are
predictable, 2) only the feature is predictable, and 3) both the
feature and the relevant parameter are unpredictable. We regard
a feature to which the subject paid attention in the 2nd phase of
the experiment as a predictable feature, and all others as
unpredictable features. We regard a parameter of a predictable
feature that was stated by the subject in the 2nd phase as a
predictable parameter.
Figure 19 shows an example of a predictable shape
solution whose viewpoint describing both a feature and its
parameter were unpredicted by another subject. The features to
which subject #7 paid attention are “the whole shape is thin”
and “the bottom diameter is large”. On the other hand, subject
#7 evaluated as effective the viewpoint offered by subject # 1
"The constricted part is slightly below the center". Subject #7
became aware that the position of the constricted part is an
important factor for fitting the design concept “stylish”. He thus
obtained a new viewpoint from another.
10 Copyright © 2008 by ASME
Figure 19 Example of predictable shape whose viewpoint
describing both feature and parameter was unpredicted but
evaluated as effective
Figure 20 shows an example of a predictable shape
solution whose viewpoint describing a parameter was
unpredicted. Subject #5 made a shape similar to the one by
subject #2. The shape feature paid attention to by both subjects
is the constricted part. The only difference is that subject #5
focused on its position and subject #2 focused on its sharpness.
Subject #5 was made aware that sharpness is an important
feature when realizing a "stylish shape".
Figure 20 Example of predictable shape whose viewpoint of
parameter was unpredicted but evaluated as effective
Table 4 shows the percentage breakdown of patterns of
effective viewpoints that describe shape features. There are very
few predictable features and parameters for predictable shape
solutions. The remaining four patterns are roughly evenly
divided, with a range of 21-26%. Thus, there are four major
patterns among effective viewpoints that describe shape
features.
Table # Percentage of pattern of effective viewpoint (feature)
Shape solution
Viewpoint
Predictable
Unpredictable
Both feature and parameter
are predictable
2.4%
-
Only feature is predictable
23.8%
26.2%
Both feature and parameter
are unpredictable
21.4%
26.2%
CONCLUSION
In this paper, we pointed out that there are three main
difficulties when trying to capture the customer's needs for a
product’s emotional quality such as product aesthetics: 1) the
difficulty of externalizing an emotional image due to its
subjectivity, 2) the diversity among sensitivities, and 3) latent
sensitivities. To address the third issue, which has been
untreated previously, we aimed to evoke the customer’s latent
sensitivity by shifting his/her fixed viewpoint for evaluating the
emotional quality of shape aesthetics. To induce such a shift,
we focused on the diversity of sensitivities existing among
people. We conducted an experiment to identify what factors
are at play when an effective information exchange takes place
among multiple subjects, causing them to shift their fixed
viewpoints. In the experiment, the subjects externalize their
image of the given emotional image "stylish" as shape solutions,
using a developed system based on their own sensitivities. The
subjects evaluate the shape solutions and three types of
viewpoints, which are adjective-based, association-based and
shape feature-based, in terms of their appropriateness for the
design concept, unexpectedness and informativeness. The
subjects evaluated each solution twice – once before and once
after reading the associated viewpoints – in order to examine
the effectiveness of presenting viewpoints. The results of the
experiment are as follows:
- Personal differences existed in terms of providing effective
solutions. We were able to divide the subjects into two groups:
solution leaders, who provided appropriate and informative
solutions, and solution followers, who evaluated the solution
leader's solutions as effective.
- We found a significant difference before and after the
viewpoints were presented, in the number of solutions
evaluated as appropriate, unexpected and informative. Thus
presenting the viewpoints is effective for increasing a solution's
effectiveness.
- We also found that there are viewpoint leaders and viewpoint
followers. Viewpoint leaders tend to be solution leaders as well.
There were exceptions, however, where some solution
followers produced more effective viewpoints expressed by
association than the solution leaders.
- By observing effective viewpoints, we found two patterns
among the adjective- and association-based viewpoints. One is
an unpredictable viewpoint for a predictable shape solution.
The other is an unpredictable viewpoint for an unpredictable
shape solution. Associational viewpoints not raised by the
majority tended to be effective ones. Among effective
viewpoints that describe shape features, we found four patterns.
The above patterns and grouping into leaders and followers
can be useful for narrowing down candidates for effective
shapes and viewpoints that can shift the customer's viewpoint.
ACKNOWLEDGEMENTS
This work is supported by Japan Society for the Promotion of
Science (No. 20686012) and Mizuho Foundation for the
Promotion of Sciences.
REFERENCES
[1] Yanagisawa, H. and Murakami, T. (2007), Emotional Shape
Generation System with Exchange of Others’ Viewpoints -
For Externalizing Customers’ Latent Sensitivity,
Proceedings of DETC'07/ASME 27th Computers and
Information in Engineering, (CD-ROM, DETC2007-34726).
[2] Yanagisawa, H., Murakami, T., Noguchi, S., Ohtomi, K.
and Hosaka, R.(2007) Quantification Method of Diverse
Kansei Quality for Emotional Design - Application of
Product Sound Design, Proceedings of DETC'07/ASME
International 27th Computers and Information in
Engineering, (CD-ROM DETC2007-34627).
- Position of
constricted part
(b) subject #1
- Thin in whole
- Large bottom
diameter
(a) subject #7
Sharpness of
constricted part
(b) subject #2
Position of
constricted part
(a) subject #5
11 Copyright © 2008 by ASME
[3] Jiao J., Zhang Y. and Helander M. (2006), A Kansei Mining
System for Affective Design. Expert Systems with
Applications, 30(4), pp. 658-673.
[4] Osgood, C.G. Suci G.J. and Tannenbaum, P., (1957) , The
Measurement of Meaning, University of Illinois Press.
[5] Thurstone, L.L., (1959), The Measurement of Values, The
University of Chicago Press.
[6] Nakamori, Y., Ryoke, M., (2004), Modeling of Fuzziness in
Multivariate Analysis, IEEE SMC conference 2004, vol.1,
294- 298.
[7] Hsiao, S.W., Huang, H.C. (2002), A neural network based
approach for product form design, Design Studies 23 (1),
pp. 67-84.
[8] Takagi, H., (1998), Interactive Evolutionary Computing –
Cooperation of computational intelligence and human
KANSEI, In proc. of 5th International conference on soft
computing and information /Intelligent system.
[9] Yanagisawa H. and Fukuda S. (2003), Interactive Design
Support System by Customer Evaluation and Genetic
Evolution: Application to Eye Glass Frame, in KES 2003,
LNAI2774, Springer-Verlag Berlin Heidelberg, pp.481-487.
[10] Smyth, S.N., Wallace D.R., (2000), Toward the
synthesis of aesthetic product form, Proc. of ASME
DETEC’00 Computers and information in Engineering
conference.
[11] Graf, J., Banzhaf, W., (1995), Interactive Evolution of
Images, 4th annual conf. On Evolutionary Programming,
San Diego, CA., pp.53-65.
[12] Yanagisawa H. and Fukuda S. (2005), Interactive
Reduct Evolutionary Computation for Aesthetic Design.
Transactions of the AMSE, Journal of Computing and
Information Science in Engineering, 5(1), pp. 1-7.
[13] Yanagisawa H. and Fukuda S. (2004), Global Feature
based Interactive Reduct Evolutional Computation for
Aesthetic Design, Proceedings of DETC'04/ASME
International 24th Computers and Information in
Engineering, (CD-ROM, DETC2004-57651).
[14] Yanagisawa, H., Murakami, T. and Fukuda, S. (2006),
Favored-Feature-Based Shape Generation : Externalization
of Latent Preference, Proceedings of DETC'06/ASME
International 26th Computers and Information in
Engineering, (CD-ROM, DETC2006-99136).
[15] Pawlak, Z.,(1982), Rough sets, International Journal
of Computer and Information Science, 11, pp.341-356.
[16] Thomas, L. S., (1977), A Scaling Method for Priorities
in Hierarchical Structures, Journal of mathematical
psychology, 15, pp. 234-281.
[17] Haupt, R., and Haupt, S.(1997), Practical Genetic
Algorithms, Wiley-Interscience.
[18] Yanagisawa, H. and Murakami, T. (2007), Extraction
of Customer's Latent Viewpoints by Mutual Evaluation,
International Conference on Engineering Design 2007.
-------------------------------------------------