Using association reasoning tool to achieve

semantic reframing of service design insight

discovery
Soe-Tsyr Daphne Yuan and Pei-Kang Hsieh, Service Science Research Center,
National Chengchi University, NO. 64, Sec. 2, ZhiNan Rd.,
Wenshan District, Taipei City 11605, Taiwan

Developing a desirable service requires in-depth understanding of customers, but

discovering customer insights has depended on designers’ experiences. It is hard
for novice designers and designers at enterprises with Goods Dominant Logic
mindsets to do this well. A partial solution rests on challenging the core design
practices of reframing and frame creation. This study presents an IT-based tool
with the knowledge in ConceptNet (a machine-readable commonsense
knowledgebase) to facilitate a variety of association reasoning methods in the
reframing and frame creation process. The concept of insight depth is proposed
for measuring the influential extent of insights. Our tool has also evidenced its
utility on improving the quality of the reframing and frame creation process for
experienced and novice designers.
Ó 2015 Elsevier Ltd. All rights reserved.

Keywords: innovation, design practice, design process, system design,

interdisciplinarity

I
n this post-digital age, not only startups but also existing enterprises
make great efforts to innovate their products and services to gain more
revenue. Only unceasing innovation can help enterprises keep up with
market trends, or even more, create trends. Service design is a means to ser-
vice innovation and is a process aiming to create new or improved (existing)
services to make them more useful, usable, desirable for clients and efficient/
effective for organizations. Discovering insights from customers and defining
design challenges are keys to service design. At the end of the insight defining
stage, there will be a clear definition of the fundamental challenge or problem
to be solved in the following design stages (Design Council 2012). Conse-
quently, if a problem is not defined with compelling insights, it is hard for cus-
tomers to accept the design no matter how well the service is developed or
delivered. This is connected to situations in which enterprises make service
value propositions that are not what customers’ desire and thus lose their cus-
Corresponding author: tomers as a result.
Soe-Tsyr Daphne
Yuan
daphneyuans@gmail. Meanwhile, both large enterprises and small businesses (SMBs) need a more
com efficient way to carry out new service development. Large enterprises with
www.elsevier.com/locate/destud
0142-694X Design Studies 40 (2015) 143e175
http://dx.doi.org/10.1016/j.destud.2015.07.001 143
Ó 2015 Elsevier Ltd. All rights reserved.
 goods-dominant mindsets have existing business models for gaining revenue
and most of them assume that they can continuously benefit from such models,
and often keep doing things as they always have. Large bureaucracies also
often make big companies slow to make decisions about new service develop-
ment. In contrast to large companies, SMBs have faster decision-making time
and are also usually more adventurous. Thus, it is much easier and they are
more willing to carry out service innovation, but they have less resources,
including manpower and capital, to put into innovations (Lindegaard, 2011).

For the above reasons, we conclude that there are two aspects that cause the
gap between knowing the importance of service design and doing it well. First,
for the aspect of management, enterprises should shift their mindsets from
Goods Dominant Logic (G-D logic) to Service Dominant Logic (S-D logic)
and focus on designing good experiences for their customers. However, enter-
prises are usually accustomed to selling what they think customers’ need,
instead of discovering what they need. However, there is not a good way to
help enterprises discover insights from customers. The second problem is the
lack of tools. Although some studies have been devoted to the expertise of de-
signers (Cross, 2004; Paton & Dorst, 2011), it is hard for novice designers and
designers at enterprises with G-D logic mindsets to discover insights from cus-
tomers. A partial answer rests on challenging core design practices of refram-
ing and frame creation. However, relevant tools to support and analyze the
framing of design activities are rare.

In this study, we argue that there should be an IT-based tool to facilitate the
process of discovering the design insights of designers, no matter if they are
experienced or novice designers. We also present a tool using the common-
sense knowledge in ConceptNet to facilitate a variety of association reasoning
methods for the reframing and frame creation process of service design insight
discovery. The knowledge used in the system comes from ConceptNet, a
famous commonsense knowledge base developed by Liu and Singh (2004a,
2004b) at MIT.

To this end, we outline the related theoretical standpoints which we pursue in

this study (Section 1). Section 2 and 3 discuss the basic concepts. Section 4 then
presents the design of our tool, followed by its evaluations and discussion of
results (Section 5). The conclusion is then provided in Section 6.

1 Characteristics of service design

Firstly, we review the theoretical foundation of service, which is essentially
different from goods. Secondly, we review the fundamental design steps and
the challenges of reframing and frame creation. We subsequently close this sec-
tion by providing the background knowledge of the proposed tool design.

144 Design Studies Vol 40 No. C September 2015

 1.1 Service-dominant logic mindset
In the current era of the service economy, changes in consumer behavior have
increased the importance of service value and competitive strategy. Thus, the
importance of these factors has surpassed those tangible physical products,
thus making the consumer market more service-oriented. Enterprises should
have the mindset of S-D logic to design their services. S-D logic stresses that
services are a process rather than an output unit; the focus thus lies in operant
resources rather than operand resources. The value of service lies in collabora-
tion involving the provider and consumer rather than the creation and delivery
process from the producer to the customer (Vargo & Lusch, 2004).

Accordingly, when designing a service, designers (both designers inside or

outside of a firm) should make their greatest efforts to understand the context
of the customers using services. The more the designers understand the cus-
tomers and their context, the better the chance the designed services will
become successful. However, the design process involves both the known
part of the knowledge and experiences emerging during the design process
and the unknown part of both the problem and the potential solution space
within which the answer of the customers’ aspired value can be sought
(Dorst, 2011).

1.2 Design thinking process and insight discovery

The core of design thinking is to understand the process and method that de-
signers use to carry out user-centered design and learn to solve problems like
them. Following the guidelines of design thinking, everyone can act like a
designer to solve problems in their daily lives. Moreover, businesses can
elevate innovation at a higher level. According to the Design Council
(2012), there are four stages of the design process e discover, define, develop
and deliver (4D’s). Designers’ approaches to divergent and convergent
thinking can be seen in each stage.

This study focuses on the discover and define stages, and the purpose is to
assist designers in identifying the design problem, opportunity and needs to
be addressed through design and also build rich knowledge resources with
inspiration and insights (Design Council 2012). This is a divergent stage in
which designers try to find all possible insights from the stakeholders. The cre-
ation of insights then rests on abductive sense making and reframing (Kolko,
2010). That is, to continuously understand the connections among people, pla-
ces and events and then attempt to find a new perspective (Klein, Moon, &
Hoffman, 2006). After having many insights, designers integrate and interpret
the insights to define the design problem.

Finding insights includes a sequence of tasks including interviews, observa-

tion, finding themes from data, framing and reframing the situation and

Semantic reframing of service design insight discovery 145

 then finally implications and insights may emerge. In addition, these tasks are
not sequential, rather, they are iterative. Designers may go back and forth to
find more compelling insights. Traditionally, the insight discovery process re-
quires lots of human effort.

In order to find themes from the qualitative data obtained through interviews
and observation, designers should analyze word repetitions, scrutinize terms,
and analyze linguistic features like metaphors or transitions; however, this
process is very time-consuming (Ryan & Bernard, 2003). Although many
design methods invoke the use of sticky notes with grouping and abstracting
techniques to facilitate the theme finding process, such methods still take up
a lot of time. Subsequently, the framing process for defining design challenges
and goals also requires a lot of time and depends on the expertise of designers
(Paton & Dorst, 2011). Expertise, though, is also a decisive factor determining
the quality of discovered insights. Novice designers play the role much more
like ‘technician’ in the design process, which means they are probably only
able to solve well-defined problems, rather than discovering and defining by
design challenges by themselves, due to their lack of experiences (Paton &
Dorst, 2011). For the sake of solving these problems, we regard the insight dis-
covery process as a new knowledge creating process which we discuss in the
following paragraph from the constructivists’ viewpoint.

From the point of view of constructivism, the idea of epistemology holds that
knowledge is constructed based on previous knowledge and constantly evolves
over time (Toulmin, 1972). On the other hand, embodied cognition emphasizes
the formative role the environment plays in the development of cognitions that
emerge from situated interactions between agents and their environments
(Wilson, 2002). That is, when facing a new and unfamiliar thing, people
tend to find some characteristics or properties of this new thing and try to
link them to one or more things that exist in their knowledge and experiences.
Novak (1998) also addressed that this kind of creation of new knowledge is a
form of meaningful learning. This meaningful learning process involves recog-
nition of new regularities in events or objects, the invention of new concepts or
extension of old concepts, and, in the most creative leaps, major restructuring
of conceptual frameworks to see new higher order relationships. This is exactly
the insight discovery process which involves observations and interviews in or-
der to gain the empathy of stakeholders; the reframing and frame creation pro-
cess refreshes the perspectives for viewing design situation; this may lead to the
discovery of potential innovative opportunities. Correspondingly, based on
this concept, Kolko (2010) proposed the notion of design synthesis. Kolko
(2010) suggests that a design insight can be seen as the combination of
problem-specific observations (I saw this) and professional experiences (I
know this). Kolko’s definition is shown in Figure 1. In view of this, a design
insight contains subjective judgment of the designer and also objective data
from the design situation itself.

146 Design Studies Vol 40 No. C September 2015

Figure 1 Generation of insight (Kolko, 2010).

1.3 Tools for reframing and frame creation

Since the insight discovery process requires intensive human efforts, this study
argues that it is important to have relevant tools to support designers in the
insight discovery process. This process is a form of meaningful learning for
creating new knowledge, which requires the ability to form and process mental
imagery. When carrying out design insight synthesis, in order to discover in-
sights, designers iteratively try to bridge what they observe (the design context)
to what the already know (their knowledge and past experience). In order to
design satisfying services, designers should understand the problem context
broadly and interpret it with their existing knowledge or past experience so
that they can frame the situation properly. It is critical to find out the incon-
sistencies or gaps between expectations and experiences, who are the stake-
holders and what they desire.

A frame is a product of mental knowledge and meaning structures (Reckwitz,

2002). As mentioned before, frames are very subjective. Designers make sub-
jective value judgments to build up their perspectives of the design situation.
Hence, frames are usually very person-specific based on the designer who in-
terviews subjects and the people who have been interviewed or observed. In
view of this, designers have to reframe the design challenge to ‘jump out of
the box’ and generate some innovative thoughts. The initial frame is the
foundation of new frames. Reframing is to construct a new frame by chang-
ing the perspective by which the design situation is viewed (Paton & Dorst,
2011).

When the system model and the context are very clear, designers may see how
the new perspective influences the situation. New frames for viewing a certain
service or action in a new context help designers discover hidden links and
chances (Kolko, 2010). Jon Kolko (2014) suggested that, in order to discover
new chances, designers can view things in new environments, from new user
perspectives and as new embodiments. For example, when designing a tooth

Semantic reframing of service design insight discovery 147

 cleaning service, designers can think about toothbrushes used in different en-
vironments like in the kitchen or on an airplane. Also, different users have
different purposes for cleaning their teeth. For instance, the elderly may
need to clean their dentures where as children may need to focus more on pre-
venting tooth decay. Furthermore, we can also think about new embodiments
of toothbrushes like mouthwash and dental floss. By reframing again and
again, designers may find some crucial implications and insights.

When framing a situation, designers can utilize different tools, such as a

concept map, to facilitate human thinking and make senses of data. Some
scholars suggest that a clear definition can be obtained only under proper facil-
itation (Ahlberg & Vuokko, 2004). A concept map is a tool for capturing
explicit and tacit knowledge and facilitating creative works; such maps
make human learning much easier under any context (Novak, 1998). Concept
maps can help designers organize data for studying, problem solving, and de-
cision-making.

In addition to concept maps, another kind of tool is called a mind map

(Buzan, 1995). These maps present the context of how people think and
help them to think logically in a visual way. Mind maps not only can make
designers arrange their thoughts logically but also can let their thinking
spread fast. Traditionally, concept maps are used to present schemas of un-
derstanding within the human mind where as mind maps are usually used to
present one’s train of thought. According to the toolkit of human-centered
design by IDEO, a successful service should hit the overlap of the lenses of
being desirable, feasible, and viable. Hence, to design a successful service, de-
signers must consider the design itself as well as the social impact of the
design, the profitability of the business model, and so on. As a result, a
concept map is suitable tool for designers to systematically frame design
situations.

However, the results of concept mapping are very subjective. Under the same
circumstances, different designers may construct totally different concept
maps based on their own experiences and points of view. Hence, with the pur-
pose of developing a desirable service, the understanding of customers is very
important when doing design synthesis. Only when the designers understand
what customers with empathy will the modeled system be insightful. In this
study, we aim to design a new tool utilizing the commonsense knowledge of
ConceptNet to recommend objective, problem-specific frame parts for de-
signers to facilitate their insight discovery process with more empathy for
customers.

1.4 ConceptNet
ConceptNet is a machine-readable commonsense knowledge base that was
structured as a network of natural language fragments that computers could

148 Design Studies Vol 40 No. C September 2015

 know about the world (Liu and Singh 2004a, 2004b). So far, it has been up-
graded to version 5 (Speer & Havasi, 2012). There are over 12.5 million asser-
tions with several types of interlingual relations, as shown in Table 1.

The representation of ConceptNet is a directed graph. Every assertion can be

seen as nodes connected by edges. Nodes are words or phrases that represent
concepts and edges are relations of linked concepts as illustrated in Figure 2.
For example, in the figure we can find that a ‘restaurant’ is used for ‘satisfying
hunger’ while a cake, which is one kind of ‘dessert’ with ‘sweet’ taste, is also
used for ‘satisfying hunger’.

The sources of ConceptNet include the Open Mind Common Sense project
(http://openmind.media.mit.edu/) by MIT media lab, WordNet (http://word-
net.princeton.edu/) by Princeton University, and Wikipedia (http://www.wiki-
pedia.org). It contains over 12.5 million edges connecting 3.9 million concepts
including concepts in English, Chinese, Japanese, Portuguese, French, Span-
ish, etc. The structure of ConceptNet is very similar to people mental imagery,
so it is able to find contexts, make analogies, gist topics, and do other cognitive
tasks just like a human.

Previous applications have been done in, for example, clothing recommen-
dation (Shen, Lieberman, & Lam, 2007), emotion detection (Cambria,
Hussain, Havasi, & Eckl, 2010), and conversational storytelling (Chi &
Lieberman, 2011). These applications utilized ConceptNet mainly to infer

Table 1 Edges in ConceptNet with example sentence frames (Speer & Havasi, 2012)

Relation Sentence pattern

IsA NP is a kind of NP.

UsedFor NP is used for VP.
HasA NP has NP.
CapableOf NP can VP.
Desires NP wants to VP.
CreatedBy You make NP by VP.
PartOf NP is part of NP.
Causes The effect of VP is NPjVP.
HasFirstSubevent The first thing you do when you VP is NPjVP.
AtLocation Somewhere NP can be is NP.
HasProperty NP is AP.
LocatedNear You are likely to find NP near NP.
DefinedAs NP is defined as NP.
SymbolOf NP represents NP.
ReceivesAction NP can be VP.
HasPrerequisite NPjVP requires NPjVP.
MotivatedByGoal You would VP because you want VP.
CausesDesire NP would make you want to VP.
MadeOf NP is made of NP.
HasSubevent One of the things you do when you VP is NPjVP.
HasLastSubevent The last thing you do when you VP is NPjVP.

Semantic reframing of service design insight discovery 149

Figure 2 Knowledge of related concepts in ConceptNet (Speer & Havasi, 2012).

contexts. The clothing recommendation system by Shen et al. (2007) used

knowledge in ConceptNet to recommend that a user should wear a suit
rather than a T-shirt to a wedding banquet since a suit has the factors of
‘formal’ and ‘stately’ which match the context of a banquet. Another appli-
cation, the SenticNet by Cambria et al. (2010), is a publicly available se-
mantic resource for opinion mining. This resource helps with semantical
analysis of users’ opinions of products or services on e-commerce sites or
their blogs and social networks. Since these opinions are very unstructured,
it utilizes the common-sense knowledge in ConceptNet to infer the polarity
of opinions and is able to support opinion mining from natural language
text at a semantic, rather than just syntactic, level. Chi and Lieberman
(2011) also utilized ConceptNet as a tool to do natural language processing.
Raconteur 2 is a system for conversational storytelling between the story-
teller and the viewer. It recommends appropriate media items for story-
tellers to make the story more engaging. It uses commonsense knowledge
in ConceptNet and a novel commonsense inference technique to identify
story patterns. With the commonsense knowledge, the systems uses a
concept vector representation that goes beyond traditional keyword match-
ing or word co-occurrence based techniques to give more precise
recommendations.

This study uses the knowledge in ConceptNet to develop an IT-based tool

that can help designers to find themes, contexts and potentially innovative
frames of the service to be designed, and conduct a meaningful sense making
process. Our choice of ConceptNet is based on its capability to represent
real-world knowledge and its proved ability to do cognitive tasks in a variety
of applications, unlike the other commonsense knowledge bases, such as

150 Design Studies Vol 40 No. C September 2015

 Weltmodell (Akbik & Michael, 2014), which are still short of enough proved
applications.

2 Association reasoning
The insight discovery process iteratively combines what you see and what you
already know, i.e., the process of framing and reframing. The ability to link
newly found phenomenon with existing knowledge is very important when
discovering insights. When framing design situations, designers are actually
manipulating their mental imagery and trying to make sense of data using as-
sociations. This is a new knowledge creating process and we examine this pro-
cess from two perspectives. The first is the ability to make associations; the
other is the synthesis process of combining what people see with what they
already know. Figure 3 shows the underlying concepts behind our tool and
the concepts and their relationships are detailed below.

To examine mental imagery processing ability, we categorize the associations

into five types based on the law of association (Dawson & Medler, 2010),
including context association, analogy association, contiguity association,
contrast association and similarity association (Table 2).

When framing and reframing the situations, designers iteratively combine

what they observe with what they already know and try to make sense of
the situation as well as to gain some implications. The ability to conduct obser-
vation and interviews is very important. The more the details that are found,
the more complete the modeling of the system. Another crucial ability is the

Figure 3 Conceptual framework of this study.

Semantic reframing of service design insight discovery 151

Table 2 Types of associations

Type of association Definition Examples

Context association The relations between two concepts with a causal Hot / Sweat
or sequential relation. Wake up / Brush Teeth
Analogy association The relations between two concepts with some Life / Drama
shared meaning. Final Exam / War
Contiguity association The relations between a series of concepts in Transportation /
contact or in proximity. Bike, Bus, Train, Ship, Plane
Contrast association The relations between two concepts with Happy / Sad
inversed properties. Hot / Cold
Similarity association The relations between two concepts that Bowl / Cup
share lots of similar properties. Hotel / B&B

ability to recall and associate with past experience. Every phenomenon found
needs to be interpreted by the designers to give meaning to it. Hence, properly
interpreting a phenomenon is one of the important expertise of designers.
Compared to novices, experts in a field of study not only have more concepts
integrated in their cognitive frameworks but also possess a broader extent of
propositional linkages between subordinate and superordinate concepts
(Novak, 1998). The ability to properly integrate a newly found concept with
existing knowledge influences the quality of derived insights.

3 Insight quality
After design synthesis, the derived insights have different qualities. In this
study, we will use both macro and micro views to examine the qualities of
the derived insights and propose a measurement metric. From the macro
view, we examine the extents of the influence of the insights. On the other
hand, the micro view considers the perceived satisfaction of the designers
when examining the insight quality from a micro view. This not only provides
a more quantitative way to measure insight qualities but also helps us to eval-
uate whether our IT-based tool can really facilitate the insight discovery pro-
cess and further increase the qualities of derived insights.

3.1 From the macro view: insight depth

The purpose of discovering insights is to develop new service concepts satis-
fying customers’ latent needs, drawing upon the capacities and influences of
service providers. In this study, we adapt the 3-dimension value proposition
space proposed by Kwan and Yuan (2011) for measuring the maximum influ-
ence extent of the derived insights from the macro view. The model has three
dimensions including stakeholders, value (strategy) directions, and customer
empowerment stages. Each dimension has several nominal values, and a point
or an area in the space is the extent of the influence of a potential value prop-
osition. As mentioned above, our main ideas of embedding S-D logic empha-
size the value co-creation between service providers and customers. For the
purpose of co-creating with customers to maximize the value, service providers
tend to let customers take control of variables that are conventionally pre-

152 Design Studies Vol 40 No. C September 2015

 determinant. In other words, service providers empower customers in order to
increase involvement and responsiveness. As a result, we only focus on
designing services with high customer empowerment. Furthermore, we order
the nominal values of each dimension based on the extent of horizontal influ-
ence. The adjusted model is then shown in Figure 4, which we call the insight
depth map.

There extent of influence of the strategies has four degrees. Take tablet com-
puters as an example. For the economic strategy, Kindle Fire offered by
Amazon is relatively cheap. On the other hand, Samsung launched the new
Galaxy Tab with 4G LTE technology, offering functional value for cus-
tomers. Game applications on the tablet entertain and satisfy users, thus
providing intrinsic value. According to a report from Electric Power
Research Institute (2012), Apple’s new iPad only consumes and average of
11.86 kW-hours of electricity over a year, costing only $1.36 USD. This is
a good example of providing environmental values. There are also four de-
grees to the extent of influence of stakeholders. If revenue is the only consid-
eration, the service provider itself is the only stakeholder. In another case, if
the service provider cares about the whole supply chain, the scope of the
stakeholder is extended to their partners. The user-centric design focuses
on the customers, so the customers and community are included. Further-
more, if the design is significant enough, the stakeholders may be the whole
society. Each point or area on the map represents an insight. The distance
between the insight and the origin is the insight depth, which indicates the
scope of the influence of the insights.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Insight depth ¼ S2 þ V2 ð1Þ
(S: influence extent of stakeholders/V: influence extent of values)

Figure 4 Insight depth map.

Semantic reframing of service design insight discovery 153

Table 3 Dimensions of the perceived satisfaction of insights

Dimension Definition Questions

Innovativeness The degree of originality. Did the insights discover by others before?
Integrity The degree of having all Did you consider all the possible influencing factors of the
constituents. insights?
Agreement The degree of self-approval. Do you approve the insights?

3.2 From the micro view: perceived satisfaction of insight

quality
For design challenges aiming to solve only a personal problem or a business
problem (i.e., a micro view), it is inappropriate to judge the insight quality
from a macro view such as from a societal point of view. Instead, a micro
view of insight quality could be connected to the desirability, feasibility, and
viability of a successful design (IDEO, 2010). To be desirable, a service must
provide a solution that can solve users’ problems and not appear on the market
prematurely. For feasibility, designers should consider many aspects when
developing and delivering a service. For instance, marketing strategies, partner-
ships, and training of the first-line employees should all be taken into consider-
ation. To make a product viable in the market, designers should consider a
delivered service not only as a service but also as a business. Only if a service
is operated as a business, can it evolve continuously. To engage in business
operation, in addition to considering the cost of developing a service, designers
should be very passionate and agree with insights they discover. Accordingly,
we evaluate insights from the micro view through three dimensions, including
the degree of innovativeness, the degree of integrity, and the degree of agree-
ment. Definitions and questions to examine the results are shown in Table 3.

4 The tool e Discoverþ

Our tool is called Discoverþ. This tool serves as an artifact (Figure 5) to facil-
itate insight discovery. The original input is the data obtained through inter-
views and observation. After processing, a few insights are generated. In
addition to the data retrieved from stakeholders, we also gather some external
data using Google search to ensure the data is sufficient. The sense-making
knowledge mainly comes from ConceptNet, which has been used for common-
sense reasoning for years (Liu and Singh, 2004a, 2004b; Shen et al., 2007). We
chose ConceptNet as the knowledge base of Discoverþ because its structure of
concepts connected with relations can be utilized to perform contextual
reasoning, which is the core of theme and frame recommendation of our
tool, as depicted in Figure 6.

4.1 Theme finder

The data obtained from the stakeholders and the data from the external source
are excessive and messy. The theme finder aims to find themes from this data.

154 Design Studies Vol 40 No. C September 2015

Figure 5 Discoverþ framework.

Figure 6 Interactive interface of the frame-finder of Discoverþ.

In the past, designers have to expend lengthy time to organize the data by
themselves. We effectuate this process by using the knowledge in ConceptNet
to facilitate the theme finding and the understanding of the problem domain.
We use a heuristic method to calculate the most probable themes. However,
whether or not using the theme finder to organize the data is the designers’ de-
cision making.

Semantic reframing of service design insight discovery 155

 Firstly, we identify some tags in the raw data through TF-IDF calculation.
TF-IDF is an abbreviation of ‘Term Frequency-Inverted Document Fre-
quency’, a numerical statistic method for identifying important terms within
documents in a corpus (Salton & Buckley, 1988). The identified terms are
the input into the module. Each term has a corresponding node in Concept-
Net. The idea is to calculate the closeness of each pair of corresponding nodes.
We start the traversal from one node, the neighbor nodes (depth-1) get one
point, the neighbors’ neighbor (depth-2) get 0.5 points, the nodes of depth-3
get 0.25 points, and the other nodes higher than depth-4 do not get points. Af-
ter traversing all of the nodes, we derive a total score of the potentiality of each
and determine the theme.

For example, Figure 7 demonstrates the ‘topic-gisting’ function in Con-

ceptNet API (Liu and Singh, 2004a, 2004b). The rectangular nodes repre-
sent the concept of the identified terms from the inputs, where as the oval
nodes correspond to the concepts related to them. Starting from the
concept of ‘gun’ in ConceptNet, the nodes beside it, for instance, ‘weapon’
and ‘bullet’, both get 1 point. The nodes next to ‘bullet’, for example, ‘rob-
bery’, gets 0.5 points in this round. After counting all the relations, in this
case, the most probable themes are ‘crime’ (3 points), ‘police’ (2.75 points),
‘robbery’ (2.25 points) and ‘bullet’ (2.25 points). The algorithm is depicted
below.

Figure 7 Theme finding in ConceptNet (Liu and Singh, 2004a, 2004b).

156 Design Studies Vol 40 No. C September 2015

 4.2 Frame finder
The frame finder aims to facilitate the identification of frames used for framing
and reframing the design challenge. Frames are a response to a problem situ-
ation. When facing a new design challenge, experienced designers keep search-
ing for the key points that make the problem hard to solve (Dorst, 2011). On
the other hand, the point of reframing is to view the situation from a new
perspective.

Traditionally, framing and reframing have been done by designers based on

their personal experiences. Accordingly, the frames generated might be subjec-
tive. Moreover, this step may be very difficult for novice designers since they
lack experience. The frame finder aims to find and construct frames that are
relevant to themes and objectives from the stakeholders’ point of view in terms
of the facilitation of different kinds of association reasoning. We divide the
concept of framing into four parts e user perspectives, environments, embodi-
ments and user goals, as suggested by Kolko (2010). These various parts are
described in detail below:

4.2.1 User perspective

Different users may have different concerns. Using users to conduct reframing
help designers to think about these different concerns. A well-defined user
perspective may prompt designers’ empathy to the users which makes the
insight closer to the users’ real needs.

For the example of thinking about a design challenge to design a new type of
hotel services, an important task is to refine the experience of using bathroom
in hotel. We can use the ‘Desires’ edges, which represent the affective connec-
tions between users and their wants (Liu and Singh, 2004a, 2004b), in Con-
ceptNet in order to find prospective users. For example, ‘gym’ is connected
with ‘hotel’ through the ‘At Location’ edge, and ‘athlete’ is connected with
‘gym’ by the ‘Desires’ relation. As a result, we can infer that athlete may be
a potential user of hotels. Besides, when thinking of athletes, we may also think
of children who love sports since they may have similar requirements of hotel
facilities. In contrast, we should also take handicapped people who move with
difficulties into consideration when designing the services in hotels. By making
similarity and contrast associations, designers can discover different types of
possible user perspectives. On the other hand, using contiguity associations
also helps designers to find more potential user perspectives. For example,
when viewing the service from the perspectives of a student, designers may
also think about the perspectives of a teacher, a secretary, or a librarian.

4.2.2 Environment perspective

Using environment as an aspect to conduct reframing helps designers to think
about various scenarios in different environments. In continuation of the
above example of a design challenge, to find potential user environments,

Semantic reframing of service design insight discovery 157

 we search for ‘AtLocation’ and ‘LocatedNear’ edges in ConceptNet, which
show the spatial relationship between things and their ordinary locations
(Liu and Singh, 2004a, 2004b).

For example, if we are designing the experience of using a bathroom in a hotel,

we can find that ‘bathroom’ is spatially connected to ‘school’, ‘theater’, ‘gym’,
‘plane’, ’library’, etc., at the ‘AtLocation’ edge, which means bathrooms usu-
ally can be found in these places. We can think of a lot of places with bath-
rooms like these. This utilizes our abilities of making similarity associations
to find potential new environments. Hence, before designing a bathroom for
a hotel, designers can think about experiences of using bathrooms in these
different environments to find some inspiration. The system generates a list
of potential use environments to simulate designers’ reframing process.

4.2.3 Embodiment
For finding new embodiment, the idea is to find different things for the same
goal. The ‘UsedFor’ edges are the source of this knowledge. This type of edge
reflects the functional connections of things and their functions, namely, their
purposes when people use them (Liu and Singh, 2004a, 2004b).

For example, one of the goals of using a ‘bathroom’ is to ‘relax’ while ‘ener-
getic music’, ‘movie’, ‘party’, ‘vacation’ are also connected to ‘relax’. Hence,
when re-designing the bathroom use experience, designers can add some ener-
getic music or films to build up the relaxing atmosphere. Furthermore, refram-
ing the experience of using a bathroom to be an experience like having a party
or having a trip could also offer inspiration. Designers can make some analogy
associations, connecting the source with a target with similar meaning, to find
more new embodiments.

4.2.4 User goal

In order to discover insights, identifying the users’ goals is also very important.
If the target users are businessmen, knowing what they desire may be impor-
tant information for an insight. This can be aided by finding the relations of
‘desires’ in ConceptNet, since the ‘desire’ edges are the relations between users
and things they desire. Previously we only used the relations of ‘desires’ iden-
tify different user perspectives. Here we put more emphasis on identifying what
these potential users want to achieve and want to prevent in a certain context.
For example, designers may find that businessmen want a meeting space in a
hotel and do not want the room to be too noisy. Knowing this helps designers
to achieve better designs. To discover more user goals, designers may also uti-
lize context associations to reveal the real reason behind people’s behavior.
However, our tool doesn’t address or resolve the possibly conflicting insights
(attained from the identified frames) because of the conflicts between needs,
wants, etc.

158 Design Studies Vol 40 No. C September 2015

 4.3 Crawler
The goal of service design is to deliver a service that is desired by the targeted
customers (Goldstein, Johnston, Duffy, & Rao, 2002). Since the people who
have been interviewed or observed are just samples representing the whole
target segment, the data obtained from interviews or observation might be
limited and can be further extended. In view of this, some external data
may be helpful for sensing the market. The crawler helps fetch data on the
Internet by using Google Search API. News articles, blog posts, or forum dis-
cussions are the target data. These articles may contain some market trends
and customer insights. The system uses the recommended themes or frame
parts as keywords to do the search. For example, using
‘businessman þ hotel’ as the keyword, the search result contains some terms
such as ‘meeting room’, ‘free parking’, or ‘assistant’ that might not appear in
the original input. The crawler also uses the TF-IDF method to extract
important terms in these data (Salton & Buckley, 1988). The newly found
terms are added into the theme finder and the frame finder as inputs in the
beginning of these modules. This can broaden the designers’ viewpoint
when discovering insights.

4.4 Insight encoder

The final step is to generate insights. After framing and reframing the situa-
tion, designers may get some inspiration by seeing who the target audiences
are, what they really need, as well as why they need it. A good insight should
contain these three parts e the user, their needs and the reason they need
something. Users may have the recommended user perspective. The things
users need can be found based on the recommended user goal, environment
and embodiment. As for the reason, designers can be inspired by the recom-
mended user goals. According to this, designers are suggested to follow the
format below to complete the insights.

{user perspective} needs {design challenge} because {user goal, environment,

embodiment}.

For example, an insight generated state:

“Travelling businessmen need to have a good bathroom use experience in a

hotel because a party-like bathroom with energetic music can eliminate their
tiredness after a day of conferences.”

The insight encoder module concludes the result of the above processes and
helps designers come up with proper insights. The tool recommends several
frame parts to help designer to complete the discovered insights.

5 Evaluation of Discoverþ
We design and conduct a series of exploratory experiments to inspect whether
our tool can really facilitate the insight discovery process for service designers.

Semantic reframing of service design insight discovery 159

 Accordingly, we enlist the propositions of the attempted utilities our tool in-
tends to achieve. Next, we make some assumptions to eliminate the gap be-
tween the experiment and the real environment. At last, we analyze the
collected data to justify the propositions and also to find interesting implica-
tions. The propositions are listed as follows:

Proposition 1The ability to process mental imagery, which is used by service

designers in framing and reframing the design situation to discover insights,
contains five types of association abilities, namely, context association,
analogy association, contiguity association, contrast association, and simi-
larity association.
Proposition 2The tool we propose to facilitate the insight discovery process
can increase the quality of derived insights.
Proposition 2-AThe tool we propose to facilitate the insight discovery process
can increase the insight depth of derived insights.

Proposition 2-BThe tool we propose to facilitate the insight discovery process

can increase the perceived satisfactions of derived insights.

Proposition 3Engaging in the field to conduct observation and interviews as

well as combining seen phenomenon are all important parts of design
synthesis.

Thinking of the requirements of service design in different fields, the subjects

involved in the exploratory experiments include 15 people who are studying
or studied in design schools and another 15 people in fields such as business,
computer science, and information systems. These people are either studying
in graduate school or graduated in the past three years and are now working
in startups or design departments. All of the subjects are aged from 22 to 28
and have basic knowledge and practical experience in service design. For
simplicity, we will call the first group of 15 people as ‘designers’ and the second
group as ‘no-designers’ to differentiate their different levels of expertise on ser-
vice design.

5.1 Exploratory experiment design

In order to get the data for justifying our propositions, the plan of exploratory
experiments has four phases (Figure 8) from the experimental subjects’ point
of view.

Phase 1: Framing based on experiences

At first, the subjects were given a case as the brief of a design challenge
and were asked to do concept mapping to frame the situation based on
their own experiences. The aim of this phase is to derive one or more insights.

160 Design Studies Vol 40 No. C September 2015

Figure 8 Exploratory experiment process from the subjects’ point of view.

Phase 2: Framing with inspirations provided by DiscoverD and Google

In this phase, the subjects can use Discoverþ and Google Search to seek inspi-
rations to reframe the situations. They can freely switch between the two sys-
tems and add frame parts or new connections to their concept map as they
wish. In order to recognize which parts and connections are newly added in
this phase, they are requested to use another color to revise the concept
maps. At the end of this phase, they come up with new insights which are
the revised versions of previous ones or totally new ones. Moreover, they
are asked to record two kinds of data. The first is the proportions of the usage
of Discoverþ and Google Search in this phase. They give a percentage, for
example, 70% for Discoverþ and 30% for Google Search. Other data to be
recorded includes the proportions of the sources of inspirations in these two
phases. They will give a percentage as well, for instance, 40% from Discoverþ,
20% from Google Search, and 20% from one’s own experience. As a result, we
can do further analysis of the facilitation capability of Discoverþ.

Phase 3: Labeling the types of associations

After two rounds of concept mapping, there will be a concept map with lots of
nodes and lines representing the relations between concepts. In phase three,
the main purpose is to find out the proportions of use of every kind of associ-
ation by the subjects in design synthesis. First, we give the definitions and ex-
amples of the five types of associations. And then the subjects are asked to
label the types of associations of all the connections on their concept maps
if they think the relationship belongs to one or more type of association.
The reason why we ask the subjects to do the labeling task instead of doing
it by ourselves is because the person who builds up the map should be the
one who understands the map. Rather than guessing, it is better to let de-
signers label them by themselves.

Phase 4: Filling in the questionnaire

In the final phase, the subjects are requested to fill in questionnaires. There are
three purposes of the questionnaire. The first one is to identify the profile of the
subjects in order to analyze the impact of different professions. Second, we

Semantic reframing of service design insight discovery 161

 want to analyze what the subjects think about their method of doing design
synthesis. For instance, we ask the subjects ‘Do you think it is important to
engage yourself in the environment when framing a design context?’ At last,
we ask some questions to examine the facilitation ability of the system as
well as some open questions to gain some feedback above the system.

5.2 Evaluation results

5.2.1 Results of proposition 1
As mentioned in Section 2, designers’ ability to process mental imagery can be
categorized into five types. To verify whether these five types include all of the
association abilities that service designers use in design synthesis, we ask the
subjects to label the connections between the concepts on their concept
maps. According to the collected data, all of the associations can be catego-
rized into at least one of these five types of associations of designer compe-
tence. Since there is no edge on the concept maps that does not fall into
these five types of associations, we argue that the abilities required to form
and process mental imagery all fall within these five types of association.

As for the composition of the usage of these abilities, a statistic is shown

Figure 9. The most used association is context association. It accounts for
about half of the associations. Moreover, similarity and contiguity association
are also commonly used. The percentages are 16.3% and 21.6%. At last, anal-
ogy and contrast association are seldom used. Only 5.7% of relations are anal-
ogy associations and another 2.5% are contrast associations.

Furthermore, for the purpose of understanding whether there are differences in

the association competence between designers and non-designers, we analyze
the proportions of subjects using each type of associations. The percentages
represent how many of the subjects use that type of association (Figure 10).
For example, 6 out of 15 subjects in the designer group used contiguity

Figure 9 Composition of each

type of associations.

162 Design Studies Vol 40 No. C September 2015

Figure 10 Percentages of associations used by different groups of subjects.

association, so the percentage is 40%. We can find that there is no significant

difference in context and similarity associations between the two groups of sub-
jects. However, non-designers like to list series of concepts to make their views
more holistic so that their percentage of using contiguity association is higher
than that of designers. For example, when considering transportation in travel,
they try to list all the possible transportation tools like airplanes, trains, buses,
cars, scooters, bikes, etc. On the other hand, more designers use analogy and
contrast associations to try to change perspectives as to view the design context.
For instance, when thinking of luxury travel, they may also think of budget
travelers. Also, when thinking of different types of tourism, they may draw
an analogy from the movie ‘Up in the Air’ to facilitate their thinking.

By looking into the composition of associations on the concept maps

(Figure 11), we can also derive the same conclusion that designers use more
analogy and contrast comparison than do non-designers while non-designers
use more contiguity associations than do designers.

We infer that because designers have been trained more to have empathy, they
tend to view things perceptually from many different perspectives. In contrast,
the non-designers in this experiment include people studied business and engi-
neering. These individuals are more likely to stick with their original way of
thinking and try to make their thoughts more logical and comprehensive.
With perceptual mindsets, designers use more analogy to make their imagina-
tions more concrete, as in the earlier example of using a movie to analogize a
type of travel. Also, designers think more about the emotions of people, so
they use more contrast associations to think about different experiences, for
example, the happiest experience and the worst experience when traveling.
On the other hand, due to the expertise in business or engineering, non-
designers tend to have realistic thinking. They use more contiguity associations
to list all the existing related factors and examine the feasibility.

Semantic reframing of service design insight discovery 163

Figure 11 Composition of
each type of associations for
different groups of subjects.

5.2.2 Results of proposition 2

To justify proposition 2, we investigate the insight qualities from both the
macro view and the micro view.

For proposition 2-A, we examine the insight qualities from the macro view
with the insight depth map. First, we encode the insights derived from the ex-
periments and fit them into the insight depth map. For example, an insight e
‘Travel agencies need to provide a flexible travel package because it can attract
more backpackers’ is coded into (Functional, Service Provider) since it de-
scribes the economic value from the provider’s perspectives. Similarly, another
insight e ‘Young people need to travel often, because travel can make people
grow up’ is coded into (Intrinsic, Customer). The mapping results of the first
insights derived from phase one of the experiments are shown in Figure 12.

Figure 12 The insight depth

map of the insights in phase 1.

164 Design Studies Vol 40 No. C September 2015

 The numbers represent the subjects’ sequence number in the experiment. We
can find that most people start from the customers’ perspective while some
start from the providers’ perspective.

We also encode the insights derived from phase 2 of the experiments and fit
them into another insight depth map, as shown in Figure 13. The bolded
numbers with stars represent the subjects whose insights in phase 2 are deeper
than those in phase 1. Namely, these insights have a larger scope of influence in
terms of the value or stakeholders’ perspective. We can find that 56.7% of sub-
jects (17 out of 30) enhanced their insight depth after reframing in phase 2
while no subjects attained a smaller scope of influence.

There seems to be no significant difference in the numbers of subjects whose

insights go deeper between the subject group of designers (9 subjects) and
non-designers (8 subjects). However, analyzing their insights more closely,
we can find that most of the deep insights of the non-designer group come
from functional values to intrinsic values (5 out of 8) while the insights of
the designer group are more diversified, for instance, from economic values
to functional values or from providers’ perspective to customers’ perspective.
Moreover, we also found that most of non-designers derived enhanced insights
in phase 2 based on some context in phase 1 (for example, the same user per-
spectives or similar environments). On the other hand, most of designers found
totally new insights in phase 2.

To look into the causes, we can find some clues in the way the two groups do
concept mapping. Designers tend to develop their concept maps evenly over
many aspects. They do not start to derive insights until their concept maps
are almost complete, as in a breadth-first strategy in a graph search.
Conversely, non-designers construct concept maps in a depth-first manner.
They tend to have some intuitive assumptions or know the direction in which

Figure 13 The insight depth

map of the insights in phase 2.

Semantic reframing of service design insight discovery 165

 they want to develop the service right after they read the design brief. They first
follow their assumptions when doing concept mapping. Therefore most of
their derived insights are related to their topics of interest. Moreover, most
of their reframed insights in phase 2 are enhanced insights based on phase 1.
We find that a high proportion of designers’ concept maps are evenly spread
while many of the non-designers’ maps are skewed.

For Proposition 2-B, to examine the insight quality from the macro view, we ask
the subjects to score the qualities of their insights based on their perceptions.
They scored the insight qualities in three dimensions including the degree of
innovativeness, the degree of integrity, and the degree of agreement, as discussed
above in Section 3. For each dimension, the subjects chose their perceptions of
the derived insights in phase 2 in comparison with the ones derived in phase 1
of the experiments. The options are ‘Better’, ‘No difference’ and ‘Worse’, repre-
sented by scores of 3, 2, and 1, respectively. For example, if a subject feels his
insight in phase 2 is more innovative than in phase 1, he chooses ‘Better’ for inno-
vativeness and gets a score of 3 points. Under this evaluation, we have the
following hypothesis: ‘If the mean score is greater than 2, in this dimension,
the quality of the newly derived (with designed artifacts) insights is better than
the quality of the derived insights based on subjects’ own experience’.

We use statistic software e PASW (Predictive Analytics Software), which was

renamed SPSS (Statistical Product and Service Solutions) and acquired by IBM
in 2009, to facilitate calculation. Table 4 shows the results of the descriptive sta-
tistics of the data. We can find that the mean scores of the three dimensions are
2.50, 2.80 and 2.73 with the standard error are 0.093, 0.074 and 0.082 and stan-
dard deviation of 0.509, 0.407 and 0.450. Furthermore, the minimum scores of
these three dimensions are all equal to 2 (Table 4). In other words, all of the sub-
jects gave scores of at least 2 points. As a result, our proposition can be justified.

In addition, following the instructions of the experiments, the subjects can use
both Google Search and Discoverþ to seek inspiration. In order to examine
the facilitation of Discoverþ, we also collect statistics the subjects’ use. The
subjects write down the percentages of the sources of inspirations and the re-
sults are shown in Figure 14.

Table 4 Descriptive statistics for proposition 2

N Minimum Maximum Mean

Statistic Statistic Statistic Statistic Std. Error

Innovativeness 30 2 3 2.50 0.093

Integrity 30 2 3 2.80 0.074
Agreement 30 2 3 2.73 0.082
Valid N (listwise) 30

166 Design Studies Vol 40 No. C September 2015

Figure 14 Sources of inspira-
tions (insight qualities from
the macro view) (phase 2).

According to the figure, we can find that an average of 85% of the inspirations
comes from Discoverþ and the other 15% come from Google Search. Further-
more, we also examine the averages of subjects whose insights go deeper and do
not go deeper in phase 2. Discoverþ inspired 92% of the subjects who derived
deeper insights while Google Search only accounted for 8%. On the other hand,
for the subjects whose insights did not go deeper in phase 2, 75% were inspired
by Discoverþ and 25% were inspired by Google Search. We found that the sub-
jects who deepened their insight depths in phase 2 perceived more inspirations
from Discoverþ than those whose insights did not go deeper in phase 2.

Furthermore, from the micro view, we also examine the difference between the
subjects who have higher perceived satisfaction and the ones with lower
perceived satisfaction. Since all the subjects gave at least two points in all of
the three dimensions, which means they think the new insights in phase 2
were not worse than in phase 1, we set the standard of 7 points for the total
perceived satisfaction score in all three dimensions. In other words, subjects
with a total score above 7 points, which means that the subjects have better
satisfactions in at least two dimensions, are seen as the group that have a
higher perceived satisfaction. On the other hand, the subjects with the score
below 7 are seen as the group that did not perceive better satisfaction with
insight quality. The results (Figure 15) show that the subjects with higher
perceived satisfaction were inspired by Discoverþ 91% of the time and by
Google Search only 9% of the time. As for the subjects with lower satisfaction,
66% were inspired by Discoverþ and 34% were inspired by Google Search. It
can thus be inferred that the more the subjects were inspired by Discoverþ, the
better the quality of the insights.

Furthermore, Figure 16 and Figure 17 show the total proportions of sources of

inspirations for the whole concept mapping process (both phase 1 and phase 2
of the experiment). We also can find that the subjects who have better insight

Semantic reframing of service design insight discovery 167

Figure 15 Sources of inspira-
tions (insight qualities from
the micro view) (phase 2).

Figure 16 Source of inspira-

tions (insight qualities from
the macro view) (phase 1
and 2).

quality (no matter from the macro view or micro view) were inspired more by
Discoverþ than Google search. At the same time, the other subjects whose
insight qualities do not get better were inspired less by the proposed tool
and more by Google search or their own experiences.

Figure 17 Sources of inspira-

tions (insight qualities from
the micro view) (phase 1
and 2).

168 Design Studies Vol 40 No. C September 2015

Figure 18 Sources of inspira-
tions for designers and non-
designers.

Therefore, we can infer that Discoverþ is useful in facilitating service designers

to conduct design synthesis. Again, we also found some interesting differences
between the subject groups of designers and non-designers. In Figure 18, we
see that non-designers spent more time using Google Search than did de-
signers. As we discussed before, we found that non-designers tend frame the
design context based on their own assumptions and preferences. They have
a clear idea of what data they need to support their design synthesis process.
Hence, once they cannot find the data they need on Discoverþ, they are
more likely to try to search for it on Google.

There are also some clues supporting this when we look into the click logs of
Discoverþ. As we mentioned in Section 4, when using Discoverþ, there are
many ways to spread thoughts by clicking different types of concepts that
our tool recommends. ‘Theme’ represents the recognized themes of the design
brief provided by the theme finder. ‘Search’ means that users use the search bar
to search for particular concepts they want to browse while ‘Raw’ means that
users directly click the related concepts of the concept they are browsing. At
last, ‘recommendation (d1)’ and ‘recommendation (d2)’ are recommended
frame parts of degree 1 and degree 2 from the frame finder. Table 5 shows
the number clicks on each type of concept in the system.

We find that there is no significant difference between these two categories of

subjects, except for in the number of searches. Non-designers searched six
times more than designers did. We can infer that non-designers tend to have

Table 5 Clicks on each type of the system recommendations

Theme Search Raw Recommendation (D1) Recommendation (D2) Total

Designers 54 3 84 11 14 166
Non-designers 42 20 81 18 15 176

Semantic reframing of service design insight discovery 169

Table 6 Descriptive statistics for proposition 3

N Minimum Maximum Mean

Statistic Statistic Statistic Statistic Std. Error

ENVIRON 30 3 5 4.6333 0.10152

PASTEXP 30 4 5 4.7000 0.08510
Valid N (listwise) 30

certain assumptions so they directly search for concepts which come to mind
after reading the design brief.

5.2.3 Result for proposition 3

To justify whether engaging in the environment and combining seen phenom-
ena with past experiences are important for service designers when doing
design synthesis, we asked the subjects to share their perspectives. There are
two questions in the questionnaire. The first one is ‘When doing design synthe-
sis, engaging in the environment and conducting observation and interviews
are important’. Another is ‘When doing design synthesis, past experiences
and existing knowledge are important’. The subjects answered these two ques-
tions based on a five-point Likert scale. The higher the score is, the more the
subjects agree with the statement. The statistics are shown in Table 6. ‘ENVI-
RON’ stands for the first question while ‘PASTEXP’ stands for the second.

Since we have 30 simple random samples and the standard deviation is un-
known, again, we use one-sample T test to see whether the subjects agree
with these two statements.

We set the null hypothesis and the alternative hypothesis as shown in

Figure 19. If the population mean m  3, we can say that subjects agree
with the statement. As follows, we use PASW to facilitate T testing. We use
a 95% confidence interval of the difference (a ¼ 0.05) to carry out testing
and the result is shown in Table 7.

We use the critical value method and confidence interval method to test the hy-
pothesis. First, for the critical value method, we look up for the critical value c
from the table of the critical value of t distribution and derive
c ¼ ta; n1 ¼ t0:05;29 ¼ 1:699. As a result, we reject the null hypothesis for
both statements because the t values of the two statements (16.089 and

Figure 19 Testing hypothesis

for proposition 3.

170 Design Studies Vol 40 No. C September 2015

Table 7 The result of One-Sample T test for proposition 3

Test value ¼ 3

t df Significance (1-tail) Mean difference 95% Confidence interval of the difference

Lower Upper

ENVIRON 16.089 29 0.00000000000000054387 1.63333 1.4257 1.8410

PASTEXP 19.977 29 0.00000000000000000169 1.70000 1.5260 1.8740

19.977) (see Table 7) are greater than the c value (1.699). Hence, we can say
that the subjects agree with the two statements.

Next, we further use the confidence interval method to test the hypothesis. For
the first statement, according to the formula, we derive that
pffiffiffi
ta; n1 )s= n ¼ 0:20763. The sample mean (x bar) is 4.63333 (Table 6). Using
pffiffiffi
the sample mean (4.63333) minus ta; n1 )s= nð0:20763Þ, we derive a ¼ 4.4257.
So, H0 is rejected and H1 is established since m0 ð3Þ < að4:4257Þ. On the other
hand, for the second statement, we can derive the value of
pffiffiffi
ta; n1 )s= n ¼ 0:174 by calculating 1.7 (mean difference) e 1.526 (Lower
Bound under 95% confidence interval). And the value of a is 4.526 since the
sample mean is 4.7. Also, subjects agree with the statement because
m0 ð3Þ < að4:526Þ.

Next, to verify the level of significance, we examine whether the p-value is

greater than a (0.05). The p-values of these two statements are
0.00000000000000054387 and 0.00000000000000000169, which are much
smaller than a (0.05). As a result, the result of the testing is very significant.

In conclusion, the proposition 3 e ‘Engaging into the field to conduct observa-

tion and interviews as well as combining seen phenomenon are both important
when doing design synthesis’ is therefore justified. This also echoes what
Kolko (2010) argued, ‘I saw this þ I know this ¼ Insight’.

5.3 Discussion of findings

The purpose of this study is to design an IT-based tool to facilitate the insight
discovery process in service design. In the previous sections, we describe the
experiments and evaluate the propositions as well as provide explanations
for the results of the collected data. Here we offer a deeper discussion of the
findings.

First, in proposition 1 the ability of processing mental imagery refers to service

designers framing and reframing the design situation to discover insights. We
examine the competences required to form and process mental imagery and
find that all the associations the subjects used fall into the five types of ability

Semantic reframing of service design insight discovery 171

 mentioned above. As a result, proposition 1 is established. In this study, the
processing mental imagery involves only words (no pictures) based on our
tool using the word-based commonsense knowledge of ConceptNet. When
considering to add visual images into our tool, they would be still regarded
as information/imagery on which the association reasoning methods could
draw upon for the reframing and frame creation process.

Furthermore, we find that there are some differences between the designer
group of subjects and the non-designer group of subjects. According to their
usage of different kinds of associations, it can be inferred that designers
have more perceptual mindsets while most of the non-designers are more
logical. We can also find some support based on the inspiration seeking
behavior of each group of subjects. As we mentioned above, non-designers
tend to have logical assumptions after reading the design brief. Therefore, their
usages of Google Search in phase 2 are higher than designers since they want to
seek information exactly matching their assumptions.

Second, we want to examine whether the designed artifact e Discoverþ is able

to increase the quality of derived insights. We evaluate this question form both
macro and micro views. From the macro view, we use insight depth maps to
compare the depth of the insights derived in phase 1 with the ones in phase
2. We find that more than half of the subjects (17 out of 30) deepen their in-
sights with the facilitation of our tool. On the other hand, from micro view,
we evaluate the perceived satisfaction of insight quality of the subjects. We
examine whether the subjects were satisfied with three dimensions and the
result is positive. That is, the subjects thought their insight quality achieved
higher innovativeness and integrity. We also find that the percentage of sub-
jects using Discoverþ in phase 2 is 85% compared to 15% using Google
Search. Furthermore, for the subjects who increased their insight quality,
the percentage of those using Discoverþ went up to 90%. Hence, we argue
that, compared with Google Search, Discoverþ provides a better source of
inspiration for insight discovery since its data and representations are more
logical and suitable for supporting a process involving a lot of manipulations
of mental imagery.

At last, proposition 3 aims to examine whether engaging in the environment

and recalling of past experience are both important when doing design synthe-
sis. We used one-sample T test to evaluate this issue and found that these two
factors are both very important for insight discovery.

There are additional results considering subjects’ work experiences that are
not shown in this paper. However, we still find an interesting difference be-
tween subjects with and without work experience. There were 10 subjects
(7 non-designers and 3 designers) with one to four years of full-time work
experience while the other 20 subjects did not have experience. We find

172 Design Studies Vol 40 No. C September 2015

 that in most aspects of analysis, subjects with work experience meet our ex-
pectations; that is, they behave more like non-designers (i.e., their concept
maps are more skewed) since there were more non-designers in the group
of subjects with work experience. Nevertheless, there is one interesting point
that is noteworthy. On average, contiguity associations account for 27% of
the association usage of subjects with work experience while these associa-
tions account for only 11% for those without experience. This disparity is
more than the difference between designers and non-designers. We infer
that is because once people begin to work in an industry, they put more
emphasis on feasibility and completeness of their thoughts to avoid unneces-
sary costs.

6 Conclusion
This study presents an IT-based tool named Discoverþ that is designed to
facilitate the insight discovery process in service design. It can be applied to
many contexts, including big enterprises, SMEs, startups or even non-profit
organizations. Service designers in all of these contexts can take advantage
of Discoverþ to make their tasks easier. The evaluation results illustrate the
ability of our tool to make derived insights more innovative and complete.
Once the insight quality is good, the service to be developed can become
more feasible, desirable and viable. The proposed insight depth map can
also be adopted in practice to measure the scope of influence of insights. Ser-
vice designers, especially those in SMEs and startups with less resources, can
formulate strategies including marketing, pricing, alliances, etc., if they are
able to grasp the insight depth earlier, before services are delivered to the mar-
ket. In addition, it can also help to position the service to be developed in the
market and recognize competitors. In addition, the evaluation results show
that the most used associations are context association, contiguity association
and similarity association. Another two types of associationeanalogy associ-
ation and contrast association are used much less, especially by people who are
not from design schools. Therefore, if we can urge service designers to try to
use these two types of association, it could refresh their perspectives and stim-
ulate new thoughts they have not thought of before.

This study still has some limitations and shortcomings First of all, due to time
constraints, we only studied a small number of subjects. Hence, for the pur-
pose of generalizing the results for larger groups and to discover more impli-
cations, future research should engage more participants in different fields with
various kinds of work experience. The knowledge used in the system comes
from ConceptNet. Although it contains a large amount of assertions, the
knowledge is not optimized for service design and may contain some noise
that could influence users. If we want to provide a better use experience, a
new knowledge could be developed specifically for service design. Finally,
some of qualitative results of the experiment could be further analyzed and

Semantic reframing of service design insight discovery 173

 interpreted by different interpreters in order to increase the interpretation
objectivity.

References
Ahlberg, M., & Vuokko, A. (2004). Six years of design experiments using concept
mapping-at the beginning and at the end of each of 23 learning projects. No.
2004. Concept maps: Theory, methodology, technology. Proceedings of the first
international conference on concept mapping, Vol. 1000 (pp. 45e51).
Akbik, A., & Michael, T. (2014). The Weltmodell: a data-driven commonsense
knowledge base. In Proceedings of the ninth international conference on lan-
guage resources and evaluation (LREC’14), Reykjavik, Iceland.
Buzan, T. (1995). The mind map book (2nd ed.). London: BBC Books.
Cambria, E., Hussain, A., Havasi, C., & Eckl, C. (2010). SenticSpace: visualizing
opinions and sentiments in a multi-dimensional vector space. In Knowledge-
based and intelligent information and engineering systems (pp. 385e393). Berlin
Heidelberg: Springer.
Chi, P. Y., & Lieberman, H. (2011). Intelligent assistance for conversational sto-
rytelling using story patterns. In Proceedings of the 16th international confer-
ence on intelligent user interfaces (IUI ’11) (pp. 217e226), New York, USA.
Cross, N. (2004). Expertise in design: an overview. Design Studies, 25(5),
427e441.
Dawson, M. R. W., & Medler, D. A. (2010). Laws of association. Dictionary of
Cognitive Science. http://www.bcp.psych.ualbertaca/wmike/Pearl_Stree/Dic-
tionary/contents/L/lawsofassoc.html.
Design Council. (2012). The double diamond design process model. Accessed in
October, 2014. http://www.designcouncil.org.uk/designprocess.
Dorst, K. (2011). The core of ‘design thinking’ and its application. Design Studies,
32(6), 521e532.
Goldstein, S. M., Johnston, R., Duffy, J., & Rao, J. (2002). The service concept:
the missing link in service design research? Journal of Operations Management,
20(2), 121e134.
Klein, G., Moon, B., & Hoffman, R. R. (2006). Making sense of sensemaking 2: a
macrocognitive model. Intelligent Systems, IEEE, 21(5), 88e92.
Kolko, J. (2014). Methods of design systhesis. Accessed October, 2014. http://jon-
kolko.com/projectFiles/preso/kolko_2009_02_methodsOfSynthesis.pdf.
Kolko, J. (2010). Abductive thinking and sensemaking: the drivers of design syn-
thesis. Design Issues, 26(1), 15e28.
Kwan, S. K., & Yuan, S. T. (2011). Customer-driven value co-creation in service
networks. In The science of service systems (pp. 189e206). US: Springer.
Lindegaard, S. (2011). Making open innovation work. CreateSpace Independent
Publishing Platform.
Liu, H., & Singh, P. (2004a). ConceptNet: a practical commonsense reasoning
toolkit. BT Technology Journal, 22(4), 211e226.
Liu, H., & Singh, P. (2004b). Commonsense reasoning in and over natural lan-
guage. In 8th conference on knowledge-based intelligent information & engineer-
ing systems (KES-04) (pp. 293e306).
Novak, J. D. (1998). Learning, creating, and using knowledge: Concept maps as
facilitative tools in schools and corporations. Mahwah, NJ: Lawrence Erlbaum
Associates.
Paton, B., & Dorst, K. (2011). Briefing and reframing: a situated practice. Design
Studies, 32(6), 573e587.

174 Design Studies Vol 40 No. C September 2015

 Reckwitz, A. (2002). Toward a theory of social practices: a development in cultur-
alist theorizing. European Journal of Social Theory, 5(2), 243e263.
Ryan, G. W., & Bernard, H. R. (2003). Techniques to identify themes. Field
Methods, 15(1), 85e109.
Salton, G., & Buckley, C. (1988). Term weighting approaches in automatic text
retrieval. Information Processing and Management, 24(5), 513e523.
Shen, E., Lieberman, H., & Lam, F. (2007). What am I gonna wear?: scenario-
oriented recommendation. In Proceedings of the 12th international conference
on intelligent user interfaces(IUI ’07) (pp. 365e368).
Speer, R., & Havasi, C. (2012). Representing general relational knowledge in
ConceptNet 5. In International conference on language resources and evaluation
(LREC’12), Istanbul, Turket.
Toulmin, S. (1972). Human understanding: The collective use and evolution of con-
cepts. Princeton, NJ: Princeton University Press.
Vargo, S. L., & Lusch, R. F. (2004). Evolving to a new dominant logic for mar-
keting. Journal of Marketing, 68(1), 1e17.
Wilson, M. (2002). Six views of embodied cognition. Psychonomic Bulletin & Re-
view, 9(4), 625e636.

Semantic reframing of service design insight discovery 175