View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Royal College of Art Research Repository

REASONING IN DESIGN: IDEA GENERATION CONDITION EFFECTS

ON REASONING PROCESSES AND EVALUATION OF IDEAS

Claus L. Cramer-Petersen Saeema Ahmed-Kristensen

[email protected] [email protected]

Technical University of Denmark

DTU Management Engineering, Design Engineering and Innovation
Produktionstorvet, building 426, DK-2800, Kgs. Lyngby

ABSTRACT
Reasoning is at the core of design activity and thinking. Thus, understanding and
explaining reasoning in design is fundamental to understand and support design
practice. This paper investigates reasoning in design and its relationship to varying
foci at the stage of idea generation and subsequent performance of ideas developed.
Understanding reasoning in design and its relationship to the performance of ideas
generated is important to understand design activity, which can be used to develop
tools or methods that can improve the effectiveness of design teams. Protocol analyses
were conducted to investigate idea generation sessions of two industry cases.
Reasoning was found to appear in sequences of alternating reasoning types where the
initiating reasoning type was decisive. The study found that abductive reasoning led to
more radical ideas, whereas deductive reasoning led to ideas being for project
requirements, but having a higher proportion being rejected as not valuable. The study
sheds light on the conditions that promote these reasoning types. The study is one of
the first of its kind and advances an understanding of reasoning in design by empirical
means and suggests a relationship between reasoning and idea performance. Findings
of the study further allows for a way to analyse and improve the performance of idea
generation in design teams.

INTRODUCTION
Reasoning is a cognitive activity that dictates how humans respond to situations in
every aspect of their lives. Design activity relies on the reasoning processes of
designers. Therefore, understanding the role that reasoning plays in design is critical
to understand how design takes place. Rittel (1987) states that “only at the microlevel
can we identify patterns of reasoning corresponding to [the design process]” and thus
understand design activity. Furthermore, reasoning in design between designers is
proposed to take an argumentative form and thus possible to study through design
activity between designers (Polk and Newell, 1995).
Reasoning in design has been theorised from a logical perspective according to
abductive, deductive and inductive reasoning types (Dorst, 2011; Roozenburg, 1993),
suggesting that abductive reasoning is dominant in design. An empirical study in
relation to the evaluation of design ideas was conducted, finding that not only
abductive reasoning is important to design activity (Dong et al, 2012). Hence, a gap is
identified in literature on how to empirically study reasoning as it occurs over the
course of a design process involving several designers. Furthermore, investigating
how reasoning is related to the performance of design outcomes is relevant to explain
the effects of different ways of reasoning.
Therefore, this paper seeks to contribute to existing literature by empirically
analysing reasoning processes involved in the generation of ideas in real life industrial
development projects. Additionally, the relationship between reasoning processes and
the value of ideas resulting from these processes is investigated.
First, the paper reviews and presents existing theories on reasoning and problem
solving in relation to design activity and design models. Second, protocol analyses of
verbal concurrent data from industry are suggested as a viable method to understand
reasoning and relations to performance of idea generation. Third, results are presented
and discussed, using qualitative and quantitative analyses to advance theory on
reasoning in design. Fourth, the implications of the study contributions are discussed
and put into perspective.

BACKGROUND AND THEORY

In the following sections, the background of the study is explained and motivated
on the basis of theories and findings from relevant literature. First, reasoning is
defined as a cognitive concept and from the perspective of formal logic. Second,
reasoning as studied in design is presented through selected models of problem
solving and design activity. Third, creative processes involved in idea generation and
how such processes can be evaluated are presented. Fourth and finally, similar
empirical studies of reasoning and design activity are reviewed.

Cognitive reasoning and mental models

Reasoning is a cognitive activity that decides how humans respond to situations in
every aspect of their lives. It consists of trains of thought, including deliberation,
arguing and logical inferences. Some reasoning is argued to be largely unconscious,
while reasoning also exists in a verbal, argumentative form (Rittel, 1987). Reasoning
processes are suggested to be independent from domain or context, but the beliefs and
knowledge underpinning reasoning in a certain situation relies on the mental model(s)
held in a context (Johnson-Laird, 2006).
Mental models are individually constructed working models that allow to integrate
information and make predictions about the world (Badke-Schaub et al, 2007).
Consequently, reasoning is initiated by mental models, but the results of reasoning
processes are evaluated according to whether they adhere to one or more mental
models (Johnson-Laird, 2006). Thus, while reasoning is initiated in response to a
certain context, the result of good reasoning expands the context, resulting in
knowledge that influences other mental models (Rouse and Morris, 1986).

Logical reasoning types

Since the works of C.S. Peirce, logical reasoning has been formulated as being of
either deductive, inductive or abductive types (March, 1976). Even though their
definitions are debated in philosophy, the three types remain a reference point (Kroll
and Koskela, 2014). In the following paragraphs, the three types are explained to offer
a logical definition as well as interpretation in regards to how such reasoning can be
interpreted as cognitive processes.

Deduction
Deduction is the inference of a result from a rule and a case; if a rule known to be
true (a law or theory) and a known case (the circumstances that apply) results can be
predicted as an effect brought about by a cause (Roozenburg, 1993).
Deductive reasoning is tautological as it allows us to arrive at a conclusion from
the logical implication of two or more propositions asserted to be true (Magnani,
1995; March, 1976; Reichertz, 2010). Consequently, deduction is truth-conveying and
proves something must be.
Hence, deductive reasoning is analytical and predictive, as it is an explication of an
already known or accepted relationship. It is “the process of establishing that a
conclusion is a valid inference from the premises” (Johnson-Laird, 2006).

Induction
Induction is the inference of a rule from a case and an effect. In the context of
research the rule is a hypothesis to be tested, the case comprises the experimentally
produced conditions derived from a hypothesis and the result is the outcome of the
experiment that confirms the prediction (Roozenburg, 1993).
Inductive reasoning is the process of deriving plausible conclusions that go beyond
information in the premises (Johnson-Laird, 2006). Inductive reasoning is tautological
like deductive reasoning in that it infers concepts only from available data within a
model or frame of reference (Magnani, 1995; Reichertz, 2010; Schurz, 2007).
However, inductive reasoning differs from deductive reasoning in that it tests to show
whether something is actually operative or true within a model (March, 1976).

Abduction
Abduction is the use of a known principle, law or theory for the purpose of a causal
explanation. The rule is given as a premise and allows us to reason from an effect (the
results) to the cause (the case). The conclusion of this inference is a hypothesis for a
cause to the effect that has to be explained (Roozenburg, 1993).
Abductive reasoning is a process of conjecture that yields an explanation to an
event. An abduction is the preliminary estimate that introduces plausible hypotheses
and informs where to first enquire by choosing the best candidate among a multitude
of possible explanations (Magnani, 1995; Schurz, 2007).
Abduction thus differs from deductive and inductive reasoning in that abductions
involve guess work and (sometimes unfounded) assumptions as the basis for
reasoning. Within the design field, abductive reasoning has in certain cases also been
suggested to involve the conception of new rules or types of relationships to explain
an intended outcome. This variation of reasoning has been termed innovative
abductive reasoning by Roozenburg (1993) to emphasize the creative outcome of such
reasoning (Dorst, 2011; Habermas, 1972; Reichertz, 2010).
Following Pierce, it is suggested that abduction is the only form of inference that
introduces new ideas, because induction does nothing but determine a value and
deduction merely evolves the necessary consequences of pure hypothesis. Therefore,
abductive reasoning is suggested to be the dominant type of reasoning in design
activity (Dorst, 2011; Roozenburg, 1993).

Reasoning in design
Reasoning in design has been theorised from the perspective of logics and the
terms of deductive, inductive and abductive reasoning (Dorst, 2011; Kroll and
Koskela, 2014; March, 1976; Roozenburg, 1993). These theories consider design and
its reasoning processes in terms of how design reasoning can be perceived as a logical
phenomenon, and are thus not based on empirical data or design in actual practice.
Roozenburg (1993) suggests that truly innovative design is dominated by
innovative abductive reasoning. He acknowledges that that while routine design often
relies on ideas from e.g. precedent designs, innovative design solutions comprise both
descriptions of form and prescriptions of actuation. He exemplifies this with the
imagined first development of a kettle to boil water, which necessitated inventing both
a container to hold water in place in parallel with a way to facilitate the user to fill the
just developed container with water and heating the water. Roozenburg (1993) defines
this synthetic action of both suggesting a new structure and a (supposedly) new
principle of heating water on a burner in order to achieve a desired function to be the
kernel of the design process. Therefore, innovative design requires both abductive
reasoning and a principle that is new to the world.
Dorst (2011) makes a similar argument on the relationship between abductive
reasoning and design thinking. He defines design reasoning as the generation of
proposals for a ‘thing’ and a related working principle to achieve an aspired value.
This creative feat is done by framing the problem after Schön (1959) to develop a
perspective from which the thing and it’s working principle can be developed in
conjunction to achieve the aspired value.

Reasoning in problem solving

While the previously described contributions on reasoning in design acknowledge
that other types of reasoning is present in design (e.g. deductive, inductive) they focus
mainly on the reasoning activity of entirely new concepts. Thus, it is relevant to
review theories of problem solving and models of design that more fully describe the
process of reasoning. This is done in the following.
From the field of cognitive science, Johnson-Laird (2006) describes a generic
problem solving cycle as the “…use [of] some constraints to generate a putative
solution, and other constraints, such as the goal of the problem, to criticise and
amend the results". This definition has clear parallels to earlier conception of the
types of reasoning involved in the three stages of inquiry: "Abduction invents or
proposes a hypothesis; it is the initial proposal of a hypothesis on probation to
account for the facts. Deduction explicates hypotheses, deducing from them the
necessary consequences which may be tested. Induction consists in the process of
testing hypotheses" (Fann, 1970, after Peirce).
Thus, the process of reasoning in problem solving can be interpreted as a three
stage process involving; 1) an abduction that leads to a certain framing, explicitly or
implicitly from mental models, followed by 2) deductions that concretise and predict a
solution or effect under the conjectured framing, and finally 3) an inductive reference
to principles or accepted facts (possibly 'outside' the framing) that evaluates and tests,
leading to a new iteration if the result is not satisfactory.

Models of design
Similar processes to those defined above have been proposed in the following three
models of design.
March (1976) suggests the Production-Deduction-Induction (PDI) model, from the
works of Peirce, as a rational design process of cyclic iterative procedures
characterised by three different types of reasoning:
a) Productive reasoning is the design or composition of something novel. It
suggests that something may be, and is analogue to abductive reasoning.
b) Deductive reasoning is a decomposition comprising prediction of performance
characteristics of a design that emerges from analysis of the composition. It
proves that something must be.
c) Inductive reasoning is a supposition from the accumulation of knowledge and
the establishment of values evolving from the prior productive and deductive
reasoning. It tests whether something actually is.
 Gero and Kannengiesser (2004) argues in their Function Behaviour Structure
(FBS) framework for eight processes to be fundamental to all designing. Of these, the
first four processes describe a cycle similar to descriptions of reasoning in design:
a) Formulation specifies an initial state interpreted from design requirements by
transforming a function to behaviour expected to enable that function.
b) Synthesis transforms expected behaviour into solution structures that can
realise the desired behaviour.
c) Analysis derives an actual behaviour from the structure
d) Evaluation compares actual behaviour with intended behaviour to make a
decision possible.
There is a direct relation between these four steps to the three logical reasoning
types.
An alternative model of design is proposed by Hatchuel and Weil (2008). The
Concept-Knowledge (C-K) theory assumes “design as an interplay between two
interdependent spaces with different structures and logics”. C-K theory strongly
emphasises the interplay between what is conjectured or unknown and what is known
or in existence and describe four operators between concept and knowledge:
a) C-K operators search attributes in K-space that can be used to partition
concepts in C. They contribute to the generation of propositions in K, and can
be understood as ‘design solutions’.
b) K-C operators are symmetrical to C-K operators. They generate tentative
concepts from K attributes and thereby assess those new concepts.
c) C-C operators are virtual, and therefore implicit or tacit, and are of importance
to the formation of C-K operators.
d) K-K operators comprise logical types of reasoning (deduction, induction,
abduction). These operators are possible to run as a program following a
formal logic.
Although the three reasoning types are not explicitly mentioned, similarity is
interpreted in that operators a and c are abductive, b is inductive and d is deductive
reasoning.
The three presented theories emphasise that design thinking features mental
activity that concerns (a) the notion something novel and useful which is (b)
concretised and explored and (c) evaluated or used to amend the original notion or
concept. While the PDI model and the FBS framework presents this as an iterative or
cyclic process, C-K theory does not make a-priori distinction between the sequences
of the different operators presented by the theory.

Reasoning is argumentative
Rittel (1987) argues that design reasoning is disorderly due to the nature of design
problems, where learning about the problem is the problem. He further states that by
perceiving design as a process of argumentation between people, competing positions
and other issues, problems and potential solutions are understood to be debated and
developed in parallel (Dorst and Cross, 2001; Rittel, 1987). Analysing such a debate,
or dialogue, between groups of people engaged in design therefore holds the potential
to understand and explain reasoning as verbal reasoning is the deployment of
linguistic processes to satisfy the demands of a cognitive reasoning task (Polk and
Newell, 1995). Thus, a key assumption of the study presented in this paper is that
reasoning can be identified and analysed through verbal utterances between people.
Even though some reasoning is unconscious, the part of reasoning that is put
forward in verbal utterances is suggested to be argumentative (Hogan et al, 1999;
Johnson-Laird, 2006). Verbal utterances are suggested to have an illocutionary force,
meaning that an utterance holds implicit, culturally defined meaning (Searle, 1979). A
study of engineering design conversations conclude that the skill of constructing
effective arguments are important, alongside technical or objective grounds, to convey
convincing contributions to design activity (Lloyd and Busby, 2001).
Furthermore, studies conclude that the use of initial framing from relevant issues in
a design situation guides the following activity (Valkenburg and Dorst, 1998).
Therefore, it is important to acknowledge that the way in which a reasoning process
starts, is important for the following dialogue.

Creative processes and idea generation

The early stages of idea generation in engineering design designate an important
part of the design process. The stage of idea generation is characterised by fewer
decisions having been made and thus a greater openness for exploring new and
creative ideas. Thus, investigating the reasoning processes of the stage of idea
generation is pertinent to understand and support the overall design process.
Creativity and the generation of creative ideas are key for the on-going design process
and are characteristic of a successful process of idea generation (Cross, 2001).
The standard definition of creativity proposed by Runco and Jaeger (2012), defines
creative outcomes to be both original and useful. Thus, it is relevant to evaluate
developed ideas according to their usefulness and originality understood as the
practicality of meeting the needs at hand (Keshwani et al, 2013; Ward and Kolomyts,
2010).
In engineering design, a common consideration is the trade-off between issues of
the key product features (Ahmed et al, 2003). Therefore, it is of relevance to
investigate the reasoning process in correlation to how ideas are evaluated and under
which conditions and constraints the ideas are generated.

Similar studies of reasoning

Similar empirical studies of reasoning in design were reviewed.
Dong et al (2012) analysed concurrent verbal protocols of reasoning processes
between participants evaluating design ideas and concepts in terms of deductive,
inductive and abductive reasoning, defined as: Deductive reasoning as the drawing of
conclusion from explicit premises; inductive reasoning as the generalisation from a
specific instance: and abductive reasoning as the reframing of users’ needs. They
conclude that abductive reasoning in evaluating ideas lead to fewer rejected ideas. The
opposite was found for deductive reasoning, where more ideas were rejected.
A study by Lloyd and Scott (1994) analysed think aloud protocols of engineering
designers for generative, deductive and evaluative reasoning and conclude that
experience leads to more generative reasoning and less deductive reasoning, while the
opposite is true for novice designers. They interpret reasoning as follows: Deductive
reasoning is understanding of and specification of a problem; generative reasoning is
presence of a new solution; and evaluative reasoning is reflection on intent and
strategy.
A similar conclusion was reached by Ahmed et al (2003) in analysing think aloud
protocols of novice and expert designers. Additionally, they found an overall
reasoning pattern adopted by designers to generate, implement and evaluate decisions,
where experienced designers were able to do a preliminary evaluation of decisions
before implementation.
 Christensen and Schunn (2009) studied the role of mental simulations in design
from protocols of concurrent verbalisation of design teams. Mental simulation,
interpreted as a reasoning process, was found to reduce uncertainty by turning the
uncertainty into approximate answers.
Valkenburg and Dorst (1998) applied Schön’s (1959) reflective practice theory to
analyse the concurrent protocols of groups of engineering students engaged in design
activity. They analyse four primary activities based on naming, framing, moving and
reflecting according to Schön’s theory and conclude that the reasoning process of the
groups switches and iterates between reasoning about sub-functions and their
integration.
From the literature, reasoning has been reviewed from both a logical perspective
and from empirical studies of design. Reasoning from logical literature focus on three
primary reasoning types, namely abductive, deductive and inductive reasoning. The
reviewed literature on reasoning and cognitive processes in design studies highlights a
gap in utilising these three reasoning types to describe design activity. Dong et al
(2012) have employed the approach, however only on the evaluation of ideas and not
on idea generation. Thus, this paper contributes to literature and building of theory
through investigating design activity using the perspective of the three reasoning
types. In addition, the study focuses on real life development project with design
practitioners and not students.

STUDY AIMS
The study aimed to understand how reasoning can be identified and described
empirically. The study aimed to investigate and understand the relationship between
different conditions for idea generation and the reasoning processes of designers. The
third aim was to investigate the influence of idea generation conditions on the
evaluation of ideas that are the outcome of these conditions.
Based on the review of literature, the following two sets hypotheses were
proposed:
 Hypothesis 1a: When idea generation conditions are less constrained, a higher
proportion of abductive reasoning is expected to initiate ideas.
 Hypothesis 1b: In contrast, when idea generation conditions are more
constrained, a higher proportion of deductive reasoning is expected to initiate
ideas.
Ideas generated under less constrained conditions are expected to motivate
abductive reasoning, since such reasoning relies on possibility and imagination. Idea
generated under more constrained conditions are expected to motivate deductive
reasoning since the focus is likely be on variations of known solution types and
technical principles.
 Hypothesis 2a: Ideas generated under less constrained conditions are more
likely to be evaluated as being potentially highly useful, but too radical for an
on-going project.
 Hypothesis 2b: Ideas generated under more constrained conditions are likely to
be evaluated as being directly applicable to the development project.
Ideas developed when under less constrained conditions are expected to be
dominated by abductive reasoning, as per hypothesis 1a. Thus, resulting ideas are
likely to be too radical in relation to the on-going development project. Ideas
generated under more constrained condition, are expected to be evaluated as being
directly applicable to the project since deductive reasoning is expected to dominate, as
per hypothesis 1b.
METHOD
Data was collected through idea generation sessions with real world industry
development projects and participants. The data was collected from two SMEs, each
of which had already undertaken research into 1) user and customer needs through
observations and interviews and 2) product cost factors through benchmarking of
competitor products and/or product teardowns. A workshop was conducted with
varying idea generation conditions provided in three, half-hour periods. The inputs to
these conditions were derived from previous analyses of user and customer needs and
cost factors. Generated ideas were evaluated by the participants after the session.
At the time of data collection, both companies were at the stage of early idea
generation of on-going development projects, one of these was the design of a support
tool for the construction industry, the other a system to compress waste. The
workshop designed for data collection was therefore part of the actual product
development process and not an artificial task defined for the purpose of the study.
Prior to the workshops, the author met with the companies to discuss the primary
conclusions of prior analyses and decided on a set of primary set of product
requirements from which the idea generation could proceed. Table 1 summarises the
main characteristics of the participating companies.

Company Product Number of Participant roles in company Team size

type employees
1 Construction ~10 Project manager, design engineer, 4
tools industrial designer , engineering
intern
2 Waste ~80 Head of development, design 8
management engineer, production engineer,
supply- and logistics manager,
mechanical engineering consultant,
head of sales
Table 1: Case company details

The idea generation session was designed and facilitated by the author and
consisted of three rounds of 30 minutes of idea generation under different conditions,
simulating varying constraints to the process. These were:
 Round 1 focused on an open brainstorm where existing and new ideas were
developed.
 Round 2 focused on meeting user and customer needs without being
concerned with cost or other aspects.
 Round 3 focused on reducing cost as much as possible while only not adding
functionality in the product or removing not valuable functionality.
The workshops were both audio and video recorded with company participants
offered anonymity. The workshops were facilitated to allow the participants to
generate many, quick ideas documented on post-its as keywords and/or sketches.
These were accompanied by a brief verbal presentation to the other participants.
Each round started with a few minutes spent on individual brainstorming followed
by presentation of ideas to the group and later the building on each other’s ideas. The
facilitator ensured that everyone was heard, that ideas were presented in a sufficient
amount of detail and recorded on single post-its. Finally, the facilitator ensured that
time was kept and to move the focus of the idea generation across the three rounds
and also in the individual rounds by directing focus on particular components or
aspects of the product that had been deemed important from the prior user and cost
analyses.
Following the idea generation rounds, the developed ideas were evaluated by the
participants. The ideas were sorted in two matrices according to the usefulness and
originality of each idea in relation to the on-going development project. Ideas and
their evaluation were documented by photographs for later analysis. Table 2
summarises the workshop flow for the generation of ideas under varying conditions
and the following evaluation.

Activity Idea generation Evaluation

Open brainstorm Focus on user and Focus on cost Evaluation of
customer needs reduction ideas
Example of Documenting and Easy detachment How to reduce Sorting ideas
activity presenting from pulley number of unique by two matrices
existing ideas parts
Time 30 minutes 30 minutes 30 minutes N/A
Table 2: Idea generation session and evaluation overview

Analysis method
Protocol analyses of concurrent verbalisation in the recordings were conducted.
Protocol analysis of design activity is a way to understand underlying cognitive
processes, e.g. reasoning, with minimal interruption of the recorded process (Ericsson
and Simon, 1993; Christensen, 2009).
Protocol Analysis as a method has been recognised in the promise to capture
aspects of design activity in great detail given a proper match between data collection
design and outcome protocol data (Dorst, 1995). Consequently, verbal protocol
analyses of real life industrial development projects is relevant and expected to be
highly representative of design cognition found in practice (Ahmed et al., 2003; Chi,
1997; Christensen, 2009). In this case, as the observations were in groups no
additional verbalisation is required, hence a minimum of interference with thought
processes.

Coding scheme development

The protocols were coded according to a coding scheme drawing from existing
studies and adapted to the protocols of similar data-set also collected by the authors.
The protocols were segmented and highlighted for indicator words. In the first
round in the coding all segments were coded for the presence of an idea or and idea
aspect (Badke-schaub et al , 2007). Next, the segments were coded as abductive,
deductive, inductive or no reasoning. Table 3 below shows the coding flow, before
describing the coding process in further detail.

Coding step 1 – identifying ideas 2 – classifying reasoning

Code list IDEA ABDUCTION
IDEA ASPECT DEDUCION
INDUCTION
Table 3: Coding process overview
 To ensure that ideas and reasoning processes could be captured in the coding
process, the entire protocol was segmented and analysed (Chi, 1997). The
segmentation was conducted to ensure the data analysis was manageable in relation to
both size and intention of the participants to allow coding of ideas and reasoning
processes (Christensen and Schunn, 2009; Goldschmidt, 1991). Hence, it was
possible to analyse the utterances by the shortest practical elements of meaning. The
reasoning steps are elaborated in the following sections.

Coding idea and idea aspect

The first step of analysing the protocols involves the identifying of ideas uttered by
the groups. In the online dictionaries, ideas are commonly defined as concerning: 1) A
course of action, 2) a basis on something believed valid and 3) an imagined outcome
(Merriam-Webster; Oxford).
In engineering design research, ideas are similarly defined as: an external
perspective: “an idea as a solution to a single function” (Linsey et al., 2011), “idea
as notion related to a task” (Goldschmidt and Tatsa, 2005), as a solution for a
problem or sub-problem (Badke-schaub et al., 2007). These perspectives have in
common that an idea is a proposition for a solution to a function, involving intentional
work (the task).
As ideas involve solutions and sub-solutions (Badke-schaub et al, 2007), it is
necessary to perceive ideas as being put forward in a distributed manner, and at
different levels of abstraction (Voss, 2006). Furthermore, ideas are not necessarily put
forward by a single person over the course of conversation, but can be elaborated by
other persons than the one initiating it. Consequently, the protocols will not be
distinguished by a complete uninterrupted utterance, but rather a group of utterances
relating to an idea put forward and related aspects of that idea. This group of
utterances, and belonging segments, are hereafter referred to as idea episodes (Chi,
1997). These definitions were the basis of coding ideas and idea aspects.

Reasoning definitions and indicators

An initial coding scheme utilising case, rule and result from a logical perspective
of reasoning suggested by Roozenburg (1993) was first employed, but abandoned due
to poor coder reliability because definitions were unclear in the verbal protocol.
Instead, definitions of what characterises the different types of reasoning were
derived from theories of reasoning processes and models of design activity presented
in the earlier sections. The definitions used were oriented towards the suggested role
or function that the three types of reasoning serve in reasoning processes. Refer to
table 4 below for the full definitions used for coding.

Reasoning type Coding definitions

Abduction • A hypothesis or assumption to account for what is observed or
what is desired or intended
• Creating ideas (to a problem) from imagination
• A belief held without proof or certain knowledge
• Preliminary guess to introduce hypotheses
Deduction • Definitive and certain conclusion
• Explicating hypothesis by suggested consequences
• Prediction of result in a given frame
• Proves something must be
• Explores consequences of an abduction
Induction • Tests a hypothesis with available data (predictions)
• Generalises from specific instance or idea
• Evaluates if something is operative
• Inferring from observed to unobserved
• Inferring about future courses of events
References (Fann et al, 1970; Johnson-Laird, 2006; Magnani, 1995; March, 1976;
Reichertz, 2010; Roozenburg, 1993; Schurz, 2007)
Table 4: Reasoning type and definitions for coding

To support the coding of reasoning types, a data-driven approach was taken by

reviewing the transcripts for the presence of any common features of segments
initially evaluated to belong to the different type of reasoning. This process resulted in
the identification of three groups of indicator words that signify the three different
types of reasoning. A similar approach was adopted, amongst others, by Christensen
and Schunn (2009) in the analysis of verbal concurrent protocols. The words were
derived in Danish, as the data was in Danish and all analysis was completed in the
original language. The full list of indicator words is translated to English for the
purpose of reporting in table 5.

Reasoning Abduction Deduction Induction

Indicator could, maybe, think, so, then, therefore, that I, me, you, they, we,
words could be, imagine, is, must be, as, can them
probably, likely
Table 5: Reasoning indicator words.

The common characteristics for the indicator words were as follows:

 Abductive reasoning: Conveys uncertainty, possibility and serves to frame the
elaboration of an idea on the remaining segments of an idea episode.
 Deductive reasoning: Co-occur with segments and words which convey a
conviction, justified belief or consequence in response to a situation.
 Inductive reasoning: Often comes after the idea episode is finished and tends
to co-occur with the use of pronouns (e.g. I, you, we) as a way for a person to
judge or qualify an idea.

Coding reasoning types

The segments coded for idea or idea aspects were then coded for using abductive,
deductive or inductive reasoning. Induction codes could also be applied to segments
outside the idea episodes, as they were often realised as delayed evaluations of an
idea.
To check for reliability of the coding scheme, a second researcher coded the
segments and an inter rater reliability score was calculated using Cohen’s Kappa. The
coding of idea episodes reached a Kappa of 0,71 based on 460 segments and coding
of reasoning type reached a Kappa of 0,61 based on 353 segments. Both scores are
deemed good and hence justify the application of the coding scheme in the present
study.

Idea evaluation
After the generation of ideas, they were evaluated by the participants. The frame of
reference for evaluating the ideas was the on-going development project.
 Each idea was evaluated in a two-by-two matrix. The criteria were a high/low fit to
the primary product requirements (on the vertical axis) and whether the idea would
provide a high or a low value for the customer/user (on the horizontal axis). Ideas
evaluated high in both axes (i.e. the top right quadrant) were moved to the second
matrix for further evaluation.
The second two-by-two matrix distinguished whether the idea had a high/low value
to the company (on the vertical axis), and according to low/high risk, complexity
and/or development effort (on the horizontal axis). Ideas evaluated positively on both
axes were accepted directly for the further design process.
The first matrix evaluated the value to the user/customer. By evaluating according
to the on-going project, ideas were filtered to only those with immediate potential for
realisation and thus innovative potential (Ward and Kolomyts, 2010). The second
matrix evaluated ideas according to their usefulness to the development project by
filtering ideas that would require too much further analysis or ideas that had a low
value to the company in general.

RESULTS AND FINDINGS

A total of 3760 segments and 155 idea episodes were coded and it was found that
reasoning occurred in 51% of all segments and in 89% of segments coded for the
presence of ideas or idea aspects. Other non-coded segments could be explanations,
team coherence, informal talk, etc., but these were not coded or analysed. Table 6
shows the total distribution of reasoning of the two protocols.

Reasoning type Abduction Deduction Induction

Coded reasoning Count 248 1367 321
Frequency 13% 71% 16%
Reasoning in start Count 66 72 15
of idea episode Frequency 43% 47% 10%
Table 6: Distribution of reasoning in both analysed protocols

Reasoning in design
Perhaps surprising is that abductive reasoning is the least frequent type of
reasoning in the protocols. However, when investigating idea episodes, it is clear that
abductive reasoning only requires few statements to hypothesise and introduce new
frames of understanding the problem. Consequently, the high proportion of deductive
reasoning is explained from the observation that deductive reasoning often come in
series of several statements about the structure of an idea, often following directly
from abductive reasoning statements. This is particularly evident when observing the
higher proportion of ideas initiated by abductive reasoning, as also shown in table 6.
Inductive reasoning was found to occur in more segments than abductive reasoning,
but is often embedded in an idea episode or follows after episode end.
In the following, two examples of idea episodes are presented to illustrate and
further discuss the data. Both examples are translated from Danish to English for the
purpose of reporting.

IDEA REASONING
Speaker Segment IDEA ASPECT TYPE
A Then I am thinking,
A if you could minimise the entire pulley x ABDUCTION
 ABDUCTION
A or then just have a reel or a caster x ABDUCTION
A that you find on the American solutions, x DEDUCTION
A but then you just do a pre… x DEDUCTION
A use a bit more to prepare x DEDUCTION
so you drive it to the window, find the right
A
distance, mount it x DEDUCTION
A and then you just have to lift it 3-4 cm x DEDUCTION
A and then you have the adjustment and lift it again x DEDUCTION
so you minimise the entire phase of pulling and
A
lifting x DEDUCTION
A so you just do it manually x DEDUCTION
It could also be that you used the pulley to drive
B
the wheel, x ABDUCTION
B so you extend it, attach the hook [inaudible] x DEDUCTION
B oh wait no, but, well… x
B it is silly as it is now INDUCTION
B but it could be with the same motor x DEDUCTION
B when it is attached to the cart base x DEDUCTION
B then there is some sort of gearing to the wheel, x DEDUCTION
B same engine drives and pulls… x DEDUCTION
Table 7: Example of idea episode initiated by abductive reasoning

The example presented in table 7 was taken from the protocol of company 1 and
was presented under condition 3, with a focus on cost reduction. The example
illustrates how multiple instances of reasoning take place during the generation and
elaboration of an idea. The idea is initiated by abductive reasoning that states an
intention to achieve a function. Following the abductive reasoning, deductive
reasoning is used to explore and suggest how to achieve the intention of the abductive
reasoning. The deductive reasoning is switched to abductive reasoning again when a
different participant offers an alternative function of the pulley. Following this is
further deductive reasoning and an instance of inductive reasoning in the form of a
negative evaluation of an existing product.

IDEA REASONING
Speaker Segment IDEA ASPECT TYPE
I have written that we can bolt it together instead
C of welding x DEDUCTION
C I know we had one before,
that was split in two and then, pfff, [gesticulates
C assembling the two parts] and so forth. x DEDUCTION
For instance today, the hinges, they are welded
C on, x DEDUCTION
C The ones we have on the B3 and B5. x DEDUCTION
C But the hinges we have on the newer models, x DEDUCTION
C they are actually bolted on. x DEDUCTION
It's an advantage in relation to changes right and
C left x INDUCTION
 C And it gives advantages in relation to adjustment x INDUCTION
D And it will exempt it from that part of assembly. x INDUCTION
Table 8: Example of idea episode initiated by deductive reasoning

The example presented in table 8 was taken from the protocol of company 2 and
was presented under condition 3.The example illustrates how ideas are also initiated
by deductive reasoning. This idea follows a different pattern than the one above
because it first states, deductively, how one fastening principle is better than an
existing one. Then follows more deductive reasoning similar to the previous example,
and finally a number of inductive statements are used to justify the initially proposed
idea by referring to a customer need (the ability to easily switch the direction the door
opens) and an advantage in the assembly of the product (easier adjustment).
Comparing the two examples, it is evident that both use a series of deductive
reasoning to argue for and detail some technical principle. What differs between them
is that the first example starts from stating what the idea should achieve (i.e. a
function) and then goes on to search for ways of solving it. The second example, on
the other hand, starts without mentioning what the idea is supposed to achieve in an
explicit sense, but rather moves on to argue and evaluate the idea function at the end
of the reasoning process.

Reasoning in design activity

The examples show that the first instance of reasoning in an idea episode
functioned to set a frame that either relied on an assumption in the first example, or a
definite statement about preference in the second example. By reviewing multiple
examples as the ones just given, it is argued that the first reasoning of an idea episode
has a determining effect on how the remaining episode progresses. This finding is
comparable to the notions on framing and primary generator in design activity, albeit
at a micro-level of reasoning (Darke, 1979; Valkenburg and Dorst, 1998).
Reasoning was found to occur in sequences of alternating reasoning types. An
implication of this is that empirical indications are found that problem solving can be
said to follow patterns similar to those suggested by models of design and problem
solving (Ahmed and Aurisicchio, 2007; Gero and Kannengiesser, 2004; Hatchuel and
Weil, 2008; Johnson-Laird, 2006; March, 1976).
Taking a different perspective on the ideas developed, the protocols also suggest
that the ideas generated follow different patterns. Some ideas are indeed dominated by
abductive reasoning and go through several abductions to hypothesise how to go
about solving a problem. Oppositely, ideas were also found to be developed strictly by
applying deductive reasoning. This shows different approaches as being similar to
findings by Ball and Ormerod (1995) suggesting that problem solving follows both
breadth-first and depth-first strategies to the development of solutions. The finding
also highlights that reasoning in design is not an orderly process, but proceeds
following varying sequential patterns, as also suggested by Rittel (1987), and C-K
theory (Hatchuel & Weil, 2008).

Reasoning under varying conditions

In this section, results on how reasoning patterns change during the different
conditions of idea generation are presented and discussed.
To reiterate, the data was collected from ideas generated under three different
conditions:
 
Condition 1 was without a particular focus and encouraged the participants to
document existing ideas, and generally to develop ideas without further
instructions
 Condition 2 emphasised to focus on user and customer needs by developing
ideas that sought to fulfil those needs in the best way possible without regard
for cost or other uncertainties
 Condition 3 emphasised focusing ideas generated on reducing the cost-price of
the product while only maintaining a minimum of functionality.
Figure 1 below shows the proportion of reasoning types used to initiate idea
episodes across the three different conditions. A one-way ANOVA analysis was
completed for each condition (p-values reported in figure 1). Confidence intervals
from the ANOVA (p = 0.05) were added and displayed as error bars. Sample size is
displayed above each bar. As expected from hypotheses 1a, there is a trend, however
not significant, in the data that a higher proportion of abductive reasoning is used to
initiate ideas under condition 2, while there is a significantly higher proportion of
ideas initiated using deductive reasoning under condition 3, thus supporting
hypothesis 1b.

30
0,70
29 3
0,60 24
74 22
Proportional distribution

67
0,50 4 20
16

0,40
7
Abductive
0,30 Deductive
14
Inductive
0,20

0,10

0,00
Total (p=0.000) Condition 1 ‐ Condition 2 ‐ Condition 3 ‐
Open (p=0.014) User (p=0.000) Cost (p=0.000)

Reasoning type to initiate idea episode
Figure 1: Comparison of reasoning type under the different conditions

The differences in proportions under condition 2 were found to be close to

significant given the relatively small sample size. Therefore, it is expected that adding
more cases to the study will result in significance. Furthermore, it is observed that
ideas developed under condition 1 are almost exactly equal to the total means, thus
emphasising that conditions 2 and 3 alone account for the changes in reasoning
patterns of the participants, and in different ways.
Investigation of the qualitative difference in reasoning under the different
conditions suggests that under condition 2, where user and customers are in focus,
ideas introduce solutions based on more radical principles and assumptions,
promoting abductive reasoning. Also included are ideas that are known from products
of other types, but not previously applied within the company’s industry. Under
condition 3, where cost reduction is in focus, ideas were often developed by drawing
on already known techniques and approaches to production, manufacturing and
material selection, promoting deductive reasoning. Surprisingly, only few of the ideas
generated suggested to lower the level of functionality in relation to the user as a
means to reduce product cost.

Idea evaluation
Finally, an analysis of the relationship between idea generation condition and idea
evaluation was conducted.
A total of 155 idea episodes, possibly containing several idea aspects, were coded
and 213 ideas were evaluated. This difference was explained by the coding
methodology, where sometimes one idea episode could be documented on several
post-it’s that would later be evaluated individually. This could distort the results if
there are strong correlations between the average episode length and reasoning type,
condition or idea evaluation. However, as table 9 below shows, the differences in
average idea episode length and the ratio of coded ideas to evaluated ideas were not
dramatic, and are not a feature of one single condition.

Condition 1 2 3
Ratio of coded idea 81% 68% 71%
to evaluated idea
Average episode 25,2 segments 23,6 segments 24,5 segments
length
Table 9: Differences between conditions in relation to episode length and ideas

Idea generation condition effects on idea evaluation

The data was analysed to investigate the relationship between idea generation
conditions and idea evaluation. Table 10 presents an overview of how all of the ideas
were evaluated. As can be seen in the table, ideas evaluated to be rejected in the
second matrix and ideas evaluated as maybe relevant only occurred in few numbers
and only under condition 1.

First matrix Second matrix

Idea Maybe Total
evaluation Reject Radical Reject relevant Analyse Accept ideas
Condition 1 3 14 3 1 9 31 61
Condition 2 15 27 0 0 11 30 83
Condition 3 20 7 0 0 11 31 69
Table 10: Overview of evaluation of ideas

To allow for a better overview of the distribution of idea evaluation under each
condition, figure 2 presents the four kinds of evaluation that accounted for 98% of all
evaluated ideas. A one-way ANOVA analysis was completed for each evaluation type
(p-values reported in figure 2). Confidence intervals from the ANOVA (p = 0.05)
were added and displayed as error bars only on ‘Reject’ and ‘Radical’ evaluation
since the proportions of ‘Analyse’ and ‘Accept’ evaluations were not significantly
different as per the ANOVA. This is explained by the almost equal proportions in
those groups. Sample size is displayed above each bar.
 The relation between ideas evaluated as being too radical shows a trend towards
what was expected from hypothesis 2a, stating that idea from condition 2 are more
likely to be evaluated as being highly valuable, but too radical for the current
development project. However, due to low sample size, the difference was not
significant. For hypothesis 2b, it was not found that these ideas developed under
condition 3 had a higher proportion of acceptance.

0,60
3 31
0,50 31
27
20 30
Proportion of ideas

0,40
14
15 7
0,30 Condition 1 ‐ Open

0,20 9 11 11 Condition 2 ‐ User
Condition 3 ‐ Cost
0,10

0,00
Reject Radical Analyse Accept
(p=0.048) (p=0.026) (p=0.876) (p=0.066)
Idea evaluation

Figure 2: Significant differences in how ideas are evaluated across the different
constraints.

From the analysis it was found that there is an opposite relation, however not
significant, when it comes to the proportion of ideas being rejected. Here, ideas from
condition 3 show a trend of having a higher proportion of ideas rejected. This can be
explained as a consequence of the lower level of originality displayed by ideas under
condition 3, which result in ideas being rejected. There is a very low proportion of
ideas from condition 1 rejected, and this is attributed to the earlier described
interpretation that this condition was dominated by existing ideas, which are more
likely to be accepted as they are already known to the participants.
Despite the different reasons for ideas not accepted, the overall proportion of ideas
from the different conditions being accepted are not significantly different in
proportion, and very equal in amount of accepted ideas (see figure 2). Instead, the
analysis unveils that ideas are rejected for different reasons.
Bringing these results together with the previous analyses highlights that the focus
upon customer and user promotes abductive reasoning, which in turn leads to more
valuable but too radical ideas. In contrast, is shown that when focus is on cost
reduction deductive reasoning is promoted, which in turn leads to more ideas being
rejected because they do not add value to the project. A similar result was found by
Dong et al (2012) where deductive reasoning lead to higher rejection rate in the
evaluation of ideas.

DISCUSSION AND CONCLUSIONS

A study of design activity, as part of real life product development processes in
two companies, was conducted. Reasoning patterns of idea generation processes
observed under three conditions for generating ideas and the evaluation of the
generated ideas were investigated.
Reasoning patterns involved in the generation of new ideas were identified to
utilise abductive, deductive and inductive reasoning. Additionally, reasoning was
shown to occur in sequences alternating between the different reasoning types to form
arguments for an idea in generation. Hypothesis 1a stating that a higher proportion of
ideas developed under less constrained conditions was not significantly supported, but
showed a trend in the data that would likely find support for the hypothesis given a
larger sample size. Hypothesis 1b stating that a higher proportion of ideas developed
under more constrained conditions was significantly supported. Hypothesis 2a stating
that ideas developed under less constrained condition were more likely to be
evaluated to be too radical for the on-going development project showed a trend in the
data, but was not significant. Hypothesis 2b stating that ideas developed under more
constrained conditions were more likely to be evaluated to directly applicable to the
on-going development project was not supported. However, it was found that a higher
proportion of these ideas were evaluated as not being valuable to the project.
The study reported here is one of the first studies of the kind for two primary
reasons. First, it is shown empirically how reasoning in design follows different
patterns in the form of sequences between alternating types of reasoning. Second, a
relationship between reasoning patterns and the evaluated value of developed ideas is
suggested to be mediated by the constraints and conditions under which the ideas are
generated.
Three of the four hypotheses did not show significant differences. This is attributed
to the low sample size, and therefore highlights the importance of conducting more
cases, which is a first priority in future research.

Implications and further research

From a managerial perspective, understanding how reasoning and constraints
influence design activity support the management of creative processes in
organisations. For example, further development of the study could be the underlying
basis for:
 The application of creative methods to increase the output of idea generation
sessions in industry
 The recruitment and assessment of employees by testing or training reasoning
skills
In relation to design education, the findings of the study could be used in the teaching
of students to improve reasoning skills to improve design competences.
From a research perspective, an understanding of the relationship between
reasoning, idea generation conditions and evaluation serves as a foundation for
understanding design activity and human behaviour in design. Directions for future
research could include:
 Adding more cases to the analysis to further investigate the hypotheses of the
present study (this is in progress)
 Understanding the direct relationship between reasoning patterns and idea
evaluation by analysing reasoning pattern and idea on a one-to-one basis
 Understanding how to increase number of generated ideas that are accepted
and valuable to a development project.
REFERENCES
Ahmed, S., and Aurisicchio, M. (2007). Understanding Engineering Design Activity
through Questioning. In Connected 2007 International Conference on Design
Education (pp. 1–7).
Ahmed, S., Wallace, K., and Blessing, L. (2003). Understanding the differences between
how novice and experienced designers approach design tasks. Research in
Engineering Design, 14, 1–11.
Badke-schaub, P., Lauche, K., Neumann, A., and Ahmed, S. (2007). Task – Team –
Process: Assessment and Analysis of the Development of Shared Representations in
an Engineering Team. In DTRS ’07 (pp. 1–12).
Badke-Schaub, P., Neumann, A., Lauche, K., and Mohammed, S. (2007). Mental models
in design teams: a valid approach to performance in design collaboration? CoDesign,
3(1), 5–20.
Ball, L., and Ormerod, T. (1995). Structured and opportunistic processing in design: A
critical discussion. International Journal of Human-Computer Studies, (43), 131–151.
Chi, M. T. H. (1997). Quantifying Qualitative Analyses of Verbal Data : A Practical
Guide, 6(3), 271–315.
Christensen, B. T., and Schunn, C. D. (2009). The Role and Impact of Mental Simulation
in Design. Applied Cognitive Psychology, 344(May 2008), 327–344.
Christensen, B. T. (2009). A methodology for studying design cognition in the real world.
In Nordic Design Research Conference.
Cross, N. (2001). Design Cognition: Results From Protocol And Other Empirical Studies
Of Design Activity. In Design Knowing and Learning: Cognition in Design Education
(pp. 79–103). Oxford: Elsevier.
Darke, J. (1979). The primary generator and the design process. Design Studies 1(1), 36–
44.
Dong, A., Mounarath, R., and Lovallo, D. (2012). THE LANGUAGE OF ABDUCTION
IN CHOOSING INNOVATION. Proceedings of DCC'14
Dorst, K. (1995). Analysing design activity: new directions in protocol analysis. Design
Studies, 16(2), 139–142.
Dorst, K. (2011). The core of “design thinking” and its application. Design Studies, 32(6),
521–532.
Dorst, K., and Cross, N. (2001). Creativity in the design process: co-evolution of
problem–solution. Design Studies, 22(5), 425-437
Ericsson, K. A., and Simon, H. A. (1993). Protocol Analysis (Revised Ed., pp. 1–443).
Cambridge, MA: MIT Press.
Fann, K. T. (1970). PEIRCE ’ S THEORY OF ABDUCTION (pp. 1-). The Hague:
Martinus Nijhoff.
Gero, J. S., and Kannengiesser, U. (2004). The situated function–behaviour–structure
framework. Design Studies, 25(4), 373–391.
Goldschmidt, G. (1991). The dialects of sketching. Creativity Research Journal, 4, 123–
143.
Goldschmidt, G., and Tatsa, D. (2005). How good are good ideas? Correlates of design
creativity. Design Studies, 26(6), 593–611.
Habermas, J. (1972). Knowledge and Human Interests (pp. 1–356). Franfurt am Main:
Suhrkamp Verlag.
Hatchuel, A., and Weil, B. (2008). C-K design theory: an advanced formulation. Research
in Engineering Design, 19(4), 181–192.
Hogan, K., Nastasi, B., and Pressley, M. (1999). Discourse patterns and collaborative
scientific reasoning in peer and teacher-guided discussions. Cognition and Instruction,
17(4).
Johnson-Laird, P. N. (2006). How We Reason (pp. 1–573). Oxford University Press.
Keshwani, S., Lenau, T. A., Ahmed-Kristensen, S., and Chakrabarti, A. (2013).
Benchmarking Bio-Inspired Designs With Brainstorming in Terms of Novelty of
Design Outcomes. ICED13: 19th International Conference on Engineering Design
(pp. 1–10).
Kroll, E., and Koskela, L. (2014). On Abduction in Design. In Preprints of Design
Computing and Cognition ’14 (pp. 357–376).
Linsey, J. S., Clauss, E. F., Kurtoglu, T., Murphy, J. T., Wood, K. L., and Markman, a. B.
(2011). An Experimental Study of Group Idea Generation Techniques: Understanding
the Roles of Idea Representation and Viewing Methods. Journal of Mechanical
Design, 133(3), 031008.
Lloyd, P., and Busby, J. (2001). Softening up the facts: engineers in design meetings.
Design Issues, 17(3).
Lloyd, P., and Scott, P. (1994). Discovering the design problem. Design Studies, 16, 125–
140.
Magnani, L. (1995). CREATIVE PROCESSES IN SCIENTIFIC DISCOVERY.
European Journal for High Ability, 6, 160–169.
March, L. (1976). The logic of design. In L. March (Ed.), The Architecture of Form (pp.
16–40). Cambridge University Press.
Merriam-Webster. (n.d.). Merriam Webster. Retrieved March 25, 2015.
Oxford. (n.d.). Oxford. Retrieved March 25, 2015.
Polk, T. a., and Newell, A. (1995). Deduction as verbal reasoning. Psychological Review,
102(3), 533–566.
Reichertz, J. (2010). Abduction: The Logic of Discovery of Grounded Theory. Forum:
Qualitative Social Research, 11(1).
Rittel, H. W. J. (1987). The reasoning of designers. In International Congress on
Planning and Design Theory (pp. 1–9). Boston.
Roozenburg, N. F. M. (1993). On the pattern of reasoning in innovative design. Design
Studies, 14(1), 4–18.
Rouse, W. B., and Morris, N. M. (1986). On looking into the black box: Prospects and
limits in the search for mental models. Psychological Bulletin, 100(3), 349–363.
Runco, M., and Jaeger, G. (2012). The standard definition of creativity. Creativity
Research Journal, 24(1), 92–96.
Schurz, G. (2007). Patterns of abduction. Synthese, 164(2), 201–234.
Schön, D. A. (1959). The Reflective Practitioner (pp. 1–384). Basic Books.
Searle, J. R. (1979). Expression and Meaning: Studies in the Theory of Speech Acts.
Essay Collection (Vol. 49, p. 187).
Valkenburg, R., and Dorst, K. (1998). The reflective practice of design teams. Design
Studies, 19(3), 249–271.
Voss, J. F. (2006). Toulmin’s Model and the Solving of Ill-Structured Problems.
Argumentation, 19(3), 321–329.
Ward, T. B., and Kolomyts, Y. (2010). Cognition and Creativity. In J. C. Kaufman and R.
J. Sternberg (Eds.), Cambridge Handbook of Creativity (pp. 93–112). Cambridge:
Cambridge University Press.