The effect of abductive reasoning on concept

selection decisions
Andy Dong, Faculty of Engineering and Information Technologies,
University of Sydney, Sydney, NSW, 2006, Australia
Dan Lovallo and Ronny Mounarath, The University of Sydney Business
School, University of Sydney, Sydney, NSW, 2006, Australia

This study examines the effect of the form of logical reasoning on concept
selection decisions. An experiment was conducted with members of a committee
called upon to decide whether or not to invest in new product concepts. Under an
abductive reasoning frame manipulation, which induced an inclination toward a
form of logical reasoning that introduces hypotheses to explain given
observations, individuals were more likely to accept concepts whereas under a
deductive reasoning manipulation they were more likely to reject concepts. We
recommend that when committees aim to increase the likelihood of design
concepts being accepted, decision makers should employ innovative abduction to
think creatively about new ways to frame the proposed concepts and to explore
new working principles underpinning them.
Ó 2015 Elsevier Ltd. All rights reserved.

Keywords: concept selection, design thinking, behavioural economics, innova-

tion, creativity

D
esign concept selection is a fundamental aspect of the design pro-
cess, especially in new product development. Long before a new
product hits the market, decision-makers were trying to decide
whether to further develop, consolidate, or abandon one or more product
concepts. Individuals engaged in decision-making applications of this type
are making strategic choices about potentially attractive design concepts
such that after the decision is taken, they or their company would devote
considerable and irrevocable time, money or both. This problem has become
more widespread as an increasing number of companies implement an ‘Inno-
vation Time Off’ policy to elicit new product concepts from employees or
implement an open innovation policy of ‘crowdsourcing’ product or service
ideas (Bayus, 2013). In other words, companies are facing a large stream of
ideas, and are struggling with the filtering and selection process. Ideally,
they need to balance the number of projects that make it through the early
selection stages without taking on poor projects but equally without filtering
Corresponding author:
Andy Dong out projects that may turn out to be valuable innovations. In companies
andy.dong@sydney. failing to innovate due to high levels of risk aversion, they may simply
edu.au need selection processes that get more projects through the early stages of
www.elsevier.com/locate/destud
0142-694X Design Studies 37 (2015) 37e58
http://dx.doi.org/10.1016/j.destud.2014.12.004 37
Ó 2015 Elsevier Ltd. All rights reserved.
 filtering. This paper takes on the question of the effect of the form of logical
reasoning on design concept selection.

A concept selection process (Ulrich & Eppinger, 2004) or an Idea Screen

(Cooper, Edgett, & Kleinschmidt, 2002) takes place toward the beginning of
a product development process. During the selection process, decision makers
perform design evaluation, by which we mean the determination of the quality
(value or worth) of a design concept against established objectives as a func-
tion of one or more its attributes (Thurston, 1991). Current design evaluation
methods call for deductive logic, which is a form of logical reasoning that aims
to guarantee the truth of the conclusion if the premise of the argument is
observed to be true. In other words, the evaluation methods aim to prove or
disprove the merits of the concept. Empirical evidence in industry for this
type of decision context describes decision-makers as tending to apply vari-
ables amenable to deductive analysis including product timing, staffing, and
platform when evaluating innovative projects (Krishnan & Ulrich, 2001; van
Riel, Semeijn, Hammedi, & Henseler, 2011). Product concept evaluation tech-
niques as a consequence employ highly deductive analysis requiring a substan-
tial amount of information aiming to prove or disprove premises established
by precedence (Udell, 1989). Empirical metrics applied in assessing creativity
in design projects even at an early stage similarly require deductive analysis,
such as in quantifying novelty by comparing an idea to a universe of ideas
(Maher, 2010; Shah, Smith, & Vargas-Hernandez, 2003). In logical notation,
in which p is a premise and q is a conclusion, the deductive logic applied by a
decision maker in quantifying novelty could be:
p/q IF positive novelty evaluation metric (p) / THEN creative (q)
p Design concept has a novelty metric greater than zero (p).
q The design concept is creative (q).

Similarly, the evaluation techniques can call for inductive reasoning, in which
the premises provide a degree of support for the conclusion. Based upon an
observation made about the proposed design concept, a decision maker estab-
lishes a hypothesis to explain the observation. The hypothesis is a generaliza-
tion of the observation; it does not introduce any new information explaining
the particular observation. Induction can only establish truth relative to the
current observations. As more evidence in support of the premises become
available, the degree of strength of the conclusion increases. With inductive
reasoning, the goal is to accumulate evidence to support or refute the hypoth-
esis. An inductive reasoning example could be:
p1 / q1 Product 1 employs voice-based input and is inadequate.
p2 / q2 Product 2 employs voice-based input and is inadequate.
pn / qn Product n employs voice-based input and is inadequate.
p/q IF voice-based input (p) / THEN inadequate (q)

38 Design Studies Vol 37 No. C March 2015

 The downside is that deductive and inductive reasoning can potentially
lead to pattern recognition biases and decision myopia (Lovallo & Sibony,
2010).

In contrast to deductive and inductive reasoning, design researchers generally

promote abductive logic as the form of logical reasoning that is the lifeblood of
creative design (Dorst, 2011; March, 1976; Roozenburg, 1993; Zeng & Cheng,
1991). Charles Peirce’s seminal thoughts about abduction state that it is the
only logical operation that introduces new ideas (Peirce, 1932). In classical
logical reasoning, abduction proposes the most parsimonious explanation
for observations. In design, abductive reasoning is implicated in at least two
important situations: in synthesizing complex and contradictory information
to generate insight (Kolko, 2010) and in reasoning toward new solutions for
design problems, from function to form (Kroll & Koskela, 2014; March,
1976; Roozenburg, 1993; Zeng & Cheng, 1991). The latter form of abduction
has been referred to as innovative abduction whereas the former has been
labelled explanatory abduction (Roozenburg, 1993). In innovative abduction,
scholars have theorized that designers invent inter alia a concept, a set of so-
lution principles, and a form that ‘explains’ and logically connects a form to
the desired function; however, the demarcation between inferences, the num-
ber of inferences from function to form, and whether the inferences must pro-
ceed from function remain to be debated. Abductive reasoning is nonetheless a
cognitive strategy that has been argued as being the kernel of design thinking
(Roozenburg, 1993), that is, one of the ‘quite specific and deliberate ways of
reasoning’ (Dorst, 2011, p. 531) associated with the practice of design. Abduc-
tive reasoning is certainly not unique to design; in scientific discovery, scholars
contrast selective (explanatory) and creative (innovative) abduction, with the
latter being central to the growth of scientific knowledge given its emphasis
on generating new plausible hypotheses that can be tested (Magnani, 1995).
Simply stated, abductive reasoning introduces hypotheses and theories to
explain given facts.

We propose that design concept selection should entail two aspects. First, it
should entail evaluating the merits of a design concept as it stands, which re-
quires deductive analysis. Further, it should be intrinsically forward-looking,
requiring a long-term perspective of ‘what might be’ rather than ‘what is’.
Since market conditions could change during the development cycle,
decision-makers may wish to identify alternative opportunities enabled by
the proposed design concept. At different decision stages in the product devel-
opment cycle, they may wish to emphasize one form of logical reasoning over
another, such as more innovative abduction at the start and more deductive
logic toward the end. Concept selection transcends merely selecting from
clearly defined options. In short, design concept selection should not only be
about the evaluation of the design concept as it is; it should also be about infer-
ring what it could be, which requires innovative abductive reasoning as

The effect of abductive reasoning 39

 described above. However, if designers apply ‘too much’ deductive reasoning,
the form of logical reasoning suggested by the evaluation metrics described
previously, could they inadvertently eliminate potentially fruitful concepts
as an unintended consequence?

If innovative abduction, the form of abductive reasoning that we consider in

this paper, is central to introducing new ideas, can deductive reasoning during
design concept selection stymie potentially innovative design concepts from
ever ‘getting off the ground’?

This article explores how the form of logical reasoning interacts with design
concept selection. It is hypothesized that the direction of project acceptance
will shift in a predictable direction according to whether an abductive
reasoning frame or a deductive reasoning frame was taken by a decision-
maker. We define a reasoning frame as a type of framing effect in which there
is a belief set or attitude that places more emphasis on one form of logical
reasoning to shape the manner in which decision makers formulate the deci-
sion task, make judgments, and construct rationale to support their decision.
All forms of logical reasoning can appear in a decision-making context in
various proportions. This research investigates the influence of the reasoning
frame on the direction of the decision. Because both deductive and inductive
reasoning seek to eliminate (deductive) or reduce (inductive) uncertainty and
neither introduces new knowledge, in this study we treat both of these forms
of logical reasoning as if they were a single type of reasoning frame (explained
below). Priors were tested using a controlled laboratory experiment with ma-
nipulations to the reasoning frame. Further, we test the effect in the context of
committee-based decision-making since it accurately reflects the commercial
realities of internal innovation management (or R&D investment). We em-
ployed a multi-methods approach to analyse the experimental data. First, to
understand the reasoning frame in committees, we performed a content anal-
ysis of the committee decision-making discussion. Then, we performed statis-
tical analyses to understand the influence of the reasoning frame on individual
decisions to accept or reject a project. This analysis provides an understanding
of how the form of logical reasoning can produce variance and the potential
effect of inducing an abductive or deductive/inductive reasoning frame on
committee decision-making depending upon whether an organization wishes
to accept or reject more projects, respectively.

1 Experimental design:
An experimental and quantitative methodology was chosen for this investiga-
tion as it provided the opportunity to create the right conditions to test the hy-
pothesis of the research question posited. The experiment was a small group
decision-making exercise where a committee of 5 participants were asked to
decide which projects from a pool of 7 were worthy of investment. The exper-
imental design is a between-group design with each group receiving a

40 Design Studies Vol 37 No. C March 2015

 manipulation to their reasoning frame (RF): either a deductive/inductive or
abductive frame. We hypothesize that in the context of innovation-oriented
committee decision-making for design concept selection:

H1: The reasoning frame influences the direction of project acceptance.

The null hypothesis is that the form of logical reasoning has no influence on
the direction of project acceptance.

1.1 Participants:
Participants were recruited using the Online Recruitment System for Eco-
nomic Experiments (ORSEE) developed by experimental economists at the
University of Sydney Business School. The system sends out invitations to
all registered students (undergraduate and postgraduate) across the university
and from all disciplines. The Faculty of Economics and Business Ethics Com-
mittee approved selection, recruitment, experiment design, and honoraria.
Recruitment took advantage of ORSEE’s random selection process and did
not select participants on the basis of their year, discipline or experience to
minimize systematic bias. The experiments were conducted over 6 experi-
mental sessions lasting about 1 h each plus 3 pilot studies. Committees were
randomly allocated 5 individuals with experimenters checking that no one
knew each other to remove as much selection and systematic bias as possible.
Conditions were alternated between successive experiments to try and achieve
a balance of subject variables between groups and conditions. In total, 6
groups consisting of 5 persons per group were recruited. Three groups were
allocated per condition. Since each group evaluated 7 projects, a total of
210 individual decisions (5 persons
3 groups
7 projects
2
conditions ¼ 210) were obtained, which is a sufficiently large sample size for
the statistical analysis.

1.2 Procedure
First, we outline the experiment. Then, we elaborate each of the following
steps:

1. The participants randomly select a role. Two are founding directors and
the other three are directors. All roles have the same voting rights.
2. The experiment facilitator explains the experiment (Section 1.3) and the
payout structure (Section 1.4).
3. The participants complete the payout comprehension test. If any partic-
ipants do not pass, the experiment facilitator provides verbal instructions
and then a verbal test of comprehension.
4. The participants are given the 7 project briefs. They are given 20 min to
read all of the project briefs.
5. The founding directors read the statements written on their role descrip-
tion sheet. This statement serves as the reasoning frame manipulation.

The effect of abductive reasoning 41

 6. Group discussions begin. Projects are discussed for exactly five minutes.
Even if participants stop talking, they are not allowed to move onto the
discussion of the next project until the five minutes is completed.
7. The experiment facilitator distributes assessment sheets and the partici-
pants complete them with no further discussion permitted. Participants
are not allowed to share their votes.
8. The participants proceed onto the next project and repeat from Step 6 un-
til all projects have been discussed.
9. The experiment facilitator tallies the votes in favour of each project
through a show of hands.
10. The experiment facilitator collects the assessment sheets.
11. The experiment ends.

1.3 Instructions to the participants

Upon commencement, and after the participants have randomly selected their
role in the experiment, the experiment facilitator explains the experiment. The
participants were informed that they were directors on an investment commit-
tee with equal standing and voting power. Participants were told that they
work for a non-profit venture capital firm that invests in new business ventures
and new innovative technologies. An innovative project is one that would be
adopted by the market. They were instructed that: ‘the firm is implementing
an open innovation strategy, which is directed towards improving people’s lives
by investing in the development of innovative mobile applications.” The experi-
menter then proceeded to explain the aims of the meeting as follows:

 You are at this meeting today to help the firm decide which projects the firm
should invest in for further product development or refinement.
 The objective is to make the correct decisions and invest in good projects and
screen out bad projects.
 The projects can be chosen based either on profit potential or on whether they
will improve people’s lives. The goal of the non-profit venture capital firm is to
do both or either one.
 It is not expected that projects will be chosen based solely on their profit po-
tential. This is because the projects are in their early stages of development
and it is difficult to forecast the financial performance of these projects.
 The committee may accept all projects; OR
 They may reject all projects; OR
 They may accept and reject projects in varying proportions.

1.4 Payoff structure

The experiment facilitator then explains the payoff structure. An incentive
compatible payoff mechanism was developed so that participants would be
motivated to vote honestly and take the experiment seriously, that is, to iden-
tify design concepts that had innovation potential. The committee was advised
that experts had previously chosen some, none, or all design concepts as

42 Design Studies Vol 37 No. C March 2015

 worthy of resourcing. If the committee chose as the experts did, they received a
payoff bonus of AUD15, making their total payoff twice the base payoff. This
inducement addresses two key precepts of microeconomic experiments
without introducing new criteria (Smith, 1982). First, it increases the saliency
by increasing the good outcome (every individual decision made had the po-
tential to earn an AUD30 payout if the group chose correctly) while decreasing
the bad outcomes of the experiment (all participants walked away with a base
rate amount of AUD15). Second, it mitigates nonsatiation because partici-
pants are motivated to earn additional payout. A payoff test was included
to mitigate random error stemming from the possibility that some participants
would not comprehend the payoff mechanism. All 30 participants passed the
test. This was important to indoctrinate incentive compatibility and maintain
the internal validity of the data set collected. We note that there was no evi-
dence in the committees’ discussions of their trying to guess what the experts
had chosen.

1.5 The project briefs

Participants were provided a set of design concept briefs in randomized order.
Participants were instructed to read all project briefs individually before com-
mittee deliberation began. It was crucial that participants read all projects first
so that the first project they read was not used as a base line measure. This
would have introduced a systematic anchoring bias (Tversky & Kahneman,
1974).

Decisions can be affected by the manner in which choices are presented, which
is known as the framing effect (Tversky & Kahneman, 1981). We minimized
this framing effect by having a standard template for the presentation of the
projects consisting of approximately one page of text describing ‘what’ the
concept is and ‘how’ it works followed by a visual scenario demonstrating
its use through line drawings of similar quality and completeness as practi-
cable. For consistency, the authors wrote the briefs based upon content pro-
vided by the project’s designer. Excerpts from design concept briefs are
presented in Figure 1.

The 7 projects were a collection of submissions from students enrolled in a

final-year capstone design studio in the Bachelor of Design Computing in
2010 at the University of Sydney who elected to participate in this study.
The projects were:

1. A daily medication box that reminds patients to take medication by SMS

2. A mobile phone application that assists the visually impaired to navigate
using Google Maps
3. A child’s necklace that helps parents to track where their child is and with
whom

The effect of abductive reasoning 43

Figure 1 Example of presentation of design concepts to committee for (a) a wearable navigation jacket and (b) a device that activates appli-
ances using gestures.

4. A beer holder that monitors alcoholic consumption rate to avoid (or

detect) inebriation
5. A mobile phone application to assist in tracking urban re-vegetation
6. A jacket with sewn-in electromechanical navigation aids using data pro-
vided by Google Maps
7. A device that activates appliances using gestures and wireless
communication

These 7 projects were selected from an initial set of 15 as the most fully devel-
oped and well-communicated projects with a level of technical complexity that
the committees could understand. Criteria for selection included a coherent
design concept, a reasonably developed form, a sufficiently elaborated mode
of use, a defined end-user, and sufficient level of technical development to sug-
gest feasibility. They were also selected for being at similar levels of academic
quality, that is, no clear outliers, so that all would appear attractive.1 Projects
were aimed at new mobile phone based products with the motivation of
designing an innovation that does not yet exist or has not been mass-
marketed. The stimulus materials are geared towards innovative and novel
design concepts that are not currently available on the market at the time of
the experiment.

44 Design Studies Vol 37 No. C March 2015

 1.6 Decision-making
To obtain individual decisions on project acceptance, each individual was
given assessment sheets. The assessment sheet consisted of five-level Likert
scales and a sixth question for the individual’s accept/reject decision. The ques-
tions were as follows:

1. I think this project is novel

2. I think this project is creative
3. I think consumers will be accepting of this product
4. I think this project has market potential
5. I think this project is technically feasible
6. I think this project should be accepted

For questions 1 to 5, a score was allocated to each response. The scores for
each response were:

 Strongly disagree ¼ 1
 Disagree ¼ 2
 Neutral ¼ 3
 Agree ¼ 4
 Strongly agree ¼ 5

The response to Question 6 is binary, yes (accept project) or no (reject project).

Only individual decisions to support or reject a project will form the basis of
the statistical analysis since our hypothesis is that the form of logical reasoning
influences project acceptance.

1.7 Reasoning frame manipulation

Two of the 5 participants were randomly assigned the role of founding direc-
tor, which entailed one extra responsibility, to address the committee at the
start of the experiment. The founding directors, unknowingly, were given
the special role of indoctrinating the abductive or deductive/inductive
reasoning frame. This framing manipulation was implicit as one of the found-
ing directors was instructed to read out the address to the board at the
commencement of discussion.

For the deductive/inductive reasoning manipulation, one of the directors read

the following statement:

The objective of today’s meeting is to capture and invest in new product

innovation. For the purposes of this meeting we determine whether each
project matches people’s needs with what is technically feasible and what
a viable business strategy can convert into market opportunity and/or
customer value.

The effect of abductive reasoning 45

 The statement was written to induce a strong inclination toward evaluating pro-
jects as they were described in the briefs with participants anchored on criteria
amenable to deductive or inductive logic to answer. As an example of deductive
logic, a premise would be that projects that are not technically feasible should not
be supported. Then, the committee would analyse whether a project has sufficient
technical feasibility, and, if not, they would deductively conclude that it should
not be supported. The use of criteria in-text increases the salience of the framing
and enhances the probability that receivers will perceive the information, discern
meaning and thus process it and store it in memory (Fiske & Taylor, 1991). We
considered the use of scoring and ranking methods for the analysis of innovative
projects (e.g. Cooper, 1985). The trade-off would have been that the participant
population would need a technical and financial background to understand and
assess many of those dimensions accurately, because these instruments include
assessments on product family, platform, strategic fit, and expected financial per-
formance. Such expertise is beyond the capability of the recruited participants.
Finally, the use of these instruments requires more information on projects
than could feasibly be provided and would likely have only increased the number
of rejected projects in the deductive/inductive reasoning frame. Furthermore,
this manipulation was gentler than it might have been, and so, if the hypothesis
holds under this construction, it would likely be obtained in a more heavy-
handed manipulation. In other words, this is a strong test.

For the abductive reasoning manipulation, one of the directors read the
following statement:

The objective of today’s meeting is to capture and invest in new product

innovation. For the purposes of this meeting, we will determine a possible
future in 2e3 year’s time wherein further development of a project will
lead to something new that becomes adopted and leads to a sustained
change in behaviour or behavioural patterns.

This statement was written to elicit an abductive reasoning frame by virtue of

the forward-looking nature of the statement. The aim was to direct the com-
mittee’s discussion toward the invention of both a plausible reframing of the
design concept and the conception of a new set of principles underpinning the
new design concept as if the committee were ipso facto introducing a new
design concept (‘something new’) which they can convert into market oppor-
tunity and customer value through ‘further development’ in ‘a possible
future’. This is thus innovative abduction (Roozenburg, 1993) applied in
the phase of design concept selection.

2 Content analysis of committee deliberations

2.1 Coding development and scheme
To understand the form of logical reasoning on the committees’ deliberations,
and to ensure that our manipulation worked, we performed a content analysis

46 Design Studies Vol 37 No. C March 2015

 of the transcripts of discussions from each of the committees in the experi-
ments. Because the committees’ discussions were limited to 5 min, almost all
of the content was relevant; there was very little idle banter. To confirm that
our manipulations worked, we would expect to find a higher proportion of ab-
ductive reasoning under the abductive framing manipulation than under the
deductive/inductive framing manipulation and vice-versa. Note also that we
do not expect to eliminate either form of logical reasoning; rather, we predict
to see them in varying proportions depending on the manipulation.

To enable the identification of these forms of logical reasoning in natural lan-

guage dialog, we postulate their forms of linguistic realization based on the
formal logic principles underlying them, a process similar to that applied by
Galle (1996) to understand how designers justify their decisions using formal
logic. We start with the most straightforward one, deduction.

Deductive logic is guaranteed to be true or false; the direction of the decision

must likewise be clear, either accept or reject. Deductive reasoning is coded
when a committee member starts from an established rule, such as a decision
criterion, and makes observations about one or more characteristics of the
proposed project as satisfying the rule in order to draw a conclusion. The
following excerpt illustrates a simple example of deductive reasoning with
explicit rejection of the project. In the excerpt, the committee member tacitly
refers to the rule that the concept should be technically feasible to be accepted.
The committee member makes the observation that the concept is not techni-
cally feasible. Therefore, the committee member rejects the project.
It doesn’t seem like it’s technically feasible . I didn’t like the second project.

This excerpt could be stated in logical notation as:

p/q IF :[technically feasible] / THEN :[accept]
p It doesn’t seem like it’s technically feasible :[technically feasible]
q I didn’t like :[accept] the second project.

Although the statistical analysis will show that deductive reasoning decreases
the acceptance of projects, deductive reasoning was also used to support a
project, as shown in the following excerpt:

I think because it’s being built for the government, there would be acceptance,
because I don’t think there’s really anything out there that’s affordable to be
taking the photos.

Transforming this excerpt into logical notation:

p/q IF [market potential] / THEN [accept]
p I think because it’s being built for the government [the government is
a consumer]
q there would be acceptance

The effect of abductive reasoning 47

 In deductive reasoning, it is crucial that committee members draw upon exist-
ing knowledge in their observations, rather than inventing new knowledge,
which would be coded as abduction. The existing knowledge could either be
stated in the project brief or through reference to widely known knowledge
or principles. For example, in the following excerpt, the committee member re-
jects the device to turn on appliances using gestures due to its similarity to ex-
isting products for remotely controlling appliances, based on the rule (criteria)
of novelty for project acceptance:

Plus it’s something that also kind of exists already. You know how you can
call home to open each, like turning on the TV recorder so you can see what’s
happening in your house, like home security, from the phone. Do you know
what I mean? Like you can access your house already from your phone.

It is important to note that we were not simply coding product appraisals

(Dong, Kleinsmann, & Valkenburg, 2009), that is, positive or negative senti-
ments about a project. We are coding instances of logical reasoning. We were
careful not to code discussions that contained an analysis of a project without
the committee member according sufficient priority to the analysis as the basis
of a decision. For example, in discussing a device to help parents track their
children, a committee member states, ‘You don’t just know where your college
kid is. You know where someone else’s kid is too. It gets a bit too e I have some
privacy issues with it.’ While there is a negative tone, it is not clear how the
analysis contributes to a decision because there is no reference, implied or
explicit, to selection criteria, and the personal judgment about privacy is
ambivalent, suggested by the qualifier ‘some’. Thus it was not coded as a
form of logical reasoning that clearly relates to a decision.

Induction involves the establishment of a general principle based on a small set

of observations. We coded instances of inductive logic when a committee
member draws a hypothesis based upon a generalization of an observation
made about a design concept. The following excerpt describes an instance of
inductive reasoning:

We’re so lazy, that anything that saves us walking up to switch the light
switch on and off is everyone’sd

Using the notation proposed by Dorst (2011) for inductive reasoning, the
‘what’ is the device to turn on and off appliances using gestures. The desired
value, as interpreted by the committee member, is saving time (i.e., being
‘lazy’). The working principle drawn from a generalization of the operation
of this device is ‘anything that saves us walking up . is everyone’s [desired
product]’. In other words, the device plus the principle that things that save
physical effort lead to the desired value, being lazy.

The process of (innovative) abductive reasoning in design emphasizes the inven-

tion of a new design concept and a new principle that would explain the

48 Design Studies Vol 37 No. C March 2015

 operation of the design concept. Dorst (2011) proposes that abduction in design
consists of creating new frames for a new ‘something’ that addresses the design
problem (the desired value, generally the function, which is the only observed
known) and a new ‘how’ or a new ‘working principle’ to account for the new
‘something’. The frames provide ways of ‘seeing’ to establish the parameters of
the design problem and its solution, or both, and set up a rationale for courses of
action undertaken. Likewise, Roozenburg (1993), citing Habermas, explains that
such abduction in design is best described as innovative abduction as opposed to
explanatory abduction, because innovative abduction entails a new, unexplained
fact (e.g., the proposed project) for which a rule is produced to explain the fact
(e.g., why the project proposed would exist). Roozenburg (1993) concluded
that innovative abduction is the only appropriate form of abductive reasoning
in design, because design entails determining the set of conditions for which
the conceptualization of the product would be true. Kroll and Koskela (2014)
proposed an extension of these models of abductive reasoning in design. Their
model consists of two-steps of reasoning, one from function to an idea, concept,
or solution principle and then from that principle to the form or, more generally,
‘how’ the principle can be realized. The idea behind this two-step model is that
innovative abduction should start with a known fact and then infer a possible
working principle, which then becomes the known fact for a second innovative
abduction, which connects the form to the working principle.

Adapting these theories about abduction to the empirical analysis of abduc-

tion in natural language, we code abduction as verbalizations that introduce
new hypotheses that are plausible explanations for the existence of the design
concept as presented in the brief. When participants propose new rules to
explain the existence of the proposed concept and the proposed function, ab-
ductive reasoning is involved. These inferences capture what Roozenburg
(1993) calls inference to a description of the form of the artefact and Dorst
(2011) labels as the ‘what’, and inference to ‘a prescription of its actuation’
(Roozenburg, 1993) or what Dorst (2011) calls the ‘how’. However, we do
not restrict our analysis of innovative abduction as being limited to seeking
rules to connect the function to form and working principle, that is, design
per se. That is, we are not coding for ‘designerly’ abductive reasoning
(Dorst, 2011; Roozenburg, 1993) alone.

Our analysis of abduction therefore follows the general form of innovative

abduction proposed by Roozenburg (1993):
q a given fact, the proposed design concept or proposed function: q
p/q a rule to be inferred first: IF p THEN q
p the conclusion: p

The possible ‘rule to be inferred first’ will be shown in Table 1. Each instance of
p 0 q is counted as an instance of abductive reasoning. Each inference may

The effect of abductive reasoning 49

Table 1 Coding scheme for reasoning frame

Reasoning frame Criterion Example

Deductive Drawing a conclusion based on implicit
or explicit premise but observation
explicit
Deductive Stating the premise and/or observation
for a deductive conclusion in relation to
established decision criteria
Deductive Personal judgment on the value of the
project if decision maker accords the
judgment sufficient priority in
determining acceptance or rejection
Inductive Generalization based on specific
instance
Abductive Reframing the product as a different
kind of product from what is actually
proposed
Abductive Modifying structural or behavioural
aspects of the product
Abductive Reframing users/users’ needs in a
different way than as proposed in the
project brief
Abductive Inventing conditions (causal
precedents) for future (im)possibility of
the project
Abductive Inventing or simulating alternative
contexts of use
Because GPS do exist today and we have seen this
sort of stuff existing already, so it’s not a
completely new idea, I guess.
I think that iPhone apps are like everyone has an
iPhone so you’re already tapping into a huge
potential clientele and then after that I mean lots of
old people like 80 percent of old people go to
nurseries and all that kind of stuff and like flowers
and plants, so they’re going to like it.
It’s just another gadget though. I don’t think that
it’s going to work. I don’t think it’s that big.
We’re so lazy, that anything that saves us walking
up to switch the light switch on and off is
everyone’s.
Let’s say it’s not a jacket anyway. I don’t think the
jacket matters right now. Because it does make a
point because you’re not looking at a full map.
You’re looking at just lights blinking.
Unless there’s a lid on there’s not even going to be
vaguely effective. If there’s a lid just that like seals.
I think this would be good for sick people who are
like alone. They don’t have any friends or families
and this helped them to remind them to take their
medicine.
You’re already looking at the necessity for a
widespread use from the very beginning to make
this work.
But you can also use it for like busy people for
terminal disease or something like that.

only be only a partial logical inference from function to form. The depth of the
logical inference depends upon the complexity of the problem and the number
of sub-problems to be resolved (Zeng & Cheng, 1991). Each partial abductive
inference generates new rules that cumulatively tie together the observed func-
tion or concept to the end-users, user needs, causal precedents for conditions
of possibility or impossibility of the concept, and contexts of use.

To demonstrate our coding scheme in formal logic notation, we present an

illustrative example. In the following excerpt, a committee member makes
an observation about a problem with the pricing for a proposed product.
To explain that this is not actually an issue, the committee member extends
the design concept itself and invents a plausible concept and end-user.
You need the gadgets though, the actual sense of. That might be a problem in
terms of pricing. Just aim for rich people. . I mean rich people like new
things because they always want to show it off to their friends and stuff.
It’s for pour me a drink. Make me a sandwich. Pretty cool.

50 Design Studies Vol 37 No. C March 2015

 The meaning of the logical derivation is as follows. If you want the function
‘turn appliances on and off”, you must discover its new concept (through
further development) as a personal assistant to ‘pour me a drink. Make me
a sandwich’. This concept is plausible only if you discover that the appropriate
end-user is ‘rich people’. Taken together, the committee member generates a
testable hypothesises that a personal assistant concept to ‘pour me a drink’
or ‘make me a sandwich’ is plausible ‘for rich people’. If true, this hypothesis
explains the existence (commercially) of the function. Using the notation for
innovative abductive reasoning proposed by Roozenburg (1993) in which the
observed fact is a function, the two rules inferred are the concept and the end-
user. Abduction proceeded in this example as follows:
q given: function ¼ turn appliances on or off
p/q first conclusion: IF It’s for pour me a drink. Make me a sandwich (concept)
THEN turn appliances on or off
p second conclusion: It’s for pour me a drink. Make me a sandwich
q given: It’s for pour me a drink. Make me a sandwich
p/q first conclusion: IF Just aim for rich people (how e new end-user)
THEN It’s for pour me a drink. Make me a sandwich.
p second conclusion: Just aim for rich people

As with deductive reasoning, abductive reasoning is not used in a single direc-

tion, that is, solely to support a project. As shown in the following excerpt, the
committee member modifies ‘what’ the device is, a device to ‘fake a hijack’ and
‘kidnap’, and the new working principle for the device by ‘robbing’ them from
kids, to arrive at a different interpretation of the function to ‘track child’ e
that is, to track a child in a misleading manner.

Yeah, and sorry for stating that, but it’s an easy way to fake a hijack and a
kidnap just robbing the watch from a kid.

In logical notation, this excerpt would be denoted as:

q given: function ¼ track child
p/q first conclusion: IF It’s an easy way to fake a hijack and a kidnap (concept)
THEN track child
p second conclusion: It’s an easy way to fake a hijack and a kidnap
q given: It’s an easy way to fake a hijack and a kidnap
p/q first conclusion: IF just robbing the watch from a kid (how e new context
of use) THEN It’s an easy way to fake a hijack and a kidnap
p second conclusion: just robbing the watch from a kid

Two of the authors read several transcripts without knowing from which
reasoning manipulation the transcript is derived. They highlighted portions
of the text realizing a deductive, inductive, or abductive reasoning frame. In
the second stage, they met to discuss the initial criteria and associated exam-
ples to determine if the coding scheme provided sufficient clarity and coverage

The effect of abductive reasoning 51

 of instances of forms of logical reasoning in natural language. Based upon this
discussion, a final set of criteria was produced to code the transcripts. A spare
transcript was coded and arbitrated with further clarification of the criteria un-
til the inter-coder reliability (based on Krippendorf’s alpha and Cohen’s
kappa) on this transcript was higher than .80, which is considered acceptable
(Lombard, Snyder-Duch, & Bracken, 2002).

The final coding scheme is shown in Table 1. The criteria for abduction are the
‘rule to be inferred first’ in the model of innovative abduction proposed by
Roozenburg (1993). The first criterion modifies the concept, the second mod-
ifies the form, and the latter three introduce new working principles.

2.2 Results of content analysis

The Krippendorf alpha (Hayes & Krippendorff, 2007) and Cohen’s kappa co-
efficients across all the transcripts were calculated after both coders completed
a transcript. When they were below the .80 threshold, the transcript was re-
coded until an acceptable level was reached, which is a stricter methodology
than generally required (Lombard et al., 2002). The final inter-coder reliability
statistics are reported in Table 2.

3 Experimental results
3.1 Manipulation check
First, to verify that our manipulation worked, we tested the proportion of ab-
ductive reasoning to deductive/inductive reasoning within each manipulation.
If the manipulations worked as expected, we expect an increase in proportion
of abductive reasoning under the abductive reasoning manipulation compared
to the deductive/inductive reasoning manipulation and an increase in propor-
tion of deductive/inductive reasoning during the deductive/inductive
reasoning manipulation compared to the abductive reasoning manipulation.
Table 3 summarizes the number of observations of abductive or deductive/
inductive reasoning. There was only 1 instance of inductive reasoning in the
transcript; the rarity of inductive reasoning was also identified in other case
study research on logical reasoning in design (Galle, 1996). Therefore, going
forward, we will refer to the deductive/inductive reasoning frame as the deduc-
tive reasoning frame due to the lack of data on inductive reasoning. The total
number of observations of abductive or deductive reasoning is similar, 104 and
106, respectively, due to the 5 min time limit on discussion. As predicted
though, the manipulations increased the proportion of abductive or deductive
reasoning, respectively. The proportion of abductive reasoning under the ab-
ductive manipulation is greater than under the deductive manipulation
(c2(1) ¼ 3.224, right one-tailed p ¼ .0363; Fisher’s exact test right one-tailed
p ¼ .0488). We can thus conclude that the manipulations produced the appro-
priate reasoning frame in the committee. We further note that all forms of
logical reasoning will appear during design concept selection, but that the

52 Design Studies Vol 37 No. C March 2015

Table 2 Inter-coder reliability for reasoning frame

Group 1 2 3 4 5 6

a .8244 .9001 .9747 .8295 .8196 .8487

k .824 .9 .975 .829 .820 .849

reasoning frame can be manipulated. We test next the effect of the reasoning
frame on the direction of project acceptances.

3.2 Hypothesis testing

Under the abductive reasoning frame manipulation, 54 of 105 possible individ-
ual decisions were to accept projects whereas under the deductive reasoning
frame manipulation, 28 of 105 individual decisions were to accept projects.
We performed a two-sample t-test between proportions from independent
samples to determine whether the proportion of projects accepted under the
abductive reasoning frame manipulation is greater than the proportion of pro-
jects accepted under the deductive reasoning frame to a significance of a ¼ .05.
We can conclude that under an abductive reasoning frame the proportion of
project acceptances is higher than under a deductive reasoning frame
(t(204.5) ¼ 3.78, right one-tailed p ¼ .0001). Similarly a Fisher’s exact test
that the proportion of acceptances under the abductive reasoning frame is
greater than the number of acceptances under the deductive reasoning frame
is significant (right one-tailed p ¼ 1.882E-04). These two statistical tests
confirm that the reasoning frame influences the direction of project accep-
tance. Project acceptance increases when the reasoning frame is abductive
and decreases when the reasoning frame is deductive.

We then performed a binary logistical regression (logit) to ascertain the prob-

ability of project acceptance. The dependent variable is whether an individual
accepted or rejected a project. The independent variables (main effects) are the
logical reasoning manipulation (deductive or abductive) (MANIPULATION)
and the ratio of counts of abductive reasoning to counts of deductive
reasoning during each five-minute block of conversation for a specific project
(ADRATIO). For example, if during the five-minute discussion for a project
we coded 4 instances of abductive reasoning and 2 instances of deductive
reasoning, the ratio is 2. If there is no instance of deductive reasoning but
the count of abductive reasoning is nonzero, then the ratio is set as 1 to indi-
cate that abductive reasoning was observed during the discussion. We included
an interaction effect between MANIPULATION and ADRATIO since Table

Table 3 Observations of abductive and deductive reasoning

Observations of abductive reasoning Observations of deductive reasoning

Abductive Reasoning Manipulation 50 38

Deductive Reasoning Manipulation 54 68
Total 104 106

The effect of abductive reasoning 53

 3 shows that the ratio is dependent upon the level of the logical reasoning
manipulation. The regression model is significant (p < .0001). Table 4 summa-
rizes the results of the regression. We can conclude that the logical reasoning
manipulation and the ratio of abductive to deductive reasoning within a proj-
ect discussion increase the probability of individual acceptance since their
regression coefficients are positive. This result provides further evidence that
the reasoning frame influences the direction of project acceptance.

As the statistical analyses showed, deductive reasoning is correlated with proj-

ect rejection and abductive reasoning is correlated with project acceptance.
This is exemplified by the following discussion between two committee mem-
bers about the mobile phone application that assists in tracking urban re-
vegetation. In the following excerpts, we italicize portions of the dialog that
realize deductive or abductive reasoning frames. In the deductive reasoning
excerpt, there is a general tendency toward rejecting the project, whereas in
the abductive reasoning excerpt the committee member tries to find a new
frame for the concept and to identify potential contexts of use and users based
upon this frame. Further, this case is representative of many discussions
wherein abductive reasoning is used to counter negative deductive reasoning
or when the discussion is trending toward rejecting the project.

Deductive: So wouldn’t all sort of the success of this hinge on lots of users
actually using it, like wanting to use it? So there has to be a demand. If
there’s not, then it’s totally useless, because it relies on users going around
and taking photos of plants. And say you’re in an area where it wasn’t tak-
ing off, like it wasn’t popular, then the whole thing just failed, because why
would you use it when there’s just like one or two plants around the whole
of Sydney? And then you wouldn’t have any sort of motivation to find .

In contrast, in the abductive reasoning excerpt, the committee member makes

one or more observations about the proposed product, but rather than reaching
a logical conclusion, the committee member attempts to explain the concept by
hypothesizing the conditions of possibility for the existence of the product as
well as framing the device as a specific type of tool for certain interested people
rather than the general vegetation mapping application as presented.

Abductive: But then I think also you’ve got to take into consideration– like
maybe around Sydney it’s a bit different when you’re in a city, but, I mean, if
you take like an entire country, for example, I think there are definitely
Table 4 Results of binary logistical regression

Parameter DF Estimate Standard Error Wald Chi-Square p-value

Intercept 1 1.9077 .3868 24.3275 <.0001

MANIPULATION 1 1.9379 .4679 17.1535 <.0001
ADRATIO 1 1.3558 .4349 9.7187 .0018
MANIPULATION * ADRATIO 1 1.3441 .4575 8.5035 .0035
)
Denotes an interaction effect between MANIPULATION and ADRATIO.

54 Design Studies Vol 37 No. C March 2015

 more specific areas where there’s obviously a lot more vegetation, plant
life. I mean, yeah, it’s not something that’s definitely for everyone, but I
just think there would be a lot of– there are a lot of people who just try
and– it could be everyday people who are just trying to find a certain plant,
and people who are interested. I don’t know. I really like the technical side
of this one, and I .

Perhaps the most important consequence of taking an abductive reasoning

frame is that abduction can change a committee member’s preference toward
a project:
In my self-evaluation I wasn’t that keen on it but now that I think about it,
homeowners that you’ve got a garden will work out what kind of plant you’ll
put in there will be a great resource.

In this excerpt, the decision maker introduces a new concept, garden tool
rather than vegetation mapper, and new end-users, home gardeners. As a
result, the decision maker ends up supporting the project.

Our conclusion is that within the confines of design concept selection, when
decision makers can invent new frames for presented design concepts, wherein
the new frame includes a new interpretation of the design concept as presented
and a new working principle explaining the concept, the decision maker is
more likely to accept the design concept.

4 Conclusion
We described research on the effects of forms of logical reasoning on an essen-
tial part of the design process e the determination whether a proposed design
concept is worthwhile to pursue. We applied statistical analysis and a set of
criteria for the analysis of forms of logical reasoning to experiments on com-
mittees selecting potentially innovative design concepts, which entailed
judging the creativity, novelty, market acceptance, technical feasibility, and
customer value of the projects. We showed that the reasoning frame influences
the likelihood of project acceptance. Under an abductive reasoning frame
manipulation, decision makers were more likely to accept projects. When there
is a deductive reasoning frame during committee deliberation, that is, evidence
of a higher proportion of deductive reasoning rather than abductive reasoning,
the likelihood of project acceptances decreased. Further, in the transcripts, we
identified instances of committee members applying abductive reasoning to
counter negative deductive logic by others. In short, when decision makers
start to apply the design thinking cognitive strategy of innovative abduction
during design concept selection, they are departing from evaluation per se
and starting ‘design thinking’. They are inventing new ways to frame the prod-
uct as something else other than as presented and exploring new working prin-
ciples underpinning their new frame. As a consequence, they are more likely to
accept a project.

The effect of abductive reasoning 55

 We neither prescribe abduction as the preferred mode of logical reasoning in
design concept selection nor suggest the elimination of deductive reasoning. As
our results show, all forms of logical reasoning occur during design concept
selection. Rather, we point to the new finding that the reasoning frame alters
project acceptance in a predictable direction. Where companies or designers
may be timid or risk averse about innovation, inculcating the core design
thinking cognitive strategy of innovative abductive reasoning into their
concept screening and evaluation committees may help to boost the rate of
concept acceptance, and thereby potentially reduce the negative opportunity
cost of letting innovation pass. If companies, or designers, wish to accept
more projects at early stages of development, they may do well to inculcate ab-
ductive forms of reasoning in the concept selection process so as not to ‘kill off’
potentially useful concepts prematurely. Where they are too risk seeking, they
may wish to enforce careful deductive analysis of the merits of the concept.
Recognizing when forms of reasoning occur may also help committees to
take opposing cognitive strategies (Gebert, Boerner, & Kearney, 2010) so as
to minimize selecting what will turn out to be a bad concept.

While we have conducted the experiment within the confines of the problem of
design concept selection, the results may have implications in other situations
when committees must choose from a set of equally attractive strategic op-
tions. While we have demonstrated the effect of the reasoning frame, other fac-
tors such as the structure of the voting rule, consensus or single vote to accept,
and the number of decision makers may also affect the decision outcome
(Bardolet, Fox, & Lovallo, 2011; Sah & Stiglitz, 1988; Yang, 2010). Given
the positive effect on getting potentially innovative ideas ‘off the ground’, it
may be that abductive reasoning is a much more valuable and core cognitive
strategy to design and other fields than previously theorized. Option evalua-
tion methods may do well to encourage abductive reasoning during early
screening phases so that decision makers do not inadvertently overlook oppor-
tunities for innovation.

Acknowledgements
This research was supported under Australian Research Council’s Discovery
Projects funding scheme (project DP130101065) to Chief Investigators Dan
Lovallo and Andy Dong. Professor Andy Dong is the recipient of an
Australian Research Council Future Fellowship (project number
FT100100376).

Notes
1. Since conducting this experiment, the authors have learned, by searching the Web, that
versions of these projects have been developed commercially or by non-profit organiza-
tions. Thus, a decision to take any, and indeed all, of these projects forward would have
been a ‘good’ decision.

56 Design Studies Vol 37 No. C March 2015

 References
Bardolet, D., Fox, C. R., & Lovallo, D. (2011). Corporate capital allocation: a
behavioral perspective. Strategic Management Journal, 32, 1465e1483.
Bayus, B. L. (2013). Crowdsourcing new product ideas over time: an analysis of
the Dell IdeaStorm community. Management Science, 59, 226e244.
Cooper, R. G. (1985). Selecting winning new product projects: using the NewProd
system. Journal of Product Innovation Management, 2, 34e44.
Cooper, R. G., Edgett, S. J., & Kleinschmidt, E. J. (2002). Optimizing the stage-
gate process: what best-practice companies do-I. Research Technology Man-
agement, 45, 21e27.
Dong, A., Kleinsmann, M., & Valkenburg, R. (2009). Affect-in-Cognition
through the language of appraisals. In J. McDonnell, & P. Lloyd (Eds.),
About: Designing - Analysing Design Meetings (pp. 119e133). London: CRC
Press.
Dorst, K. (2011). The core of ‘design thinking’ and its application. Design Studies,
32, 521e532.
Fiske, S. T., & Taylor, S. E. (1991). Social Cognition (2 ed.). New York: McGraw-
Hill.
Galle, P. (1996). Design rationalization and the logic of design: a case study.
Design Studies, 17, 253e275.
Gebert, D., Boerner, S., & Kearney, E. (2010). Fostering team innovation: why is
it important to combine opposing action strategies? Organization Science, 21,
593e608.
Hayes, A. F., & Krippendorff, K. (2007). Answering the call for a standard reli-
ability measure for coding data. Communication Methods and Measures, 1,
77e89.
Kolko, J. (2010). Abductive thinking and sensemaking: the drivers of design syn-
thesis. Design Issues, 26, 15e28.
Krishnan, V., & Ulrich, K. T. (2001). Product development decisions: a review of
the literature. Management Science, 47, 1e21.
Kroll, E., & Koskela, L. (2014). On abduction in design. In J. S. Gero, &
S. Hanna (Eds.), Design Computing and Cognition DCC’14 (pp. 357e376).
Dordrecht: Springer.
Lombard, M., Snyder-Duch, J., & Bracken, C. C. (2002). Content analysis in
mass communication: assessment and reporting of intercoder reliability. Hu-
man Communication Research, 28, 587e604.
Lovallo, D., & Sibony, O. (2010). The case for behavioral strategy. McKinsey
Quarterly 30e43.
Magnani, L. (1995). Creative processes in scientific discovery. European Journal of
High Ability, 6, 160e169.
Maher, M. L. (2010). Evaluating creativity in humans, computers, and collectively
intelligent systems. In B. T. Christensen, T. Kristensen, & S. Boztepe (Eds.),
Proceedings of the 1st DESIRE Network Conference on Creativity and Innova-
tion in Design (pp. 22e28), Aarhus, Denmark: Desire Network.
March, L. (1976). The logic of design and the question of value. In L. March
(Ed.), The Architecture of Form (pp. 1e40). Cambridge: Cambridge University
Press.
Peirce, C. S. (1932). Elements of Logic. Cambridge: Harvard University Press.
van Riel, A. C. R., Semeijn, J., Hammedi, W., & Henseler, J. (2011). Technology-
based service proposal screening and decision-making effectiveness. Manage-
ment Decision, 49, 762e783.

The effect of abductive reasoning 57

 Roozenburg, N. F. (1993). On the pattern of reasoning in innovative design.
Design Studies, 14, 4e18.
Sah, R. K., & Stiglitz, J. E. (1988). Committees, hierarchies and polyarchies. The
Economic Journal, 98, 451e470.
Shah, J. J., Smith, S. M., & Vargas-Hernandez, N. (2003). Metrics for measuring
ideation effectiveness. Design Studies, 24, 111e134.
Smith, V. L. (1982). Microeconomic systems as an experimental science. American
Economic Review, 72, 923e955.
Thurston, D. L. (1991). A formal method for subjective design evaluation with
multiple attributes. Research in Engineering Design, 3, 105e122.
Tversky, A., & Kahneman, D. (1974). Judgement under uncertainty: heuristics
and biases. Science, 185, 1124e1131.
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychol-
ogy of choice. Science, 211, 453e458.
Udell, G. G. (1989). Invention evaluation services: a review of the state of the art.
Journal of Product Innovation Management, 6, 157e168.
Ulrich, K. T., & Eppinger, S. D. (2004). Product Design and Development (3 ed.).
New York: McGraw-Hill/Irwin.
Yang, M. C. (2010). Consensus and single leader decision-making in teams using
structured design methods. Design Studies, 31, 345e362.
Zeng, Y., & Cheng, G. D. (1991). On the logic of design. Design Studies, 12,
137e141.

58 Design Studies Vol 37 No. C March 2015