Design, Science and Wicked Problems PDF
Design, Science and Wicked Problems PDF
Design, Science and Wicked Problems PDF
Robert Farrell and Cliff Hooker, School of Humanities and Social Sciences,
University of Newcastle, Callaghan, NSW 2305, Australia
We examine the claim that design is demarcated from science by having wicked
problems while science does not and argue that it is wrong. We examine each of
the ten features Rittel and Weber hold to be characteristic of wicked problems
and show that they derive from three general sources common to science and
design: agent finitude, system complexity and problem normativity, and play
analogous roles in each. This provides the basis for a common core cognitive
process to design and science. Underlying our arguments is a shift to a strategic
problem-solving conception of method in both disciplines that opens up new
opportunities for synergetic cross-disciplinary research and practice.
Ó 2013 Elsevier Ltd. All rights reserved.
A
n important class of argument intended to distinguish design from sci-
ence is focussed around the claim that design is characteristically faced
with wicked problems whereas science is not. The two kinds of prob-
lem are argued to require different skills and methods for their solution. There-
fore, design and science are distinct types of intellectual study and production.
An influential position of this kind was set out by Rittel and Weber in their
landmark paper:
Rittel and Weber label as ‘tame’, in contrast to wicked, problems they claim are
typical of science. “For any given tame problem, an exhaustive formulation can
be stated containing all the information the problem-solver needs for understand-
ing and solving the problem .” (p. 161) “. their mission [solution goal] is clear.
It is clear, in turn, whether or not the problems have been solved.” (p. 160).
At the time of writing, Rittel and Webber were responding specifically to the
disappointed expectations aroused by the new systems approaches to problem
solving that would bring the social sciences within science and engineering,
and more generally to still broader claims for computational approaches to
mind and artificial intelligence, engineering and formal management approaches
to problem solving, and the like that would permit subsumption of psychology
generally (thence economics, etc.) and so also design under the prevailing logical
conception of scientific rationality (cf. Cross, 2007). Though much has changed,
their negative response to such rationalising ambitions remains widely sup-
ported throughout the literature on design process. Cross, for example, claims
that one of the reasons for distinguishing design from science is that “design
problems are ill-defined, ill-structured, or ‘wicked’” (1982, p. 224) whereas sci-
ence problems are mere ‘puzzles’2 to be solved by applying well known rules
to the data given. In a similar vein Willem quotes Archer as saying that “if the
solution to a problem arises automatically and inevitably from the interaction
of the data, then the problem is not, by definition, a design problem”; Willem
goes on to explain that “it is not an ill-structured problem” (1990, p. 44).
In this paper we will examine the arguments presented for the claim that wick-
edness of characteristic problems divides design from science as cognitive prob-
lem-solving processes, that is, that the cognitive or problem-solving nature of
the two processes are importantly different. We will accept that it is legitimate
These days this last claim is the more important one. This is because many more
designers today would no doubt allow that there is no simple division of problems
between design and science, yet nonetheless want to hold that there are important
cognitive differences between the two. For instance, Zeisel (2005) and Koskinen,
Zimmerman, Binder, Redstrom, and Wensween (2011) hold that research occurs
within, and about, the design process, but the latter remains distinctive. Our con-
trary ambition here is to show that, at least in respect of their core problem-solving
processes, there is no arguable difference between design and science. (The differ-
ences, which remain real, must lie in their external pragmatic conditions and their
cognitive consequences in turn e see below.) It is crucial to doing this that we also
shift our conception of science, in particular of scientific method, from that which
prevailed in the 1970s when Rittel and Webber published their paper to a more
adequate one which gives a prominent place to strategic problem solving directed
at multi-valued knowledge acquisition. Such a notion was not always available.
For instance, behind Cross’ notion above, that science problems are mere puzzles
to be solved by applying well known rules to the data, lies the assumption that sci-
entific research is fixed by logic, that scientific method is a logical machine that
takes data as input and generates true or most probably true theories as output
using only sound logical inferences. But this once dominant conception of scien-
tific method is now generally agreed by scholars of science to be fundamentally
flawed3 and we are free to shift conceptions. Then, as we shall show, design and
science turn out to share the same core cognitive process. Thus, rather than
dividing design from science cognitively, the distinction helps unite them.
Finitude. The finitude of our cognitive capacity and our resources, is a pro-
found limitation on our abilities e individual, social, and as a species e to
To illustrate what is at issue here for Rittel and Webber consider a brief to
design a commercial bank. On which among the huge number of alternative
We might take the view that, while in principle all potential solutions are know-
able and mutually comparable, the complicatedness of design situations ren-
ders the best solution very difficult to discover in practice. Rittel and Webber
say about poverty, for instance, that it has many partial, context-dependent
causes and we simply don’t, and practically can’t, know enough about all,
and perhaps any, of them. In this ‘ignorance’ reading of the situation, underly-
ing our ignorance is presumably a tame or textbook problem, just one that re-
quires ideal knowledge to identify, but towards which we should try to move.
But focussing on that would be both a too-partial reading of Rittel and Webber
and, more importantly, a strategic mistake. For instance, one might try devel-
oping all options to see if the best one stands out, or try guessing which one is
the best one. But in the presence of substantial numbers of options and time
and resourcing constraints, the former is likely not possible and the latter likely
not successful. In short, such strategies are unlikely to pay off. Knowing of the
bare existence of a best bank design, even were there one, is of no practical use,
and may even be detrimental, to designers if they are ignorant of what it is. The
practical problem is how then to proceed.
The rational alternative is to cease directly pursuing the best solution and
begin a search for an accessible, at-least-satisfactory one, even if it is not the
best. And now we are already in the domain of a very different overall model
of design methodology. Here the way to proceed is to consider a few alterna-
tive proposals across a reasonable spectrum of options, each accompanied
with a proposed investigation, use of realistic resources, etc. and a rationale
as to the value and realism of the approach. There follows a cyclic process: af-
ter critical discussion, including of available knowledge and resources, there
will emerge a decision to initially pursue a very few (perhaps only one) option
to a next presentation stage and, after the results of that round are in, the
whole process can be repeated again and again, sometimes with the same op-
tions, sometimes with a suggestive new one added, until one or more at-least-
satisfactory designs emerge that also can be completed within the investigatory
knowledge and resources available. One of these can then be chosen to
execute.
To make the point, notice that feature #i has an exact analogue in scientific
method itself. It occurs, e.g., whenever a given data set has two or more
different potential theoretical explanations (cf. design options) but limited
investigatory resources make it impractical to fully research every possibility.
It was, e.g., initially unclear whether chronic fatigue syndrome was caused by a
bacterium or virus, a fungus or mould, in each case perhaps deeply embedded
in tissue, or was due to a psycho-somatic condition, with any of these options
difficult and resource demanding to pursue. Then, just as with design, the issue
It was Popper who famously pointed out that the foundations of science are
not anchored on the rock of proven truths but instead driven down into a
swamp of possibilities just deep enough to achieve sufficient stability to
continue research (Popper, 1980). It is not surprising that the same methodo-
logical process turns up in science also since, as noted above, it derives solely
from rationality in the face of finitude. In short, the cognitive differences here
between the cognitive processes of design and science are minimal and matters
of degree, not kind.
This commonality becomes still clearer whenever scientists venture into new
unexplored territories, e.g. from Newtonian into relativistic or quantum do-
mains, or where scientists are fundamentally re-evaluating previously explored
territory, e.g. re-exploring embryology in terms of cellular bio-synthetic path-
ways rather than earlier macro-physiological characterisations, in short when-
ever scientists are engaged in deep or revolutionary research. Faced with the
initial anomalous discrete spectral data that ultimately led to quantum theory,
scientists first tried various Newtonian and quasi-Newtonian approaches to
understanding the data, even to the point of giving up energy conservation
to preserve a general Newtonian conception, before standard quantum theory
was tentatively accepted. Not only is there the exploration of accessible op-
tions, there is also the question of appropriate methods. For instance, in the
transition from Newtonian to quantum mechanics, the Newtonian measuring
methods are revealed to carry small, ineliminable residual errors and comple-
mentary Newtonian measurements become mutually excluding in quantum
theory, governed by the uncertainty relations. Thus, as with design, it is neces-
sary to search for problemesolutionemethod options. To borrow from
Simon’s (1977) treatment of this situation, scientists will order the ill-
defined, ill-structured, situation by assuming a structure and then see whether
it promises to bring about a solution to the original problem. It is clear from all
this that the problems that scientists encounter are not tame problems which
The only difference between science and design here is that in science these
problems and their resolution processes can be spread across many institutions
and individuals and slowly collectively evolve, whereas in design it is more
common for an individual or single agency to take on such investigations e
though there are also public tenders involving multiple agencies and there is
indirect common learning through adaptation of tertiary education curricula
and the like. These differences are not trivial to the overall, collective charac-
ters of design and science as cultural expressions, an issue we examine else-
where as part of considering the roles of norms in the two activities (Farrell
& Hooker, 2012b). The point here is that, whatever these differences, they
do not alter the cognitive design/research process involved.
This analysis places Rittel and Webber’s feature #i firmly under our first basic
wickedness feature, finitude. It is possible, however, that some readers e
perhaps taking off from the poverty example e would consider that the pri-
mary issue concerns normative conflict in design versus none in science. In
response we make three brief remarks: (i) even if normative diversity is
involved in the example of poverty, this reading is inappropriate because it
misses the main cognitive significance of feature #i, (ii) the interpretation still
falls under our three basic conditions, but now under normativity, and (iii) as
noted above, we treat normative issues elsewhere, where we argue that they do
not divide design and science in terms of their core cognitive process.
One reason for this follows immediately from feature #i: since any given solu-
tion to a design problem has emerged as an at-least-satisfactory construct
given our finite resources, it follows that if given an increase in resources,
then we can always search problememethodesolution space more effectively
for more promising options and/or improve solution designs for the options
explored. Hence it is always possible to improve a given design solution.
This is an important cognitive feature of the design process in general, and
of wicked problem solutions in particular. But it is obviously equally true of
science as of design and for exactly the same reasons. It is always possible
to improve a given scientific solution. In both cases it lies in contrast to
tame problems where solutions are arrived at in a finite number of steps and
are complete.
But Rittel and Webber offer, not one, but three reasons for feature #ii, of
which that above is essentially the first. (Their version reads: “. the process
The third reason is “because there are no ends to the causal chains that link
interacting open systems” (p. 162). This is a separate issue where a planning
decision is envisaged as being applied in one system that is in mutual interac-
tion with several other systems. For instance, a modification of the transport
system of a city, such as installing lights or a new bridge, interacts with eco-
nomic decisions about costs of public versus private transport and with social
decisions about the safety and privacy of each. Rittel and Webber have in
mind that decisions within each of the latter, including responses to transport
design decisions, equally react back on the traffic flows within the transport
system, and so on in a never-ending flow of mutual responses. In these circum-
stances, they conclude, decision consequences in any such system will lack
well-defined boundaries and so there will also be no well-defined solution
criteria for transport problems.12 For the same reason, various problems
have scopes that overlap, so that no one of them can be tackled without
affecting all the others. A changed transport design will alter business and lei-
sure activities, socio-economic stratification and segregation, and thus the na-
ture and distribution of medical demand and criminal activity, and so on.
This third reason comes down to the fact that planners, and designers, deal
with complex systems. Our ignorance of the complex causal chains in such sys-
tems makes it exceedingly difficult to arrive at optimal solutions whereby we
can judge that the problem has been solved and we can stop further investiga-
tion. But this feature is clearly shared with science. Science has been addressing
the issue of complex systems for many decades now; for example, climate sci-
entists have developed sophisticated theories and methodologies that have
However, the focus of the text strongly suggests that the intended reading
should instead be along these lines: because of normative diversity in evalua-
tion, there are no agreed evaluation criteria, so solutions to wicked problems
are not correct or incorrect as judged by agreed evaluation criteria, but vari-
ously good-or-bad according to the norms being used. Various parties will
judge any proffered solution and their judgements “are likely to differ widely
to accord with their group or personal interests, their special value-sets, and
their ideological predilections.” (p. 163) If this is right, then this feature of
wickedness clearly falls under condition three, normativity. Either way feature
#iii is explained by one or another of our conditions.
This may be understood as a special case of feature #ii, third reason, where the
temporally extended consequences of decisions are confined to one system.
The response there applies here as well. Or it may be read as a version of
feature #ix (below) and our response there will apply here.
Despite some temptation to assume so, this is not simply another version of
feature #iii, or #iv. Instead, like feature #ii, third reason, it introduces yet
another additional, and significant, methodological consideration, path-
dependence, over and above those discussed under feature #i.
This feature of wickedness is just the issue of finitude discussed under feature
#i: “. normally, in the pursuit of a wicked planning problem, a host of poten-
tial solutions arises; and another host is never thought up. It is then a matter of
judgement whether one should try to enlarge the available set or not. And it is,
of course, a matter of judgment which of these solutions should be pursued
and implemented.” (p. 164)16 Just so. And exactly the same in science. Only
the puzzle-solving activities (Kuhn’s description) of a mature science will
look at all like Rittel and Webber’s portrayal of scientific research, and then
only insofar as it is not disturbed by hidden errors or incompatibilities arising
externally. An immature science e that is, either a science in a new domain or
one newly emerging from a revolution in an old domain e will not have an
enumerable set of potential solutions to problems, for this is what makes a
discipline ‘immature’. Nor will there be a well-described set of permissible
methods, these too will be hotly debated in an immature science. Thus imma-
ture science will have the character that Rittel and Webber ascribe to social
planning. This feature of wickedness is wholly explainable in terms of condi-
tion one.
“But by ‘essentially unique’ we mean that, despite long lists of similarities be-
tween a current problem and a previous one, there always might be an addi-
tional distinguishing property that is of overriding importance.” (p. 164)
Noting that in science, including engineering, the ideal suggested by classes
of differential equations is that of obtaining parametric characterisations of
classes of dynamics, Rittel and Webber point out that this becomes less
possible as complexity mounts. And we would add, especially in systems
whose behaviour is governed by many weak interactions rather than a few
strong ones and where there are many long-period, path-dependent processes
running. “In the more complex world of social policy planning, every situation
is likely to be one-of-a-kind. If we are right about that, the direct transference
of the physical-science and engineering thought ways into social policy might
be dysfunctional, i.e. positively harmful. “Solutions” might be applied to
seemingly familiar problems which are quite incompatible with them.” (p.
165) But while all this is true, the barriers should not be over-estimated; tech-
niques of the sort used in geo-physics (see feature #v above) mean that many
component interactions can be successfully generalised, at least to some extent,
and general models of complex systems still offer general insight into charac-
teristic kinds of dynamics with their concomitant shaping possibilities, e.g.
parametric shaping of a strange attractor despite the chaotic behaviour it sup-
ports, and so on.
Rittel and Webber contend that “Removal of that cause [of the original prob-
lem] poses another problem of which the original problem is a ‘symptom’.” (p.
165) This feature can be interpreted in two broad ways. Firstly, it could be in-
terpreted as an unending progression, a progression that can be read either
‘horizontally’ or ‘vertically’. Reading it horizontally yields a regress of means
within method: a proposed solution to the initial problem (one that removes its
cause) involves bringing about condition X, as eliminating some form of crime
may in turn require providing police; but procuring X in turn requires bringing
about Y, as policing in turn has budgeting and planning consequences, and
procuring Y in turn ., and so on. This is true, but true in some degree of
any kind of problem, not just wicked ones. And certainly equally true of sci-
entific problems. Understanding hot plasmas requires the capacities to create
and contain gases at a million degrees, and instruments to probe their inte-
riors, but these in turn . Nor, if there were any qualms over claims the regress
is endless (which a reader may well have), would these qualms be any less
appropriate to science than to design. So this feature of wicked problems
will not divide science from design (or even wicked problems from most
others). It simply expresses another aspect of the finitude of any human situ-
ation (condition one).
The second broad interpretation follows from the following quote: “Thus
‘crime in the streets’ can be considered as a symptom of general moral decay,
or permissiveness, or deficient opportunity, or wealth, or poverty, or whatever
causal explanation you happen to like best. The level at which a problem is
settled depends upon the self-confidence of the analyst and cannot be decided
on logical grounds.” Here ‘happen’ and ‘like’ in “explanation you happen to
like best” implies that selection of explanations is arbitrary, whereas it is
not, though it may be characterised by ignorance, for either science or design.
However, looking past this loose language, Rittel and Webber could be
arguing that how we interpret a problem, and how it relates to other problems,
The wording of this feature encourages the belief that it is another version of
feature #i. However, and despite further confusing text,17 it seems clear that
Rittel and Webber’s intention is to argue that there is a looseness of connection
between evidence and explanation that can be exploited to defend any given
explanation from refutation while claiming supporting evidence for it. In sci-
ence, they assert, this does not happen, for if, under conditions C, hypothesis
H implies that evidence E occurs, and E doe not occur, then H is refuted. But
in design it is possible to argue that E did actually occur, or that intervening
processes independent of H prevented E from occurring, or that E is delayed
but will still occur, and so on. “In dealing with wicked problems, the modes of
reasoning used in the argument are much richer than those permissible in the
scientific discourse. Because of the essential uniqueness of the problem (see
Proposition 7 [feature #vii]) and lacking opportunity for rigorous experimen-
tation (see Proposition 5 [feature #v]), it is not possible to put H to a crucial
test.” (p. 166).
Is this then a dividing point between design and science? No, because the real
situation in science is exactly the opposite to that of the idealised logic machine
that Rittel and Webber here assume applies. Notoriously, it is also logically
possible in science to attempt to protect a hypothesis from refutation, and
through exactly the same kinds of responses as quoted above. Just this is the
point of the famous DuhemeQuine thesis, which holds e and for science in
general, not just wicked problems e that it is always possible to avoid refuta-
tion of a hypothesis in precisely these ways. It is for this reason that Karl
Popper, the great exponent of falsification in science, enjoined the methodo-
logical policy of testing hypotheses as severely as possible in an attempt to
falsify them e precisely because doing that was not logically guaranteed e
and considered it of the essence of taking a rational stance to enquiry. And
it was an attempt to show how a modicum of hypothesis-protection in the
face of apparently adverse evidence might nonetheless rationally sit with ulti-
mate falsification that Popper’s follower Imre Lakatos introduced the dual
model of a protected core or key hypothesis surrounded by a falsifiable belt
of auxiliary hypotheses. And this is still while trying to take a formal-logical
approach to science, rather than the more strategic conception of scientific
“Planners are liable for the consequences of the actions they generate; the ef-
fects can matter a great deal to those people that are touched by those ac-
tions.” (p. 167) This wickedness-making feature is the only one of Rittel and
Webber’s ten features that doesn’t neatly fit into our three conditions for wick-
edness. However, we do not think that this undermines our analysis and this
for three reasons. Firstly, the idea that a planner/designer has no right to be
wrong only reasonably stretches to mastery of current knowledge and skills:
why should we hold someone responsible for consequences of which they,
and everyone else, were ignorant? Moreover, it is exactly the same for scien-
tists. A poorly constructed research programme wastes scarce resources,
may mislead scientists who rely on its data and conclusions, and its poor per-
formance contributes to having its approach, or even the whole domain,
2 Conclusion
There is a widely accepted argument distinguishing design from science
based upon the idea that science and design address different types of prob-
lems e wicked problems in the design case, and tame problems in the science
case. We have argued that this argument does not succeed. The wicked/tame
distinction is not an exclusive dichotomy; rather, it is a continuum upon
which all problems can be based, scientific and design alike. Along the
way we have shown what underlies Rittel and Webber’s 10 features of wick-
edness are just three conditions that generate some aspect of wickedness, to
some degree: finitude, complexity and normativity. The first two of these
concern such general conditions that they are easily seen to apply to both
design and science (and much else) leaving design and science to share the
common cognitive process to which they give rise. Upon examination the
third condition, normativity, also does not distinguish design from science,
but even if it did it would follow from our analysis here that the wickedness
or not of problems is no longer an independent argument for distinguishing
design from science.
This paper has focussed on the analysis of wickedness. What has been
learned is that design method, like scientific research method, is a product
of a common core cognitive process and management of pragmatic
complicating conditions, and that methodological procedures and skills
break into ways of progressing each of core and pragmatics and managing
their interactions. This structure can be exploited e for instance, by trans-
fer of problem-solving experience and strategies, whether across sub-
domains within design or between science and design, despite pragmatic
differences. This will, for example, facilitate (within limits) the abstraction
and transfer of engineering design theory and procedures, where well
structured characterisations of problems, procedures and pragmatics
have made possible sophisticated resolution structures, to other, more
pragmatically complicated fields.19 More generally it provides designers
a critical tool to widen their outlook and reflect on their practices and pro-
vides a common framework within which to pose and test design research
issues.
Acknowledgements
The insightful and encouraging comments of two anonymous referees and a
journal associate editor are gratefully acknowledged as contributing to a clear,
balanced and well-focussed paper.
Endnotes