This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Design science or nonscience
Philip Sargent, Quillion Systems Ltd., St. John's Innovation Centre,
Cowley Road, Cambridge CB4 4WS, UK
This paper presents a viewpoint of design as intrinsically requiring the
integration of techniques from several fundamentally incompatible world-
views: distinct 'sciences'. This implies that there can be no single 'design
science'. The paper also casts doubt on whether a design science, if it did
exist, would be useful in the sense of helping to produce better or quicker
designs. This entails a short discussion of the distinctions between
'complete', 'closed' and 'adequate' theories. The paper concludes with
explicit recommendations to make future designs better and to improve the
effectiveness of design processes by enabling design 'shared memory'.
These recommendations follow directly from the strong limits imposed by
the proposition that there is no single science of design.
1 Dasgupta, S 'The structure of design processes' Advances in Comput. Vol 28 (1989) pp 1-28 2 Konda, S, Monarch, I, Sar-gent, P M and Subrahmanlan, E 'Shared memory in design: a unifying theme for research and practice' Res. in Eng. Design Vol 4 No 1 (1992) 23-42 3 Shaw, M 'Heterogeneous de-sign idioms for software architecture' in Proc. Sixth Intl. Workshop on Software Specifica-tion and Design. Como, Italy, 25- 26 Oct. 1991: IEEE Computer Society
minimize radio-interference and provide mechanical support during mo-tion. These world-views are incommensurate because they have no useful points of correspondence in reality except where they interact within a specific designed artefact. This is an operational definition since it depends on there being no useful general correspondences between the different world-views. The term 'useful points of correspondence' is necessary because fundamental physics often provides routes of corres-pondence between different world-views, but the connection is not useful to designers. An illustration is the fundamental connection between moving bodies and electromagnetism as in the car body example: it is special relativity; not a useful body of knowledge for most designers.
1.1 Defining hypothesis This paper advances the hypothesis that reconciling incommensurate requirements is an essential aspect of design, that the design process always requires it l and designed artefacts must always resolve it — in contrast to recent work by Konda et al.2 where this aspect was held to be typical but not essential. If a single world-view is adequate to a design problem then the goals must have a common representation and thus a common measure of utility. A single utility measure can be calculated and the 'design' can be solved completely by optimization techniques.
This definition is circular: I am defining design so that it conforms to the incommensurate viewpoints hypothesis so the validity of this paper rests on whether this definition is defensible and where it is reasonable. It is a useful definition in that it allows certain conclusions to follow; whether it is 'true' in any sense is almost impossible to say because different people accept different activities as being 'design'. However, by describing in more detail what these multiple world-views involve it is possible to arrive at a number of extensible definitions of design which nevertheless conform to an overall structure. This should enable theorists to state with some precision where their positions lie and so provide a useful means for clarifying debate in design theory.
1.2 Design idioms Designers use many 'design idioms', well-known reconciliations of world-views expressed as examples of good partial designs. Examples include bearing design (materials and kinematics), electric motor design (electro-magnetism and mechanics) dam design (soil mechanics and hydraulics). Shaw3 has recently used the term 'idiom' to refer to specific architectural designs constructed from multiple techniques, but where the techniques are more similar than the world-views used here. Her idioms include `table-driven interpreters', 'layered hierarchies' and 'pipelines': all from
390 Design Studies Vol 15 No 4 October 1994
software engineering, and all have been discovered or invented by the need to solve real-world problems. Only later have the idioms been recognized, then systematized and appropriate mathematics devised to explore their behaviour.
The view of design theory taken here implies that such idioms are compound bodies of knowledge where the component disciplines have been discovered to be necessary in actual engineering design problems. Large numbers of historical problems have enabled practising engineers to collect and systematize these compound bodies of knowledge and to define the classes of designs for which these idioms are relevant. Once an idiom is discovered, a mathematical engineering analysis technique can often be developed which takes as its boundary conditions the contribut-ing world-views and which may display considerable internal elegance. However, there is nothing in the individual sciences of, say, soil mecha-nics and hydraulics which, in the absence of real, specific dam design problems, implies that any compound body of knowledge exists.
Thus most of the techniques and methods used by designers can be seen to be derived from practice and are not in principle derivable by any kind of design science although the elegance of their reformulation may present that appearance. This is a fundamental reason why the pragmatic approach of 'case-based reasoning'4 is appropriate as a method for constructing designer support tools.
4 Bardasz, T and Zeid, I 'DEJAVU: a case-based reason-ing designer's assistant shell' in Al and Design 92 1992, Camegie Mellon University, Pittsburgh, PA: Kluwer, Dordrecht 5 Petroski, H To Engineer is Hu-man: The Role of Failure in Suc-cessful Design St. Martin's Press, New York (1985) 6 Ullman, D 'Design histories: archiving the evolution of pro-ducts' in DARPA Workshop on Manufacturing, 1991, Salt Lake City, Utah
1.3 The role of failure Since any new design could in principle involve individual sciences which have not historically been used together, and since there is no way to be sure that a previously unknown interaction is not going to affect the artefact, there will inevitably be service failures. These failures are important learning situations because they ensure that designers will in their future work properly explore the new design idiom these failures represent and the possibility of failure is part of the culture of engineerings Can engineer — someone who turns specifications into malfunctions', MIT t-shirt slogan. September 1991).
The possible physical effects which will influence a new artefact are not in principle enumerable in advance. Only the designer's imagination, the appropriate use of factors of ignorance6 and exhaustive preliminary testing can help reduce the danger to the public that inevitable failures will entail.
In planning the design process, these unforseeable interactions mean that
Design science or nonscience 391
any particular multiple world-view task can take an inestimable length of
time to reach a solution of unspecifiable precision. Any 'design science' is
therefore strictly limited in its predictive power except in very mature
domains using very well established design idioms. In these cases it is
arguable whether such a routine process is 'design' or whether it is mere
form-filling.
7 Vincent', W G What En-gineers Know and How They Know It, Johns Hopkins Universi-ty Press, Baltimore (1990) 8 Levy, S, Subrahmanlan, E, Konda, S, Coyne, R, Wester-berg, A and Reich, Y An over-view of the n-Dim environment, Engineering Design Center, Car-negie Mellon University, Tech-nical Report EDRC 05-65-93, January 1993 9 Finkelstein, A, Kramer, J, Nuseibeh, B, Finkelstein, L and Goedicke, M 'ViewPoints: a framework for integrating multiple perspectives in system develop-ment' Int. J. Software Engng Knowledge Engng Vol 2 No 1 (1992) 31-58 10 Nuselbeh, B and Finkel-stein, A 'ViewPoints: a vehicle for method and tool integration' in IEEE Intl. Workshop on Compu-ter Aided Software Engineering (CASE '92)6-10 July 1992, Mon-treal, Canada 11 Hoover, C W and Jones, J B (Eds) Improving Engineering Design: Designing for Competi-tive Advantage. National Academy Press: Washington, DC, 107 (1991)
2 What is design? Much of the information that is manipulated during design derives not
from the initial specification but, as Vincenti says, 'from the internal
needs of the design itself 7. It is the implications of the specification as
manifested in potential artefacts (which exhibit several different sorts of
behaviour simultaneously) that produce the raw material on which
designers work. Real artefacts can never be completely described by a
single discipline so multiple world-views appear as soon as a potential
design solution is found. In principle there is no way that a potential
solution can be assumed to be valid. There must always be a design
analysis stage in which a good idea is evaluated and checked. Typically the
idea will arise from looking at the problem from only one or two, perhaps
novel, viewpoints, but it must be checked against all viewpoints that can
be imagined". These are often listed as the `-ilities': design for
maintability, assembleability, testability etc. or described by the term
DF(X) or 'design-for-V11. The fact that a simple, unprioritized list is
always quoted strongly implies that no fundamental classification has yet
been found. The purpose of this paper is to suggest that design processes
which attempt to incorporate many world-views are inevitably ad hoc.
This does not mean that a classification cannot be attempted, but it does
mean that any such classification would not be fundamentally based;
though even an arbitrary, standardized classification would be a conve-
nient aid to education, training and research.
2.1 Revisiting the hypothesis: exploration A circular hypothesis was advanced earlier defining design as essentially
requiring multiple viewpoints and, by implication, including all activities
which require this. How well does such a definition match activities
commonly held to be 'design' and how many other activities does it try to
force into the 'design classification'? Before we answer this question I will
put forward a potential answer: perhaps it is particular types of design
idioms which typify 'clesignerly' activities to us. Perhaps it is not just
wrestling with incommensurate problems which defines design, but a
particular style of wrestling? Or maybe we are only willing to admit
familiar idioms to the classification of 'design' and not others, irrespective
of how well they 'should' fit according to our definitions. (This last point
392 Design Studies Vol 15 No 4 October 1994
has to be true in any case since the meanings of words in English are
defined socially and not by definition.)
Certain characteristic activities recur whenever multiple world-views have
to be rationalized in a single object. The most significant of these is
exploration, another is negotiation.
Since there is no a priori way of specifying what the interactions of two
different viewpoints might be, the only way to approach the problem is to
try something suggested by one viewpoint, see how the object being
designed is affected, and then change viewpoint to see how that change is
seen from another point of view. Most of the information used in the
rationalization of viewpoints is created by this exploratory process. This is
indeed a common, perhaps universal, characteristic of activities most
people call 'design' as the phrase quoted earlier from Vincenti7 shows.
Instead of proceeding along straight-lines (as suggested by most system-
atic methods of the Germanic schools) most designers approach their task
on a broad front of exploration and retraction,12 what one might call a
hesitant `wavefrone of exploration.
Many engineers put stronger limits on what they think of as routine
design: the rationalizations between viewpoints must involve some ab-
stract formal reasoning, certainly mathematical and usually numerical.
Complexity must arise from the interactions themselves. These must
require the use of nonobvious behaviour representations, e.g. Goodman
plots, Ashby charts, phase diagrams etc., which require professional
training to use (these representations are often recognized as being
sufficiently nonobvious and important enough to warrant retaining the
name of their originator). These interactions are individually characteris-
tics of the separate world-views (fatigue initiation, materials selection and
crystal structure stability, respectively), characteristics also common to
any engineering analysis in these world-views, but not characteristics of
design.
12 Ullman, D G 'Current status of design research in the US' in International Conference on En-gineering Design Zurich (1991)
However, the idea that reasoning and exploration produces nonobvious results, which in retrospect are almost inevitable, is generally held to be
typical of good design: e.g. that making some part lighter can actually
make the whole component stronger, or that one less heat exchanger may
enhance rather than reduce operating flexibility. So we have a reasonably
common perception of the types of viewpoint interactions (idioms in my
terminology) which we are prepared to admit to belonging to engineering
`design'. There is some evidence for this in software engineering. Early
programming was too hit and miss and too simple to he worthy in most
Design science or nonscience 393
engineers' eyes of the name design. Today we have developed a very large
armoury of techniques and methods, the process of producing software
has become much more complex and only very conservative engineers
would object to the phrase 'designing a software system'.
This is partly why we have such trouble defining 'design'. Architects and
engineers use different sets of idioms, but if one abstracts all the general
techniques from those idioms then one arrives at a definition of design
that encompasses all problem solving directed at achieving goals; any
activity pursued with 'intention'. This is clearly a wider definition than
most people are prepared to accept for design.
13 Warfleld, J The Science of Generic Design: Managing Com-
plexity Through Systems Design
Vols 1 and 2, Intersystems Pub-
lishers, Seaside, CA (1990)
14 BucclareIII, L 'Reflections on engineering design practice'
Design Studies Vol 5 No 5 (1984)
185-190 15 Bucciarelll, L 'An ethnog-
raphic perspective on engineer-ing design' Design Studies Vol 9
No 3 (1988) 159-168
16 Lander, S E, Lesser, V R and Connell, M E 'Knowledge-
based conflict resolution for cooperation among expert
agents' in Computer aided coop-
erative product development (MlT-JSME Workshop, Novem-ber 1989, D Sriram, R Logcher and S Fukuda (Eds) MIT Press,
Cambridge, MA (1991) pp 253-263
17 Dawes, R M Rational Choice in an Uncertain World Harcourt
Brace Jovanovich, New York
(1988) 18 de Neufville, R Applied Sys-tems Analysis: Engineering Plan-
ning and Technology Manage-ment McGraw-Hill (1990)
19 Keeney, R L and Rafffa, H Decisions with Multiple Objec-
tives: Preferences and Value Tradeoffs John Wiley, New York (1976)
By concentrating on multiple world-views whose rationalization requires
professional training (intuitive as well as intellectual) and on idioms which
already exist because they have been shown to be necessary by practice,
we arrive at an open-ended definition which can encompass what we
commonly think of as engineering design. (It is still somewhat too broad
though.) We can agree that certain identified sets of idioms constitute
`design' but fundamentally we cannot lay out criteria for the acceptance of
other, as yet unidentified, world-view interactionsI3 because new world-
views may see the problem in an entirely different way.
As disciplines mature more idioms are developed and completely new
activities become admitted to the field of 'design' (perhaps 'polysacchar-
ide genetic design' in the near future) — but none are admitted until there
are actual practising designers. This definition of design does not depend
on abstract classification of ideal activities but on social acceptance of the
worth of professional activity. So writing text-books might be `design'
according to my multiple viewpoint principle but because society does not
recognize the idiom of text-book construction as an activity worthy of
specific professional training, it is not design.
2.2 Negotiation Bucciarelli'4.15 among others has shown conclusively the central role that
negotiation takes in design activities'''. There are two kinds: negotiation
about meaning and negotiation between competing goals.
If we have multiple world-views then each view contains within it one or
more incommensurable measures of 'goodness', such as fast and safe in
the earlier example. Decision theory 17-19 has shown conclusively that
negotiation of incommensurable goals has to take place in a social
framework.
394 Design Studies Vol 15 No 4 October 1994
Rationalizing two viewpoints also requires considerable negotiation of
meaning15 because each viewpoint classifies the universe differently20.
The term 'chip' may mean much the same thing to an electronics
manufacturing engineer as to a semiconductor designer, but the bound-
aries of the classification are different. The manufacturing engineer
considers the silicon, metal leads and plastic package as 'the chip' whereas
the semiconductor designer just considers the silicon. In a novel design, in
both mechanical and software engineering, deciding what is a part of what
— what is an accessory and what is the main goal of each component — is a
major activity.
2.3 Social theories of design processes The multiple world-view approach to design theory implies that team-
work and communication are highly important activities. These are part of
the design process but their importance comes directly from the way we
have looked at the designed artefact. This correspondence and influence
between artefact and process is not often found so clearly in design
theories.
The methods used to enhance team-work: project management, struc-
tured meetings, negotiation, room layout, workplace ergonomics etc., are
design techniques which are process-centred rather than artefact-centred.
However, there seems little point in distinguishing between these two
classes of techniques because nothing useful comes of it. Warfield"
fundamentally separates these 'soft' techniques from the more 'hard-
science' based techniques but the case for doing so is not firmly based.
Designers `do' thermal calculations but they also 'do' scheduling and
project management. If we did make a fundamental distinction, since we
already have multiple world-views for the different 'hard-science' techni-
ques, we would need to say that scheduling, for example, represented a
different world-view and that the distinction between it and kinematics
was itself qualitatively distinct from the difference between, say, kinema-
tics and materials. That seems far-fetched and, more to the point, not
useful. Thus in this paper there will be no clustering of world-views into
more- or less-similar collections. All distinct world-views have to be considered equally valid in principle (though for any specific design
problem some will be more or less relevant).
20 Sargent, P M 'Product mod-els: adequacy, annotation and negotiation' in Workshop 2 tmenciated with "Al and Design '92". Carnegie Mellon University, Pittsburgh, PA, 1992
2.4 What is a science? Given the importance accorded to multiple world-views, we need to be
clear about what we mean by a single world-view: a single 'science'. Here
it is taken to be a mostly closed, coherent and consistent body of
knowledge together with a set of techniques and a general approach. The
Design science or nonscience 395
techniques produce results (outputs) which are the same sorts of things as
the inputs. (If they produce exactly the same things then the 'science' is
exactly closed and is termed a 'calculus', as in chess where every move
takes a position as input and results in a new position.) The techniques
and the knowledge in the 'science' must be mutually reinforcing (cohe-
rent) in that the knowledge explains why the techniques work and
manipulating the techniques elaborates the knowledge.
A stress analysis of a solid component takes loads (forces) as inputs and
produces stresses as outputs, but at any boundary or feature it is trivial to
convert these back into forces. The structural analysis of a pin-jointed
structure also uses forces but the techniques are quite different. Thus
design of a space-frame uses both these viewpoints: structural for overall
elastic stability and stress-analysis to design against plastic yield in the
solid joints. These viewpoints are treated as distinct 'sciences' because
although they share the same general problem solving strategy and
although in principle they can both be derived from Newton's laws of
motion, the habits of thought in the sets of techniques are different. As an
informal proof, one can observe that in universities they are taught in
different courses.
2.5 Extensible design theories Following on from the circular multiple world-view definition advanced
earlier, we can identify different definitions of design by adding different
restrictions to this over-general definition. Identifying a variety of such
definitions is only of any use, however, if we can match each such
definition with appropriate theories (which may be partial theories, even
within a restricted definition).
The following list of aspects of design has been advanced as being useful
to the synthesis of 'design science', but I think it is more useful to view
them as defining (in an overlapping and not exclusive manner) different
proposal', Engineering Design Research Center, Carnegie Mel-
lon University, Report to NSF
(1992) 24 Staley, S M and Vora, L S 'Reconciling design theory and methodology: OFD/TIES and its
place in the domain of science model', Ford Research, Technic-al Report SR-90-41, February 1990
25 Dixon, J R On a research methodology towards a scientific theory of design, in Design
Theory '88, S L Newsome, W R Spillers and S Finger (Eds)
Springer-Verlag, Berlin (1988) pp 316-337
26 Dasgupta, S Two laws of design' Intelligent Systems Engng Winter 1992
Accepting any one, or a set of, these aspects defines a variety of design for which explanatory theories and useful tools can be devised. Clearly some are dispensable (such as 'teamwork' for the lone inventor) but others might be essential for any activity socially accepted as typifying 'design' (and I suggest that exploration of emergent information might be one such). A list such as this can he used to distinguish between commonly used categories such as 'innovative', 'creative' or 'routine' design and would work well with current attempts to classify design by three orthogonal axes to represent the product and five orthogonal axes to represent the process. (For artefact: role of geometry, number to be produced, and magnitude of system, e.g. door-knob to aircraft. For process: activity type, e.g. routine to creative, modularity, number of regulatory constraints, type of information flow, number of designers23. This is not a final taxonomy and is already being improved.)
Identifying which are dispensable aspects for specific types of design is itself useful since support for such aspects can be omitted from designer support tools, with consequent savings in tool-development effort.
3 Types of theory There are several types of design theory. A popular type is the unified `design science' which some claim to have developed13'24 and which others see as their goal11,23,25,26. Alternative theory types are the purely descrip-tive, which can be divided into: (a) those which claim to describe all of the design process (or at least be capable of describing all of it); and (b) those which only claim to describe part of it.
It is often said that we are in a `pretheory stage' of design research' which I interpret to mean that although there are many design science theories there are none which are widely accepted. This paper presents a definitive theory, one that attempts to define design (partially and informally), but is not prescriptive since it has no direct consequences directing how a designer should or should not carry out design.
It is a popular assumption that a 'proper' theory of design would be
Design science or nonscience 397
'scientific' (a very contentious and effectively meaningless term2 because of overuse) and would be more or less the same as a 'design science'. It would:
• Guide designers in producing better designs
• Help designers work faster
• Help organizations manage design and designers more effectively
There is a misconception that a valid theory would inevitably solve design
problems: 'A set of viable, proven, formal processes for solving all types of design problems will constitute a theoretical foundation for engineering
design' I I (my italics), but I hope that the multiple world-views hypothesis
shows that need not necessarily be the case. These over-ambitious
expectations form very strong constraints on the type of theory being
looked for. Dixon (who has been the most articulate in stating these
hopes) assumes that any theory which covers all design will have these
characteristics, the only grounds for supposing that they will not be
satisfied would be that of failing to find any theory which covers all design25. This is not true. It is entirely possible to develop theories of
design which are entirely general, are satisfying from a philosophical point
of view, but which are useless in practice to designers (the theory put
forward in this paper is perhaps an example). Another view that has been
expressed is that any 'proper' design theory would be able to put
individual design methods on a common metric24. This is also not true, for
the same reasons: because incommensurability is inherent.
Any theory may be adequate in that it explains all the current data, but it
still may not be correct (all theories are hypotheses, of course). However,
even if not correct, a theory still shows that there can exist theories of that type.
A theory that is closed covers all the data and all manipulations of the data performed using the theory, and, without addition, would be capable of
explaining all such data which may come to light in the future. A closed
theory has no scope for expansion (see the description of a calculus above)
because all the loose ends have been tied up (often by the wishful thinking
of an overly tidy mind). However, even a closed theory may not have
been elaborated fully in that all the implicit truths may not have yet been
made explicit. Perhaps a more extreme example of an unelaborated
closed theory is quantum electrodynamics (QED) from which the entire
science of chemistry can, in principle, be derived.
A complete theory explains all the data but does not pretend to be
398 Design Studies Vol 15 No 4 October 1994
finished, manipulations of the data using the theory produce results which
require extensions to the theory. An incomplete theory cannot explain
some of the data which, by its own terms of reference, it should be able to.
A partial theory only attempts to cover that subset of the data explicitly
within its scope.
Thus social theories of the importance of negotiation are partial, incom-
plete theories of design. Warfield's 'design science' is a complete, closed
theory. The Pahl and Beitz model of design is partial and is claimed by
some to be complete, those who think the evidence is not strong argue
that it is incomplete even over the areas it claims to cover.
Most design theorists only claim publicly that their theories are partial,
but many are clearly convinced that their theories are complete. Dissent-
ing data are often classified as being outside the domain of the theory
rather than being admitted as counter examples. However, that is because
we are assuming that our theories are the same kind of thing that are used
to describe the simple natural world (physics). Design, as a subject of
study, is not alone in this: 'Much of computer science includes phenomena
about which we do not yet have well-established scientific truths but do
have interesting observations and generalizations'3. Indeed, most sciences
are like this. Physics is an atypical, rather than an archetypal science and
comparisons of any subject with physics with a view to proclaiming it 'a
science' are usually not useful.
Brooks27 proposed that three types of 'result' in software engineering be
recognized (as quoted by Shaw')
1) Findings (well-established scientific truths)
2) Observations (reports on actual phenomena)
3) Rules-of-thumb (generalizations proposed by an individual but
perhaps not rigorously supported by data)
and further proposed that appropriate criteria for these three be, respec-
tively
27 Brookes, F P 'Grasping reality through illusion — interac-tive graphics serving science' in Proc. ACM/SIGCHI Human Fac-tors in Computer Systems Con-ference, May 1988. Association of Computing Machinery (ACM)
1) Truth and rigour
2) Interestingness
3) Usefulness
and 'freshness' for all three. Confusion between these three classes, and
especially the application of inappropriate criteria, has lead to great
confusion in design theory.
Design science or nonscience 399
Learnin • Teamwork
A SPECIFIC DESIGN
Disposition \
C • straints / •-• ...
\ / .--
/ \ \
G
Behaviour eometry .. .0' Manufacturing -. \ .' \
/ .- ....... \ ..... .....
...., -..... ....„.. ..... \ ..- . \
../ VLSI \ Optimization / "....
.." N..N._ ...•
/ \ ../ \
Figure I Design as a nonexistent subject
Function -.
Design may be like any phenomenon in the complex real world. There
may be no single theory which covers any observation. Every designed
artefact may involve a different set of not just incommensurate technical
fields, but a different set of appropriate partial design theories (see Figure
1). As for the natural world we can attempt to map out where different
theories work and where they don't, but we should not expect our maps to
be exact.
It could be that the only thing common to all design is the intention to
produce something useful. That does not mean that design theory and
methodology research ends, it means that it is unending.
4 The usefulness of this approach The view of design as necessarily composed of interacting, distinct
world-views does have some useful consequences for design research, and
secondarily for designers. It implies that design idioms are precious
because they have been hard-won by experience and because they cannot
be derived ab initio. Thus the storing and sharing of past design histories,
- Matching
400 Design Studies Vol 15 No 4 October 1994
particularly where the design rationale is recorded, assumes a primary
importance for design research. This would entail research on repre-
sentations for which indexing and retrieval are easy and selective. The
need is stronger than merely recording the rationale behind a specific
artefact. What is required is the entire 'theory of the artefact'28, the entire
range of behaviour of artefacts and partial designs that have been
explored by the design process and then discarded. This would thus
comprise as complete a description as possible of the design idioms,
including novel idioms, used in the derivation of the artefact design.
Retaining the shared knowledge of an established design team is also seen
to be of primary importance. Any methods that encourage team-work
without reducing the variety of viewpoints are valuable. This reflects
current thinking in manufacturing engineering and engineering
management29 but whereas those disciplines base their thinking on
observations, here we derive the same principle directly from thinking
about design theory.
28 Subrahmanlan, E The theory of the artefact' (Personal communication) (1992) 29 Handy, C Understanding Organizations (4th edn) Penguin, London (1993) 30 Young, R 'Multi-disciplinary modelling of interface design issues' University of Cambridge Computer Laboratory, Seminar, March 1992 31 Sargent P PA 'Specialized, self-aware tools' in Workshop 6: Tool Integration, associated with Al and Design '92 Conference. Carnegie Mellon University, Pitt-sburgh, PA, 1992 32 Ashby, M F Materials Selection in Mechanical Design Pergamon Press, Oxford (1992) 33 Bentley, J Programming Pearls (reprint) Addison-Wesley, Reading, MA (1989)
Although this paper has concentrated on teams of designers the same
considerations affect the thoughts inside in the head of a lone designer.
He or she must integrate and trade-off multiple world-views of the same
problem. So techniques should he devised which enable single designers
to manipulate (play with) ideas using different disciplines at the same
time, yet preserve some consistency in the underlying artefact description.
Consistency must be maintained (at least partially) even though modifica-
tions may be made by distinct, largely incompatible methods. There are
clear implications for software research in multiple database update
techniques, product model architectures, goal-oriented programming,
user interface design3° and design environment tool integration9•31. We should be aware, however, that incompatibilities may not be noticed or
fixed for the same reasons that failure can occur in practice: it is in
principle impossible to enumerate all possible interactions.
When attempting to devise new design idioms for new types of artefact it
is useful to try to find a small set of different disciplines that give world-views which are as different as possible while retaining as great a
similarity as possible in the underlying representations. The former gives
the expressive power, the latter means that each view informs the other
about something relevant. This may involve the search for new repre-
sentations and, typically, this is indeed how new design techniques come
to light: the Gantt chart, the Ashby materials selection chart32 or the
HeapSort algorithm33 all rely on a novel representation which can be
manipulated from multiple, different viewpoints.
Design science or nonscience 401
Lastly, this view of design theories can serve as a test for snake-oil
salesmen. Design and the 'product development process' are now seen as
industrially very significant and companies are spending very large sums
of money investing in new organizational systems. This supports a large
industry of consultants with an attendant fringe of methodologists. Some
of the more mystical and charismatic of these design theorists can actually
do much good since much of the improvements to be gained only arise
from convincing large numbers of people. However, when some practical
bundles of techniques make a bid for formality and academic respectabil-
ity, then we need some criteria by which to classify and judge them. I
propose that any methodology or model which claims to be totally unified,
coherent, consistent and complete should be treated with caution.
5 Conclusions There can be no unified 'science of design' if an essence of design is the
rationalizing and assimilating of multiple incommensurate 'sciences'.
However, the absence of any such theory is itself useful because it
indicates which areas of design study require pragmatic support.
34 Sargent, P M 'A computa-tional view of creative steps' in
2nd International Round-table
Conference on Computational Models of Creative Design,
Heron Island, Queensland, Au-stralia, 1992
Chasing after an illusionary 'design science' is observed to be more a
characteristic of engineers seeking enhanced status as physical scientists
rather than emphasizing design creativity — which this paper among
others234 show to be qualitatively different.
6 Acknowledgments I wish to thank Eswaran Subrahmanian, Suresh Konda, Ira Monarch,
Rob Coyne and David Steier particularly for many intriguing and
argumentative weeks during my visits to EDRC, Carnegie Mellon Uni-
versity; to Ken Wallace for constructing an environment so conducive to
the study of design theory within the Engineering Department at Cam-
bridge University and to David Brown, David Ullman and Richard Coyne
for conversation while they were visiting Ken's design centre. Many thanks too to Jinae Lee, Lukas Ruecker, Steven Eppinger, Don Clausing,
Louis Bucciarelli, Warren Seering, Howie Shrobe, and Hal Abelson at
MIT and Mark Cutkowsky, Larry Leifer and Fred Lakin at Stanford for
so politely answering my naive questions and lastly I wish to thank all the
participants at John Gero's Second International Round-table Conference
on Computational Models of Creative Design for many hours of fascinat-
ing discussion and to the Key Centre for Engineering Design (Sydney