KNOWLEDGE SUPPORT SYSTEMS THE AMPLIFICATION FOR CONCEPTUAL OF CREATIVITY DESIGN: From: AAAI Technical Report SS-93-01. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved. ERNEST EDMONDS and LINDA CANDY LUTClll Research Centre Department of ComputerStudies LoughboroughUniversity of Technology Lougtdgorough Leicestershire UK Abstract. The paper is concerned with computer-based support for conceptual design and, in particular, with the support of creative design. The nature of conceptual design is briefly reviewed and the lack of effective computer support noted. Recent developments in computer-based KnowledgeSupport Systems, that offer interesting possibilities, are reviewed. The study of the early design of a clearly innovative product, the Lotus bicycle, is used to inform a discussion of the requirements for KnowledgeSupport Systems that can support conceptual design. Introduction Thus, the process is as muchone of problemdefinition or selection as problemsolution: it is essentially creative. The designer generates a set of scenarios or possible prototypical solutions at this stage. A potential problem, in relation to innovation, is that they tend to eliminate options early on in the process (Lawson, 1980). One important role that computer support could play is in keeping ideas open longer by providing "external" memoryand concept managementfacilities. As Faltings (1991) puts it, "Most research in intelligent CAD systems has focused on detail design, the adaptation of an initial concept to precise specifications. Little is known about the process of conceptual design, the transition between functional specification and concept of an artifact that achieves it.". Figure 1 gives an impression of the role that computers play, at the moment, in innovation. The boundaries between these activities are of prime concern. The question is, simply, "Can the CAEsupport begin sooner and can it, as a consequence, amplify creative design activities.". The problemis that the earliest stages are characterised by uncertainty, tentative decisions and informal representations of design ideas. It is suggested that creative thought often occurs at the conceptual stage and hence any support that could enhance the designer’s performanceat that point could be extremely valuable. A secondary, but significant, issue is that the transition from conceptual to detail design often involves the manual entry of information into a CADsystem, because the conceptual data had not been captured electronically. These are the issues that the paper will aclrlre~. The concept of KnowledgeSupport Systems, in which the end user manipulates machine representations of knowledgedirectly (Shaw &Gaines, 1988), has recently been developed in LUTCHIand applied to scientific exploration (Candy et al, 1993a; Edmondset al, 1993, O’Brien et al, 1992a, 1992b). This work has clearly demonstrated a potential for supporting creative work (Candy et al, 1993b). At the same time, other workers have beganto study the approachin the specific context of design (Fischer &Nakakoji, 1992; Hori et al, 1993). Background Perhaps the most important point to note about the nature of conceptual design is that madeby Visser (1992), as result of empiricalstudies of designers: ".. a problem solver is not ’given’ problems, but ’constructs’ them." Research into the design process has suggested that many of the requirements for support tools are similar to those observed in the LUTCHI study of scientific exploration above (Visser, 1992; Fischer, 1990). For example, Visser proposes that tools to assist in the management of memory load would positively support design. Her results relate closely to those of Candyet al (1993a) in the use of Knowledge Support Systems in the science domain. Or, as Hori et al (1993) put it: "In ... creative design, humansbuild new concepts out of nebulous mental worlds. They have some seeds of a new concept, incubate them and create a new concept..." (my emphasis). 145 Exploration and Innovation concept design detail design CAE support II 1 II II m m m m m m m mmmm+ Figure1: Thedesignprocess,the designer’sexploratoryandinnovative work and the support from computers:animpressionisticview based uponthe design conceptdevelopedby MikeBurrows The central problem (figure 2). Thedesignof this bicycleis the subject of the study reported. Computerswere not used at all in the The main concern of the paper is to consider the development of the original conceptwhilst, in contrast, feasibility of applyingthe recent researchdevelopments in Lotus operate a sophisticated CADinfrastructure. The research reported is investigating the feasibility of computingdescribed aboveto the provision of computer bridgingthis gap. supportfor conceptual design. Support for creativity in practice Apragmaticapproachwas used for the investigation in whichthe development of a successful innovativeproduct was investigated. The results of the investigation are related to the practical use of advancedCAD systemsin a closelyrelatedcontext. Lotus Engineering are firmly committed to the exploitation of knowledge-basedtechnologyto support design. They have used the ICADsystem, from ICAD EngineeringAutomation Ltd, for several years in order to developflexible andefficient supportfor the earlier design stages. Theyhave beenable to demonstratethat a number of clear benefits accrue, includingreducedrimescales,the In the 1992 Olympics,Chris Boardmanwonthe rust UK gold medal, in the 4000mcycling pursuit event. Hewas riding a revolutionarybicycle built by LotusEngineering 146 effective handling of complexityand the possibility of buildinggenericdesignmodulesthat are readily available for re-use (Gregory,1992). However, so far this workhas not involved the creative concept designer in direct computeruse. The Lotus ICADusers are specifically trained andfrequentlytackle particular problemsthat are morewell definedthan the onesthat onefinds at the very initial conceptionof a product. Nevertheless,that work includesinvestigationsinto supportfor early explorations andis at the leading edgecommercially,clearly pointing towards the increased use of knowledge-basedCADin conceptualdesign. Ourconcernis, primarily, the use of systemslike ICAD to act as a bridge betweensupport for early conceptualdesignand the standard CAD system. OlympicGamesin 1992has formedthe basis of the study. The results of the LUTCHI work have identified key design requirements for KnowledgeSupport Systems. Theresulting underlyinguser interface architecture is designedto be generallyapplicableandeasily tailored for specific domainapplications. Theworkis extendingour understanding of howknowledge-based systems can supportscienceandthe intention is to nowinvestigate the same approach in design. The cognitive findings are concernedwith the nature andaccessibility of knowledge to the domainexpert. The important distinguishing characteristic of this workis that it is concernedwith complexdomains in which the expert does not have complete knowledgebut, rather, where the approach enablesthemto extendand refine their ownunderstanding. Thisissue is central to the creative professionalworkthat is the subjectof our concern. 2. Personalandprofessionalgoals intertwined: It is, therefore, postulated, with Fischer and Nakakoji (1992), that the knowledgesupport systems approach could provide helpful supportfor conceptualdesign. The authorsare exploringthis hypothesisin order to drawout the consequentialrequirementsfor such a support system in a specific caseof recentinnovativedesign. Thedesignprocess seemsto havebeencharacterized by a numberof salient factors or events identified by Mike Burrows as beingsignificant. 1. Timeduration: Alengthy gestation period existed betweenthe formation of the initial thoughtsand the final realization of the revolutionary concept. This wasa normalexperiencefor Burrowsin his design work. Burrowsis a serious amateur cyclist and monocoque recumbent racer. Personal ambitions provided the stimulus to achieve advantage.Indeed, he expresses his maingoal as havingbeen"to go faster": i.e. to satisfy a personal requirementsfor his racing in cycling events. However,he is also a professionalengineerby trade. As well as motivation,skill and knowledge,he had access to machineryfor making, for example, hand madebicycle frames. Makingwas, most probably, an important factor in the designprocess. 3. Knowledge of the field: Apart from general design and engineering knowledge, Burrows has a keen interest in the history and developmentof the bicycle itself. Heis very awareof aspects of the process,such as the fixed ideas held about bicycles, and the required series of conceptualleaps and experimentswith prototypes that can eventually lead to breakthroughs. 4. Convention and creativity The Lotus bicycle concept There are, amongstcyclists, deeply held views of the bicycle concept.Verysmall changesin the angles used in a frameare verysignificant to the perceptionof "experts". It is not clear that this expert opinionis well foundedbut it is certainly strongly held. Thus,amongstexperts, the search space of possible designs is thought to be well understood. Thusa critical factor in creative bicycle designis the ability to moveoutside the existing search space; i.e. to breakwith the conventionalexpectationsand images. MikeBurrowsis collaboratingwith the authors so that a cognitivehistory of the development of the original conceptof the newbicyclecan be re-constructed.It is pertinentto placethis history in the contextof other empiricalresearchinto early designand, in particular, Knowledge Support Systems. For example, amongstthe conclusionsof the LUTCHI workreferenced abovewere that the userrequired:. to be able to take an holistic viewof the data at any time ¯ the ability to suspendjudgementon any matter at any time ¯ to be able to readily makeunplanned deviations. Thelast point waselaborated,in the contextof design, by Visser. Theseresults providea framework for the analysis of the early designhistory of the bicycle. Asa result, therefore, preliminaryrequirementsfor computer support are postulated. Thedevelopment of the bicycle concept fromBurrows’early explorations(c1979)to the time Lotusundertakingthe productionof the bicyclefor the Moving outside of the search spaceis closely connectedto the ability to respondand exploit the unexpected:for example,the combinedideas of a lady-frameshape with the BMX small size frame, triggered by hearing of a BMX used for racing. Thechanceavailability of carbonfibre material (1982) causedthe realisation that a small, but conventional,framecould be improvedby filling it in. A changein the British RTCregulations, allowing the use of rear disk wheelsin time- trials, generatednewdesign possibilities. 147 Figure 2: Chris Boardman breakingthe world 5000m record on the Lotusbicycle. These events led progressively, over years, to the realization that increasedspeedcouldresult fromthe use of a small filled in frame, set backfromthe usual position, using a very light weightmouldablematerial anddesigned without the need to shield the rear wheel for wind resistance. Theresulting bicycleconceptwasdevelopedin practice to the extent that a workingmachineexisted. The next step was to movethat initial concept into the processesof the Lotusdesignsystem. The use of CADin the earlier stages of design within Lotusis advanced andinteresting. Nevertheless,it did not influencethe initial bicycleconcept.Particular attention is paid to the exploitation of the ICAD software, which has successfully enabled Lotus to support design in a significantly more iterative manner. The parameters within whichthe postulated additional computersupport might take place may be placed in that context. Outstandingquestionsinclude, whatare the constraints on the formof data that couldusefullybe transferredinto the ¯ current CAD systems?To whatextent could the existing systemshaveoffered support to the bicycle’s conceptual design? Whatare the future requirements of CADfor early design? 148 Implications for intelligent support to conceptual design Therequirementsfor the user to be able to take an holistic view, to be able to suspend judgement and to make unplanneddeviations, that were mentionedabove, have beenstrongly supportedby the bicycle study. Additional points havealso emerged,however. The need to keep a multitude of design processes alive over long periods of time must be supported. Thegoals of each of these processesmightbe at a very high level, such as "go faster", andso maynot, in themselves,define a search space. Domainspecific knowledgeis clearly importantbut the creative step maywell rely uponan act that "breaks set" with conventional wisdom.In other words,creativity maybe associatedwith a re-definition of the problemspace. It is concludedthat supportsystemsfor conceptualdesign mustallowthe user:.. to take an holistic viewof the data at anytime ¯ to suspendjudgementon any matter at any time ¯ to be able to readily makeunplanned deviations. ¯ to return to old ideas andgoals ¯ to formulate,as well as solve, problems ¯ to re-formulatethe problemspace It is the process of interaction between a Knowledge SupportSystemand the designer that is the key element in enablingthe amplification of the designer’s creative achievements.This process mustbe allowedto occur in a waythat, as a minimum,conformsto the above design ideals. GREGORY, A. (1992). Separating Fact from Ficti¢ Manufacturing Breakthrough.1, 6. pp 329-333. HORI, K., SUGIMOTO,M. & OHSUGA,S. (199: Applicationof Articulation Aid to Design.In: Informati. Modellingand Knowledge Bases IV, Oaakkola,H. et al (ed IOSPress, to appear. LAWSON,B. (1980) How Designers ArchitecturalPress, London. References CANDY, L., O’BRIEN,S.M. & EDMONDS, E.A. (1993a). End User Manipulation of a Knowledge-BasedSystem: A Studyof an Expert’s Practice. International Journalof ManMachineStudies. 38, 1. pp 129-145. Think. The O’BRIEN, S.M. CANDY,L., EDMONDS, E.A., FOSTE T.J. & MCDAID,E.(1992a). End User Knowled ManipulationSystems:The SpeechKnowledge Interface. ] Proc. of 2OthAnnualComputerScience Conference,Agraw J.P., Kumar V. and Wallentine, V. (eds), TheACM Press. 359-366. CANDY, L., EDMONDS, E.A. & O’BRIEN,S.M. (19931)). Amplifying Creativity: The Role of End User Knowledge Manipulation Systems. In: Artificial Intelligence and O’BRIEN, S.M., CANDY, L., EDMONDS,E.A. Creativity. Darmell, T (ed). Studies in Cognitive Systems, FOSTER,T.J. (1992b) Knowledge Acquisition Refinement Using End-User Knowledge Manipulati. KluwerAcademic,Dordrecht,to appear. Systems. In: Applications of Artificial Intelligence Systems Conference, SPIE ProceedingsV EDMONDS,E.A., O’BRIEN, S.M., BAYLEY,T. & Knowledge-Based MCDAID, E. (1993). Constructing End-User Knowledge 1707,Orlando,Florida. Biswas,G. (ed). pp 25-36. ManipulationSystems.International Journal of Man.Machine Studies. 38, 1. pp 51-70. SHAW,M. L. J. & GAINES, B. R. (1988). KITTE KnowledgeInitiation and Transfer Tools for Experts a FALTINGS, B. (1991). Qualitative Models in Conceptual Novices. In: Knowledge Acquisition Tools for Expc Systems. Boose, J. & Gaines, B. (eds). AcademicPress. Design.In: Artificial Intelligencein Design’91, Gero,G. (ed). 309-338. Butterworth-Heinemann, pp 645-663. FISCHER,G. (1990) Communications Requirements for CooperativeProblemSolving Systems. International Journal of InformationSystems. 15, 1. pp 21-36. FISCHER,G. & NAKAKOJI, K. (1992). Beyond the Macho Approach of Artificial Intelligence: EmpowerHuman Designers - Do Not Replace Them. Knowledge-Based Systems. 5, 1. pp 15-30. 149 VISSER, W.(1992). Designers’ Activities Examined at Tlu Levels: Organization, Strategies and Problem-Solvi Processes. Knowledge-Based Systems. 5, 1. pp 92-104.