From: AAAI Technical Report WS-93-07. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved. Collaborative design in system development: Whatplace for design rationale? L. Candy and E. Edmonds LUTCHIResearch Centre Deparrtment of ComputerStudies Loughborough Leicestershire LEll 3TU United Kingdom 1. Introduction This paperreports a study of the use of DesignRationalemethodsin the co-operativedesign of a software system. Structured records of the design process aim to support the understandingof decisionstaken and therebyallowdesigners to give better informedreconsiderationto themat a later stage. This can be particularly importantduring maintenance.Methodsfor capturingthe complexityof design deliberations in order to producea DesignRationale (DR)in software design are still in the early stages of research (Carroll & Moran,1991). Theapproachesadopted are diverse: for example,fromconstructing representationsfor DRthat are applied by the designeras part of the reflective process(e.g. MacLean et ai, 1991)to relating DRto design practice by examiningconcrete problemsrather than abstract issues (e.g Lewis, Riemanand Bell, 1991). Nevertheless,evidencedrawnfromactual data gathered about the process of softwaredesignin practice remainslimited. 2. Design of a Software Artifact A softwaresystemcan be viewedas an ’artifacf that embodiesimplicit theoretical constructs that realize functional and operational requirements(Carroll and Campbell,1989). Throughout the designprocess, structures are chosenbecauseof their ability to achievethe intended functionality, and such decisionsmaybe evaluatedagainst various criteria. Duringthe design processthe descriptionsare modifiedand there is a clarification and refinementof intended functions and requirements.Theremaybe additional factors arising fromthe context of the project that affect the waythe processis carried out : for example,the needto keepsight of the general applicability of the design whilst meetingdomainspecific needsin a bespokeapplication or the constraints of the givenhardwareplatformand the softwarearchitecture and tools. In particular, it is necessaryto be awareof the influenceof the differing contributionsof the team membersarising from the application of experienceand skills to the common problem addressed. 3. A Study of Collaborative Design This paper drawsuponan exercise in collaborative design of a Knowledge SupportSystem (KSS)(Edmonds and Candy,1993) in whichan attempt wasmadeto elicit design rationale during the designprocessitself. It describesthe evolutionof the systemdesign as it emerged fromthe initial requirementsdocumentto the design specification of the newsystem. The 283 collaborative design process described concentrates uponthe designer/programmers, primarily, but also takes accountof the roles of the other members of the team, suchas the user and the independentev’,duator. Thedesign process wasdocumented from the conceptual phase to the implementationof the first prototypeconstructedin the intendeddelivery vehicle. Morerigorous proceduresthan used in an earlier version of the systemwereemployedand methodsfor recording the development processin all areas of the project activities discussed.It shouldbe notedthat this wasa major re-designexercise building uponthe evaluationresults froman earlier system(Candyet al, 1993). Thusan explicit attemptto incorporateDRduring the process of interpreting requirementsinto design wasmade.Theinitial primarygoal of building DRinto the design process of the re-design of the systemwasbased uponthe hypothesis that this wouldimprove the designresults by eliciting decisionsbaseduponjustifiable criteria, enablingthe sharingof design thinking and encouragingmore’rigorous’ attention to usability issues (as defined for examplein TheGuideto Usability, 1990and Smithand Mosier, 1986). Validation procedures wereapplied at all stages of the design activities havingbeendevisedand agreedby the team. Theseproceduresrequired the designers to justify options and decisions during the process and that designguidelines fromresearch literature wouldbe applied whereappropriate. Requirementswere to be based uponthe evidencefrom previous studies. Solutions and possible options available to meetrequirementswouldbe identified and documented.Theidea of a design databasefor this purposearose fromthe discussionsthat took place. Designchoices and decisions wererecordedand a detailed accountof howa solution wasto be handledgiven. In the conceptualdesign phase, the requirementswereinterpreted by the teamas set of design options. Theseoptions werethen validated against the requirementsstatementsby the user and the evaluator. Therequirementsdocumentand the design drawings, notes, documentsand prototypes(dynamicsimulations) represented the shared understandingof the team and as such, providetangible records of the evolution of the decision making.Themonitoringof their use providedevidenceabouthowdesign of this kind actually takes place in situations constrained by time and resource limitations. Thesefactors whichemergedabout the dynamicsof the’group thinkingare identified. Twomaintasks carried out by the designersare highlighted:firstly, the joint interpretation of initial generalrequirementsand, secondly,the selectionand use of the materialsand tools of design. 4. Conclusions Thedesignrationale ideas that informedthe initial intentions of this exercise werebasedupona modelof design that is rational, deductiveand hence, can be shared. Theevidencefromputting this into practice is that the designprocessesare disruptedby the requiredanalytic considerations and that the process is strongly synthetic, based uponmodelsof knownsystems and the handlingof newsystems, during construction. Therationale that did emergeis post hoe, and, in effect, a rationalisation of designdecisionsrather than a true accountof the unfoldingof the process. Theobservationof the design processand its outcomesled us to concludethat DRas applied in the formused, did not necessarily support the co-operative designprocessitself. Theseresults suggestthat it is difficult to applyDRas a supportto the design process whilst nevertheless being able to providean analysis of whathappened.DRas something opento scientific scrutiny as a psychologicalor sociologicalarea of study is possible. However, the ingredientsof designpractice itself are not necessarilysubject to the samekind of analysis. Questionsare posedas to the kind of knowledge that is required for the support of co-operativesoftware design in practice. Thus, whilst DesignRationalemightbe valuable as a post hoe methodfor communicating betweendifferent stages of the design, developmentand maintenanceof process, other methodsare required to facilitate communication and sharing during a given design phase. 284 5. References Carroll, J. M.and Campbell,R.L.(1989).Artifacts as psychologicaltheories: the case human-computer interaction. Behaviourand Information Technology,Vol.8, No.4, pp 247256. Carroll, J.M. and Moran,T.P. (1991) Introduction to this special issue on design rationale. Human-ComputerInteraction, Volume 6, Nos 3 & 4, pp. 197-200. Lawrence Erlbaum AssociatesInc. Lewis, C., Rieman, J. and Bell B. (1991) Problem-centred design for expressiveness and facility in a graphical programmingsystem. Human-Computer Interaction, Volume6, Nos 3 & 4, pp.319-355. LawrenceErlbaumAssociates Inc. MacLean,A., Young,R.M., Bellotti, V. and Moran, T.P. (1991) Questions, options, and criteria: elements of design space analysis. Human-Computer Interaction, Volume6, Nos 3 & 4, pp.319-355. LawrenceErlbaumAssociates Inc. Maclaren, R., Hornby,P., Robson,J., O’Brien, P., Clegg, C. and & Richardson, S.System design methods- the humandimension (1991) Project IED/4/1249Report. DTI/HMSO. J., Benyon,D., Davies, G., Keller, L. and Rogers Y. (1990) A Guide to Usability. The OpenUniversity in association with the Departmentof Tradeand lndustry. Preece, Smith, S.L and Mosier, J.N. (1986). Guidelines for designing user interface software, Mitre Report 10090 ESD-TR-86-278,Bedford Massachusetts, August. 285