HUMAN-COMPUTER INTERACTION: Psychology as a Science of … · 2017. 6. 17. · Human-computer interaction(HCI) has emerged relatively recently as a highly successful area of computer

0084-6570/97/0201-0061$08.00 61

Annu.Rev. Psychol. 1997.48:61–83Copyright© 1997by AnnualReviewsInc. All rightsreserved

HUMAN-COMPUTERINTERACTION: Psychology asaScienceof Design

John M.CarrollComputer ScienceDepartmentandCenter for Human-Computer Interaction, VirginiaTech, Blacksburg, Virginia24061-0106

KEY WORDS: usabil ity, design, usermodels,appliedpsychology, humanfactors

ABSTRACT

Human-computer interaction (HCI) study is theregion of intersectionbetweenpsychologyand thesocial sciences,on theonehand, and computerscienceandtechnology, on the other. HCI researchers analyze and design specific userinterface technologies (e.g. pointing devices). They study and improve theprocessesof technology development (e.g. taskanalysis, designrationale). Theydevelop and evaluate new applications of technology (e.g. word processors,digital libraries). Throughout the past two decades, HCI has progressivelyintegrated its scientif ic concerns with the engineering goal of improving theusabili ty of computersystemsand applications,which hasresultedin abody oftechnical knowledgeand methodology. HCI continuesto provideachallengingtest domain for applyinganddevelopingpsychological andsocial theory in thecontext of technology development and use.

CONTENTSTHE EMERGENCEOFUSABILITY ..................................................................................... 62

SoftwarePsychology ........................................................................................................... 63IterativeDevelopment......................................................................................................... 64UserModels........................................................................................................................ 65

USER-CENTEREDSYSTEM DEVELOPMENT................................................................... 67Usabili ty Engineering.......................................................................................................... 68DesignRationale ................................................................................................................. 72Cooperative Activity ............................................................................................................ 74

SYNTHESIZING A DISCIPLINE ........................................................................................... 77

THE EMERGENCE OF USABILITY

Human-computer interaction(HCI) hasemergedrelatively recentlyasahighlysuccessfulareaof computerscienceresearchanddevelopmentandof appliedpsychology.Someof thereasonsfor thissuccessareclearlytechnical.HCI hasevokedmanydifficult problemsandelegantsolutions in the recenthistory ofcomputing, e.g.in work on directmanipulation interfaces,userinterfaceman-agementsystems,task-orientedhelp andinstruction, andcomputer-supportedcollaborativework. Other reasonsfor its successare broadly cultural: Theprovince of HCI is the view the nonspecialist public has of computerandinformationtechnologyandits impacton their lives; HCI is thevisible partofcomputerscience.Themostrecentreasonsfor thesuccessof HCI arecommer-cial: As the underlyingtechnologiesof computingbecomecommodities, in-scribedon genericchips, the noncommodity valueof computerproductsre-sidesin applications and userinterfaces, thatis, in HCI.

HCI has evolved rapidly in the past two decadesas it has struggledtodevelopa scientific basisandutility in systemandsoftwaredevelopment. Inthis chapter,I review theprogressionof HCI towarda scienceof design.Mytouchstoneis Simon’s (1969)provocativebookTheSciencesof theArtificial.ThebookentirelypredatesHCI, andmanyof its specific characterizationsandclaimsaboutdesignareno longerauthoritative (seeEhn1988).Nevertheless,two of Simon’s themesecho through the history of HCI and still provideguidance inchartingits continuing development.

Early in the book,Simondiscussedtheapparentlycomplexpathof an anttraversinga beach,observingthat the structureof the ant’s behaviorderiveschiefly from thebeach;theantpursuesa relativelysimplegoalandaccommo-datesto whateverthebeachpresents.Theexternalworld, includingtechnologyhumanbeingscreate,shouldbeexpectedto play a powerfulrole in structuringhumanbehaviorandexperience.Late in the book, Simonsoundsthe secondtheme:the needfor a scienceof design,a researchparadigmand universitycurriculum directedat understanding, furthering, and disseminating designknowledge.He lamentedthe tendencyof engineeringdisciplines to adoptgoals and methodologies from the natural sciences,to their detrimentwithrespect todesign.

HCI is ascienceof design.It seeksto understandandsupporthumanbeingsinteractingwith andthroughtechnology.Much of thestructureof this interac-tion derivesfrom thetechnology,andmanyof theinterventions mustbemadethroughthedesignof technology.HCI is notmerelyappliedpsychology. It hasguidedanddevelopedthebasicscienceasmuchasit hastakendirectionfromit. It illustrates possibilitiesof psychologyas a designscience.

62 CARROLL

SoftwarePsychologyThe work that constitutes the historical foundationof HCI wascalled “soft-warepsychology” in the1970s(e.g.Shneiderman1980).Thegoalthenwastoestablish the utility of abehavioralapproachto understanding softwaredesign,programming,andthe useof interactivesystems,andto motivateandguidesystemdevelopersto considerthe characteristicsof humanbeings.Softwarepsychologyhad two distinct methodological axioms.The first assumedthevalidity of a receivedview of systemandsoftwaredevelopment,namelytheso-called waterfall model of top-down decomposition and discretely se-quencedstageswith well-specifiedhand-offs(e.g.Royce1970).The secondassumedtwo centralrolesfor psychologywithin this context:(a) to producegeneraldescriptions of humanbeingsinteractingwith systems andsoftware,which could be synthesizedas guidelines for developers;and (b) to directlyverify the usability of systemsandsoftwareas(or moretypically, after) theyweredeveloped.

Softwarepsychologyinaugurateda varietyof technicalprojectspertainingto what we now call the usability of systemsand software:assessingtherelativecomplexityof syntacticconstructionsin programminglanguages(e.g.Simeet al 1973),classifying people’s errorsin specifyingqueriesandproce-dures(Mill er 1974),describingthe utili ty of mnemonicvariablenamesandin-line programcomments(Weissman1974),andexplicatinghow flowchartsserveas a programming aid (Shneidermanet al 1977). This work inspiredmany industrialhuman-factorsgroupsto expandthe scopeof their responsi-biliti es to includesupportfor programming groupsandthe usability of soft-ware.

The basicaxiomsof softwarepsychology provedto be problematic.Thewaterfall idealization of designwork is infeasibleandineffective(e.g.Brooks1975/1995).It is only observedwhen enforced,and it is bestregardedas acrudemanagementtool for very large-scale,long-termprojects.As computerresearchanddevelopmentdiversifiedin the 1970sand1980s,small anddis-tributed personal work organizations becamemore commonplace. Productdevelopmentcycleswereoftencompressedto lessthana year.

Thetwo rolesassignedto softwarepsychologistswerealsoproblematicandresultedin a division of labor.Researchers,mainly in universities,developedgeneraldescriptions of usersandframedthemasgeneralguidelines.Human-factorsspecialistsin industry tried to apply theseguidelinesin specificpro-jects. This division did not work particularly well. From the standpoint ofpracticalgoals,theresearchof this periodtendedto focuson unrepresentativesituations (e.g.undergraduatesstandingin for programmers,50-lineprogramsstandingin for businesssystems, andteletypesstandingin for displaytubes).To obtainstatistically stableresults,researchersoftencreatedoutrageouscon-

HUMAN-COMPUTER INTERACTION 63

trasts(organizedversusdisorganizedmenus,structuredversusscrambledpro-grams).The researcherssometimesunderstoodlittle aboutthe usersof theirguidelinesandprofferedsuperfluousadvice.

Psychologistsandothersplayingthehuman-factorsspecialistrole in indus-try werefrustratedtrying to useandencourageuseof theseguidelines.Theywere also frustratedin their other role of verifying the usability of finishedsystems,becausethe researchtools they had (formal experimentsaimedatspecific differences amongalternatives)were toocostly andtoo uninformativeto allow them to serveas anything more than gatekeepers.Human-factorsspecialistswereoftenseenasimposingbureaucraticobstaclesblockingheroicdevelopers.

Theoriginsof HCI in softwarepsychologyposedtwo centralproblemsforthe fieldin the 1980s.One problemwas to betterdescribe designanddevelop-mentwork and to understandhow it could be supported.The otherproblemwas to betterspecify the role that psychology, in particular,and social andbehavioralscience, more broadly,shouldplay inHCI.

IterativeDevelopmentStartingin the 1970s,empiricalstudiesof the designprocessbeganto expli-catethe difficulties of the waterfall model.Designwork is frequentlypiece-meal,concrete,anditerative.Designersmaywork on a singlerequirementat atime,embodyit in a scenarioof userinteractionto understandit, reasonaboutand developa partial solution to addressit, and then test the partial solu-tion—all quite tentatively—beforemovingon to considerotherrequirements.During this process,they sometimesradically reformulatethe fundamentalgoals and constraints ofthe problem. Ratherthan a chaos,it is a highlyinvoluted,highly structuredprocessof problemdiscoveryandclarification inthecontextof unboundedcomplexity (e.g.Carrolletal 1979,Curtisetal 1988,Malhotraet al 1980).

The leadingidea is that designersoften needto do designto adequatelyunderstanddesign problems.One prominent empirical case was Brooks’s(1975/1995)analysisof the developmentof the IBM 360 OperatingSystem,oneof the largestandmostscrupulously plannedsoftwaredesignprojectsofits era. Brooks, the project manager,concludedthat system and softwaredesignersshouldalways“plan to throw oneaway” (pp. 116–23).This wasastriking lessonto draw and carriedwith it many implications.For example,formal and comprehensive planningand specificationaids (suchas detailedflowcharts) will have limi ted utility in supporting such an iterative designprocess.

This reformulationof designcreatedanopeningfor new ideas.Noteworthyinspirationcamefrom the work of the greatindustrial designerHenry Drey-fuss,who hadpioneeredanempiricalapproachin the1940s(Dreyfuss1955).

64 CARROLL

Dreyfuss’s approach institutionalizesanaccommodationto designers’ propen-sity for concrete,incrementalreasoningandtesting. It incorporatesfour centralideas:(a) early prototyping with (b) the involvementof real users,(c) intro-duction of new functions through familiar “survival forms,” and (d) manycyclesof designiteration. Dreyfusspushedbeyondthe designer’s needforprototyping and iteration as a meansof clarifying the designproblem(alsoemphasizedby Brooks1975/1995) to the user’s knowledge,experience,andinvolvementto constraindesignsolutions.

A typical exampleis Dreyfuss’s designwork on airplaneinteriors forLockheed.Dreyfuss sent two associatesback and forth acrossthe UnitedStateson commercialflights to inventory passengerexperiences.They foundthat passengerswereoften baffledby designdetailslike watertapsthat wereunnecessarilynovel. They were impressedthat peoplewanted to think ofairplaneseatsas armchairsand not as “devices.” Initial designswereproto-typedin a Manhattanwarehouse,anda full flight of “passengers” was hiredtooccupythe mock-upfor 10 hours:to storecarry-onluggage,to eatmeals,touselavatories.Thesetestsconcretizedrequirementsfor seating,storagespace,lavatory-doorlatches,andsoforth, andpermittedlow-cost,iterativereworkingof the originaldesigns.

In the1980s,theinevitability of anempiricalorientationtowardsystemandsoftware designrapidlyevolved from asomewhat revolutionaryperspective totheestablishmentview. This developmentprovidedearlyandcritical motiva-tion anddirectionfor researchonuserinterfacemanagementsystemsto enableprototyping (Tanner& Buxton 1985),it encourageduserparticipation in de-sign (Gould & Boies 1983, Nygaard1979, Pava1983), it emphasizeduserinterfacemetaphorsfor presentingnovel functionality throughfamiliar con-cepts(Carroll & Thomas1982, Smith et al1982), and it made“rapid prototyp-ing” a standardsystemdevelopmentmethodology (Wasserman& Shewmake1982,1985).

Iterativedevelopmentshifted the focusof usability evaluationfrom sum-mativeto formative(Scriven1967).Formalexperimentsare fine fordetermin-ing whichof two designs arebetteronasetof apriori dimensions,but theyareneitherflexible nor rich enoughto guidecontinualredesign.“Thinking aloud”had beenpioneeredby deGroot(1965) and Newell & Simon (1972) in thestudy of puzzlesand games,and it had beenshown to vividly illuminatestrategicthought (Ericsson& Simon 1985). In the 1980s, thinking aloudbecamethecentralempirical,formativeevaluationmethodin HCI (e.g.Macket al 1983,Wright & Converse1992).

UserModelsThe secondproblemareabequeathedto HCI by softwarepsychologywastocharacterizea robustsciencebasethatcouldunderwritesystemdevelopment.


The cornerstonein this effort was the GOMS project of Card et al (1983).GOMS (Goals,Operators,Methods,and Selectionrules) provideda frame-work for systematically analyzingthe goals,methods, andactionsthat com-prise routine human-computer interactions.This was anadvanceon priorhuman-factors modeling,which did not addressthecognitivestructuresunder-lying manifestbehavior.Indeed,it wasan advanceon the cognitivepsychol-ogy of the time: It explicitly integratedmanycomponentsof skilled perform-anceto producepredictionsabout real tasks.At first, GOMS appearedtopromisea comprehensiveparadigmfor scientifically groundedHCI design(Newell & Card1985,but cf Carroll & Campbell1986).Theactualimpactofthesemodelshasbeenmorenarrow,althoughtheyhavebeenusefullyappliedin domainswhereuserperformanceefficiencyis thecritical usabilityvariable[e.g. modeling telephoneoperator scripts(Gray et al1992)].

Onesalientlimitationof GOMS modelsis that theydo not describelearn-ing, yet the problemsof new usersbeginning with computersystemswasperhapsthe technicalfocusof the1980s.Thelearningproblemwasconceivedasamatterof coordinatingknowledgein two domains: thetaskandthedevice.Theuserlearnsmappingsbetweengoalsin thetaskdomainandactionsin thedevicedomain,betweeneventsin the devicedomainandeffectsin the taskdomain(Moran 1983,Payneet al 1990).Much of this discussion focusedonthe “consistency”of mappings from commandsand other userinterfaceac-tions to application functions(e.g.Payne& Green1989).For example,learn-ing a commandnamed“pull” is facilitated by a complementary commandnamed “push” (versus, say, one named “grab” ); commands like “editdata_file” and “remove data_file” are mutually facilitative (seealso Esper1925).Consistencyturnedout to behighly intentional (Carroll 1985),signifi-cantly idiosyncratic (Furnaset al 1983),and evenquestionableas a generaldesignobjective(Grudin 1989).Nevertheless,promoting consistencyin userinterfaceandapplicationdesignremainsa prominentpracticalissue(Nielsen1989),and these modelsarethefoundationfor ourunderstandingof it.

The role of prior knowledgein learning to use computersystemswasanotherfocusof usermodeling work. It waswidely notedthatnewuserstriedto understandcomputersas analogicalextensionsof familiar activities andobjects(e.g.Douglas& Moran1983,Macketal 1983).Thisobservationledtoa variety of user interface“metaphors,”suchas the now pervasivedesktopinterface,anda paradigmfor userinterfacecontrol anddisplay called“directmanipulation,” whichwasgraduallyarticulatedthroughthe1980s(Hutchinsetal 1986,Shneiderman1983).It alsoled to theoriesof userinterfacemetaphor(Carroll et al 1988).

A secondlimitationof theearlyGOMS-stylecognitive modelsis that theydo notaddress nonroutine problemsolvinganderror.Studies showed that new

66 CARROLL

usersspenta third to a half of their time in error recovery(e.g. Mack et al1983).More significantly, thesestudies showed that often thedispositions thatmake peoplegood sensemakersare also causesfor characteristiclearnerproblems(Carroll 1990):Peoplewantto learnby doing,but this inclinesthemto opportunistically jump around in sometimes brittle learningsequences.They want to reasonthings out and constructtheir own understandings,buttheyarenot alwaysplanful,andtheyoftendrawincorrectinferences.Theytryto engageandextend prior knowledge andskill, which canlead tointerferenceor overgeneralization.They try to learnthrougherrordiagnosisandrecovery,but errorscanbe subtle,cantangle,andcanbecomeintractableobstaclestocomprehensionandto motivation.

This work entrainedthe conceptionof the “active user,” improvising, hy-pothesizing,trying to makesenseof a very complexenvironment. It led toemphaseson designing for learning-by-doing and for error (Carroll 1990,Lewis & Norman1986).For example,Carroll & Carrithers(1984)createda“training wheels” interfacein which attempteddeparturesfrom a correctac-tion pathareblocked.Users are informedthat theactionis notappropriateandarepermittedto try againwithout penalty.This kind of designsupportssense-makingbut not necessarilyefficient performance.The activeuserwassome-what of an alternativeto theGOMS conceptionof theuserasan informationprocessor.

USER-CENTEREDSYSTEM DEVELOPMENT

In theearlydaysof HCI, thenotionthatcomputersystemsandsoftwareshouldbe designedanddevelopedwith explicit considerationof the needs,abilities,andpreferencesof their ultimateuserswasnot takenseriously.Most writingsaboutcomputingfrom the mid-1970s are stunningly dismissive of usabilityandratherpatronizing of users. Afteronly a decade,thecomputerindustry andthedisciplineof computerscienceweretransformed.Thecasehadbeenmadefor a user-centeredsystemdevelopmentprocess,a processin which usabilitywasa primarygoal.Peoplebeganto distinguishsharplybetweentechnology-drivenexploratorydevelopment,which is now oftenaccompaniedby explicitdisclaimersaboutusability, and“real” systemdevelopment,in which empiri-cally verifiedusability is the finalarbiter.

With the adventof the 1990s,HCI researchhad becomerelatively wellintegratedin computerscience.A 1988Associationfor Computing Machinery(ACM) taskforce listedHCI asoneof ninecoreareasof the computersciencediscipline (Denninget al 1989).A joint curriculumtaskforceof theACM andthe Instituteof ElectricalandElectronicEngineers(IEEE) recommendedtheinclusion of HCI as a commonrequirementin computerscienceprograms(Tucker& Turner1991).HCI was includedasone of tenmajor sectionsof the


first Handbookof ComputerScienceand Engineering (Tucker1997). In the1990s, computer sciencestudents and the corporations thathire them aredemandingHCI coursesin universitycurricula.Severalmajor computersci-encedepartmentshavedesignatedHCI asa researchfocus.Two comprehen-sive undergraduatetextshave appeared (Dix etal 1993,Preeceet al1994).

In industry, HCIpractitionershavebecomewell integratedin systemdevel-opment.HCI specialistshave moved into a great variety of roles beyondhuman-factorsassurance.Theyhavebeenroutinely includedin customer/userinteractionsto understandthe needfor new products,productplanning,andspecification;the developmentand evaluationof prototypes; the designofdocumentation andtraining;andinstallationandusersupport.Therehasbeenanobvioustrendfor HCI specialiststo bepromotedinto projectmanagement.None of thesetrendsimpugns the psychologicalnatureof HCI; rather,theyindicatethat it hasbeena practicalsuccess.In addition,they remindus thatsuccessfulappliedwork involvesmorethanmerelyapplyinglab-basedtheo-ries and results,a theme thatcontinuesto be articulatedin thecurrent era.

HCI remainsan emergingareain computerscience.As an appliedareaofsocialandbehavioralscience,it continuesto broaden.Theissuesraisedin theearlydaysof softwarepsychologyarestill beingresolvedandelaborated:Howcaniterativedevelopmentbe supportedandimproved?How shouldwe man-ageresourcesin iterativedevelopmentto optimize costbenefit?How shouldweexpandthescopeandrichnessof theearlycognitiveusermodels?How canwe cumulateanddeveloptechnicallessonslearnedin iterativedevelopment?Whatrole canHCI play in broadening,grounding,andinvigorating thedevel-opmentof thesocialand behavioralscience base?

Theseissueshavebeenpursuedthroughrefinementsin theoriginal notionsof iterative developmentand user models, discussedbelow in the sectionsUsability Engineering,DesignRationale,and CooperativeActivity.

UsabilityEngineeringIterativedevelopmentis consistentwith the real natureof design.It empha-sizesthe discoveryof new goals,the role of prototyping andevaluation,andthe importanceof involving diversestakeholders,including users.But whatmakesiterativedevelopment more thanmerelywell-intentionedtrial ander-ror?

“Usability engineering”becamethebannerunderwhich diversemethodo-logical endeavorswerecarriedout in the1980s(Nielsen1993,Whitesideet al1988). There were three key notions: First,it was proposedthat iterativedevelopmentbe managedaccordingto explicit and measurableobjectives,called “usability specifications”(Carroll & Rosson 1985; seealso Bennett1984,Butler1985).Thus,in designing a word-processing program, onewoulditeratively designandredesign, prototypeandevaluate, include realsecretaries

68 CARROLL

in the designdeliberations, but alsomakeexplicit commitmentsto preciselyoperationalizedgoalssuchas “two thirds of userswill be able to prepareatwo-pagebusinessletter in lessthanten minuteswith fewer thanthreeerrorsafter 30 minutestraining.” Usability specifications arenow standardpracticein HCI development.

The secondkey notion in usability engineeringwasa call to broadentheempirical scopeof design.A variety of approachesand techniquesfor userparticipationwere developed,many of which emphasized“low-tech,” coop-erativeactivitiesto facilitatecollaborationbetweenusers,who bring expertiseon the work situation, and developers,who bring expertiseon technology(Greenbaum& Kyng 1991,Kuhn & Muller 1993,Schuler& Namioka1993).Thisapproachwentbeyondprior formulationsof userinvolvementby describ-ing broaderandmoreactiveroles for users.In “participatorydesign,”astheapproachis now known,usersareinvolved in settingdesigngoalsandplan-ning prototypes,insteadof becominginvolved only after initial prototypesexist.

Field-studyapproachesto characterizingtheusers’ realneedsandcircum-stancesalso becameprominent(Whiteside & Wixon 1987). This emphasiscameto beknown as “contextual design” (Wixonetal 1990).It overturnedthelaboratorybias inheritedby HCI from cognitivepsychology by arguingthatlaboratorysituationsareoftennotcontrolledsimulacrumsof realsituationsbutarebetterregardedasdistinctbut eccentricsituations in their own right. Notethatfield-studyapproachesare notequivalentto userparticipation:Oftenfieldstudiesbring to light facts in the backgroundof the contextof use,circum-stancesof which the usersthemselvesareunaware.Conversely,field studiescannotreveal the perspectivesand insights usersbring to the developmentprocess as designers.

In theearly1990s,contextualdesignconvergedwith a lineof ethnographicresearchthat had producededifying descriptionsof usecontexts,but whichhad generallyescheweddirect involvement in design(e.g. Suchman1987).This combination, sometimes called “ethnographically-informed design”(Bentleyetal 1992),hasbecomequiteprominent. Like participatorydesign,itpushesa specific methodological commitment a step further by advocatingvery detailedobservationof behaviorin real situations. In practice,ethnog-raphicallyinformeddesignis frequentlyaimedat characterizingunarticulatedpowerrelations,organizationalassumptions, andpracticalknow-howthator-ganizethe workplace(Blomberg 1995). A complementarystreamof work,sometimes called“conversationanalysis,” addressesitself to finer-grainedbehavioralstructure(Greatbatchet al 1995). The commitment to revelatoryinterpretationof situatedbehaviorsometimes engendersa behaviorismthatgivestoo little weightto what peopleexperience and report(Nardi 1995b).


The third key notion in usability engineeringwascosteffectiveness.It isexpensiveto carry out manycyclesof prototyping, evaluation,andredesign.Developersneedto employefficient methodsthroughoutandto know whenthey have reacheddiminishing returns.In the 1980s,much HCI work wasdirectedat creatingbetterprototyping toolsandenvironments.An earlythemein this work was the separation of user interfacesoftwarefrom applicationsoftware(i.e.system functionality) to modularizeredesignof theuserinterfacein user interfacemanagementsystems(e.g. Tanner& Buxton 1985). Sub-sequentwork emphasizedthe coordination of userinterfaceand applicationsoftwaredevelopmentto facilitate the creationof task-transparentuserinter-faces(Sukaviriya et al 1993, Szekelyet al 1993). Concernsabout easeofimplementation motivateda family of prototyping tools basedon thepremisethat user interface software could be directly created “by demonstration”(Cypher etal 1993).

Theissueof costeffectivenessalsoguidedmethodologicalwork on usabil-ity evaluation.Indeed,the earliest refinementof the GOMS model was akeystroke-countingapproximation (Cardetal 1983).Frequently,methodologi-cal shortcutswere groundedin the exigenciesof systemdevelopment. Forexample,thinking aloudprotocolshadbecomea standardempiricalmethod,but theyweregenerallynot analyzedat the level of detail typical in cognitivepsychology.Often,they weremerelygleanedfor critical incidents: episodesofuse in which something goesunexpectedlywell or badly (Flanagan1954,Shattuck &Woods1994).

Methodological work on usabilityevaluationsoughtto developsystematictechniquesto ensurecost effectiveness(Bias & Mayhew 1994).Good et al(1986)developed“impact analysis,”in which a variety of userperformancefactorsarerepeatedlymeasuredthroughoutthe developmentprocessto iden-tify thosefactorsthatoffer thegreatestpotentialimpactwith respectto givenusabilityspecifications. Willigesetal (1993)developedastatisticalmeta-strat-egyfor reducingthefactorialspaceof experimentalstudiesthroughsequentialestimatesof relationships amonglargesets of independentvariables.

Thecostof userstudiesbecamea centralissue.Many proposalswere madefor “heuristic” evaluation,checklistandscript-orientedapproachesto supple-ment or evenreplacedirect user testing(Nielsen& Mack 1994, Nielsen&Mollich 1990).Otherwork examinedthe empiricalparametersfor userstud-ies. Nielsen (1994) found that HCI practitionerswere able to run only anaverageof 9 subjectsin laboratoryusability evaluations. Direct cost-benefitstudiesdeterminedthatonly 4–5experimentalsubjects found 80% ofusabilityproblems(Nielsen1994,Virzi 1992).[Indeed,therehavebeenrecentdiscus-sionsof “quick and dirty”ethnography(Hugheset al 1994).]

70 CARROLL

The cost of analytical techniquesalso becamean issue.The assumptionmadein GOMS that actionsare independentand additive renderedGOMSincapableof describingtheco-articulationof concurrentactivities.Theselimi -tations were acknowledgedby Card et al (1983); for all theselimi tations,GOMS was also difficult to apply. John (1990) extendedGOMS for tasksinvolving parallel activities by incorporating the conceptof critical path.Kieras(1988)developeda simplified andexplicit methodfor building GOMSanalysesthroughtop-down,breadth-firstexpansionof goalsinto methodstothe level desired. Other simplified cognitive approaches were proposed.“Claims analysis” was suggestedto integratedesignrationale(seesectionbelow) and user modeling in analytic evaluation(Carroll & Rosson 1991).“Cognitive walkthrough” was proposedas heuristicevaluationgroundedinsimplified cognitive theory(Polsonet al 1992).Comparativecost-benefitre-searchis neededand is beginningto appear(Jeffrieset al 1991,Karat et al1992,Nielsen& Phillips 1993).

The term “usability engineering”connotesa practicethat is broaderandmore systematic than is the case.HCI is largely a first-generationfield inwhich variouspioneerstendto follow the practicesthey helpedto originate,not alwaysthe moreeclecticbestpractices.The pioneerzeitgeistof the fieldneedsto be smoothedandconsolidatedwith increasedemphasison synthesisandintegration. For example,although it hasbeenshownthatdifferentevalu-ation methodsidentify different kinds of usability problems(Jeffries et al1991),thereareno schemesfor integratingdifferent typesof evaluationdata(e.g.quantitative performancedataandqualitative protocoldata)or for inte-grating evaluationdatacollectedat different points in the iterative develop-mentprocess.Practitionersdo not alwaysdifferentiateamongthe variety ofevaluationgoalstheymustmanage (Carroll& Rosson1995).

A variety of suchintegrationquestionsremainsat the centerof usabilityengineering.How canusermodelsof themosttechnicalsort(GOMS)beusedin participatorydesignprocesses?Participatorydesignandusabilityengineer-ing aresometimesconsidered alternative technical approaches(e.g.Monk etal1993).Canbroaderuserparticipationimprove the costeffectiveness of usabil-ity engineering?Canusershelp to critique or evenwrite usability specifica-tions, heuristicevaluationchecklists,walk-throughscenarios,anddesignra-tionales?Cost-benefitanalysisoften views benefitfairly narrowly; in thelong-term,part of the benefit could be the developmentof usermodelsanddesignrationale from the specific resultsof usability evaluations. Can wedevelopmeansof estimating thesemoreprofoundbenefits?Canthedevelop-ment of usermodelsand designrationalebe (partially) automated as a by-product of usability engineeringactivities?How directly can ethnographicanalysisguidedesignwork?


DesignRationaleA computersystemdoesnot itself elucidatethe motivationsthat initiated itsdesign; the user requirementsit was intendedto address;the discussions,debates,andnegotiationsthat determinedits organization; the reasonsfor itsparticularfeatures;thereasonsagainstfeaturesit doesnot have;theweighingof trade-offs;andso forth. Suchinformationcanbe critical to the variety ofstakeholders in adesign process: customers,users,servicers, andmarketers, aswell asdesignerswhowantto build uponthesystemandtheideasit embodies.This information comprisesthe design rationale ofthe system(Moran &Carroll 1996).

Approachesto designrationalecanbe divided into thosethat describethedesignsolution and thosethat describethe designprocess.The former ap-proachseeksto position a given design in a larger context of issuesandalternativedesigns. Forexample,MacLeanetal (1991)developedanapproachthatexplicatesthedesign“space”surroundinga givendesignby enumeratingthe issuesdesignersidentified, the options they consideredin respondingtotheseissues,andthe criteria they usedin weighingthe options.Thus,a win-dow interfaceincorporatinga particulartechniquefor scrolling window con-tentscan be locatedin a designspace of otherscrollingtechniques.

Carroll & Rosson(1991)developedanapproachthatconsiderssystems tobe “embodied” social and behavioralclaims about the needs,abilities, andactivities of their users. Their approachseeks toarticulate thesocial andbehavioraltheoryimplicit in a design.Thus,a programmingenvironmentcanbeseenasembodyinga rangeof claimsaboutwhatprogrammersknow,whattheydo,andwhattheyexperience,andaboutthenatureof programmingtasksand thecontextswithin which these tasksare carried out.

Theseapproachesmakeit easyto succinctlysummarizethecritical usabil-ity trade-offsof a particularusagesituation, and to link theseissuescloselywith specificfeaturesof the computerartifact in use.For example,includinganimateddemonstrations as a self-instructionresourcein aprogrammingenvi-ronmentmayintrinsically motivatelearnersandsupportlearningaboutappro-priategoals,but the demonstrations may alsoplacelearnersin a passiveroleand suggestgoals that are too difficult or nonproductive(e.g. the goal ofalteringthe demonstration itself).

Many process-orientedapproachesto design rationaleare basedon theissue-basedinformation system(IBIS) developedby Rittel (e.g.Rittel & We-ber1973).In thisapproach,designdeliberations aredescribedaccordingto theissuesthat arise during the designprocess,the various positions raisedinresponseto the issues,andtheargumentsfor andagainsteachposition. Con-klin & Yakemovic (1991) showed howsuch a designrationale could becapturedandusedin a largecommercialproject,thoughtheyalsoemphasized

72 CARROLL

the considerablework involved in coding this rationale.Other process-ori-entedapproachesto rationaleemphasizecapturing less codedinformation,suchas sketchesand designnotes,or videotapedself-disclosuresby systemdevelopers (e.g.Carroll et al 1994).

The emergentnatureof the designprocessmakesdesignrationalebothimportantand difficult.It is a toolfor managingthe complexity of aprocess inwhich everything,including the problemdefinition, constantly changes.De-sign rationalecantrack the considerationof varioussubproblems, trade-offs,andpartial solutions,the reasonsfor changesanddecisions,andthestatusofassumptionsandissues.It canhelpdesignteamsstructuretheir problem-solv-ing efforts andavoid thrashing.But how muchof the designprocesscanbeusefully represented,and who will do the work of creatingand maintainingthis designrationale?Oftendesignrationaleis mostusefulto thosewho werenot even involved with the project when the rationalewas created(Grudin1994).Shouldeverydeveloperhaveananalystnotingeveryhypotheticalandeverytrain of thought,videotaping everychanceencounterin thehall, collect-ing every sketch andevery note?

Currentwork on designrationaleis concernedwith assessingandsupport-ing its efficacy.Empirical studiesof designersandproject teamsareinvesti-gatinghow designrationaletechniquescan belearned and used(e.g. Bucking-hamShum1996).Otherwork is experimenting with softwaretools to supportthecreationof andaccessto rationales(Conklin & Begeman1988,Fischeretal 1991,Rosson& Carroll 1995).

Design rationale integratesadvancesin iterative developmentand usermodels.Making the processand outcomesof designmore explicit allowsiterativedevelopmentto bemoresystematicandmoremanageable.However,it alsocreatesan explicit designrepresentation,a “theory” of theartifactandits usetestedby formativeandsummative evaluationthroughoutthe iterativedevelopmentprocess.This theory is not a classicuser model; it describesspecificsituationsof use,not purportedlygeneralinformation processesandcognitivestructures.However,this specificitymakesit morepowerfulwithintheimmediatedesigncontext,andschemeshavebeenproposedfor generaliz-ing such situatedtheories(e.g. Carroll et al 1992). This integrative role ofdesign rationale exemplifies what Scriven (1967) called “mediated evalu-ation,” an approach inwhich design analysis ofimplicit goals andpositions oninherenttrade-offsamonggoals is usedto guide designevaluationof userperformance and experience.

Designrationalecanbea languagefor stakeholdersin thedesign,but thesedifferent stakeholdersoften speakdifferent disciplinary languages,aremoti-vatedby different values,andseedifferent technicalissueswhen looking atthe“same”designproblem.Caneveryoneusethesamedesignrationaleinfor-


mation?Will gatheringandcodifyingrationalealterthedesignprocess?Will itinterfere?Who should have accessto the designrationalethat is created?Therearepotentialconflictsbetweenprivacyrightsandaccessrightsof stake-holders.For example,it is usefulfor developersto becandidabouttrade-offsthey managed,butare users entitledto thisinformation?

CooperativeActivityA trendto thesocialhasgraduallydevelopedin HCI over thepast15 years,adevelopmentthat hasmarkedlyaccelerated,deepened,anddiversified in thepastfive (Hiltz & Turoff 1978/1993).This recenttrendis actuallyanexusof atleastfour logically independentdevelopments.First, therewasa clearconsen-sus by 1990 that the cognitive modeling approachhad failed to provide acomprehensiveparadigm.Second,manyvoicessuggesteda moresocially ororganizationallyorientedapproachwasrequiredto supplementor replacethecognitiveparadigm.Third, thegrowingtechnicalprominenceof HCI attracteda sociopolitical critiqueof usability asa potentialapologyfor de-skilling andotherunpleasantaspectsof industrialrestructuring.Fourth,new technologiesfor communication and collaborativeactivity swept through the computingindustryandraised significantly new challengesand opportunities for HCI.

It was somewhatan accidentof historythattheoriginal foundationsof HCIaresostrongly in cognitivepsychology. Theresearchagendaopenedupby thepioneersof softwarepsychology,togetherwith the new opportunities occa-sionedby personaland distributed computing, attracteda core of cognitivepsychologists,who attractedevenmore.HCI emergedfrom whatin retrospectseemsto havebeenthe evangelicalheydayof the cognitive paradigm.Theinitial euphoriawith modelsof idealperformancetime wassoondisplacedbyfrustrationwith the limitationsof these models,particularlyregardinglearningin context, error and error recovery,preference,fatigue, work context andcollaboration,andindividualdifferences.In 1990and1991themajor interna-tional conferencesin HCI featuredpanelsaddressedto the failure of theory(Monk et al 1990,Sutcliffe et al 1991).Recentreformulations of the role ofcognitive usermodeling position it moreeclecticallywithin usabilityengineer-ing (Nielsen 1993,Preeceetal 1994).

In the 1990s,new voices enteredthe HCI discussion,urging a strongersocialandcontextualorientation.Anthropologistsandsociologists joinedwhathadbeenlargelya cognitivepsychologyproject(Bowkeret al 1995,Thomas1995).Europeanperspectiveson technologyandwork beganto penetratethediscourseof whathadbeenasomewhatinsularAmericanandBritish endeavor(Bjerknes et al 1987, Greenbaum & Kyng 1991). Conceptsfrom activitytheory,work psychology, andthe labormovementbecamewell known(Bød-ker 1991,Carroll 1991,Ehn1988,Nardi 1995a).More researchattention wasdirectedatunderstanding situatedanddistributedcognition(Carroll& Rosson

74 CARROLL

1992,Norman1991,Suchman1995,Zuboff 1988).As partof a largerparadig-maticrestructuringof socialandbehavioralscience,traditionsthathadsoughtto studyindividualsin isolation from their contexts,andsocialphenomenainisolationfrom individuals,weredeclining (Knorr-Cetina1981).

By far the most theoreticallyrich alternateparadigmis activity theory,derivedfrom thework of Vygotsky (Nardi 1995a,Wertsch1985).Theobjectof descriptionin this approachis an“activity system,” the ensemble oftechno-logical factorswith socialfactorsandof individual attitudes,experiences,andactionswith community practices,traditions,andvalues.Activity theoryem-phasizesthat theseensemblesare inherently contingentand changing,thathumanactivities aremediatedand transformedby humancreations,suchastechnologies,andthatpeoplemakethemselvesthroughtheir useof tools.Thetensionsandcontradictionswithin anactivity systemat a givenpoint definea“zone of proximal development” within which people can effect changes(Engestrom1993).Activity theory shifts attentionfrom characterizingstaticandindividual competenciesto characterizinghow peoplecannegotiatewiththe socialandtechnologicalenvironment to solveproblemsandlearn,whichsubsumesmanyof theissuesof situatedanddistributedcognition.

Kuutti & Arvonen(1992)showedhow thedesignof a medicalinformationsystemreintegratedthe activity systemof healthprofessionalsin a medicalcenter,a hospital, andthroughoutthe largercommunity by enablinga sharedconceptof total patientcare,supportedby electronictools andnew practices(see alsoBødker 1991). Activity theory subsumesmany ofthe methodologicalissues about participatory design. Theusercommunity needsto bereceptivetochange;thetensionsandinconsistenciesin thecurrentactivity systemmustbemadevisible andquestioned,that is, broughtinto thezoneof proximal devel-opment.It subsumesethnography:Onemuststudycommunitiesof practiceinsitu andin detail to understandtheir zoneof proximaldevelopment.A criticalunansweredquestionaboutactivity theory is whetherit canbe codified asa(predictive)engineeringmodel (e.g. Blackler 1995), or whetherpart of thethrust of activity theory is to emphasizeaspectsof usability that arenotsusceptibleto engineeringmodels.

Thesocialandorganizational perspectivebroughtwith it a newcritiqueoftechnologydevelopmentin HCI. Onethemeis that different perspectivesareinevitableandthatconflict mustbesystemicallyaccommodatedand managed.HCI conceptsand techniquesthrough the 1980stendedto be directedat anaivenotion of engineeringoptimality. The1990s broughttheview that stake-holder interestsoften conflict profoundly (e.g.Suchman1995).Someof themost significant conflicts and misunderstandingsare not betweenusersanddesigners—thehallmarkconcernof participatorydesign—butbetweendiffer-ent constituencies of users(e.g. Blomberg 1995). Theseconflicts typically


involve powerrelationsbetweenhigher-andlower-statusemployees,betweenmanagersandtheir subordinates,andbetweenteachersandstudents.Conflictscanof coursebesalutaryfor groupdecision-making(Kling 1991);thepoint isthatthey mustbe addressed.

Thesecondthemeof thesociopolitical critiquewasthepotentialfor usabil-ity to unwittingly becomea vehiclefor de-skilling anddisempoweringwork-ers.Making someone’s work easierreducesthe skill requiredto performthework; in agivenorganizationalcontext,it mayreducestatus,pay,andevenjobsecurity.The effectscan be subtleand complexbecausethe workersthem-selvesmay not recognizethe potential for suchan outcomeandmay partici-patein their own de-skilling. This level of analysisreceivedvery littl e atten-tion in the 1980s by the mainstream involved inHCI. Thus, the malaiseregarding cognitive user models arose indoubts that the paradigmcouldachievemore than performancemodeling, not in worries that minimizinglow-level actionsmight be fundamentallythe wrong goal to pursue.Recentwork hasexploredthe dynamicco-designof new technologyandnew socialand organizationalstructures (Gasser 1986,Kling & Jewett1994).

New HCI technologiesto supportmeetingsandothercollaborativeactivityhaveencouragedthe developmentof socialandcontextualanalysis.Throughthe 1980s,electronicmail becamean increasinglycommoncommunicationtool, andother internettools like newsgroups,multi-userdomains,andreal-time chat becameless arcane.A variety of environmentsfor collaborativeproblemsolving were investigated,including electronicmeetingrooms(Nu-namakeret al 1991), two-way video “media-spaces”(Bly et al 1993), andthree-dimensional“virtual reality” simulations(Leigh et al 1996).Theportraitof a solitary user finding and creating information in a computerbecamebackgroundto the portrait of severalto many peopleworking togetherat avariety of timesandplaces.Speculationaboutnew paradigmsfor education,work, andleisureactivity havebecomerampantin thefield andin thecultureat large.

Theshift of focusin HCI towardcooperativeactivity raisesmanymethodo-logical issues.As suggestedabove,the conceptof usability becomesmoremultifarious and therefore susceptible to a multitude of distortions. Agre(1995)discusseda casein which a usabilityanalysisdevolvedinto a studyofthe malleability of userperceptionsaboutprivacy. Grudin (1994) raisedtheissuethat thepeoplewho do thework in a group-orientedsystemmaynot betheoneswhoenjoythebenefits.Most ofwhatweknowof usability pertains toinformation creationand managementtasks,but in future systemsinterper-sonalcommunicationtasksmay becomemore centralthan information access.How arevarioustypesof cooperativeactivity affectedby variousparadigmsfor computer-mediatedcommunication?Mostof ourknowledgepertainsto the

76 CARROLL

relativelyprimitive paradigmsbasedon electronicmail (Hiltz & Turoff 1978/1993, Sproull & Kiesler 1991). What kinds of humancommunities will befavoredby a computernetworkinfrastructure;what kinds will be weakened?Will worldwide interestgroupssupplantphysicalneighborhoods as our pri-mary communities?Will our neighborhoodsbecometransformedinto localnetworks(Schuler1994)?

SYNTHESIZING A DISCIPLINE

In Sciencesof theArtificial, Simonwrotethat“the properstudyof mankind isthe scienceof design.”The designof humanactivities and the technologiesthatsupportthemarea specialcase of “design”in thisbroad sense.

HCI has madesteadyand sometimes dramaticprogressas a scienceofdesign.It hasbecomea major researchareain computerscienceandtheveryfulcrum of informationtechnologydevelopment. However,the emergenceofHCI is ongoing.Perhapsthemostimpressive currentfeatureof theareais itsfragmentation.Theparadigmaticconsensusof 1970ssoftware psychology andof 1980scognitive HCI is gone.This is not necessarilybad. Someof thecurrent riftsmay helpset theagenda for thefuture.Forexample, a strong formof contextualism assertsthat thereis no role for controlledresearchin HCI(Carroll 1989,Whiteside& Wixon 1987).This is a potentially constructivechallengeto the HCI researchcommunity. Other rifts are more a matterofmutualneglect.For example,aftera brief periodof confrontationin themid-1980s,the proponentsof cognitiveusermodeling havelargely disconnectedfrom programmatic discussionsof activity theory,andconversely as well.

Mainly, this fragmentationreflectsthe difficulty of assimilating the greatvarietyof methodologies,theoreticalperspectives,driving problems,andpeo-ple that havebecomepartof HCI. Today’s HCI researchersandpractitionersare, after all, immigrants from other disciplines and backgrounds.It is notsurprising that they often continue to favor what theyknow how to do.Younger peoplenow enteringthe field will bring a broaderfoundationofknowledge andskill, andit is likely thatthepotential forabroaderHCI will beadvanced throughthem.

The recentpastsuggeststhreespecific technicalthemesfor the near-termfuture. First, the engineeringscopeof HCI will continueto broadenbeyonduserinterfaceinteractions. Designinghumanactivitiesand appropriatetech-nology to supportthemis likely to becomea moreintegratedendeavor.Sec-ond, the successof HCI in genericapplicationsis likely to bolster furtherdomain-specificwork, in particular complex applicationareas.Finally, theimpact of HCI on psychologyitself, perhapsthe strongestfulfillm ent of Si-mon’s vision of a designscience,is likely to progressmoreconcertedly,given


recentdevelopments in ecologicalandrationalpsychology,andgivena politi-cal contextthatincreasinglyquestionsthe useof science.

Activity theoryemphasizesthat the purview of usability is the totality ofwhat peopledo and experience,and that the diversefacetsof usability areinterdependentandco-evolvewith technology.It hasbegunto play a role inguiding the envisionmentof new technology(e.g.Kuutti & Arvonen1992).However,it needs tobe cultivated as amore comprehensive foundationfor thedevelopmentprocess:requirementsgathering,participatoryinteraction,soft-wareanalysisanddesign,specificationandprototyping, systemimplementa-tion, documentation and instructiondesign, and usabilityevaluation.Scenario-baseddesign,a fusion of activity theorywith object-orientedsoftwareengi-neeringin whichnarrativesof usesituationsrepresentsoftwareandsystems, isone proposalfor howthis mightbeachieved (Carroll1995).

HCI hasfocusedon commondenominatortechnicaldomains:mass-marketapplicationslike word processorsand spreadsheets,standardgraphicaluserinterfaces, like the Macintosh and Windows,and generictechniques like menuhierarchies,direct manipulation, and programmingby demonstration. In the1980stherewere so many studiesof user issuesin text editing that it wascalledthe“white rat” of HCI. As in thepsychologyof learning,paradigmaticfocusyieldsa technicalcoherence,but at thepriceof confoundingeccentrici-ties. In the near future, there is likely to be more HCI researchdirectedatspecifictechnicaldomains;for example,userinterfacesfor typesetters(Bød-ker et al 1987)or petroleumengineers(Brooks1993),andsoftwaretoolsandenvironmentsfor teachersandstudents(Solowayet al 1994)or for scientistsand engineers(Gallopoulosetal 1994).

Theemergenceof HCI in thepasttwo decadesillustratesthepossibility ofpsychologicalinquiry in the contextof systemdevelopmentandof progresswith fundamentalissuesjoined with engineeringdesign.It demonstrates,forexample,how the complexproblem-solving of systemdevelopmentis not astraightforward scaling of laboratory-basedstudiesof puzzles.Laboratorysituations, afterall, aremodelsof realsituations in which peoplearedeprivedof socialandtool resourcesthatconstitute thosesituations.As emphasizedbyactivity theory,humanbehaviorandexperienceboth adaptto and transformsocial and technologicalcontext.Humanactivities motivate the creationofnew tools, but thesein turn alter activities,which in time motivates furthertools.Thecontextof behaviorandexperienceis changingmorerapidly,moreincessantly,perhapsmoreprofoundlythaneverbeforein history.HCI designprovidesan opportunity fundamentallyto expandtraditional ecologicalandrecentrationalapproachesto psychology. These approachesconsider theenvi-ronmenta causalfactor in psychological explanation,but theyhavegenerally

78 CARROLL

ignoredthe evolution of the environmentitself (Anderson1990,Brunswick1956,Gibson1979).

It is excitingto see thattheemergingroleof socialandcognitive scienceincomputerscienceandthecomputerindustry is far morediverse,pervasive,andcritical thanimagined in the 1970s.As it hasturnedout, that role wasnot tosupporta receivedview of designbut to help overturnit andhelp clarify thereal nature of design.Nor wasthat roleto recastpsychological laws ashuman-factorsguidelines;it wasindeedto playapartin driving theevolution of socialandcognitive science,andthe recognitionthat computerscanbe deliberatelydesignedto facilitate humanactivity and experienceonly when social andcognitiverequirementsdrive thedesignprocessthroughout.Thereis unprece-dentedpotential for interdisciplinary synergyhere.Socialsciencehasalwaysbornethe vision of what humansocietymight become,but it has typicallylacked the means to be constructive. Computer science—quite the con-verse—cannotavoid causingsubstantial social restructuring.An integratedandeffectiveHCI canbe a turning point in both disciplinesand,perhaps,inhumanhistory.

ACKNOWLEDGMENTS

Somepartsof the sectionon The Emergenceof Usability are adaptedfromCarroll (1992). Somepartsof the sectionon Design Rationaleare adaptedfrom theIntroductionto Moran& Carroll (1996).

LiteratureCited

Agre P. 1995. Conceptionsof the userin com-puter systems design. SeeThomas 1995,pp. 67–106

AndersonJR.1990. TheAdaptiveCharacterofThought. Hil lsdale,NJ:Erlbaum

Bennett JB. 1984. Managing to meetusabil ityrequirements: establishing and meetingsoftwaredevelopment goals. InVisual Dis-play Terminals, ed. J Bennett, D Case,JSandelin, M Smith, pp. 161–84. Engle-wood Cliffs, NJ:Prentice-Hall

Bentley R, HughesJA, Randall D, RoddenT,Sawyer P,et al. 1992. Ethnographically-in-formedsystemdesign for air traffic control.Proc. CSCW 1994 Conf. Comput.-SupportCoop. Work, Toronto, pp. 123–29. NewYork: Assoc.Comput. Mach.

BiasRG,Mayhew DJ,eds.1994. Cost-Justify-ing Usabil ity. Boston: Academic

BjerknesG, Ehn P, Kyng M, eds.1987. Com-puters and Democracy: A ScandinavianChallenge.Brookfield, VT: Avebury

Blackler F. 1995. Activity theory, CSCWandorganizations. SeeMonk & Gilbert 1995,pp. 223–48

Blomberg JL. 1995. Ethnography: aligningfield studies of work and systemdesign.SeeMonk & Gilbert1995, pp. 175–97

Bly S, Harrison S, Irwin S. 1993. Mediaspaces:bringing people together in a video,audio, and computing environment. Com-mun. Assoc.Comput. Mach.36(1):28–47

Bødker S. 1991. Through the Interface: A Hu-man Activity Approach to User InterfaceDesign. Hil lsdale,NJ:Erlbaum

Bødker S, Ehn P, KammersgaardJ, Kyng M,Sundblad Y. 1987. A utopian experience.SeeBjerknes etal 1987, pp. 251–78

Bowker G, Star SL, GasserL, Turner B, eds.1995. Social ScienceResearch, TechnicalSystemsand Cooperative Work. Mahwah,NJ:Erlbaum

Brooks FP. 1975/1995. The Mythical Man-Month: Essayson Software Engineering.Reading, MA: Addison-Wesley

Brooks R. 1993. The casefor the specializedinterface.IEEE Softw. 10(2):86–88

Brunswick E. 1956. Perception and the Repre-sentative Design of Psychological Experi-ments.Berkeley:Univ. Calif. Press


BuckinghamShum S. 1996. Analyzing the us-ability of adesignrationalenotation. In De-signRationale: Concepts, Techniques,andUse, ed. TP Moran, JM Carrol l , pp.185–215. Mahwah, NJ:Erlbaum

Butler KA. 1985. Connecting theoryandprac-tice: a case study of achieving usabil itygoals. Proc. CHI’85 Hum. Factors Com-put. Syst., San Francisco, pp. 85–88. NewYork: Assoc.Comput. Mach.

CardSK, MoranTP,NewellA. 1983. ThePsy-chology of Human-Computer Interaction.Hil lsdale,NJ:Erlbaum

Carroll JM. 1985. What’s in a Name?AnEssayin the Psychology ofReference.New York:Freeman

Carroll JM. 1989. Evaluation, description andinvention: paradigms for human-computerinteraction. In Advances in Computers,ed. M Yovits, 29:47–77. SanDiego: Aca-demic

Carroll JM. 1990. The Nurnberg Funnel: De-signing Minimalist Instruction for Practi-cal Computer Skill . Cambridge, MA: MITPress

Carroll JM, ed. 1991. Designing Interaction:Psychology at the Human-Computer Inter-face.NewYork: CambridgeUniv. Press

Carroll JM. 1992. Creating a design scienceofhuman-computer interaction. In FutureTendenciesin Computer Science,Control,and Applied Mathematics: 25th Anniver-saryof INRIA, ed.A Bensoussan, J-PVer-jus, pp. 205–15. New York: Springer-Ver-lag

Carroll JM, AlpertSR,KaratJ,VanDeusenM,Rosson MB. 1994. Capturing design his-tory and rationalein multimedia narratives.In Proc. CHI’94: Hum. Factors Comput.Syst., Boston, pp. 192–97. New York: As-soc.Comput. Mach./Addison-Wesley

Carroll JM, Campbell RL. 1986. Softening uphard science:reply to Newell and Card.Hum.-Comput. Interact. 2:227–49

Carrol l JM, Carrithers C. 1984. Blockinglearner errors in a training wheelssystem.Hum.Factors26(4):377–89

Carroll JM, Mack RL, Kellogg WA. 1988. In-terface metaphorsand user interface de-sign. SeeHelander1988, pp. 67–85

Carroll JM, RossonMB. 1985. Usabili ty speci-ficationsasa tool in iterative development.In Advancesin Human-Computer Interac-tion, ed. HR Hartson, 1:1–28. Norwood,NJ:Ablex

Carroll JM, Rosson MB. 1991.Deliberatedevolution: stalking theView Matcher in de-sign space. Hum.-Comput. Interact. 6:281–318

Carroll JM, RossonMB. 1992. Getting aroundthe task-artifact cycle: how to makeclaimsand design by scenario. Assoc. Comput.Mach.Trans. Inf. Syst.10:181–212

Carroll JM, Rosson MB. 1995. Managingevaluation goalsfor training. Commun. As-soc.Comput. Mach. 38(7):40–48

Carroll JM, Singley MK, RossonMB. 1992.Integratingtheorydevelopment with designevaluation. Behav. Inf. Technol . 11:247–255

Carroll JM, ThomasJC. 1982. Metaphor andthe cognitive representation of computingsystems. IEEE Trans. Syst., Man Cybern.12:107–16

Carroll JM, ThomasJC, Malhotra A. 1979. Aclinical-experimental analysis of designproblemsolving. DesignStud. 1:84–92

Conklin J, BegemanM. 1988. gIBIS: a hy-pertext tool for exploratory policy discus-sion. Assoc.Comput. Mach.Trans.Off. Inf.Syst. 6(4):303–31

Conklin J, Yakemovic KCB. 1991. A process-oriented approach to design rationale.Hum.-Comput. Interact.6:357–91

Curtis B, KrasnerH, Iscoe N. 1988. A fieldstudy of the software design processforlarge systems.Commun. Assoc.Comput.Mach. 31:1268–87

Cypher A, Halbert D, Kurlander D, LiebermanH, Maulsby D, et al, eds. 1993. WatchWhat I Do: Programming by Demonstra-tion. Cambridge,MA: MIT Press

deGroot AD. 1965. Thought and Choice inChess.Paris:Mouton

Denning PJ,ComerDE, GriesD, Mulder MC,Tucker AB, et al. 1989. Computing as adiscipl ine. Commun. Assoc. Comput.Mach. 32:9–23

Dix A, Finlay J, Abowd G, Beale R. 1993.Human-Computer Interaction. EnglewoodCliff s, NJ:Prentice-Hall

Douglas SA, Moran TP. 1983. Learning texteditor semanticsby analogy. Proc.CHI’83Conf. Hum.Factors Comput. Syst., Boston,pp. 207–11. New York: Assoc. Comput.Mach.

DreyfussH. 1955. Designing for People. NewYork: Simon& Schuster

Ehn P. 1988. Work-Oriented Designof Com-puter Artifacts. Stockholm: Arbetslivscen-trum

Engestrom Y. 1993. Developmental work re-search: reconstructing expertise throughexpansive learning. In Human Jobs andComputer Interfaces,ed. M Nurminen, GWeir. Amsterdam:Elsevier

Ericsson KA, Simon HA. 1985. ProtocolAnalysis: Verbal Reports as Data. Cam-bridge,MA: MIT Press

EsperEA. 1925. A technique for the experi-mental investigation of associative interfer-encein artifi cial linguistic material. Lang.Monogr. 1:1–47

Fischer G, Lemke AC, McCall R, Morch AI.1991. Making argumentation servedesign.Hum.-Comput. Interact.6(3/4):393–419

80 CARROLL

FlanaganJC. 1954. The critical incident tech-nique.Psychol. Bull . 51:28–35

FurnasG, LandauerTK, GomezL, Dumais S.1983. Statistical semantics: analysis of thepotential performance ofkeyword informa-tion systems. Bel l Syst. Tech. J. 62:1753–806

Gallopoulos E, Houstis E, Rice JR. 1994.Computer asthinker/doer: problem-solvingenvironments for computational science.IEEEComput. Sci.Eng. 1(1):11–23

GasserL. 1986. The integration of computingand routine work. Assoc.Comput. Mach.Trans. Off. Inf. Syst.4(3):205–25

GibsonJJ.1979. The Ecological Approach toVisual Perception. Boston: Houghton Mif-fli n

Good M, Spine TM, Whiteside J, George P.1986. User-derived impact analysis as atool for usabi l i ty engineering. Proc.CHI’86 Conf. Hum.Factors Comput. Syst.,Boston, pp. 241–46. New York: Assoc.Comput. Mach.

Gould JD,BoiesSJ.1983. Humanfactorschal-lenges increatinga principalsupport officesystem:the speechfi ling approach. Assoc.Comput. Mach. Trans. Off. Inf. Syst. 1:273–98

Gray WD, John BE, Atwood ME. 1992. Theprécis of project Ernestine, or, An over-view of a val idation of GOMS. Proc.CHI’92 Conf. Hum.Factors Comput. Syst.,Monterey,CA,pp. 307–12. NewYork: As-soc.Comput. Mach.

Greatbatch D, Heath C, Luff P, Campion P.1995. Conversation analysis: human-com-puter interaction and the general practiceconsultation. See Monk & Gilbert 1995,pp. 199–222

Greenbaum J, Kyng M, eds.1991. Design atWork: Cooperative Design of ComputerSystems.Hil lsdale,NJ:Erlbaum

Grudin J.1989. The caseagainstuserinterfaceconsistency. Commun. Assoc. Comput.Mach.32:1164–73

Grudin J. 1994. Groupwareandsocialdynam-ics: eight challenges for developers.Com-mun. Assoc.Comput. Mach.37(1):92–105

Helander M, ed. 1988. Handbook of Human-Computer Interaction. Amsterdam: North-Holland

Hil tz SR, Turoff M. 1978/1993. The NetworkNation: Human Communication via Com-puter. Cambridge:MIT Press

Hughes J, King V, Rodden T, AndersonH.1994. Moving out from the control room:ethnography in system design. Proc.CSCW’94 Conf. Comput.-Support. Coop.Work, Chapel Hil l, NC, pp. 429–39. NewYork: Assoc.Comput. Mach.

Hutchins E, Hollan J, Norman DA. 1986. Di-rect manipulation interfaces.SeeNorman& Draper1986, pp. 87–124

JeffriesRJ,Mil ler JR,Wharton C, Uyeda KM.1991. User interfaceevaluation in the realworld: a comparison of four techniques.Proc.CHI’91 Hum.Factors Comput. Syst.,NewOrleans,pp. 119–24. New York: As-soc.Comput. Mach.

JohnBE. 1990. Extensions of GOMSanalysesto expert performancerequiring perceptionof dynamic visual and auditory informa-tion. Proc. CHI’90 Hum.Factors Comput.Syst., Seattle, pp. 107–15. New York: As-soc.Comput. Mach.

Karat CM, Campbell RL, Fiegel T. 1992.Comparisonof empirical testing andwalk-through methods in user interfaceevalu-ation. Proc.CHI’92 Hum.FactorsComput.Syst., Monterrey, CA, pp. 397–404. NewYork: Assoc.Comput. Mach.

KierasDE. 1988. Towardsa practical GOMSmodel methodology for user interfacede-sign. See Helander1988, pp. 135–57

Kli ng R. 1991. Cooperation, coordination andcontrol in computer-supportedwork. Com-mun. Assoc.Comput. Mach. 34(12):83–88

Kli ng R, JewettT. 1994. The socialdesign ofworkli fe with computers and networks: anaturalsystemsperspective.In AdvancesinComputers, ed. M Yovits, 39:239–93. Or-lando: Academic

Knorr-Cetina KD. 1981. The micro-sociologi-cal challenge of macro-sociology: towardsa reconstruction of socialtheoryandmeth-odology. In Advancesin Social Theory andMethodology: Towards an Integration ofMicro- and Macro-Sociologies, ed. KDKnorr-Cetina, A Cicourel. London: Rout-ledge& KeganPaul

Kuhn S, Muller MJ, eds.1993. Specialsectionon participatory design. Commun. Assoc.Comput. Mach.36(4):24–103

Kuutti K, ArvonenT. 1992. Identifying CSCWapplications by means of activity theoryconcepts:a caseexample. Proc. CSCW’92Comput.-Support. Coop. Work, Toronto,pp. 233–40. New York: Assoc. Comput.Mach.

Leigh J, JohnsonAE, Vasilakis CA, DeFantiTA. 1996. Multi-perspective collaborativedesign in persistentnetworkedvirtual envi-ronments. Proc. IEEE Vir tual Real i tyAnnu. Int. Symp., VRAIS96, Santa Clara,CA. In press

Lewis CH, NormanDA. 1986. Designing forerror. See Norman & Draper 1986, pp.411–32

MackRL, LewisCH, Carroll JM. 1983. Learn-ing to use office systems:problems andprospects. Assoc. Comput. Mach. Trans.Off. Inf. Syst. 1:254–71

MacLeanA, YoungRM, Bellotti VME, MoranTP. 1991. Questions, options, andcriteria:elements of design spaceanalysis. Hum.-Comput. Interact.6:201–50


Malhotra A, ThomasJC, Carroll JM, MillerLA. 1980. Cognitive processesin design.Int. J. Man-Mach.Interact. 12:119–40

Mil ler LA. 1974. Programming by nonpro-grammers. Int. J. Man-Mach. Stud. 6:237–60

Monk AM, Carroll JM, Harrison M, Long J,Young RM. 1990. New approachesto the-ory in HCI: How should we judge theiracceptabil ity? Proc. INTERACT’90 Hum.-Comput. Interact., Cambr idge, pp.1055–58. Amsterdam:North-Holland

Monk AM, Gilbert N, eds.1995. Perspectiveson HCI: Diverse Approaches. London:Academic

Monk AM, Nardi B, Gilbert N, Mantei M,McCarthy J. 1993. Mixing oil and water?Ethnography versusexperimentalpsychol-ogy in the study of computer-mediatedcommunication. Proc. INTERCHI ’ 93Hum. Factors Comput. Syst., Amsterdam,pp. 3–6. NewYork: Assoc.Comput. Mach.

Moran TP. 1983. Getting into a system:exter-nal-internal task mapping analysis. Proc.CHI’83 Hum.Factors Comput. Syst., Bos-ton, pp. 45–49. NewYork: Assoc.Comput.Mach.

Moran TP, Carroll JM, eds.1996. DesignRa-tionale: Concepts, Techniques,and Use.Mahwah, NJ:Erlbaum

Nardi BA, ed.1995a. Context and Conscious-ness: Activity Theory and Human-Com-puter Interaction. Cambridge, MA: MITPress

Nardi BA. 1995b. Studying context: a com-parison of activity theory, situated actionmodels and distributed cognition. SeeNardi1995a,pp. 69–102

Newell A, Card SK. 1985. The prospectsforpsychological sciencein Human-ComputerInteraction. Hum.-Comput. Interact. 1:209–42

Newell A, Simon HA. 1972. Human ProblemSolving. Englewood Cliffs, NJ: Prentice-Hall

NielsenJ, ed. 1989. Coordinating User Inter-faces for Consistency. New York: Aca-demic

Nielsen J. 1993. Usabil ity Engineering. Bos-ton: Academic

Nielsen J. 1994. Estimating the numberof sub-jectsneededfor a thinkingaloudtest.Int. J.Hum.-Comput. Stud. 41:385–97

NielsenJ, Mack RL, eds.1994. Usabili ty In-spectionMethods.NewYork: Wiley

Nielsen J, Mollich R. 1990. Heuristic evalu-ation of user interfaces. Proc. CHI’90Conf. Hum.Factors Comput. Syst., Seattle,pp. 249–56. New York: Assoc. Comput.Mach.

Nielsen J, Phill ips VL. 1993. Estimating therelative usabili ty of two interfaces:heuris-tic, formal and empirical methods com-

pared. Proc. INTERCHI’93 Conf. Hum.Factors Comput. Syst., Amsterdam, pp.214–21. NewYork: Assoc.Comput. Mach.

Norman DA. 1991. Cognitive artifacts. SeeCarroll 1991, pp. 17–38

NormanDA, DraperSW,eds.1986. UserCen-tered System Design. Hi l l sdale, NJ:Erlbaum

NunamakerJ, Dennis A, ValacichJ, Vogel D,George J.1991. Electronic meeting sys-temsto support group work. Commun. As-soc.Comput. Mach.34(7):40–61

NygaardK. 1979. The iron andmetalproject:trade union participation. In Computers Di-viding Man and Work: RecentScandina-vian Researchon Planning and Computersfrom a Trade Union Perspective, ed. ASandberg, p. 98. Malmö: Arbetslivscen-trum

PavaCHP.1983. Managing NewOffice Tech-nology: An Organizational Strategy. NewYork: FreePress

Payne SJ, Green TRG. 1989. Task-ActionGrammars: the model and its develop-ments. In TaskAnalysis for Human-Com-puter Interaction, ed. D Diaper, pp. 75–107. Chichester:Ellis Horwood

Payne SJ,Squibb H, HowesA. 1990. The na-ture of device models: the yoked statespace hypothesis and some experimentswith text editors. Hum.-Comput. Interact.5:415–44

PolsonPG, Lewis CH, RiemanJ, Wharton C.1992. Cognitive walkthroughs: a methodfor theory-basedevaluation of user inter-faces.Int. J. Man-Mach.Stud. 36:741–73

PreeceJ, RogersY, SharpH, Benyon D, Hol-land S, CareyT. 1994. Human-ComputerInteraction. Readi ng, M A : A ddi son-Wesley

Rittel H, Webber M.1973. Dilemmasin a gen-eral theory of planning. Policy Sci. 4:155–69

Rosson MB, Carroll JM. 1995. Narrowing thespecifi cation-implementation gap in sce-nario-baseddesign. In Scenario-BasedDe-sign: Envisioning Workand Technology inSystemDevelopment, ed. JM Carroll , pp.247–78. New York: Wiley

RoyceWW. 1970. Managing the developmentof large software systems: conceptsandtechniques. Proc. West. Electr . ShowConv., WESTCON, Los Angeles, pp.(A/1)1-9. Reprinted in Proc. Int. Conf.Softw. Eng., 11th, Pittsburgh, May 1989,pp. 328–38

Schuler D. 1994. Community networks: build-ing a newparticipatory medium. Commun.Assoc.Comput. Mach. 37(1):39–51

Schuler D, Namioka A, eds.1993. Participa-tory Design: Principles and Practices.Hil lsdale,NJ:Erlbaum

Scriven M. 1967. The methodology of evalu-

82 CARROLL

ation. In Perspectivesof CurriculumEvalu-ation, ed.R Tyler,R Gagne,M Scriven, pp.39–83. Chicago: Rand McNally

Shattuck L, Woods D. 1994. The critical inci-denttechnique: 40 yearslater. Proc. Hum.Factors Ergon. Soc. Annu. Meet., 38th,Nashville, pp. 1080–84. Santa Monica,CA: Hum.FactorsErgon. Soc.

ShneidermanB. 1980. Software Psychology:Human Factors in Computer and Informa-tionSystems.Cambridge,MA: Winthrop

ShneidermanB. 1983. Direct manipulation: astepbeyond programminglanguages.IEEEComput. 16(8):57–62

Shneiderman B,MayerR, McKayD, Heller P.1977. Experimental investigations of theutil ity of detailed flowcharts in program-ming. Commun. Assoc.Comput. Mach.20:373–81

Sime ME, GreenTRG, GuestDJ. 1973. Psy-chological evaluation of two conditionalconstructions usedin computer languages.Int. J. Man-Mach.Stud. 5:105–13

Simon HA. 1969. The Sciencesof the Artifi-cial. Cambridge:MIT Press

Smith DC, Irby C, Kimball R, Verplank B,HarslemE. 1982. Designing the Star userinterface.Byte7(4):242–82

Soloway E, Guzdial M, Hay KE. 1994.Learner-centereddesign: the challenge forHCI in the 21st Century. Interactions1(2):36–48

Sproull L, Kiesler S. 1991. Connections: NewWaysof Working in A NetworkedOrgani-zation. Cambridge:MIT Press

SuchmanLA. 1987. Plans and Situated Ac-tions: The Problem of Human-MachineCommunication. New York: CambridgeUniv. Press

SuchmanLA, ed.1995. Specialsection onrep-resentations of work. Commun. Assoc.Comput. Mach. 38(9):33–68

Sukaviriya P, Foley JD, Griff ith T. 1993. Asecond-generationuserinterfacedesignen-vironment: the model andruntimearchitec-ture.Proc. INTERCHI’93 Conf. Hum.Fac-tors Comput. Syst., Amsterdam, pp. 375–82. New York: Assoc.Comput. Mach.

Sutcliffe A, Carroll JM, Young RM, Long J.1991. HCI theory on trial. Proc. INTER-CHI’93 Conf. Hum.Factors Comput. Syst.,Amsterdam, pp. 375–82. New York: As-soc.Comput. Mach.

Szekely P, Luo P, NechesR. 1993. Beyondinterface builders: model-based interfacetools. Proc. CHI’91 Conf. Hum. FactorsComput. Syst., NewOrleans, pp. 399–401.New York: Assoc.Comput. Mach.

Tanner PP,Buxton WAS. 1985. Someissuesin futureuserinterfacemanagement system(UIMS) development. In User Interface

Management Systems: Proc. WorkshopUser Interface Management Syst., See-hiem, Ger., 1983, ed. GE Pfaff , 67–79.NewYork: Springer-Verlag

ThomasPJ,ed.1995. TheSocial and Interac-tional Dimensionsof Human-Computer In-teract. New York: CambridgeUniv. Press

Tucker AB, ed.1997. TheHandbook of Com-puter Scienceand Engineering. Boca Ra-ton, FL: CRCPress

Tucker AB, Turner AJ. 1991. A summaryofthe ACM/IEEE-CS Joint Curriculum TaskForce Report: computing Curricula 1991.Commun. Assoc.Comput. Mach. 34:68–84

Virzi RA. 1992. Refining the testphaseof us-ability evaluation: how many subjects isenough?Hum.Factors 34:457–68

WassermanAI, ShewmakeDT. 1982. Rapidprototyping of interactive information sys-tems. Assoc.Comput. Mach. Softw. Eng.Notes7:171–80

Wasserman AI, Shewmake DT. 1985. The roleof prototypes in the User Software Engi-neering (USE) methodology. In Advancesin Human-Computer Interaction, ed. HRHartson, 1:191–209. Norwood, NJ:Ablex

WeissmanL. 1974. A methodology for study-ing the psychological complexity of com-puter programs. PhD thesi s. Uni v.Toronto, Toronto

WertschJ. 1985. Vygotskyand the Social For-mation of Mind. Cambridge,MA: HarvardUniv. Press

Whiteside J, BennettJ, HoltzblattK. 1988. Us-abil ity engineering: our experience andevolution. SeeHelander1988, pp. 791–817

WhitesideJ, Wixon D. 1987. Improving hu-man-computer interaction: a questfor cog-nitive science. In Interfacing Thought:Cognitive Aspectsof Human-Computer In-teraction, ed. JM Carrol l , pp. 337–52.Cambridge,MA: Bradford/MIT Press

Will igesRC, Will igesBH, HanSH. 1993. Se-quential experimentation in human-com-puter interfacedesign. In Advancesin Hu-man-Computer Interaction, ed. HR Hart-son, D Hix, 4:1–30. Norwood, NJ:Ablex

Wixon D, Holtzblatt K, Knox S.1990. Contex-tual design: an emergent view of systemdesign. In Proc. CHI ‘90: Hum. FactorsComput. Sys.,Seattle, pp. 329–36.NewYork: Assoc.Comput. Mach.

Wright RB, ConverseSA. 1992. Method biasandconcurrent verbal protocol in softwareusabil ity testing. Proc. Hum. Factors Er-gon. Soc. Annu. Meet., 36th, Atlanta, pp.891–912. SantaMonica,CA: Hum. FactorsErgon. Soc.

Zuboff Z. 1988. In the Age of the Smart Ma-chine: The Future of Work and Power.NewYork: Basic Books


Documents

HUMAN-COMPUTER INTERACTION: Psychology as a Science of … · 2017. 6. 17. · Human-computer interaction(HCI) has emerged relatively recently as a highly successful area of computer