how organizational practices can compensate for …faculties are limited. instead of being infinite...

COGNITIVE REPAIRS:HOW ORGANIZATIONAL PRACTICESCAN

COMPENSATEFORINDIVIDUAL SHORTCOMINGS

ChipHeath,RichardP.Larrick, andJoshuaKlayman

ABSTRACT

The literature in cognitivepsychologyhasdescribedavariety of shortcomingsthatpreventindividuals from learningeffectively. We reviewthis literatureandprovideexamplesofanumberoforganizationalpracticesthatmayeffectivelyrepairthecog-nitive shortcomingsof individuals. We call thesepracticescognitiverepairs. Wethen discusssix tradeoffsthat affect the successof cognitive repairs.We closebyconsideringhow a cognitiveperspectivemight benefit thosewho study organiza-tional learningandthosewho manageit.

Researchin Organizational Behavior, Volume 20, pages1-37.Copyright © 1998 hy JAI PressInc.All rights of reproduction in any form reserved.ISBN: 0-7623-0366-2

1

CHIPHEATH, RICHARD P. LARRICK, andJOSHUAKLAYMAN CognitiveRepairs 3

INTRODUCTION

In afamousspeech,Hamletdeclares,“What a pieceof work is man.How nobleinreason,how infinite in faculties” (Hamlet, II, 2). Anobserverwho summarizedthepsychologyof the late twentiethcenturywould probably choosevery differentphrasesto describethehumancondition—perhaps,“What fools thesemortalsbe”(MidsummerNight’s Dream, III, 2).

Are people“infinite in faculties”and “noble in reason”?HerbertSimon won aNobelprize for arguingthat social sciencemustunderstandthewaysthat humanfacultiesare limited. Insteadof being infinite in faculties,Simon’s humanscouldbe only “boundedlyrational” becausetheir cognitive abilities—theirability toperceive,remember,andprocessinformation—wererestricted.Well, then,if peo-ple are not infinite in faculties,a.re they“noble in reason”?Cognitivepsycholo-gistshavespent30 yearsexaminingthe actualprocessesthat peopleusewhentheycollect information,combineit, anddrawinferencesabouttheir world (Nis-bett & Ross,1980;Kahneman,Slovic, & Tversky, 1982;Holland,Holyoak,Nis-bett,& Thagard,1986).Insteadof depictingpeopleas“noble” (or magnificent)inreason,this researchhasarguedthat peoplereasonin ways that producesystem-atic errors.A pessimisticmodernHamletmightcombinetheobservationsof thesetwo researchstreamsanddescribehumansas equippedwith primitive hardwnreandbuggy software.

However, outsidershave not always acceptedthe pessimisticdescriptionofhumanfacultiesandreasonthat is foundin theresearchliterature.As oneskepticput it, “If weare sostupid,how did wegetto themoon?”(Nisbett& Ross,1980).

How shouldweresolvetheapparentdiscrepancybetweenthe pessimisticliter-atureon humanshortcomingsandtheoptimistic evidenceof humanaccomplish-ment?One way is to dismiss the laboratoryresearch.Someresearchershavearguedthat theshortcomingsthathavebeendocumentedin the lab are so minorthat they do not constitutemistakesof any real consequence(Funder, 1987;Cohen,1981).Othershavearguedthat individualsareless likely to makeerrorsinnaturalenvironmentsthan in contrivedlaboratoryexperiments(Anderson,1991;Gigerenzer,1996;Cheng& Holyoak, 1985;Hilton, 1995).

We proposeanotherway to resolvethe discrepancy.Unlike someresearchers,wedonotdismisstheexamplesof limitations,errors,andbiasesreportedin thelit-erature;we assumethat individuals are limited, their decisionprocessesarebiased,and that they often make costly mistakeson importantdecisions.Weresolvethe apparentdiscrepancybetweenevidenceof individual shortcomingsandtheempirical factof moonwalksby observingthat individualsdid notmakeitto themoon,NASAdid.

Organizationslike NASA mayhavediscoveredwaysto avoid~orrepairtheindi-vidual shortcomingsthathavebeendocumentedby cognitiveresearchers.Orga-nizations may develop such repairs through deliberate analysis, learn themthroughtrial anderror,or discoverthem throughserendipitousaccident.In some

cases,repairsmay derivefrom formalacademicdisciplineslike economicsor sta-tistics(e.g.,Nisbett,1992;Nisbett,Krantz,Jepson,& Kunda,1983;Larrick,Mor-gan, & Nisbett, 1990), but in most casesthey will not: They will be ad hoc,intuitive rulesthat emergefrom day-to-daypractice.Our thesis,then,is that indi-vidualsindeedfacecognitivelimitationsandshortcomings,but thatorganizationscanprovide individualswith normsandproceduresthat mitigatetheir limitationsand reducetheir shortcomings.

In this paperwedescribea varietyof potentiallyseriousshortcomingsthathavebeendocumentedin researchon humanjudgmentandreasoning.Wefocusin par-ticular on learningandhypothesistesting,that is, how peopleuseinformation todevelopandrevisetheir mentalmodelof the world.Foreachcognitiveshortcom-ing wediscuss,we provideexamplesof organizationalpracticesthatmay repairthis shortcoming.We call thesepracticescognitiverepairsto emphasizethe factthat they correct somecognitive processthat wasinitially flawed andin needofrepair.

We identify potentialcognitiverepairsto spurresearchersto considerhow suchrepairsmightlook andfunction.Althoughatthispoint, wecanmakeonly ananec-dotal casefor interpretingcertainpracticesas “repairs,”wehopethat,by pointingout someplausibleexamplesof suchrepairs,wewill promptresearchersin bothpsychologyandorganizationsto considermoresystematicallyhow suchrepairsmight function.

More generally,theconceptof organizationalcognitive repairsillustratesthatresearchersmay find interestingrelationshipsbetweenindividual cognitionandorganizationalpractice.Theserelationshipshavenot receivedthe attentiontheydeserve.On the oneside, researchin cognitive psychologyhas largely treatedindividual learnersas “rugged individualists” who face a difficult environmentalone, equippedonly with their own, flawed cognitive strategies.On the otherside,organizationalresearchhaslargely ignoredtheliteratureon individualcog-nition, focusinginsteadon issuesof motivationor incentives.By studyingorgani-zationalsourcesof cognitive repairs,we bring togethertwo frequentlydisparateliteraturesanddemonstratehow knowledgeat onelevel of analysiscaninform theother.

By reviewing individual shortcomingsand identifying potential cognitiverepairs,we alsohopeto contributeto theacademicandpopularliteratureonorga-nizational learning (Epple, Argote, & Devadas,1991; Deming, 1982; Senge,1990;Cohen, 1991;Miner & Mezias, 1996).Oneimportantmeansto facilitatelearningby organizationsis to developprocessesthatovercometheshortcomingsof individualswithin theorganization.

Below, we start with a briefexampleof thekinds of repairsthatweconsiderinthis paper.Thenwe introducea frameworkthatdescribesdifferent stagesin thelearningprocess,andweuseit to reviewtheliteratureon individual shortcomingsand to suggestpotential repairs.As a preliminary reply to Hamlet,we say thatevenif ruggedindividualsare unlikely to be infinite in facultiesandnoblein rea-

4 CHIP HEATH, RICHARD P. LARRICK, and JOSHUA KLAYMAN CognitiveRepairs

son, individualswho haveaccessto organizationalandcultural repairsmaysome-timesappearso.

An Exampleof OurApproach

Consideronestudythat might beregardedas anominousindicationof ignoblereasoningby individualexperts.HynesandVanmarcke(1976)askedseven“inter-nationally known” civil engineersto predict theheight of a structurethat wouldcausea foundationto fail; they alsoaskedtheengineersto set a 50 percentconfi-denceinterval aroundtheir predictionso that their confidenceinterval was wideenoughto havea50percentchanceof enclosingthetruefailureheight.Theresultswerequite sobering:notoneengineercorrectlypredictedthe truefailure heightwithin hisor her confidenceinterval.

Evidently, the civil engineersthoughtthey knew more than they did—if theyhadbeenawareof the limitationsof their analysis,theywould havesetwidercon-fidenceintervalsandwould havepredictedthe truefailure height morecorrectly.In the psychologicalliterature this kind of finding hasbeenlabeled“overconfi-dence,”andit is notanaberration.Similarresultshavebeenobservedwith anum-berof individualprofessionals(e.g.,Russo& Schoemaker,1992).In summarizingthe evidence,Griffin andTversky (1992)quippedthat expertsare “often wrongbut rarely in doubt.”

To illustratewhy this studypaintsan ominouspictureof individual reasoning,considerthat (unlessyou arereadingthis paperoutside)you are sitting in abuild-ing that wasconstructedby civil engineerswho weresubstantiallyless accom-plishedthanthe internationallyknown expertsin thestudy. Your civil engineersmadenumerousdecisionsto ensurethestability andsafetyof your building; theydecidedhow strong to makeits roofsupportsandhow stableto makeits founda-tton. If evenexpertengineersare overconfident,shouldyou be concernedaboutyour safety?

The answer,we believe, is no. Fortunately,the engineeringprofessionhasdevelopeda particularrepair,called“safety factors,” that mitigatethe overconfi-dent reasoningof individual engineers.In an actualassignmentcivil engineerswould preciselycalculatetheamountandstrengthof foundationmaterialsneces-saryto hold a structureof a particularheight,then they would multiply their pre-ciseanswerby a safetyfactor(i.e., a numberbetweenthreeandeight),andusethelarger figure to build the foundation.Were theconfidenceintervalsof the engi-neerstoo narrow?Yes. Weretheytoo narrow by a factorof three?No.

Safetyfactorsareanexampleof thekindof cognitiverepairweconsiderin thispaper.An organization(e.g.,anengineeringfirm or theengineeringprofessionatlarge) provides its memberswith a repair that helps combata systematicandpotentially seriousbias in individual judgment.As a result, the organizationshieldsindividualswithin the organizationfrom actingon their flawed decisions,

5

and it shieldsindividuals inside andoutsidethe organizationfrom sufferingtheconsequences.

Whatis a CognitiveRepair?

Organizationalrepairscanroughlybedivided into two classes:(1) motivationalrepairsincreasetheenergy andenthusiasmwith which individualspursuea taskand(2) cognitiverepairsimprovethementalproceduresindividualsuseto decidewhich taskto pursueandhow to pursueit. Organizationalresearchonmotivationand incentivescanbe regardedasthe studyof motivationalrepairs(Milgrom &Roberts,1992;Eisenhardt,1989).Organizationsmayneedto repairmotivationalproblemsin order to encourageindividuals to learn(e.g., seeHeath, Knez, &

Camerer,1993).Forexample,individualsmay notbewilling to experimentwithnewtasksbecausethey havebecomeendowedwith thebenefitsassociatedwiththeold task.

Althoughpreviousworkhasrecognizedtheimportanceof motivationalrepairs,it hasneglectedcognitiverepairs.Evenwhenindividualshavetheright incentivesandresources,they may not learnfrom their experienceif they usethe wrongmentalprocessto generatehypotheses,collectinformation,anddrawconclusions.Thecivil engineerswho misestimatedthestability of theclay embankment.wereadequatelymotivatedto get the right answer.However,they did not on their.owninvoke the kind of correctives(e.g., safety factors)that might havemadetheirguessesmoreappropriatelycautious.

REVIEW OF INDIVIDUAL LEARNING ANDEXAMPLES OF COGNITIVE REPAIRS

In this section weorganizetheliteratureon learninganddecisionmakingaroundthreedifferentstagesof the learningprocess.Effectivelearnersmust (1) generatehypothesesthatexplainthecausalstructureof theworld, (2)collectinformationtodistinguishamongtheirhypotheses,(3) drawconclusionsthatareappropriateand~autious.The boundariesbetweenthesestagesare fuzzy—theyare interrelated~ndinterconnected(Klayman, 1995). However, we distinguish among them

ecausethey involve differentpsychologicalprocesses.Our strategythroughoutthe review is to considerfirst the individual thenthe‘ganization. For eachstageof learning, we describehow an ideal individualirner might reason,and review psychologicalresearchshowinghow realmdi-luals departfrom this ideal.Then, we describepotentialcognitive repairsbyich organizationsmight correctthe individual shortcomingin question.

6 CHIPHEATH, RICHARD P. LARRICK, andJOSHUAKLAYMAN CognitiveRepairs 7

GeneratingHypotheses

In thefirst stageof the learningprocessindividualsmustgeneratehypothesesabouttherelationshipsamongevents.Subjectto constraintsof time andinforma-tion, individuals shouldgeneratehypothesesthat are deep(i.e., by consideringcausesthat are moregeneralor systemic)andbroad(i.e., by consideringa largernumberof potentialcauses).However,a greatdealof psychologicalresearchsug-geststhat individualsdevelophypothesesthat are shallowandnarrow.

IndividualsGenerateHypothesesthat areShallowRatherthanDeep

Individuals Searchfor ExplanationsthatMakeThemselvesLook Good

Individuals often conductshallow searcheswhentheytry to explainsuccessorfailure becausethey searchin a self-servingway(i.e., in a way thatallows themto feel good aboutthemselves).In a meta-analysisof 91 testsof this self-servingbias, Mullen and Riordan(1988) show that individuals typically concludethattheir successesresultedfrom stable,internal factors(e.g.,ability), but that theirJailures resultedfrom unstable,environmentalfactors(e.g., the difficulty of theenvironment, insufficient effort, or bad luck) (see also Fiske & Taylor, 1991,pp. 78-82).

How might organizationsrepairself-servingbiases?Somerepairsmay bequitesimple:Traderson Wall Streetarewarned,“Don’t confusebrainsanda bull mar-ket” (Odean,1996).This compactphrasepromptsindividual tradersto considerthebaserateof successin the market,and it makesit moredifficult for them toindulge in self-servingexplanationsfor their success.

At FloridaPowerandLightemployeesdevelopeda newwayto fight self-serv-ing biasesafteran incidentthatprominentlyfeaturedaJapaneseinspectorfor theDeming Prizewho laterbecamea folk herowithin the company(Walton, 1990,p. 61). To impressthe inspector,FP&L managerstookhim to visit a new facilitythathadbeenconstructedfasterandmoreeconomicallythan any facility in thehistory of the industry.However,theDeminginspectordid not simply accepttheresultsatfacevalueandcongratulatethemon their “quality” projectmanagement;instead,heaskedanumberof questionsto determinewhytheywereso successful.Themanagers’answersweresoinadequatethat it soonbecameclearthattheydidnot understandenoughabouttheir successto recreateit in the future. The inspec-tor dismissedtheir “success” in his Japanese-accentedEnglish—”you wererucky.” Lateron his phrase,completewith accent,becamea commonrepair forself-servinginterpretationsof success.

The Deming inspectordeflateda self-servingbias by consideringalternativehypothesesfor success(e.g.,luckratherthanskill or knowledge).TraditionallyatFP&L, managerswerenotquestionedaslongastheyachievedgoodresults.Afterthis incident managerswere much more likely to be askedto explain their sue-

cesses.If they couldnot do so, theverdictwould be delivered:“you wererucky”(Walton, 1990,p.61).Thestrategicuseof theaccentwasdesignedto remindman-agersaboutthe earlier incident whereluck produceddramaticresultsthat wereunlikely to be repeated.

Individuals Focuson PeopleRatherthan Situations

Individuals also generatea shallow set of hypothesesbecausesocial settingstendto highlight peopleascauses.In Westerncultureindividualstypically chooseto explaineventsin termsof people’sactionsandtraits ratherthansituationalfac-tors (Gilbert & Malone, 1995;Ross, 1977; Ross& Nisbett, 1991). In a recentstudy observersheardanotherstudentgive a pro-life or pro-choicespeechonabortion.Afterward, observersassumedspeakersheld attitudesconsistentwiththeir speecheseventhoughthe speecheswerederivedfrom scriptswritten by theexperimentersandeventhoughthe observersthemselvestold thespeakerswhichposition to speakfor (Gilbert& Jones,1986).Similarly, Deming(1982)describesa companythatuseda varietyof flammableproductsin their productionprocess.After analyzingthe dataon fires, Deming foundthat thefires werea stableandpredictableoutcomeof theproductionprocess.However, accordingto Deming,thecompanypresidentfocusedhis attentionselsewhere.He “senta letterto everyoneof the 10,500employeesof thecompanyto pleadwith themtosetfewerfires”(p. 325).

People’sactionsare frequentlymoreobvious than their situations.Therefore,when individuals generatehypothesesaboutwhy an eventoccurred,their firsthypothesisis likely to bethat somepersoncausedit (e.g.,Ross& Nisbett,1991).Thistendencyto focuson peopleratherthansituationshasbeendocumentedby-somanyinvestigatorsin so many situationsthat it hasbeencalledthefundamentalattributionerror(Ross,1977; for recentreviewsseeRoss& Nisbett,1991;Gilbert& Malone, 1995).

Organizationsmightrepairthefundamentalattributionerrorby remindingiridi-vidualsto considercausesotherthanpeople,especiallythepeoplewho arelikelyto be closestto any problem: front-line workers.For example,an old militaryadagesays,“Thereare no suchthingsasbadtroops,only badofficers” (Cohen&Gooch, 1990, p. 228). Parallel repairs are found in total quality management(TQM). Ishikawa says,“whenevermistakesoccur, two-thirds to four-fifths ofresponsibilityrestswith management”(Ishikawa,1985,p. ix). Suchmaximsmaypartially repairthefundamentalattributionerrorbecausethey encourageindivid-ualsto lookbeyondthe front line. On theotherhand,they maysimply focustheerroron peopleat a higherlevel. Thus,a betterrepairmaybe onefrom Deming,who tellsmanagersthatof theproblemshehasseen,“94% belongto thesystem”(Deming, 1982,p. 315). IshikawaandDemingboth usevivid statisticsto over-comethe fundamentalattributionerroreventhough it is unlikely thateitherhas


conducteda preciseempirical analysis.Deming’s “94%” is particularlynotewor-thy becauseof its apparentprecision.

Individuals StopSearchingasSoonas TheyGenerateOneHypothesis

Self-servingbiasesandthe fundamentalattributionerror arespecialcasesof amuchbroadertendency:Individuals tend to stopsearchingfor a causeas soonasthey locatea plausible candidatehypothesis(Gregory, Cialdini, & Carpenter,1982;Hoch, 1984).

To counterthis generaltendency,organizationshavedevelopedsome repairsthat arewidely applicableacrossa numberof domains.In onetechniqueknownasthe “Five Whys,” workersat Toyota learnedto ask “why?” five timesbeforetheystoppedgeneratinghypotheses.Whenthey did so,they weremorelikely to find aroot causeratherthan a superficialone.Imai (1986) illustratesthetechniquewiththe following example:

Question1:Answer I:Question2:Answer2:Question3:Answer3:Question4:Answer4:Question5:AnswerS:

Why did the machinestop?Becausethe fuseblew dueto an overload.Why wastherean overload?Becausethe bearinglubricationwas inadequate.Why wasthe lubricationinadequate?Becausethe lubricationpumpwas not functioningright.Why wasn’tthe lubricatingpumpworking right’?Becausethe pumpaxleworeout.Why was it worn out?Becausesludgegot in.

Imai arguesthat by asking “why” five times,workers identified“the realcauseand thereforethe real solution: attachinga strainerto the lubricating pump. Ifworkershadnot gonethrough suchrepetitivequestions,they might havesettledwith anintermediatecountermeasure,suchasreplacingthe fuse” (Imai, 1986,p.50). Another illustration of the Five Whys dealsdirectly with the fundamentalattributionerror: “Problem: He doesn’t managewell. (1) Why?He’s not on thefloor. (2)Why?He’sin theequipmentroom.(3) Why?Thenewestequipmentisn’tworking. (4) Why?Purchasinggavethesuppliera shortleadtime. (5) Why?Poorplanningsystem”(Forum, 1992,p. 54). In general,whenindividualsask “why”thefirst time, they are likely to developanswersthat invokesomesalient,recent,or proximal event(e.g., someperson’sactions).Subsequentwhys are likely tocauseindividualsto think morebroadlyandsituationally.

Although the Five Whys is an admirablecognitiverepairbecauseof its powerandsimplicity, individualsmay find it difficult to executeby themselves.Whenindividualshaveonegood hypothesisin mind, that hypothesisoftenblockstheirability to seealternatives(Gregory,Cialdini, & Carpenter,1982;Gnepp& Klay-man,1992;Mynatt, Doherty,& Dragan,1993).For example,Hoch (1984)found

that subjectswho generatedpro reasonsfor buying a producthadmoredifficultygeneratingconreasonsimmediatelyafterward.

If individualsfind it difficult to generatealternatehypotheseson their own,thenorganizationsmay repairshallow searchby confronting individualswith otherswho are expertin askingquestionsthat reveal deepcauses.At Microsoft, BillGateshasby personalexample,encouragedaculturethat relieson relentlessques-tioning. SaysoneWindows manager,“you go into the meetingsandyou comeoutjust sweatingbecause,if thereis anyflaw, hewill landon it immediatelyandpickitto bits” (Cusumano& Selby,1995,p. 25).Employees“overuse”termsborrowedfrom Gates,like “drill down” asa euphemismfor “going into moredetail” (“WhatBill GatesReally Wants,” 1995).

A similarcognitiverepairis foundin theorganizationthatadministersthe Dem-ing quality prize. Here, official Deming inspectorsexaminemanagersusing atechniquecalled“single-caseborequestions.”They beginwith broadexploratoryqueriesand then relentlesslydelve down into weaknessesandomissionsin theanswersthey receive.Single-caseborequestionssometimesidentify causesthatarequitedeep.Forexample,FloridaPowerandLightoftenhadto dealwith poweroutagesthat occurredwhena treefell on a powerline andseveredit. To improvethereliability of its service,FP&L organizeda unit to trim all thetreesin siteswheredamagehadoccurred,andthuspreventfutureoutages.Managersat FP&Lcongratulatedthemselvesfor creatingaprocedurethat preventedfutureproblems.However,theDeminginspectorswerenotsatisfiedwith theproceduresince~itpre-ventedproblemsonly in areasthathadalreadyexperiencedacrisis.Theysearchedfor a solutionat adeeperlevel, andaskedmanagersa numberof questionsaboutwhat might be consideredforestry! What kind of treesgrow in the region?Dopalmsgrowfasteror slower than oaks?Managersat FP&L realizedtheydid notknow the answersto these questions,and that they had not searcheddeeplyenoughto solvetheir problems.After their experiencewith single-caseboreques-tions,FP&L managersconsultedwith forestersanddevelopeda regularmainte-nanceprocedureto trim treesbasedon their growth ratesandacrossthe entireregion, notjust in areaswheretreeshadpreviouslyseveredlines After participat-ing in sessionsof this kind with the Deming inspectors,managersat the firmlearnedto asksingle-caseborequestionsin their own internaldiscussions,thusinstitutionalizingthis cognitiverepair(Walton, 1990,pp. 57-63).

Individuals GenerateHypothesesthatareNarrowRatherThanBroad

In an ideal world individual learnerswould notonly generatedeeperhypothe-ses;they would alsoconsidera broadratherthan narrow set of potentialhypothe-ses. However, even when individuals generatealternative hypotheses,their“alternatives”often differ only slightly from one another,andall lie within thesamegeneralframe.For example,participantsin oneexperimentwereaskedtoconsiderthe seriousparking problem facedby their university, and they were

10 CHIP HEATH, RICHARD P. LARRICK, andJOSHUAKLAYMAN CognitiveRepairs 11

giventimeto generateas manysolutionsastheycould(Gettyset al., 1987).Com-bined,participantsgeneratedabout300 solutionsthat researcherswere laterableto classify into aboutsevenmajor categories.One category,for example,sug-gestedwaysto reducedemandfor parking(e.g.,by increasingparking fees)andanothersuggestedwaystouseparkingmoreefficiently (e.g.,by segregatingpark-ing slots accordingto size).The averageparticipantproposedabout11 solutionsbut these11 solutionsrepresentedonly aboutthreeof the sevenpossiblecatego-ries.Theauthorsaskedan independentpanelof expertsto compilea completelistof high-qualitysolutions,and they usedthis completelist to assesshow manysolutionsweremissedby eachindividual. The typical participantmissedfrom 70to 80 percentof the high-quality solutions.However, whenasked, individualsbelievedtheyhadmissedonly 25 percent.

Evenexpertsfail toconsiderahrQadrangeof alternativehypotheses.Forexam-ple, onegroup of researchersshowedprofessionalauto mechanicsa “fault tree”that listed a numberof hypothesesaboutwhy a car might not start (e.g.,battery,starting system,fuel system,ignition). Somemechanicswerepresentedwith a“full tree” that containedsevenspecific hypotheses,otherswere given a ‘~prunedtree” that omitted some importanthypotheses(e.g. the ignition system).Theresultsindicatedthat whenhypotheseswereprunedoff the tree,mechanicsdid notadequatelyconsiderthem(Fischhoff,Slovic, & Lichtenstein,1978).

How might organizationsrepair narrow searchby individuals?Individualsmight searchmorebroadlyif they arecuedtothink aboutaproblemfrom differentperspectives.At Sharp,employeesare told to be “dragonfliesbut not flatfish.”Dragonflieshavecompoundeyes andseethings from multiple perspectivesatonce,but flatfish havelargeeyesthatonly lookin onedirection(Nonaka& Takeu-chi, 1995).

The “dragonfly” repair exhortsindividuals to considerdifferent perspectives,but this may bedifficult for individualsto do by themselves.Organizationsmightrepairnarrow searchmoreeffectivelyby encouragingindividualsto- recruitotherswho havedifferent perspectives.A goodexampleof this is providedby Bridge-stoneTire, which conducts“kokai watches”to generatealternativehypothesesforimprovingwork practices.During akokai watcha groupof up to a dozenpeople,from differentareasof a factory,gatherfor a few hoursto watchotherswork. Inone four-hour watcha dozenpeopleidentified 63 potentialdangerswith a newmachine(Walton, 1990,pp.200-201).

The kokai watchhasa numberof featuresthatensurethatwatchersgenerateabroadarrayof hypotheses.First, it mandatesa largenumberof watchers(up to12). Second,it selectswatchersfrom a variety of different areas—inonekokaiwatchthat examineddie andmaterialchanges,watchersincludeda plant visitor,a memberof the humanresourcesstaff, a chemist,anda project manager.“Theideawasthatpeoplecouldobservea process,eventhosewho werestrangersto it,with fresheyes,seeingthings that closelyinvolved workersmight not” (Walton,1990,p. 200).Third, it ensuresthat watchersdo not discardhypothesesprema-

turely. The watchersare instructedto “write downanything,‘Hey, looks like theguy is walkingtoo much,’ or ‘Looks like he’snothandlingthe knife right”’ (Wal-ton, 1990,p. 201).Only afterwatchersgeneratehypothesesindependentlyaretheresultscombinedandfiltered.

Otherorganizationalproceduresalsorepairnarrow individual searchby ensur-ing that individuals generatehypothesesindependently.For example, whenMotorola forms cross-functionalteamsto evaluatenew products,they do notallow employeeswho haveparticipatedin one product team to participate inanotherteamwith a similar product.This prohibition limits the pool of potentialteammembersin a costlyway; evaluationteamsinvolve six to nine peopleandspendtwo to threemonthsto developa businessplanfor thenew product.How-ever, by consciouslydisregardingprevious experience,Motorola allows newteamsto developrecommendationsindependentlyfrom previousteams.At thesametime, Motorola avoids losing the knowledgeof previous“veterans”—theyserveasa “review team” that evaluatestherecommendationsof thenewestteam.2

Otherrepairsensure that a broad rangeof alternativesare consideredsimulta-neously.Somecompaniesdivide aproductdevelopmentteaminto competingsub-groupswhich developseparateproject proposals,andonly later recombinetodebatetheadvantagesanddisadvantagesof theindependentproposals.Again,thisstrategyis costly becauseit is redundant.However, it may have advantagesbecausethebuilt-in independenceensuresthat different subgroupswill approacha problemfrom differentperspectives(Nonaka& Takeuchi,1995,p. 14).

CollectingInformation

In the secondstageof the learningprocessideallearnerscollectinformationtotestandrevisetheir hypotheses.Therearetwo main sourcesof suchinformation:theinformationthatindividualsalreadyhavein their memoryandtheinformationthat they collect from theenvironment.Both kinds of information havepotentialflaws,but individualsmight minimizetheseflaws if theycollectedinformationina rigorousway. However, learnersdo not alwaysactas thoughthey are awareofthe potential flaws in their information—theyfrequently collect only a small,biasedsample.

IndividualsOftenCollectSmallSamplesof Information

Individuals oftencollect only a limited sampleof information becausethey areconstrainedby time or attention.In a classicstudy, Payne(1976) askedhis sub-jectsto chooseone apartmentout of a numberof alternatives,eachof which wasdescribedon severaldifferentdimensions(e.g., rent,cleanliness,landlord quality,noiselevel). Whensubjectschoseamongonlytwo apartments,theytendedto con-sider all of the information beforethey decided.However, individualssearchedasmallerandsmallerpercentageof informationasmore informationbecameavail-


able. Forexample,onesubject,who was decidingamong 12 apartmentscharac-terized on eight different dimensions,looked at only about 25 percentof theinformation beforemakinga final choice.

It would be reasonablefor individual learnersto collect only a smallsampleof

information if theyperformedacost/benefitanalysisanddecidedthat collectingalarge samplewastoo costly. However, thereis evidencethat individualscollectonly a smallsampleof informationbecausethey systematicallyunderestimatethebenefitsof largersamples.TverskyandKahneman(1971)arguethat individualstypically believe that small sampleswill be quite similar to the populationfromwhichtheyare drawn.Theylabeledthis beliefthe“law of smallnumbers”tohigh-light that it contradictsthestatistical“law of largenumbers,”which arguesthatsamplescanyield an accuratepictureof a populationwhenthey are sufficientlylarge.Whenindividualsbelievein thelaw of smallnumbers,they assumethat anysamplewill be sufficient,no matterhow small.

At the extreme,individualsmay not collect any information from theexternalenvironmentbecausethey believe that they alreadyhaveadequateinformationstoredin their head.Organizationsmayovercomethis tendencyby encouragingorrequiring individualsto collectlargersamples.This kind of repairis pervasiveinwritings on TQM. “In promotingstatisticalqualitycontrol,we haveusedthe slo-gan, ‘Let ustalk with data”’ (Ishikawa,1985,p. 200).At manyTQM companiesoneof themainprinciplesof thequalityeffort is “Managementby Fact” (Walton,1990,p. 37).

AndTQM notonly talks aboutdata,it providesindividualswith toolsthat helpthemcollectandanalyzedata.Forexample,six of the“SevenTools” of TQM pro-vide waysto collectdata(e.g.,checksheets)or to simplify anddisplaylargequan-tities of data (e.g., histograms,scatterplots, Paretodiagrams,control charts)(Deming, 1982;Imai, 1986;Ishikawa,1982, 1985;Juran,1992).

IndividualsCollectBiasedSamplesof Information

Individual learnersnotonly collectsmallsamplesof information,theyalsotendto collect samplesthat are biased (i.e., that are unrepresentativeof the largerworld). Considerthecommonclaims that “the other line alwaysmovesfaster” or“it only rains afterI washmy car.” Unlesswe want to believethat a malevolent

spirit is in chargeof suchharassment,theseexamplesdemonstratethatour mem-oriesdo not storearandomsampleof all waiting timesor all rainstorms—wearemorelikely toremembertherainstormsthatspoil thefinish on-ourfreshlywashed

car.Evenwhenindividualscollectinformationfromtheoutsidework1(rath~vtha~ifrommemory),theydonotalwaysattendto themostrelevantandimportantinfor-mation.Below,wediscussanumberof factorsthatmight leadindividual learnersto collect biasedsamples.

Individuals Only ConsiderAvailable Information

As indicatedby thecarwashexample,individualsoftencollectbiasedsamplesbecausetheycollect informationthat is easilyavailable in memory,for example,becauseit is especiallyvivid or recent.The problemis that individualstypicallyassumethat the information that is availableis also mostfrequent,probable,andcausally important (Tversky & Kahneman,1973). This assumptionis oftenwrong. Individuals dramaticallyoverestimatethe likelihood of vivid causesofdeathlike accidentsor homicides,and theyunderestimatethe likelihood of lessvivid causeslike diseaseor strokes.Individuals estimatethataccidentscausedasmany deathsas diseasesand that homicideswere as commonas strokes.In fact,diseasestake 16 timesmorelives than accidentsandstrokestake 11 timesmorelives than homicides. Individuals also overweight recent information. Theyassumethat themostrecentfloodprovidesanupperboundon possibleflood dam-age,andthepurchaseof earthquakeinsurance“increasessharplyafteraquakeandthen decreasessteadily as memoriesfade” (Slovic, Fischhoff, & Lichtenstein,1982,p. 465).

Many organizationsrepair individuals’ tendencyto rely on biased,availableinformationby institutingaprocessthatcollectsinformation moresystematically.At a Motorola division that developsequipmentfor cellularphonesystems,onegrouprealizedthat an availability biaswascausingit to overlookcertaincustom-erswhenit evaluatednew products.The unit assignedaccountmanagersonly tolargeaccounts,sowhenmanagersevaluatednewproducts,theyprimarily consid-eredtheneedsandrequirementsof only largecustomers.However, the unit alsoserveda numberof smallercustomersthat did not havetheir own accountman-ager. Together,thesesmall customersaccountedfor a largepercentageof reve-nues. Motorola overcame the availability bias by developing a FeaturePrioritizationProcess;they surveyedcustomersup to four timesa year and thenweightedall of the inputsbasedon customervolumeandpriority.3

Hospitalsalsohavea varietyof proceduresto forceindividuals to collectinfor-mationmoresystematically.Traumaphysiciansareoften confrontedby vivid butpotentiallymisleadinginformation.Onedoctorstatesthat, contraryto what onemight expect,stabbingsandbulletwoundsare “relatively straightforwardaffairs”becausethey leave“cleartrackson thebody.” Otherinjuries aremoredifficult totreatbecausethey leave no visiblecues. “It would beall too humanto focuson alaceratedscalp—agory butbasicallyinsignificant injury—and miss a fracturedthighbonethathadinvisibly severedamajorartery” (Rosenthal,1994,p. 48). Themedicalprofessionhasdevelopeda seriesof strict protocolsfor traumasituationsthat allow doctors to quickly collect all the relevant information, not just thatwhich is salient. Forexample,whena patientfirst entersthe emergencyroom,physiciansfollow the “ABCs”; they establishairway, thenbreathing,thencircu-lation.4 Forsituationsthat are morecritical, suchascardiacemergencies,proto-cols areevenmorerigorousandspecific.

14 CHIP HEATH, RICHARD P. LARRICK, andJOSHUAKLAYMAN 15Cognitive Repairs

If individuals tend to focuson information that is highly available,it is notter-ribly surprising that they are frequentlyunawareof missing information. Evenwheninformationis present,learnersdo notpayasmuchattentionto whatdoesn’thappenas what does (Agostinelli, Sherman,Fazio, & Hearst, 1986; Newman,Wolff, & Hearst,1980).

Certain professionsand organizationshave learnedto repair the tendencytoignoremissing information. Homicide detectiveslearn to notice the absenceofitemsat murderscenes,sincemany murdererstakebacksomethingthatbelongstothemaftercommitting thecrime.“You look atwhat’s beentakenandyou find outwho it belongedto originally” (Fletcher,1990,p. 75).

A particularlyimportantform of missinginformation is theabsenceof experi-encewith highly unusualevents.Bankexaminersrarely seea bank fail, nucleartechniciansrarelyseeameltdown,airlinepersonnelrarelywitnessa crash(March,Sproull, & Tamuz, 1991; Perrow, 1984). Certain organizationsinstitutionalizeproceduresthatencourageindividualsto payattentionto suchinformation:desjtitothe factthat sucheventsare unlikely to be availablein their own experience.Forexample,attheFederalReserveBank,whichcertifiesthesecurityof banks,seniorbankexaminersdeliberatelyrecountstoriesof failedbanksto keepjunior exam-iners aware that they should be vigilant.5 At one bank’s commercial lendingdepartment,seniorcredit officers would hold seminarsand informal brown-baglunchesto discusspast lending mistakes,particularly in areascharacterizedbyunusualor rareevents(e.g., “problems with highly leveragedcompanies,real

6estate,environmentalliability on contaminatedproperty”). By forcingindividu-als to rehearsesuchinformation,organizationshelp individualslearnfrom vicar-iousexperiencesthat arerarebut highly informative.Furthermore,organizationsremind individuals of potentially painful information that self-servingbiaseswould makethempreferto ignore.

IndividualsCollect BiasedInformationBasedon TheirPreexistingTheories

Researchsuggeststhat individuals tend to think of “facts, experiences,andargumentsthat supportacurrenthypothesismorereadilythan thosethatrefuteit”(Klayman, 1995; seealso,Baron, 1988; Kunda, 1990; Nisbett& Ross, 1980).Thus, when individuals collect information from memory, they may focus oninformation thatsupportstheir preexistingtheories.Individuals may also do thiswhenthey collectinformation from theexternalenvironment.Forexample,whenindividuals collect information from others,they often ask specific, directivequestionsthatare likely to elicit the answertheyexpect(Hodgins& Zuckerman,1993;Zuckerman,Knee,Hodgins, & Miyake, 1995).

The ChicagoBoardof Tradehasa staffof in-houseinvestigatorswhoscrutinizetradesthatmay violate exchangerules.In theseinvestigations,which are obvi-ouslyquitesensitive,it is very importantthat investigatorsdo notcollectinforma-tion that isbiasedby their initial theories.To repairthis tendency,the investigators

are trainedto avoid questionsthatcanbe answeredwith a yes or no response.“This forcesan investigatorto askopen-endedquestionsand allows her to drawout as much informationaboutthe situationaspossible.”By askingopen-endedquestions,investigatorsavoid thepossibility of directing the interview in a waythatelicitsonly informationthat is consistentwith their preexistingtheories.7

Someorganizationshavedevelopedmaxims thatseemdesignedto encourageindividualsto collectdataratherthanrelying on their (potentiallybiased)theories.At BridgestoneTire employeesusetwo Japaneseterms:genbutsu(actualproduct)andgenba(actualplace)(Walton, 1990,p. 194).Thesetermsremind employeesnot to rely on their own theories,but to actuallygo out andinvestigatethe actualproduct in the actualplace where the problemsarose.Anothergroup (Forum,1992)usesa similar cognitive repair theycall thethreeactualrule: (I) Go to theactualplace;(2) Seetheactualproblem;(3) Talkto theactualpeopleinvolved.

Individuals ConsiderOnly Part of theRelevantInformation

Finally, individual learnersmaycollectbiasedsamplesbecausetheytendto col-lect information from only onesmall cornerof theuniverseof information. Thisarisesfrom basiccognitive processes.Memory is associative—whenindividualsretrieve onepiece of information,they tend to think of other information that islinked to it by strong associations,commonfeatures,or similar meaning.Evenwhen individuals collect information from the external environment,they arelikely to collect information basedon the same kind of associativeprocess.Researchin cognitivepsychologyhasshownthat individualsattendto andprocessinformation morecomprehensivelywhen they havea mental schemathat tellsthemwhatinformationis neededin a givensituationandwhereto find it (Ander-son,1995).

Accordingly,organizationscanrepairbiasedinformationcollectionby provid-ing individualswitha schemathat remindsthemof thefull rangeof relevantinfor-mation.Many schemasof this kind canbefoundin thefinancialservicesindustry,whereindividualsmustassessa widevariety of informationto determinewhetherto buy,sell,or lend.At theFederalReserveBank of New York, theBankExami-nationsgroupprotectstheFDIC insurancefund by ensuringthat individualbanksarein soundfinancialcondition.Whenreviewingeachbank,examinersusea rat-ing systemknownas CAMEL: they reviewCapitaladequacy,Assetquality, Man-agement, Earnings, and Liquidity.8 In another bank’s commercial loandepartment,creditanalystsusethe“Five Csof Credit”: Collateral,Capacity,Cap-ital, Conditions,andCharacter.9

Organizationalschemaslike CAMEL andthe Five Cs are likely to encourageindividuals to collect a broaderrangeof information than they would normally

collect.It wouldbeveryeasyfor individual learnersto collectinformationon onlyon themostsalientfactors(suchas cashflow in a loandecision).Although cashflow is certainlyimportant, it canalsobemisleadingor unreliable,particularly in

16 CHIPHEATH, RICHARD P. LARRICK, andJOSHUAKLAYMAN Cognitive Repairs 17

an environmentwhereconditionsare changing.By emphasizingthe Five Cs, abankcanrepair the tendencyof individual analyststo neglectinformation aboutimportantvariablesthatare less obviousor are harderto assess.Forexample,theFive Cs remindsloan officers to considercharacter—Whatare themanagementskills of the owners?Do they havegood personalcredit records?Although theanswersto suchquestionsare quiteimportant,individualanalystsmight forgettoask them in a numbers-orientedenvironmentlike a bank, without a cognitiverepairlike theFive Cs.

Individuals Who Collect BiasedInformationFail to Correctfor Bias

We havediscusseda numberof factorsthat might leadindividual learnerstocollect biasedinformation.However,evenif learnerscollectbiasedinformation,theymight still beableto draweffectiveconclusionaslongastheyrecognizedthebiasandcorrectedfor it. Forexample,supposeanindividualmadethe statement,“the otherline alwaysmovesfaster,”but thenremindedherselfthatsuchsituationsmight beoverly availablein her memory.This kind of correctionimprovestheconclusionsdrawnfrom evena biasedsample.On theotherhand,evenif individ-uals are awarethat they havecollectedbiasedinformation, they may not knowhow to correctfor biasesafterthefact. Forexample,after individualsaskbiasedquestionsandthereforereceivebiasedanswers,theydo not takeinto accounthowmuchtheanswerswerebiasedby their initial questions(Zuckerman,Knee,Hod-gins, & Miyake, 1995).

Becauseindividualsdo notalways correcttheir information for biases,someorganizationsattemptto ensurethatindividualscollectunbiasedsamplesfrom-thestart. Microsoft requiresits softwaredevelopersto usethe sameprogramsandmachinesthat areusedby their customers.Forexample,programmerswho weredevelopingthe new Windows NT operatingsystemranthecurrentday’s version

of theprogramasthey programmedthe nextday’sversion.At Microsoftthispro-cessis known as “eating your own dog food.” It ensuresthat developerscollect alarge,unbiasedsampleof information aboutthe current stateof theprogram.IfWindows NT crashedwhile a developerwasdesigninga new printer driver, hehad to fix theproblemwith NT beforehecould returnto his driver(Cusumano&Selby, 1995,p. 331).Microsoft alsorequiresdevelopersto usethesamemachine

usedby customers,a requirementthathas been“controversialat times” becausedeveloperslike to have the fastest,coolestmachineson their desks. However,whendevelopershavebettertechnologythan the averagecustomerthey collectbiasedinformation abouthow well their softwareprogramsperform.Oneman-agersaid,“every time I’ve hada projectwherethe developershadhardwarethatwasa generationbeyondwhatcustomershad,the [software]alwayshadperfor-manceproblems”(Cusumano& Selby,1995,p. 347).By requiringdeveloperstousethe samemachinesas their customers,Microsoft forcesthem to collect an

unbiasedsampleof information abouttheoperatingspeedandmemorydemandsof thesoftwarethey are developing.

DrawingConclusions

After generatinghypothesesandcollecting information,ideal learnersshouldevaluatethe informationtheyhavecollectedanddrawconclusionsthatareappro-priate andcautious.Researchershavesuggestedthreemain classesof problemsthat real individuals face whenthey interpretevidence. First, they often weighinformation in a way that is not statistically appropriate—forexample, theyemphasizethe importanceof extremeevidencebut they do notemphasizetherel-ative amount of extremeversusnon-extremeevidence.A second,even moreinsidiousproblemis that individualsusetheir initial theoriesto interpretthe evi-dence.While individualsmay readily acceptinformation that is consistentwiththeir initial hypothesis,they casta critical eyeon information thatcontradictsit.Third, asa resultof thetwo previousprocessesandothers,individuals frequentlydrawconclusionsthat areoverconfidentandoverlyoptimistic.

Individuals WeighVivid and ExtremeEvidenceMoreHeavily

Onceindividuals havecollectedinformation,how shouldthey combineit andweighit? An ideal learnerwould weigh information basedon the quality of theinformation.However,actuallearnersdo notalwaysassignappropriateweightstoall aspectsof thedecision.For example,they tend to weigh morevivid, easilyimaginedinformationmoreheavily(Keller & McGill, 1994).They alsofocusonthe extremity or strengthof the availableinformation (e.g.,the warmthof a rec-ommendationletter) withoutadequatelyattendingto the amountor weightof theevidence(e.g., thewriter’s amountof contactwith the recommendee)(Griffin &Tversky, 1992).

If individuals tend to overemphasizevivid or extremeinformation, organiza-tionsmight preventthis by requiring individualsto consciouslyclassifyinforma-tion accordingto its appropriateweight. Many companieshave internal auditgroupsthatexaminetherecordsof companydivisionsto ensurethattheyareusingproperaccountingproceduresand spendingmoneyon legitimateexpenses.Anaudit usuallyuncoversa variety of major andminor“exceptions”(i.e., situationswherecorrectprocedureswerenot followed).Oneauditorsaysthat auditorsmustbecarefulnot to “place toomuchemphasison memorableerrors,e.g.,anerror inthepresident’sexpensereportor the misuseof the companycar.” Oneauditinggroup repairedthis temptationby first classifying eachexceptionas major or

inminorthenconsciouslyignoring theminor issues.

Consistentwith the tendencyto overweightthe extremityof information andignore the amount,individuals frequentlyplace higherweight on onevivid casethan on a much larger sampleof information.JosephStalin is reportedto have

18 CHIPHEATH, RICHARDP. LARRICK, andJOSHUA KLAYMAN CognitiveRepairs 19

said,“The deathof a singleRussiansoldieris a tragedy.A million deathsis asta-tistic” (Nisbett& Ross,1980,p.43). In a studythat supportsthis observation,Bor-gida and Nisbett(1977) showedsome studentsa statistical summaryof howdozensof studentshadratedvariouscoursesin the previousterm.Otherstudentsattendeda panel discussion,during which two or threeupper-divisionstudentsratedeachcourseon a numericalscaleandprovidedsomegeneric,uninformativecomments.Despitethefact that the statisticalsummaryprovidedstudentswith alarger amountof information, individualswho heardthe small sampleof vividinformationweremore likely to changethecoursesthey selected.

Microsoft alsodiscoveredthat individualsdiscountlargesamplesof statisticalinformation.At onepoint, Microsoft startedsurveyingusersto seehow many ofthem foundit easyto usea particularfeature.Softwaredevelopersoftenrefusedtobelievethe statistics.“Theusability groupwould tell thedevelopmentgroup‘Sixout of ten couldn’t dothis.’ And thedeveloper’sreactionwould be, ‘Where’d youfind six dumbpeople?”’(Cusumano& Selby,1995,p. 379).In orderto repairthistendencyto ignorebaserate information,Microsoft madetheinformation morevivid. It built a “usability testlab” wheredeveloperscanwatchrealusersstrugglewith new productsfrom behinda one-waymirror. Insteadof presentingdevelop-erswith pallid statistics,thetestlab presentsthemwith realpeople(albeita mxtchsmallersample).The lab managersaysthat whendevelopersseea user,“twentyideasjust immediatelycometo mind. Firstof all, you immediatelyempathizewiththeperson.Theusual nonsenseanswer‘Well, they canjust look in themanualifthey don’t know how to useit,’ or ‘My ideais brilliant; youjust foundten stupidpeople’...thatkind of stuffjust goesout thedoor...” (Cusumano& Selby, 1995,p.379).This cognitive repair is interestingbecauseit usesonekind of bias (over-weighting of extreme,or vivid information) to fight another(underweightingofstatisticalinformation).

IndividualsUse Their PreexistingTheoriesto Interpret the Evidence

Individuals not only weigh information inappropriately,they also havediffi-culty interpretinginformationindependentlyof their preexistingtheories.Insteadof usingthe information to testtheir theories,they use their theoriesto testtheirinformation.This often leadsthem to discountinformation that disagreeswiththeir preexistingbeliefs.

In a classicdemonstrationof suchdiscounting,Lord, Ross,andLepper(1979)selectedundergraduateswho strongly supportedor opposedcapital punishmentandpresentedthem with two purportedacademicstudiesthat evaluatedcapitalpunishment’seffectivenessusing very different methods.A study using onemethodfound that capitalpunishmentwaseffectiveanda studyusing theothermethodfound it was ineffective(the researcherscounterbalancedwhich methodwasassociatedwith which result). Participantsapplaudedthepositiveaspectsofwhichevermethodsupportedtheir own preexistingtheory, andthey critiqued the

“design flaws” in the other. In fact, after receivingmixed resultsfrom the twostudies,subjectsbecamemoreconvincedof thevalidity of their original position.Seemingly, they regardedthe evidenceas “one good study that supportsmybeliefs,andone lousy studythat drawsthe wrong conclusions.”Individual sub-jects thusfailedto evaluatetheincominginformationseparatelyfrom their preex-isting theories.Unfortunately,similar resultshavebeennotedwith professionalscientists(Mahoney,1976;Koehler, 1993).

Onebankhelped its loan officers repair the way they interpretevidencebyencouragingthemto considera nonstandardtheoryof lending.In mortgagelend-ing, loanofficersoftenlookfor reasonsto denyloansbecauseloansaredifficult tomake(they aresubjectto a mountainof regulations)andpotentiallyquite costly(e.g.,foreclosureon a badloan may costup to 20% of the propertyvalue).Thus,the initial hypothesisin manyloan decisionsis that anapplicantshouldbedenieda loanunlessprovenotherwise.Onemortgageloan departmentgrew at anannualrateof 30 percentby forcingloanofficersto consideranalternativetothestandardhypothesis.Insteadof askingwhetheran applicantshouldbedenieda mortgageloan, it askedwhetherthe applicantshouldbe approved.This reversalled thedepartmentto developspecial programsfor qualified applicantswho had lowincomesor otherspecialcircumstances.11

Individuals use their theoriestodevelopexpectationsaboutwhatis normal,andthey frequentlylabel unexpectedeventsas“problems”or “failures.” Theselabelsmay be misleading,however, particularly in researchand developmentwhereunexpectedeventsmay point theway to importantbreakthroughs.Oneresearchorganizationhasdevelopeda repair that discouragesindividuals from thinkingthat unexpectedeventsarefailures(Sapolsky,1997).JacksonLaboratoriesbreedsmice that exhibitphysiologicalor behavioraltraits that areof interestto medicalresearchers.For example,it sells mice that lack growth hormoneto researcherswho are interestedin understandingthe biology of mammaliangrowth. It foundthat theanimaltechnicians(e.g.,the peoplewho cleanedthe cages)often noticedunusualbehaviorthat wasscientifically important.The mice that lackedgrowthhormonewerediscoveredby a technicianwho noticed a particularmousethatdidn’t grow at a normal rate. Anothertechniciannoticed a mousethat didn’trespondnormally to the loud noisesthatoccurredwhenthecageswerecleaned--its offspring were found to be susceptibleto hereditarydeafness.After severalexperienceslike this whereunexpectedbehaviorproducedimportantdiscoveries,thecompanystartedholding regularmeetingswith animaltechniciansto inquirewhetherthey havespottedanything unusual.Theseforums for highlighting theimportanceof unexpectedeventsare called“deviantsearches.”

CRSS,an architecturalfirm, developeda specialpositionto repairtheproblemof theory-basedinterpretationof evidence.“Most designerslove to draw,to make‘thumbnail sketches’,” says onemanager,but this rush to draw conclusionsisoften premature.CRSScreateda uniquejob description,the “programmer,” toensurethat somemembersof its designteamswerenotallowing their own theo-


ries to dominatetheway they evaluatedinformation from clients. Programmersarenot in chargeof designingor problemsolving, they are in chargeof “problemseeking.”They are trainedto use techniquesthat helpthem to resistprematureconclusions,and thus listenmorecarefullyto clients.“The experienced,creativeI programmerjwithholdsjudgment,resistspre-conceivedsolutionsandthe pres-sureto synthesize...herefusesto makesketchesuntil he knows theclient’sprob-lem” (Peters,1992,p. 402).

Often, organizationsensurethat individuals weigh information effectively byforcing them to interactwith otherswho might weighthe informationdifferently.Oneresearcherhasexploredwhethertrainingasa scientistcurestheproblemsthatotherindividualshavein evaluatingevidence(Dunbar,1995).The answeris no.Forexample,scientists,especiallyyoung ones,often believethat a singleexperi-mentalresult hasjust resolvedan importantproblem.However, when Dunbarstudieda setof microbiologylabsthat hadbeenparticularlysuccessful,hefoundthat they placedmoreemphasison grouplabmeetings.At thesemeetingsanindi-vidual scientistpresentedhis or her resultsto a variety of skeptical,uninvolvedpeers.Whenthe individual scientistpresenteda striking new pieceof evidence(e.g., I havedetectedEnzymeZ in a biological processwhere it has neverbeenobservedbefore),the individual’s peerswere typically quite willing to proposealternateways of interpretingthe evidence(e.g., the samplewas contaminatedwith residualEnzymeZ from a prior procedure).In successfullabs, evenwhenindividual scientistsfailed to weigha particularpieceof evidenceappropriately,their peersdid so for them.Moreover,the most successfullabs were those thatincludedmemberswith different trainingandbackgrounds.Such “lab meetings”arenot limitedto successfulmolecularbiology labs;similarmeetingstakeplaceatventurecapital firms wherefirms decide whetherto allocatemoneyto new ven-tures(Kaplan, 1995).

IndividualsDraw Conclusionsthatare Overconfidentand Overly Optimistic

Imagine that individualshavegenerateda set of hypotheses,collectedsomenew information,and interpretedthe relevanceof the information for the initialhypotheses.How muchconfidenceshouldtheyplacein theconclusionstheyhavedrawn?If individual learnerswereadequatelycautious,their conclusionswouldreflect thedegreeof uncertaintyin the dataon which they are based.Over theyears,researchhas documentedthat individualsoften expressmorecertaintyintheir conclusionsthan is warrantedby the facts available to them (or by theiractualperformance).This kind of problemhasbeendocumentedextensivelyinlaboratorystudies,but alsoin field studiesof individualjudgmentin a varietyofprofessions,like thecivil engineersin theintroduction(Griffin & Tversky, 1992;Lichtenstein,Fischhoff,& Phillips, 1982;Russo& Schoemaker,1992).

Individualsoftenexhibita particularkind of overconfidencethat wemightlabela planningfallacy (Buehler,Griffin, & Ross,1994)or anoptimismbias.Thisopti-

mismbias is pervasivein work environments.Softwaredevelopersat Microsoftoften experienceburnoutbecausethey “grossly underestimate”how long it willtakethem to accomplishcertaintasks(Cusumano& Selby,1995,p. 94). Organi-zationsdo not alwayssuccessfullyovercomethis individual bias.A studyof pio-neerprocessplantsrevealedthat thetypical plantexperiencedactualconstructioncoststhat werealmostdoubletheoriginal projections;similarly, a studyof start-ups showedthat more than 80 percentfell shortof their projectedmarketshare(Davis, 1985). Theseexamplessuggestthat individuals drawconclusionsthatunderestimatethe amountof uncertaintyanderror in their predictions,but theytendto do it asymmetrically—theyrarelyoverestimateaproject’scostor time tocompletion.

Of course,individualsmaydisplayanoptimismbiasbecausetheyconfrontmis-alignedincentives.Perhapsif engineerscorrectlyestimatedthetruecostof a newenergy plant, decisionmakersmight choosenot to build it. However, the realcausesof theoptimismbiasseemto becognitive,sinceindividualsareoverconfi-dent by the samemagnitudeeven in lab experimentsthat rewardaccuracy.Forexample,individualstypically assumethat their predictionsaremoreprecisethantheyare.Whentheyareaskedtosetconfidenceintervalsaroundaquantity,sothattheirconfidenceintervalhasa98 percentchanceof includingthetrueanswer,theyare typically surprisedby the true answernot 2 percentof the time,but 20 to 50percent(Lichtenstein,Fischhoff,& Phillips, 1982;Russo& Schoemaker,1992).

How might organizationsrepairindividual tendenciestoward optimism biasandoverconfidence?Onestrategyis to allow individuals to makeoverconfidentpredictions,thenadjustthem overtly. This wasthe strategypursuedby the engi-neeringprofessionwith its safetyfactors.Microsoft usesa similarstrategyto cor-rect the overly optimistic projectionsof individual softwaredevelopers:It hasrulesabouttheamountof buffertimethatshouldbeaddedto projects.Forreason-ably well-understoodprogrammingchallenges,suchasapplicationsprogramslikeExcel andWord, Microsoft typically addsbuffer time that constitutes30 percentof the schedule.However,for operatingsystemslike Windows,wheredevelopersmustcreatea systemthat has to mesheffectively with numerouspiecesof hard-ware and software, Microsoft may add buffer time that reaches50 percent(Cusumano& Selby, 1995).Similar repairshaveevolvedin otherindustries.AtoneBig Six accountingfirm, whereteamsmustprepareformal plansfor a con-sultingengagement,projectleadersdeveloptheir bestestimatesof time,expense,

t2andcontingencycosts,thenincreasethefinal numberby 15 percent. Thisrepairhas evolveddespitethe fact that this environmentprovidessomeincentivestounderestimatecosts—bidsthat are toohigh may notbeaccepted.

WhenMicrosoft addsbuffer time to a schedule,it correctsthe predictionsofoverconfidentindividuals by overriding them. However, it has also developedproceduresthathelp individualdevelopersdecreasetheir initial level of overcon-fidence.Forexample,thecompanyhasimprovedits schedulesby requiringdevel-opersto createa detailedwork plan that specifieswhich tasksthey will perform

22 CHIPHEATH, RICHARD P. LARRICK, andJOSHUAKLAYMAN CognitiveRepairs

during specific windows of time.Saysonemanager,“The classicexampleis youaska developerhow long it will takehim to do somethingandhe’ll say a month,becausea monthequalsan infinite amountof time.And you say, ‘Okay, a monthhas22 working daysin it. What arethe22 thingsyou’re goingto do during those22 days?’And the guy will say, 0h,well, maybeit will taketwo months.’ Evenby breakingit downinto 22 taskshe realizes,‘Oh, it’s a lot harderthanI thought”’

(Cusumano& Selby,1995,p. 254).Someorganizationsrepair overconfidenceby forcing individuals to interact

with otherswhoare trainedto questiontheir conclusions.Forexample,the Penta-gon for manyyears had what they calledthe “murder board,”a group of experi-encedofficers that reviewedthe plans for importantmissions,with the goal ofkilling themission.Accordingto Pentagonlore, the failedIranian hostagerescueduring theCarteryearswasnotvettedby this boardbecausehighgovernmentoffi-cialsweretooconcernedaboutsecurityleaks.13

Otherorganizationshavedevelopednorms of frank feedbackto ensurethatindividualsquestionothers’conclusionshonestlyandopenly.In its featureanima-tion unit, Disney regularly holds “Gong Shows” where personnel(includingdepartmentsecretaries)canpitch ideasto a group of seniorexecutives.GongShowsmay attract40 peoplewho presenttheir ideato the executivesandotherpresentersfor three to five minutes. The senior executivesare careful to giveexceptionallyfrank feedbackattheendof thesession,highlightingbothgood andbadaspectsof the presentations.“Somebodymay havea greatconcept,but thestory may notbeverygood.[Wecan’tsay] ‘Oh, that’s fabulous.Greatpitchguys!’andwhen they leave,mumble, ‘That wasawful!’....We don’t pull our punches.[Eventually]peoplebeginto understandthat no matterhow good,bad,or indiffer-entthe idea, it canbeexpressed,accepted,andthoughtabout”(McGowan, 1996).

GENERAL DISCUSSION

In this paperwehavereviewedtheliteratureon individual learningusingasimpleframeworkthatconsidersthreebroadstagesof the learning process.We arguedthat ideal learnerswould generatea broadanddeepset of hypotheses,testthemwith a large,unbiasedset of information,anddrawconclusionsin a cautiousandbalancedway. Thepsychologicalliteratureindicates,however,that real individu-als are not ideal learners;they think andact in ways that reducetheir ability tolearneffectively.

Fortunately,individual learnersdo nothaveto go it alone.We havearguedthatorganizationsfrequentlyrepair the shortcomingsof individual learnersthroughthe use of sayings,informal routines, and formal procedures.We believe theexampleswe haveoffered illustrate the tremendouspromise of organizationalsourcesof cognitiverepairs.

23

Nevertheless,wedo not think that cognitive repairswill overcomeevery indi-vidual problem.Cognitiverepairsare heuristics—likethemental processestheyrepair,they arepragmaticandoften efficient, but also approximateand inexact.For example,they may solve 75 percentof individual shortcomingswhile incur-ringonly one-thirdof thecostsof optimalprocedures(e.g.,from economicsor sta-tistics).However, theyare unlikely to be perfect.

Considerthe five whys. It undoubtedlyprompts individuals to think moredeeplyaboutcauses,but it is only a roughheuristic.Why five questionsandnotthreeor seven?And which questions?“Problem: He doesn’tmanagewell.” (1)Why?He doesn’tmanageconflict well. (2) Why?Hegrew up in a dysfunctionalfamily. (3) Why?His parentswerealcoholics...”In this example,theanswerstookan unhelpfuldetouraway from potentialsolutionssometimearoundanswer2.

Evenwhenrepairsarereasonablyeffective, they maystill leave room for fur-therrepair.Consider,forexample,themilitary’s partialrepairfor thefundamentalattributionerror: “There are no bad troops,only badofficers.” This adagemayrepairtendenciesto attributeblameto the peoplewho areclosestto aproblem(thetroopswho areon thebattlefield);however,it merelyfocusesattentionon anothergroup of people.Thus, it may preventindividuals from fixing systemsor proce-duresthat havebasicflaws(Cohen& Gooch,1990).A moreeffectiverepairmightsay, “Therearenobadpeople,only badsystems.”

Otherrepairsmay beimperfectbecausethey fix oneproblemwell, butexacer-bate others.Forexample,the Five Cs mayhelp individual loan officers collectmore kinds of information but they may createsecondaryproblems.First, byemphasizingcharacter,theFiveCsmayprovokethefundamentalattributionerror.Second,althoughthey expandthe set of factorsloan officers will considerin aloan decision,they may alsoinstitutionalizeany tendencythat they mayhavetoignoreotherpotentiallyrelevantfactors.Third, theymayhelp loanofficerscollectinformation,but they do not necessarilyhelptheminterpretit. They seemto indi-cate that eachC shouldbe weightedequally, whereasan ideal statisticalmodelwould weigh someCsmoreheavilythan others.

As thesecaveatsillustrate,cognitive repairsare unlikely to completelyrepairthe shortcomingsof individual learners.Thus,when we assesswhethera givencognitive repair is successful,we must considerthe costs and benefits of therepair.Below,weconsidersix dimensionsthatmayaffectthecostsandbenefitsofrepairs,andthereforetheir success.

TradeoffsAssociatedwith SuccessfulRepairs

In order to be successful,a cognitiverepair mustbe effective—it must mend

some individualshortcomingand improvelearningrelativeto the statusquo. Tobe truly successful,however,a cognitiverepairmustalsobeacceptedin theorga-nization andactivelyused.A repairthatis not implementedis nota repair.

24 CHIPHEATH, RICHARD P. LARRICK, andJOSHUA KLAYMAN CognitiveRepairs

Cognitive repairsare a kind of innovation,and as such,their usewill undoubt-edly be affectedby many of the characteristicsthat havepreviouslybeenmen-tioned in literatureson diffusion andadoption (Rogers, 1995; Scott, 1995). Wewill focuson innovationcharacteristicsthat areparticularlyrelevantfor cognitiverepairs.Cognitiveshortcomingsnot only createthe needfor a repair,they alsolimit what repairsmay succeed.

Below,we considersix dimensions that affect whether a repair will be success-ful: simpleversuscomplex,domain-specificversusdomain-general,familiar ver-susnovel, correctiveversuspreventative,social versusindividual, and top-downversusbottom-up.(We will typically focuson theendpointsof thesedimensions,but they shouldberegardedas continuousratherthandichotomous.)Mostdimen-sionsinvolve tradeoffs.Forexample,qualitiesthat makea repairmoreeffectiveinsolving an individual shortcomingsometimesreducethe chancesthat it will beacceptedandusedby individuals.In the absenceof perfectlyeffectiveandaccept-ablerepairs,we must recognizeandunderstandthe tradeoffsthat makerepairsmore or lesssuccessful.

SimpleversusComplex

Oneobviousdimensionalongwhich cognitiverepairs vary is whetherthey arerelatively simpleorcomplex.Many of therepairswehavediscussedin this paperarestrikingly simple—theyrequirean individual to rememberandapplyaproce-durethat is only a few stepslong(e.g., thefive whys or thephysiciansABCs). Incontrast,many of the proceduresthat are taughtas formal repairsin academicenvironmentsare quite complex,and involve many stagesof sorting, arranging,andcalculating(e.g., formal financialor statisticalanalysis).

Simple repairshaveprofoundadvantagesovercomplexrepairs.First, they aremorelikely to beusedbecausethecostsaresmall; individualswill find it easiertolearnand implementshorterprocedures.By contrast,complexrepairstypicallyrequireextensivebackgroundknowledgeand tax basiccognitive resourceslikeattention andmemory (Bell, Raiffa, & Tversky, 1988; Nisbett& Ross,1980).Thus,whenindividualsencounteracomplexrepair,theyarelikely to perceivethecostsof learning it aslargeand immediate,and thebenefitsof using it as small,uncertain,anddelayed.

Second,simplerepairsare easierto rememberandreconstructthan complexrepairs,and this increasesthe probability that individualswill accuratelyapplythem and accuratelytransmitthemto others.Becausecomplexrepairs requireindividualsto remembera numberof stages,they are more likely to be distortedwhentheyaretransmittedfrom individual to individual. Thisproblemwill bepar-ticularly pronouncedin situationsthat requirelearningby observationandimita-tion (DiMaggio & Powell, 1983). Individuals who learn a repair throughobservationmay find it difficult to infer thecompleterules of behaviorfor com-

25

plex repairsbecauseinformationabouttherulesis incomplete,unavailable,or dis-tributed acrosstimein a way that makeslearningdifficult.

Althoughsimplerepairshaveprofoundadvantagesovercomplexrepairs,theyalso havesome disadvantages.Fundamentally,thetradeoffbetweensimple andcomplexrepairsis a tradeoffbetweeneaseof useandaccuracy.Complexproce-duresare often complicatedbecausethey attemptto be precise.Simple repairsgain easeof useby sacrificing accuracy.Forexample,a simpleaphorismsuchas“don’t confusebrainsanda bull market” suggeststhe correctdirection to adjustone’sjudgment,butprovidesno guidanceaboutexactlyhow muchoneshoulddis-credit the successof an individual trader. To precisely estimatethe amount ofcreditdueto brainsversusthemarket,anindividualwould haveto perform amorecomplexprocedure,suchascalculatingtheoverall marketperformanceandmea-suringan individual’s performancerelativeto thedispersionandcentral tendencyof the market.

Domain-SpeczflcversusDomain-General

Cognitiverepairsalsovary in therangeof contextsto which theycanbeapplied(Nisbett, 1992), with some repairs being relatively more domain-specificandsomebeing moredomain-general.Domain-specificrepairsaretailorednarrowlyfor a specific context(e.g., the FeaturePrioritizationProcessat Motorola or theFive Cs of Credit). Domain-generalrepairs are describedso generally,andabstractlythat theyapplyacrossmostjudgmenttasks(e.g.,theFiveWhysor mosteconomicor statisticalprinciples).

Domain-specificruleshaveat leasttwo advantagesoverdomain-generalrules.First, individuals find it easierto recognizethata domainspecific rule is relevantbecausethesituationitself remindsthem to usetherule (e.g., acreditanalystwhohaslearnedto think aboutthe FiveCsof Creditwill find it difficult to think aboutlendingdecisionswithoutconsideringall five categoriesof information). Second,individualsmay find it easierto applydomain-specificthandomain-generalrules.Consider,for examplealoanofficer whois tryingto applya generalrule like “cal-culatethenetpresentvalue(NPV)” of makingthe loan.Thisdomain-generalruleappliesto many morefinancialdecisionsthantheFiveCs; but it containsnohintsabouthowit shouldbeappliedto aloan decision.In contrast,theFiveCspointoutspecificaspectsof the loandecisionthat might affectthe loan’squality.Similarly,securitiestradersmight find it hardto benefitfrom a domain-generalwarningagainstself-servingbiases(e.g., “pay attentionto situationaldeterminantsof suc-cess,anddon’t over-attributeachievementto personalcharacteristics”).In con-trast,they areunlikely to missthepoint of a moredomain-specificwarningnot toconfusebrainsanda bull market.

Although domain-specificruleshaveadvantages,they also havelimits. Theirspecificcontentwill makethem morelikely to spreadwithin their domain,but itmay also preventthem from spreadingacrossdomains.For example,engineers

26 CHIPHEATH, RICHARD P. LARRICK, andJOSHUA KLAYMAN CognitiveRepairs 27

havesafetyfactorsandsoftwaredevelopershavebuffer time, but knowing aboutsafetyfactorsdoesnotautomaticallysuggesttheneedfor buffertime.Andevenasingleindividualmay usea ruleeffectivelyin onedomainbut fail to-seeits impli-cationsfor another.Auditorsareoftenquitegoodat ignoringtheirpreexistingThe-onesabout a client’s financialhealth whenthey investigateauditingproblems.However, they arelesslikely to doso whentheyconfront similarproblemsoutside

theauditingdomain,evenif theproblemrelatesto their otherprofessionalactivi-ties(Smith& Kida, 1991).

A secondlimitation of domain-specificrepairsis thattheyaretightly tailoredtofit a particulartaskandenvironment.Becauseof this tight fit, they may be lesssuccessfulthandomain-generalrepairswhenthe task environmentis in flux. Abuffer factor designedduring a specific period of time—”multiply all time-to-delivery estimatesby 1.5—maylose its effectivenesswhentechnologicaloreco-nomic conditionschange.Considerthat Microsoft hadto developseparatebufferfactorsto repairoverconfidencein applicationsandoperating-systents?.In general,domain-specificruleswill be helpful in companiesor divisions wherethe tasksandenvironmentsare stableover time,while domain-generalapproacheswill behelpful in situationswhere tasksand environmentschangefrequently(e.g., athigher levels in anorganizationwheretasksanddecisionsareless routine).

A potential method of combining the advantagesof domain-specific anddomain-generalrulesmaybeto give individualsadomain-specificrepairandthentrain themto generalizethat repairto otherdomains(Fong& Nisbett, 1990;Lar-rick, Morgan,& Nisbett,1990).Individualstypically find it easierto generalizebyanalogyfrom onespecificinstanceto anotherthanto mapfrom generaLprinciplesdown to specifics(Bassok,1990).For example,peoplewho learnto ignoresunkcostsin financial investmentscorrectly recognizethat this rule appliesto invest-mentsof timeaswell (Larrick, Morgan,& Nisbett, 1990).Similarly, amanagerinindustrymay find it easierto apply the specific military adageabouttheir beingno suchthing as badtroops” thanto applya moregenerallessonaboutthe fun-

damentalattributionerror.

CorrectiveversusPreventative

Cognitiverepairsalsodiffer in whethertheypreventorcorrecttheshortcomings

of individuals. Correctiverepairsinterveneduring or aftera particularcognitiveprocess(e.g.,theaccountingteamthatcorrectstheir tendencyto overweightvividexceptionsby forcingthemselvesto consciouslyclassifyeachexceptionas majoror minor).At theextreme,a correctiverepairmight only interveneat thevery endof a processto correctthe overall outcome(e.g., safety factors). Preventativerepairs interveneearly in a cognitive processbefore shortcomingshavehad achanceto act. Microsoft preventsdevelopersfrom acquiring a biasedsampleaboutthe speedof their programsby forcing them to developprogramson thesamemachinesusedby customers.

Someshortcomingsareeasierto correctthanothers.Forexample,whenashort-coming arisesbecauseindividualshavethewrong rule, theymay not find it diffi-cult to substitutea different rule (Wilson & Brekke, 1994).Traumaphysiciansmay learnto checkairwaybeforebreathing,andaccountantsmay learnto ignorevivid butminorexceptions.In general,correctiverepairswill beappropriatewhenindividualsaccepttheneedfor a repair andthey understandhow to executethecorrection.

However,whena shortcomingarisesbecauseof somebasiccognitiveprocess,organizationsmay needto intervenemore forcefully by bypassingor eliminatingthe faulty process(Arkes, 1991;Wilson & Brekke, 1994).Forexample,individu-als may find it difficult to generatea broadand independentset of hypothesesbecauseassociativememoryleadsthemto considerthesamealternativestheypre-viously considered.Theoretically,Motorola could instruct individuals who aredevelopinga new consumerproductto “ignore whatyou’vedonein thepastandapproachthis problemcreatively.”However, individualsmight find it difficult toignoretheir previoussolutions.Thus,Motorolapreventsthe problemby prohibit-ing themfrom servingon morethan oneproductdevelopmentteam.Similarly,theChicagoBoardof Tradecould warnits investigatorsto discountthe answerstheyreceivewhenthey ask leadingquestions.Insteadit preventsindividual investiga-tors from askingyesor noquestions,andthusensuresthattheyreceivelessbiasedinformation in thefirst place.

Familiar versusNovel

Repairsmayalsovary in theextentto which theyarenovel ratherthan familiar.Novel repairsrequireindividuals to changetheir assumptionsor to learna proce-durefrom scratch.Forexamplethe“programmers”attheCRSSarchitecturalfirmhad to learnto resist their tendencyto sketchsolutionsbeforeevaluatingall theinformationfrom aclient. On theotherhand,familiar repairsbuild on preexistingknowledge(e.g.,theCAMEL schemafor bankexaminersor theABCs for traumaphysicians).Theserepairshavefamiliarcontent-traumaphysiciansknow that theyshouldattendto breathingandcirculation,andbankexaminersknow they shouldpay attentionto capital andearnings.They also havea familiar form—theyareorganizedby a simple acronymthat individuals alreadyhave as a part of theirmentallexicon.

Familiarrepairsmaybeat anadvantageovernovel repairsbecausetheyarelesscostlyto useand their benefitsmay be moreapparent.CAMEL and the ABCsreducecostsby usinga familiar acronymto remindindividuals to collecttypesofinformation that they know they shouldbecollecting. In contrast,the CRSSpro-grammershad to work hardto overcomethebehaviorstheyhad learnedas archi-tects,andthey may havequestionedthebenefitsof theelaboratenew proceduresthey werebeing taught.Familiarrepairsare alsoless likely to provokeresistance


thannovel repairs.Anythingthatrequiresindividualstothrow outold practicesoradoptnewbeliefsmay be technicallyandpsychologicallydifficult.

However, familiar repairsmay sometimesbe too familiar for their own good.First, theymay be lesslikely to createenthusiasm.If individualsthink that a newrepairdiffers only trivially from currentpractice,they may seeno advantageto it.Becauseindividualsoftenignorethefamiliar, would-bechangeagentsoftenstriveto createthe perceptionthat their programsare novel andunique(Abrahamson,1996).Second,familiar repairsmaybesubjectto distortion.If a repairseemspar-tially familiar, individuals may neglectits unfamiliar aspectsor force them tomimic themore familiar aspects(a processthat psychologistscall assimilation).Forexample,thepropertechniquefor brainstormingrequiresa specificsequenceof steps:first, a creative, idea-generationstagewhich doesnot allow criticism,then a stagewhereideasare evaluatedand selected.Although organizationsfre-quently reportthat they usebrainstorming,carefulexaminationrevealsthat theorganizationsare actually engagedin a more familiar activity: a basic businessmeeting(Zbaracki,in press).The novel aspectsof thebrainstormingprocedure,suchas theseparationof stagesandthe“no criticism” rule, areoften lostas brain-storming is assimilatedto the more familiar practice of the standardbusinessmeeting.In theend,only the attractivelabelremains.In situationswhereassimt-lation is a problem,repairsthat are novel may be less likely to suffer distortionthanrepairsthat are morefamiliar becausenovel repairsdo not evokethe preex-isting knowledgethat leadsto assimilation.

SocialversusIndividual

Manyof the cognitiverepairswehaveconsideredaresocial;theywork becauseindividuals interactwith others (e.g., single-caseborequestionsin the DemingPrizeorganization,or the murderhoard at the Pentagon).Otherrepairsare indi-vidual; individualsapply themto their own learningprocesseswithout theinter-vention of others(e.g., individuals learn to usethe Five Whys, and individualinvestigatorsat theBoardof Tradelearnto avoid“yes or no” questions).

In general,we suspectthatmanysuccessfulrepairswill be socialbecauseindi-vidualsmaynotrecognizetheneedto repairthemselves.Theverycognitiveshort-comings that organizationsmight hope to repair will make it difficult forindividualsto seetheirown flaws. As wehavediscussed,individualstendtoretaincurrentassumptionsin the faceof conflicting data(Klayman, 1995). Also, theyinterpreteventsin waysthat protecttheir self-image;theyavoidpotentiallythreat-ening feedback(Frey, 1986;Larrick, 1993)andattributetheir poor outcomestoluck or forcesoutsidetheir control.Although individualsmay find it hardto rec-ognizetheir own biases,they may find it easierto recognizethebiasesof others.Many of the repairswedocumenthavethefeelof (friendly) socialgamesmanship.Forexample,learnersatFP&L did notconsiderthehypothesisthatthey hadbeen“rucky”—their colleaguesconsideredit for them.Similarly, duringweeklymicro-

biology lab meetings,researchersdid nothaveto suggestalternativewaysof inter-pretingtheir evidence,their peersdid so. Social competitionamongindividualsaids the spreadof repairsevenwhenindividuals are overconfidentand believethey would havedonejustaswell without therepair.

Social repairsdo haveto overcomesomedisadvantages.Forexample,individ-uals may not appreciateotherswho attemptto repair their biases,and they maydismisstherepairattemptsas theproductof picky or abrasivepersonalities.Thus,socialrepairsmaybemoresuccessfulwhenan individual understandsthat hisorhertormentorsareplayinganimportant,formalrole. Individualsmay find it easierto entertainan antagonist’scritiqueswhenhe orsheis labeledasa “devil’s advo-cate,”or whenthe individual isappearingbeforethe“murderboard.”Disneyclar-ified the roleof the evaluatorsandtheoccasionby establishingits norm of frankfeedbackandby labelingits tryoutsas “The Gong Show.”

Eventually,social repairsmaybetransformedinto individual repairsasindivid-ualslearnto imitate the patternsof analysisforcedon themby others.In orderforindividualsto learn,they needvivid, immediatefeedback.Social encountersarelikely to provide a key sourceof such feedback.For example,when Demingexaminersask single-case-borequestions,or when lab colleaguestry to shootholes in a lab presentation,individual learnersmay eventually learnto engageinpreemptiveself-criticisminorderto lookbetterduringsocialencounters(Tetlock,1992).(Many academicpapersarebetterbecauseauthorslearnto mentallysimu-latepotentialreviewer’scomments.)Such repairsinvoke socialforcesat two dif-ferentlevels: individualswho anticipatesocial interactionmay bemoreawareofsomeof their own shortcomings,andthenactualsocial interactionmayovercomeadditionalshortcomingsthat individualscannotrepairon their own.

Top-DownversusBottom-Up

Somecognitiverepairsoriginatefrom “top-down” within an organization.Typ-ically theserepairsare deliberatelydesignedand implementedby managersoroutsideexperts.Othersarisefrombottom-upthroughinformal observationorser-endipitousdiscoveryfrom the peoplewho aredoingthework.

The sourceof the repair is importantbecauseit is likely to affect its form. Ingeneral,bottom up repairs,such as organizationaladages,will be simpler andmore domain-specificthan top-down repairsdesignedby technicallysophisti-catedengineers,statisticians,ormanagementgurus.Thelocal originof bottom-uprepairs may also makethem feel more familiar andacceptablethan top-downrepairs.Thus,the origin of a repair will be highly correlatedwith many of thetradeoffswehavealreadydiscussed.

More importantly, theorigin of therepair is alsolikely to affect how potentialadoptersperceiveit. Top-downrepairsmay beperceivedwith suspicionor hostil-ity preciselybecausethey originateoutsidetheorganizationor are imposedfromabove.Front-line workers may doubtthat outsiderslike consultantsunderstand

30 CHIPHEATH, RICHARDP. LARRICK, andJOSHUAKLAYMAN Cognitive Repairs 31

their situationwell enoughto makewiserecommendations.Whenmanagerssug-gesta repair,theyseemto imply thatemployeeshavebeenperformingpoorly orcannotbe trustedto perform theirjob correctly. If so, thenindividualsmay resistadoptinga repairbecauseof thesamekind of self-servingbiaseswediscussedear-lier.

Top-downrepairsmayalsoberesistedbecausethey will beperceivedasdrivenby politics or fashionandnotby thedemandsof thetask.Sometop-downrepairsmay be resistedbecausethey seemtoo political. Particularly when top-downrepairsrely onfixed procedures,they mayprovokeresistancebecauseindividualsmaythink thatthey aredesignedto centralizecontrolor removeindividual-discre-tion. Othertop-downrepairsmay be resistedbecausethey seemto be driven byfashion.Institutional theoristscontendthatorganizationsadoptnew practicesforreasonsother than greaterefficacy. Organizationalmembersmay sharesimilar,cynical intuitions(Adams,1996),andwill resist repairsthat theyseeasmerewin-dow dressingor as this year’s fad. Whenindividualshave trouble recognizingtheir shortcomings,then they may be particularly likely to attribute top-downrepairs to politics or fashion becausethey will not recognizethe repair’s truevalue.

Bottom-uprepairswill oftenbenefitfrom theirlocal, homegrownorigins.Localrepairs havea meaningfulhistory that makesthem memorableandappealing.Even a repair that is potentially threatening,suchas “you were rucky,” maybemoreacceptableif organizationalmembersseeit as their own invention.Justaslawyersareentitled to tell lawyer jokes,organizationalmembersare entitled todevelopself-criticalrepairsandto conveytheir insider statusby usingiliem. Andhomegrownrepairsevokea strongersenseof ownership;at the sametime thattheycall attentionto a potentialshortcoming,they also give then’~ercreditfor fix-ing it.

Conclusions:SuccessfulRepairs

Wehaveconsideredsix differentdimensionsalongwhich cognitiverepairscanbe classified.For example,the physician’sABCs are simple, domain-specific,corrective,familiar, individual, andtop-down.Although wehavesuggestedsomeadvantagesanddisadvantagesofeachendpointof eachdimension,webelieve-thatour discussionsuggestsat leastsomepreliminaryconclusionsaboutsuccessfulrepairs.For example,becauseindividuals have limited faculties, organizationswho wishindividuals to learncomplex,domain-generalrepairswill find them-selvesdevotinga greatdealof scarcetime,money,andeffort to ensurethat suchrepairs are learnedandused. Similarly, becauseindividuals are overconfidentabout their own conclusions,they may not spontaneouslyexecuteindividualrepairsto correcttheir own biases.Basedon theadvantagesanddisadvantageswehave considered,we suspectthat the most successfulrepairs will be simple,domain-specific,socially administered,andevolvedfrom bottom-upratherthan

developedfrom top-down.Wefind this conclusionintriguing becauseit describesrepairsthat differ sharplyfrom thosethat are recommendedin academiclitera-tureson decisionanalysis,statistics,andeconomics.

Implications for Research

Cognitivepsychologistsoften think of peopleasruggedcognitive individual-ists, constrainedby their own cognitive limitations anda poor environmentinwhichto learn.Cognitiveresearcherscontinueto argueoverhow well individualsactuallyperform thecognitivetasksthey encounterin their lives (seeAnderson,1991, and accompanyingcommentaries,and the debate betweenGigerenzer,1996; andKahneman& Tversky, 1996). However, it is importantto rememberthatmuchof whatpeopledo, including muchof their cognition,takesplacein thecontextof organizationsandothersocialstructures.

Somerecentapproachesin psychologydo explorecultural andsocial contribu-tions to individual learning.For example,work on “transactivememory” (Weg-ner, Erber, & Raymond, 1991; Liang, Moreland, & Argote, 1995) shows howindividuals reducetheir memorylimitations by distributing memory-heavytasksacrossmultiple people.Thus,therecanbecollectivememorythatdoesnot residein anyindividual. Ourconceptof organizationalrepairsisin thesamespirit, but itdealswith “higherorder” cognition:Reasoninganddecisionmakingcan alsobeimprovedthroughsocialstructureandculturalbootstrapping.

Wealsobelievethat organizationalpsychologistscouldbetterunderstandorga-nizationalprocessesif they thoughtmoreaboutthecognitiveprocessesof the indi-vidualswho makeup the organization.Researchtying individual psychologytoorganizationalbehaviorcertainly hasa long and venerablepedigree(March &Simon, 1958), but recently,some researchershaveexpressedconcernthat thatapproachis still underutilized.For example,in thecontextof institutionaldiffu-sionprocesses,Zucker(1991)haswarnedthat “without a solid cognitive, micro-level foundation,werisk treatinginstitutionalizationasa blackbox attheorgani-zationallevel, focusing on contentat the exclusionof developinga systematicexplanatorytheoryof process”(p. 105).And Miner (1994)warns,“most evolu-tionarydiscussionsof organizationalchangediscussroutinesasthoughtheyexistindependentof individualhumanbeings”and“evoke imagesof disembodiedenti-tiesremovedfrom day-to-dayhumaninteraction” (p.87). Knowledgeof individ-ual cognitioncanbecrucial to understandingwhy thingshappenastheydoiri-anorganization.Forexample,why do engineeringfirms usea systemby which thebestengineersmaketheir bestestimates,only to havethefirm second-guessthemby addingahugesafetyfactor?Any explanationwould beincompletewithout anunderstandingof individualoverconfidence.Understandingtheabilities andcon-straintsof individualspermits a kind of cognitive archaeologyof organizationalpracticesthatmay allow organizationalresearchersto betterunderstandwhy cer-tain rules,norms,andproceduresdevelopin organizations,andwhy othersfail.

32 CHIPHEATH, RICHARD P.LARRICK, andJOSHUAKLAYMAN CognitiveRepairs 33

Implicationsfor Practice

Managerswho think explicitly aboutcognitiverepairswill, we think, be in abetterposition to foster improvementsin their organizations.Managersalreadythink aboutfactorssuchasincentivesystemsandinformation-technoLQgyastoolsto fosterlearningandinnovation.Webelievethatcognitiverepairswill bea usefuladditionto thetoolbox.Managerswho consciouslyconsiderindividualcognitionmay beableto recognizea larger numberof repairopportunitiesandmay targettop-downrepairsmoreeffectively. Furthermore,they may designmoreeffectiverepairsif they takea cognitiveapproachandconsiderrepairdimensionslike thesix we discussedearlier.

Evenwhenrepairsare developedbottom-upratherthantop-down,a managerwho is informedabout individualcognitionmight havea positive influence.Ashas beenobservedby researcherswho think aboutevolutionaryapproachestoorganizations,oneof thecritical componentsof organizationallearningis to startwith arich andvariedpoolof alternativepractices(Levitt & March, 1988;Miner,1994).Savvymanagerscanenhancethe poolof alternativesthatareavailablebyteachingorganizationmembersabouttheconceptof repairsandby encouragingthem to identify existingrepairsand to seeknew repair opportunities.Managerscanrecognizeandrewardindividualswho discovercognitiverepairs,and-they-candisseminateeffective repairs via demonstration,training, communication,androtationof personnel.

Managersmayalsofind it very importantto think aboutcognitiverepairswhenthey evaluateexisting organizationalpractices. Consider, for example, a newexecutivewho discoversthather developmentgroup is split up into six separateteams,eachtrying to solvethe sameproblemwithout any communicationwitheachotheruntil well into the process.Thismight seemlike a paradigmof bureau-cratic inefficiency. Yet, this is the kind of systemthat Motorola has found to beeffectivein generatinga broaderset of options.Without understandingthepoten-tial valueof this repair,thenew executivemight be sadly surprisedby theresultsof streamliningtheprocess.

Managersmight also benefit from a cognitive approachbecausecognitiverepairs, like other innovations,may suffer from the law of unintendedconse-quences.A repair that is intendedto fix oneproblem can well endup creatinganother.Recall, for example,Microsoft’s laudableattemptto makethecustomermoresalientto programdevelopersby havingthemwatchlive-customersattemptto use their productsin theusability testlab.The testlab repairedthedeveloper’stendencyto beunmovedby cold statistics,but it probablyexacerbatedtheir ten-dencyto believethat small sampleswere reliably representative.In responsetoseeingonecustomerin thetestlab,developersmight wastetimealteringa featurethat would havebeenokay for mostcustomers.Managerswho takea cognitiveapproachwould, we hope,be more likely to avoid unwantedside effects, or atleast be in a betterpositionto recognizeandcopewith them.

FinalWords

In contrastto Hamlet’senthusiasm,we argue that thereis good reasonto beawareof the limitations of individual learners.Peoplearenot “infinite in facul-ties” or “noble in reason.”As individuals,wemakesystematicmistakesthatcom-promiseour ability to learnfrom our experiencesandto understandtheworld.Onthe otherhand,wemortalsarenot all fools.We areableto form socialstructuresthat havethe potentialto magnify someof our abilities andto minimize someofour failings. In this paperwe haveconcentratedon demonstratingthat effectiveorganizationalrepairshappen.We do not meanto imply that organizationscureany and all individual shortcomings,nor even that organizationsalways makethingsbetterratherthanworse.Nevertheless,wedobelievethattheorganizationsin which we work canprovideus with norms, reminders,protocols,andproce-duresthat helpus movebeyondour individualshortcomings.

ACKNOWLEDGMENTS

We thanktheGraduateSchoolof Business,Universityof ChicagoandtheFuquaSchoolof

Business,Duke University, for researchsupport.For helpful commentson this project,we

thank participantsatRod Kramer’s 1996 conferenceon OrganizationalLearning; we alsothank Jim March, Cade Massey, Elizabeth McSweeny, Sim Sitkin, Dick Thaler, Elke

Weber, andJohnWright.

NOTES

1 We primarilywantto distinguishproblemsof incentivesfrom problemsof mentalprocess.Inthis reviewwe will not distinguishbetweenmentalerrorsthat arisefrom “motivatedreasoning”andthosethat arise from “colder” processes.

2. Personalcommunication,Abhay Joshi.3. Personalcommunication,RobertAlanFisch.4. Personalcommunication,JanElsbach.5. Personalcommunication,FranciscoBayron.6. Personalcommunication,DeanSiewert.7. Personalcommunication,JustinScottBradley.8. Personalcommunication,FranciscoBayron.9. Personalcommunication,DeanSiewert.

10. Personalcommunication,Ken Myszkowski.

II. Personalcommunication,Leo Tucker.12. Personalcommunication,RaymondStukel.13. Personalcommunication,JohnPayne

REFERENCES

Abrahamson,E. (1996).Managementfashion.Academy ofManagementReview, 21, 254-285.Adams,S. (1996).The Dilbert principle. New York: HarperBusiness.

34 CHIPHEATH, RICHARDP. LARRICK, andJOSHUAKLAYMANCognitiveRepairs 35

Agostinelli,0, Sherman,S. J.,Fazio,R.H., & Hearst,E. 5. (1986).Detectingandidentifyingchange:Additionversusdeletion.Journal ofExperimental Psychology. Human Perception and Perfor-mance, 12, 445-454.

Anderson,J.R. (1991).Ishumancognitionadaptive? Behavioral and Brain Sciences,14,471-517.Anderson,JR.(1995).Cognitivepsychologyand its implications.New York: W.H. Freeman& Co.Arkes, H. R. (199 I). Costsandbenefitsofjudgmenterrors: Implicationsfor debiasing.Psychological

Bulletin, 110, 486-498.Baron,J. (1988).Thinking and deciding. Cambridge,UK: CambridgeUniversityPress.Bassok,M. (1990).Transferof domain-specificproblem-solvingprocedures.Journal ofExperimental

Psychology: Learning, Memory, and Cognition, 16, 522-533.Bell, D. E., Raiffa,H., & Tversky,A. (1988).Decision making: Descriptive, normative, and prescrip-

tive interactions.Cambridge:CambridgeUniversityPress.Borgida,E., & Nisbett, R. E. (1977). The differential impactof abstractvs. concreteinformationon

decisions.JournalofAppliedSocialPsychology7, 258-271.Buehler,R., Griffin, 0., & Ross,M. (1994).Exploring the“planning fallacy”: Why peopleunderesti-

matetheir taskcompletiontimes. Journalof PersonalityandSocial Psychology67, 366-381.Cheng,P. VV., & Holyoak,K. J. (1985).Pragmaticreasoningschemas.CognitivePsychology,17, 391-

416.Cohen,L. 1. (1981).Canhumanirrationality beexperimentallydemonstrated?Behavioral andBrain

Sciences, 4, 317-331.Cohen,M. D. (1991). Individual learningandoganizationalroutine:Emergingconnections.Organi-

zation Science, 2, 135-139.Cohen,E.A., & Gooch,J. (1990).Military misfortunes:The anatomyoffailurein war. New York: Free

Press.Cusumano,M. A., & Selby,R. W. (1995).Microsoftsecrets.New York: FreePress.Davis,0. (1985,March-April). New projects:Bewareof false economies.Har’.’ard BusinessReview,

95-11)1.Deming, W. E. (1982).Out of thecrisis. Boston:MassachusettsInstituteof TechnologyCenterfor

AdvancedEngineeringStudy.

DiMaggio, P.J., & Powell Xv. W. (1983).The iron cagerevisited: Institutionalisomorphismandcol-lective rationality in organizationalfields.AmericanSociologicalReview,48, 1,47-160.

Dunbar,K. (1995).Howscientistsreally reason:Scientific reasoningin real-world laboratories.In R.J. Steinberg& J. Davidson(Eds.),The natureof insight.Cambridge,MA: MIT Press.

Eisenhardt,K. M. (1989). Agency theory: An assessmentand review. Academyof ManagementReview, 14(1),57-74.

Eppel,D., Argote, L., & Devadas,R. (1991).Organizationallearningcurves:A methodfor investigat-ing intra-planttransferof knowledgeacquiredthroughReamingby doing. OrganizationalSci-ence, 2, 58-70.

Fischhoff,B., Slavic,P.. & Lichtenstein,5. (1978).Fault trees:Sensitivityof estimatedfailure proba-bilities to problemrepresentation.Journal of ExperimentalPsychology:Human PerceptionandPerformance,4, 330-344.

Fiske, S. T.,& Taylor,S. E. (1991).Social cognition(2nded.). New York: McGraw-Hill.Fletcher,C. (1990).Whatcopsknow.New York: PocketBooks.Fong, G. T., & Nisbett,R. E. (1990).Immediateanddelayedtransferof trainingeffectsin statistical

reasoning.Journalof Experimental Psychology:General, 120, 34-45.ForumCorporation(1992).P.R.O.G.R.E.S.S.tool kit. ForumCorporation.Frey, D. (1986). Recentresearchon selectiveexposureto information. Advancesin Experitnental

SocialPsychology19, 41-80.Funder,D.C. (1987).Errorsandmistakes:Evaluatingtheaccuracyof socialjudgment.Psychological

Bulletin, 101, 75-90.

Gettys,C.F. etal. (1987).An evaluationofhumanactgenerationperformance.OrganizationalBehav-ior andHumanDecisionProcesses,39, 23-51.

Gigerenzer,G. (1996).On narrow norms and vague heuristics:A reply to KahnemanandTversky(1996).PsychologicalReview,103, 592-596.

Gilbert, D. T., & Jones,E. E. (1986). Perceiver-inducedconstraint: Interpretationsof self-generatedreality. Journalof Personalityand SocialPsychology50, 269-280.

Gilbert, D. T.,& Malone,P.S.(1995).The correspondencebias.PsychologicalBulletin, 117, 21-38.Gnepp,J., & Klayman,J. (1992).Recognitionof uncertaintyin emotionalinferences:Reasoningabout

emotionallyequivocalsituations.DevelopmentalPsychology28, 145-158.Gregory, Xv. L., Cialdini, R. B., & Carpenter,K. M. (1982).Self-relevantscenariosas mediatorsof

likelihoodestimatesandcompliance:Doesimaginingmakeit so?Journalof PersonalityandSocial Psychology43, 89-99.

Griffin, 0., & Tversky,A. (1992).Theweighingof evidenceandthedeterminantsof confidence.Cog-nitive Psychology24,411-435.

Heath,C.,Knez,M., & Camerer,C. (1993).The strategicmanagementoftheentitlementprocessin theemploymentrelationship.StrategicManagementJournal, 14, 75-93.

Hilton, D.J.(1995).The socialcontextof reasoning:Conversationalinferenceandrationaljudgment.Psychological Bulletin, 118, 248-271.

Hoch,S. J. (1984).Availability andinterferencein predictivejudgment.JournalofExperimentalPsy-chology:Learning, Memory,andCognition, 10, 649-662.

Hodgins,H. S.,& Zuckerman,M. (1993).Beyondselectinginformation: Biasesin spontaneousques-tionsandresultantconclusions.Journalof ExperimentalSocial Psychology29, 387-407.

Holland,J. H., Holyoak,K. J., Nisbett, R. E., & Thagard,P. T. (1986).Induction: Processesof infer-ence,learning, anddiscovery.Cambridge,MA: BradfordBooks/MITPress.

Hynes,M., & Vanmareke,E. (1976).Reliabilityof embankmentperformancepredictions.Proceedingsof theASCEEngineeringMechanicsDivision SpecialtyConference.Waterloo,Ontario,Can-ada:Universityof WaterlooPress.

lmai, M. (1986).Kaizen: The keyto Japan’scompetitivesuccess.New York: McGraw-Hill.Ishikawa,K. (1982).Guideto quality control (2ndrevised ed.). HongKong: Asia Productivity Orga-

nization.Ishikawa,K. (1985).Whatis total qualiry control? TheJapaneseway.EnglewoodCliffs, NJ: Prentice-

Hall.Juran,J.M. (1992).Juranon quality bydesign:The newstepsfor planningquality into goodsandser-

vices.New York: FreePress.Kahneman,D., Slovic,P., & Tversky,A. (1982).Judgmentunderuncertainty:Heuristicsand biases.

Cambridge,England:CambridgeUniversityPress.Kahneman,D., & Tversky,A. (1996).On therealityof cognitiveillusions. PsychologicalReview,103,

582-591.Kaplan, J. (1995).Start-up:A silicon valiry adventure.Boston:Houghton,Muffin Co.Keller, P. A., & McGill, A. L. (1994).Differencesin therelative influenceof productattributesunder

altemativeprocessingconditions: Attribute importanceversusattributeeaseof imagability.Journalof ConsumerPsychology3, 29-49.

Koehler,I. J. (1993).The influenceof priorbeliefson scientificjudgmentsofevidencequality. Orga-nizational Behavior and Human Decision Processes,56, 28-55.

Klayman,J. (1995). Varietiesof confirmationbias. The Psychologyof Learningand Motivation, 32,385-418.

Kunda, Z. (1991)).The casefor motivatedreasoning.PsychologicalBulleti,t, 108, 480-498.Larrick, R. P.(1993).Motivationalfactorsin decisiontheories:Theroleof self-protection.Psycholog-

ical Bulletin, 113, 440-450.Larrick, R. P., Morgan, J. N., & Nisbett, R. E. (1990). Teachingthe useof cost-benefitreasoningin

everydaylife. PyschologicalScience,1, 362-370.

36 CHIPHEATH, RICHARD P. LARRICK, andJOSHUA KLAYMAN CognitiveRepairs 37

Levitt, B., & March,J.G. (1988).Organizationalleaming.AnnualReviewofSociology14, 319-40.Liang, D. Xv., Moreland,R., & Argote, L. (1995). Groupversusindividual trainingandgroupperfor-

mance:The mediatingfactorof transactivememory.Personality& SocialPsychologyBulletin,21, 384-393.

Lichtenstein,S.,Fischhoff,B., & Phillips,L. D. (1982).Calibrationofprobabilities:The stateoftheartto 1980. In A. Tversky,D. Kahneman,& P. Slovic (Eds.),Judgmentunderuncertainty:Heu-risticsandbiases(pp. 306-334).New York: Cambridge.

Lord, C., Ross,L., & Lepper,M. R. (1979). Biasedassimilationandattitudepolarization:Theeffectsofprior theorieson subsequentlyconsideredevidence.Journalof PersonaliryandSocialPsy-chology,37, 2098-2109.

Mahoney,M. J.(1976).Scientistassubject:Thepsychologicalimperative.Cambridge,MA: Ballinger.Making quality morethana fad. (1992,May 18). Fortune,pp. 12-13.March,J. G., Sproull,L. S.,& Tamuz,M. (1991).Learningfromsamplesofoneor fewer. Organization

Science,2, 1-13.March,J. G.,& Simon,H. A. (1958). Organizations.New York: JohnWiley.McGowan,J. (1996,April 1). How Disneykeepsideascoming. Fortune,pp. 131-133.Milgrom, P., & Roberts,1. (1992).Economics,organization,andmanagement.EnglewoodCliffs, NJ:

PrenticeHall.Miner, A. 5. (1994).Seekingadaptiveadvantage:Evolutionary theoryandmanagerialaction.In J. A.

C. Baum & J. V. Siogh(Eds.),Evolutionarydynamicsof organizations (pp.76-89).New York:OxfordUniversity Press.

Miner, A. S.& Mezias,S.J.(1996).Ugly ducklingnomore:Pastsandfuturesof organizationalleam-ing research.OrganizationScience,7, 88-99.

Mullen, B., & Riordan,C. A. (1988).Self-servingattributionsfor performancein naturalisticsettings~A meta-analyticreview.JournalofAppliedSocialPsychology18, 3-22.

Mynatt, C. R,Doherty,M. E., & Dragan,Xv. (1993).Informationrelevance,working memory,andtheconsiderationof altematives.QuaterlyJournalof ExperimentalPsychology46A,759-778.

Newman,J., Wolff, Xv. T., & Hearst,E. (1980).The featurepositiveeffect in adult humansubjects.JournalofExperimentalPsychology:HumanLearningandMemory, 6, 630-650.

Nisbett, R., & Ross,L. (1980). Human inference:Strategiesand shortcomingsof social judgment.EnglewoodCliffs, NJ: Prentice-Hall.

Nisbett,R. E. (1992).Reasoning,abstraction,andtheprejudicesof 20th centurypsychology.In R. E.Nisbett (Ed),Rulesfor reasoning(pp. 1-12).Hillsdale, NJ: Erlbaum.

Nisbett,R. E.,.Krantz,D. H.,Jepson,D., & Kunda,Z. (1983).The useof statisticalheuristicsin every-day reasoning.PsychologicalReview,90, 339-363.

Nonaka,I., & Takeuchi,H. (1995).The knowledge-creatingcompany:HowJapanesecompaniescre-ate thedynamicsof innovation.New York: OxfordUniversity Press.

Odean,T. (1996). Volume, volatility, price and profit when all tradersareabove average.Workingpaper.Universityof California,Berkeley.

Payne,J. Xv. (1976).Taskcomplexityandcontingentprocessingin decisionmaking: An informationsearchandprotocolanalysis.OrganizationalBehaviorandHumanPerformance,16,366-387.

Perrow,C. (1984).Normal accidents:Living with high-risktechnologies.New York: Basic.Peters,T. (1992).Liberation management:Necessarydisorganizationfrr the nanosecondnineties.

New York: Alfred A. Knopf.Rogers,E. (1995). Thediffusion of innovations(4th ed.). New York: FreePress.Rosenthal,E. (1994,May). Invisible injuries. Discover,pp. 48-51.Ross,L.,& Nisbett,R. E. (1991).Thepersonandthesituation:Perspectivesofsocialpsychology.New

York: McGraw-Hill.Ross,L. (1977).The intuitive psychologistandhis shortcomings.In L. Berkowitz(Ed.),Advancesin

experimentalsocialpsychology(Vol. 10). New York: AcademicPress.

Russo,J. E.,& Schoemaker,P.J. H. (1992).Managingoverconfidence.SloanManagementReview,33,7-17.

Sapolsky,R. (1997,November).On theroleof upholsteryin cardiovascularphysiology.Discover,pp.58-66.

Scott,Xv. R. (1995).Institutionsand organizations.ThousandOaks,CA: Sage.Senge,P.M. (1990).Thefifth discipline.New York: DoubledayCurrency.Slovic,P., Fischhoff,B., & Lichtenstein,5. (1982).Factsversusfears:Understandingperceivedrisk.

In A. Tversky,D. Kahneman,& P.Slovic (Eds.),Judgmentunderuncertainty:Heuristicsandbiases(pp. 463-492).New York: Cambridge.

Smith, J.F., & Kida, T. (1991).Heuristicsandbiases:Expertiseandtaskrealismin auditing.Psycho-logical Bulletin, 109,472-489.

Tetlock, P. E. (1992).The impactof accountabilityon judgmentandchoice:Towarda socialcontin-gency model.Advancesin ExperimentalSocial Psychology25, 331-376.

Tversky,A., & Kahneman,D. (1971).Belief in the law of smallnumbers.PsychologicalBulletin, 76,105-110.

Tversky,A., & Kahneman,D. (1973).Availability: A heuristicforjudging frequencyandprobability.CognitivePsychology5, 207-232.

Walton, M. (1990).Demingmanagementat work. New York: Putnam.Wegner,D. M., Erber,R., & Raymond,P. (1991).Transactivememoryin closerelationships.Journal

of Personalityand SocialPsychology61, 923-929.What Bill Gatesreally wants.(1995,January16). Fortune,p. 34.Wilson, T. D., & Brekie, N. (1994). Mentalcontaminationand mentalcorrection:Unwantedinflu-

enceson judgmentsandevaluations.PsychologicalBulletin, 116, 117-142.Zbaracki,M. (in press).The rhetoric andreality of total quality management.AdministrativeScience

Quarterly.Zucker, L. G. (1991).Postscript:Microfoundationsof institutional thought.In Xv. Xv. Powell & P. J.

DiMaggio(Eds.),The newinstitutionalismin organizationalanalysis(pp. 103-107).Chicago:Universityof Chicago.

Zuckerman,M., Knee,C. R., Hodgins,H. S.,& Miyake,K. (1995).Hypothesisconfirmation:Thejointeffectof positivetest strategyandacquiescenceresponseset.JournalofPersonalityandSocialPsychology68, 52-60.

how organizational practices can compensate for …faculties are limited. instead of being infinite...

Documents