+ All Categories
Home > Documents > Judgmental Errors

Judgmental Errors

Date post: 03-Apr-2018
Category:
Upload: lefki-panteli
View: 220 times
Download: 0 times
Share this document with a friend

of 13

Transcript
  • 7/28/2019 Judgmental Errors

    1/13

    Psychological Bulletin1991, VbUlO, No. 3,486-498 Copyright 1991by the American Psychological Association. Inc.0033-2909/91/J3.00

    Costsand Benefits of Judgment Errors: Implications for DebiasingHal R. ArkesOhio University

    Someauthors questioned the ecologicalvalidity of judgmental biases demonstrated in the labora-tory. One objection to these demonstrations is that evolutionary pressures would have renderedsuch maladaptivebehaviors extinct ifthey hadanyimpact in the "realworld." Iattempt to show thateven beneficial adaptations may have costs. 1extend this argument to propose three types ofjudgment errorsstrategy-based errors, association-based errors, and psychophysical basederrorseach of which is acost of a highly adaptive system. This taxonomy of judgment behaviorsisused toadvancehypotheses as towhich debiasingtechniquesare likely tosucceed in each category.

    During the last two decades, cognitive psychologists docu-mented many types of judgment and decision-making errors.Spurred in good measure by the work of Tversky and Kahne-man (1974), the research area has come to be known as "judg-ment under uncertainty." The relatively poor performance ofsubjects in many of these judgment experiments has causedsome researchers to question the ecological validity of suchstudies (e.g., Berkeley & Humphreys, 1982; Edwards, 1983;Funder, 1987; Hogarth, 1981; Phillips, 1983). The reasoningseems to be that because subjects' performance is so poor inthese experiments, it may not be representative of their behav-ior in more naturalistic environments in which people seemquite competent.

    One purpose of the present article is to argue that even suc-cessful adaptationscan havecosts. This position is common inbiology. Itsapplication to the areas of judgment and decisionmaking will, I hope, help explain whyparticular judgment be-haviors persist despite their obvious drawbacks insome situa-tions. Mygoal is to show that the costsof otherwise beneficialcognitive adaptations are the consequence of appropriate re-sponses toenvironmental demands.

    A second goal of this article is to propose a taxonomy ofjudgment behaviors based on the nature of the adaptationalcosts. The use of this taxonomy may help suggestwhat type ofdebiasing techniques may be effective ineach categoryofjudg-ment behavior.

    Costs and Benefits From an EvolutionaryPerspectiveTo examine maladaptive judgment behaviors, we need to

    consider what makes anycharacteristicadaptive. Viewed froman evolutionary perspective, adaptive behaviors contribute toreproductive success, although the route from good judgmentperformance to reproductive success may be quite indirect. Be-cause it is not obvious howseriousjudgment errors couldpossi-

    I am grateful to Bruce Carlson and Robyn Dawes for their helpfulsuggestions on an earlier draft of this article. Daniel Kahneman andtwo anonymous reviewers alsoprovided very constructive comments.

    Correspondence concerning this articleshould be addressed to HalR. Arkes, Department of Psychology,Ohio University Athens, Ohio45701.

    bly beadaptive, there is reason forpsychologists toassume thatthese maladaptive behaviors exist mainly in artificial labora-tory environments and not in naturalistic ecologies.

    However, Archer (1988) pointed out that successful adapta-tions have costs as well as benefits. Consider first a physiologi-cal example. In dangerous situations, the body mobilizes for afight or flight response. This is highly adaptive. However, pro-longed stress will result in seriousphysicaldeterioration. Thisisa maladaptive long-term consequenceof aresponse that isgen-erally beneficial. Hence, stomach ulceration is not grounds fordeeming the alarm reaction to be maladaptive.

    A phylogenetic example is the emergenceof upright gait. Al-though upright gait has resulted in epidemic levels of lowerback pain in humans, it has resulted in substantial adaptiveconsequences (e.g, freeing the hands for tool use).The benefitoutweighs the cost.

    Apsychological example is provided by the costsand bene-fits of expertise (Arkes &Freedman, 1984; Arkes &Harkness,1980). Experts have substantial background knowledge thattheycan draw on to instantiate the missingslotsof incompleteschemata. They subsequently demonstrate a tendencytorecallthe instantiated information as having been presented when infact it wasnot. Forexample, Arkes and Harkness(1980)showedthat speech therapy students whomade a diagnosis of Down'ssyndrome tended to remember having seen the symptom "fis-sured tongue." However, this common symptom of Down's syn-drome had not been presented in the actual list of symptoms.Because nonexperts have less background knowledge, they areless likely to make this type of error. Thus, even though we allagree that expertise is beneficial, it does have its costs. In ananalogous way, the presence of widespread, maladaptive judg-ment strategies is not necessarily contrary to the principles ofevolution. (See also Enhorn &Hogarth, 1981, p. 58.)

    I divide the judgment and decision-making errors docu-mented in the literature into three broad categories. Eachcate-gory of errors is a cost of an otherwise adaptive system. First, Ipresent a brief overview before describing each category inmore detail.

    Strategy-based errors occur when subjects use a suboptimalstrategy; the extra effort required to use a more sophisticatedstrategy is a cost that often outweighs the potential benefit of

    486

  • 7/28/2019 Judgmental Errors

    2/13

    COSTS AND BENEFITS OF JUDGMENT ERRORS 487

    enhanced accuracy. Hence, decision makers remain satisfiedwith the suboptimal strategy in low-stakes situations.

    Association-based errors are costs of otherwise highly adap-tive system of associations within semantic memory. The auto-maticity of such associations, generally of enormous benefit,becomes a cost when judgmental!}' irrelevant or counterpro-ductive semantic associations are brought to bear on the deci-sion or judgment.

    Psychophysically basederrorsresult from the nonlinear map-ping of physical stimuli onto psychological responses. Sucherrors representcosts incurred in less frequent stimulus rangeswherevery highand very lowstimulus magnitudes are located.These costs are more than offset by sensitivity gains in the morefrequent stimulus ranges located in the central portion of thestimulus spectrum.

    I now present a more detailed description of each of thesethree categories ofjudgment errors.

    Three Types of Judgment ErrorsStrategy-Based Judgment Errors

    Despite the fact that poor judgment performance has beendemonstrated in a large number of situations (Kahneman,Slo-vic, & Tversky, 1982), evidence exists that some suboptimalbehaviors may be adaptive in a larger sense. Suppose a personadoptsaquickand dirty strategy tosolvea problem. Because itisquick, it iseasy toexecute. This is abenefit. Because it isdirty,itresults in more errors than a more meticulous strategy. This isa cost. Although the choice of this strategy mayresult in fewercorrect answers compared with the other strategy,thiscost maybe outweighedby the benefit of time and effort saved.

    Thorngate (1980) and Johnson and Payne (1985) comparedthe performance of various decision strategies with regard totheir ability to select alternatives with the highest expectedvalue. Some strategies were quite rudimentary. For example,some completely ignored the probability of an outcome andonly considered the average payoff for each possible choice. Itwas found that many of the 10 decision strategies performedwell. Even the primitive ones selected alternatives with the high-est expected value under some circumstances and almost neverselected alternatives with the lowest. The use of such elemen-tary strategies would be drastically less taxing to the humaninformation-processing system than would the use of morecomplete but complicated ones. Hence, ifasuboptimal strategywere to be used, the large savings in cognitiveeffort might faroutweigh the small loss in potential outcomes. Thispoint wasstressed a number of years ago by Beach and Mitchell (1978).When subjects know that the stakes are high, they often canchange from asuboptimal strategy to abetter one. It isworthitfor them to do so. For example, Harkness, DeBono, and Bor-gida (1985)asked undergraduates toperform a covariation esti-mation task. Female subjects examined data that describedother women whom Tom did or did not want to date andwhethereach of these women possessed a particular character-istic.The presented data could besummarized asentries intoa2 X 2 matrix, such as the one presented in Figure 1. Forexam-ple, the rowsof the matrix mightbe labeled "The woman hasa

    Tomdoeswant todatethiswoman.

    Tomdoes notwant todatethiswoman.

    Thewoman hasa good senseof humor.

    The woman doesnot have agoodsense of humor.

    cell a

    8

    cede8

    cellb

    4

    celld

    4

    Figure 1. Example of information presented to subjects in the studyby Harkness, DeBono, and Borgida (1985). (From"Personal Involve-ment and Strategies for Making Contingency Judgments: AStake inthe DatingGame Makesa Difference" by A. R. Harkness, K. G. De-Bono, and E. Borgida, 1985, Journal of Personality and Social Psychol-ogy, 49, P- 25.Copyright 1985 by the AmericanPsychological Associa-tion. Adapted by permission.)

    good senseof humor" and "The woman does not have a goodsense of humor." The columns might be labeled "Tom doeswant todate this woman"and "Tom does not want todate thiswoman." Subjects considered these data to decide howmuchToms likingfora woman covaried with a characteristic such asher sense of humor. Using a procedure developed by Shakleeand Tucker (1980), Harkness et al. (1985) were able to deter-minethe complexityof the strategy used by the female subjectsas they performed this covariation estimation task. In somegroups, the strategy used was rather elementary. This was notthe case, however, if the man whose preferences were beingexamined was someone the female subject would be going outwith for the next 3 to 5 weeks. In this group, the women usedcomplex covariation estimation strategies significantly moreoften. This finding is consistent with Payne's (1982)descriptionof contingent decision behavior. The decision behavior iscon-tingenton such factorsas the reward forhigh levelsof accuracy.Researchers who are optimistic about human judgment anddecision-making performance point out that sensitivity to suchfactors as incentive, task complexity (Billings & Marcus, 1983;Paquette & Kida, 1988), and time pressure (Christensen-Sza-lanski, 1980; Payne, Bettman, & Johnson, 1988) is highlyadap-

    Association-BasedJudgment ErrorsExperiments in which semantic memory has been primed

    have become very common during the last 20years. The prin-cipal result of such studies is that priming causes an activationof concepts related to the prime. A number of models, such asHAM (Anderson& Bower, 1973)and ACT* (Anderson, 1983)among many others, posited spreading activation as a funda-mental characteristicof semantic memory.

    Consider a study by Kubovy (1977). When subjects wereasked to report "the first digit that comes to mind,"only 2.2%

  • 7/28/2019 Judgmental Errors

    3/13

    488 HAL R. ARKESchose the digit 1. When subjects were asked to report the "firstone-digit number that comes to mind," 18%chose 1. This resultisconsistent with the tenets ofspreading activation. Thesecondgroup of subjects is much more likely to respond with a 1 be-cause that digit was primed by the request to report a one-digitnumber.

    Ifweconsider the firstgroupto be the control group becausetheywere asked the question in a more neutral manner, shouldwe then consider the response of the second group to be amanifestation of bias? Have these subjects made a judgmenterror?

    The relatively high probability of reporting the digit 1 as thefirst one-digit number to come to mind is a consequence of thespreading activation characteristic of semantic memory. Thefact that related concepts in semantic memory influence eachother through this activation is essential to normal cognitivefunctioning.1 The benefits of spreading activation include someof the most fundamental cognitive tasks: stimulus generaliza-tion, inference, and transfer of training, for example. Thesesubstantial benefits of spreading activation are accompanied bya cost, which Kubovy (1977) demonstrated, namely, the inabil-ity of humans to prevent associated items from influencingtheir cognition evenwhen those related itemsare irrelevant orcounterproductive to judgmental accuracy

    An experiment byGilovich(1981) servesas anexample fromthe judgment literature. Newspaper sportswriters rated the po-tential of various hypothetical college players to become profes-sional football players. Ifa college player wassaid tohavecomefrom the same hometown as a current professional footballplayer, thecollege playerwasrated much higher than if hegrewup in some other town. If weassume that one hometown haslittle to do with one's potential as a professional football player,then we must attribute the higher rating to the fact that themere association between the to-be-rated player and the suc-cessful professional playerwasresponsible for thehigher rating.

    Another example is provided by Gregory, Cialdini, and Car-penter (1982). People who were instructed to imagine experi-encingcertain events subsequently rated those events as morelikely to occur to them compared with subjects who did notpreviously imagine them. Gregory et al. (1982) explained theirresults in terms of availability (Tversky &Kahneman, 1973).Through the activity of imagining, items are made more avail-able in long-term memory. As a result such items are judged tobe more probable. Whereas Kubovy (1977) increased theavail-abilityof the digit1 bymentioning it in an unobtrusive manner,Gregoryet al. were able to increase the availability of scenariosby blatantly asking subjects to imagine their occurrence. Inboth cases the experimenters exploited the normal working ofthe memory system to heighten the retrieval of some item,thereby creating an "error."

    Application of this same principle can result in other judg-ment errors, aswhen Tverskyand Kahneman (1973) presenteda list of famous women and not-so-famous men to a group ofsubjects. Although the list contained more men than women,the subjects erroneously claimed that the list contained morewomen.The notoriety of the women heightened their availabil-ity in memory and, thus, their retrievability. The many judg-ment errors that have been demonstrated in this manner are

    certainly not adaptive if what we mean by "adaptive" impliescorrespondence to reality (e.g, the actual number of men andwomen in the list). However, I believe that such errors are aconsequence of the normal operation of long-term memory:Priming content with related items or by asking the person toperform a cognitive activity will result in heightened retriev-ability, which can result innonveridical estimates of frequencyand probability.

    Again, such errors are a cost of a memory system whoseprinciplesof association and retrieval produce benefits far inexcess of these costs.2 I very briefly enumerate several sucherrors.

    Explanation bias. Ross, Lepper, Strack, and Steinmetz(1977) asked subjects to read a short scenario about a personand then explain why this individual might have eventuallydone some specified behavior, such as contributing money tothe Peace Corps or committing suicide. Subjects were assuredthat the to-be-explained outcome was entirely hypothetical, be-cause it was not knownwhat actually happened to this individ-ual. Participants subsequently rated the probability that theperson actually did each of several behaviors. Ross et al. (1977)found thatsubjects assigned higherprobabilities to theoutcomethat they had explained. Making an option more available canmake the option seem more probable.

    Hindsight bias. Ajudgment error closelyrelated toavailabil-ity is the hindsight bias (Fischhoff, 1975). In hindsightwe tendto exaggerate the likelihood that we would have been able topredict the event beforehand.Ofcourse, the event that did oc-cur and its possible causes are far more available than eventsthat never occurred. For example, after the election has takenplace, subjects saythat they had assigned higher probability tothe winner's prospects than they actually had assigned beforethe election occurred (Powell, 1988).

    Ignoring P(D\H). Consider a former who wishes to deter-mine if there is a relation betweencloudseeding andrain.Avail-able evidence includes the entries in a 2 x 2 table, the rows ofwhich are "seeding" and "no seeding" and the columns ofwhichare"rain" and "no rain." Of course, the farmer needs toexamine the numbersofentries in each cell to arrive at acorrectconclusion. However, many investigators found that subjectswhoare trying to determine the relation between cloud seedingand rain often do not consider as relevant the evidence in the"no-seeding" row(e.g, Arkes &Harkness, 1983). Similarly, stu-dentsareusually astonished tolearn that todetermine the rela-tion betweenasymptomand a disease oneneeds tocollect dataon thelikelihood of asymptomwhen thedisease is not present.Fischhoff and Beyth-Marom (1983) deemed tobea"meta-bias"the general tendency to ignoredatawhen the hypothesis is nottrue or when the possible antecedent cause is not present.

    1 Ratcliffand McKoon (1981) questioned the validity of spreadingactivation theories of semantic memory They did not question thevalidity of the findings that have spawned such theories, however. Ibelieve the judgment errors I attribute to associative mechanismscouldbeexplainedby either the spreadingactivationor compound cue(Ratcliff& McKoon, 1981) theories.

    2 TVersky and Kahneman (1974) also noted that heuristics such asavailability have benefits as well as costs.

  • 7/28/2019 Judgmental Errors

    4/13

    COSTS AND BENEFITS OF JUDGMENT ERRORS 489The reason whythis meta-biasexists isthat thehypothesized

    cause, but not its absence, primes relevant data. If I believe thatdisease D causes symptoms S, it seems obvious to ascertain thestatus of iS whenD ispresent. Theabsenceof Ddoes not primeS; as a result, many people do not believe that the status of Sneeds to be ascertained in such cases.

    Confirmation bias. I define confirmation bias as a selectivesearch, recollection, or assimilation of information in a waythat lends spurious support to a hypothesis under consider-ation. Some authors put this term in quotation marks to denotethat itrefers to arather loosely related groupof findings. (Fisch-hoff & Beyth-Marom, 1983, even suggested abandoning thetermbecause of its imprecise referent.)

    Confirmation bias was demonstrated by Chapman andChapman (1967) using the Draw-a-Person Test. This is a pro-jective instrument in which a patient draws a picture of a per-son, and a clinician then examines the picture for particularcues that supposedly are associated with various types ofpsy-chopathology. Because there was negligible evidence favorableto the validity of this technique, the Chapmans thought thatassociations between the features of the drawings and the pur-ported diagnosis must be entirely illusory.Totest this hypothe-sis, drawings of people were randomly paired with personalitytraits presumably characteristic of the person who did thedrawings. Clinicians and undergraduates who viewed thesedrawings perceived correlations between certain drawing fea-tures and the personality traits of the person who drew thefigure.For example, subjects claimed that drawings containinglarge eyes were frequently done by people whowere said to besuspicious. Drawings with muscular figures were frequentlysaid to be doneby men whowereconcerned about their manli-ness. The Chapmans concluded that because there was no realcorrelation between these drawing features and personalitytraits, subjects must be relyingon preexisting associations inperceiving this association.Totest this hypothesis,theChapmans performed a follow-upstudy. Undergraduateswereasked torate the strengthofseman-tic association between the body parts emphasized in thevariousdrawingsand the personality traits saidto be character-istic of the drawers. In this follow-up study, the subjects ratedeyesasclosely associated with suspiciousness, forexample. Thisis precisely the illusory correlation detected by subjects in thefirst study: They had incorrectly reported that suspiciousnesswas characteristic of the persons whodrew figures with largeeyes. This is an illustrationof the confirmation bias because thesubjects assimilated the evidencein abiased waybased on theirpreconceived association between eyes and suspiciousness.This study isquitesimilar to the one byGilovich (1 98 1) inthat aprior association results in an inappropriate consideration ofthe evidence. In this case, the inappropriate considerationserves to bolster the prior association.

    Pseudodiagnosticity. Bayes's theorem may be expressed inthe following way:- P(H,)P(D I\H 1)where H and D signify the hypotheses and data, respectively,and the subscript i indexes a set of data. Assume that the twohypotheses are mutually exclusiveand exhaustive.

    Suppose subjects are given the choice of examining onepairofdata todetermine /"(H,). They may choose to learn P(/>,IHand P(Di |H2); they maychoose P(D2|H,)andP(D2|H2);or theymay choose P(D, |H,) and P(D2|H). It may be seen by examin-ing Bayes's theorem that to infer the probability of H,, thechoice of either of the first two pairs would be helpful. Choos-ing the last pair will generallynot provide diagnostic informa-tion. However, the members of the last pairP(Dl |H,) andP(D1\H)are both strongly cued when H, isconsidered. Thismay be why these two nondiagnostic data are selected by somany subjects in a study by Doherty, Mynatt, Tweney, andSchiavo (1979). These investigators termed this nonoptimaljudgment behavior pseudodiagnosticity.

    Overconfidence. One of the most robust findings in thejudgment and decision-making literature is overconndence(Lichtenstein, Fischhoff, & Phillips, 1982). Koriat, Lichten-stein, and Fischhoff (1980) suggested that aprimary reason forunwarranted confidenceis that subjects can generate support-ing reasons for their decisions much more readily than contra-dictory ones. The supporting reasons are more strongly cued.Forexample, suppose I amasked whether OsloorLeningrad isfurther north, and Ianswer "Oslo." Now I amasked toassignaconfidence level to my answer. Tocomplete this task, I searchmy semantic memoryfor the information that made Oslo seemlike the correct answer. Items pertaining to Oslo's nearby gla-ciers and fjords are much more strongly cued than informationconcerning Oslo's summer warmth. The evidence I am mostlikely to retrieve thus is an unrepresentative sampleof all avail-able evidence, and my confidence is thereby inappropriatelyinflated. Becauseof the increase in the confidence with whichan opinion is held, the process leading to Overconfidence isrelated to the confirmation bias.

    Representativeness. To appreciate the nature of this heuris-tic, it may be instructive first to consider the term overgeneml-ization (Slobin, 1971). Weadmire the intelligence of the childwhogeneralizes the past tense verbending"ed"to the unfamil-iar verb "revel," thereby making"reveled."Wethinkless highlyof the child whogeneralizesthe same past tense endingto theverb "do," thereby making "doed." Wecall the latter behaviorovergeneralization, even though it seems to be amanifestationof the same very fundamental principle wecall generalization.Of course, those of us who areawareof the existenceof irregu-lar verbs can be arrogant with children about what constitutesovergeneralization of an inferential strategy outside its domainof appropriate application. Overgeneralization is a judgmenterror, but again I think this is a small cost of an otherwiseadaptive associationisticsystem.

    The representativeness heuristic (Tversky & Kahneman,1974) provides an example of such overgeneralization. Thisheuristic refers to the fact that people often judge probabilitieson the basisof similarity, or representativeness. Forexample, injudging whether Instance A belongs to Class B, people oftenrely on the extent to which A seems representative of B. Ofcourse, the probability that Abelongs to B can be influencedbymany factors that have no bearing on representativeness. Forexample, basing one's decision solelyon representativenesswillresult in the underutilization of base rates (Kahneman &Tversky, 1973), thereby resulting in errors.

  • 7/28/2019 Judgmental Errors

    5/13

    490 HAL R. ARKESTheories of category classification as old as that of Hull

    (1920) are based on the principle that decisions concerning thecategorymembership of an exemplar are based on the degree ofsimilarity between the exemplar and the category More recentmodels, such as the feature-comparison model of Smith, Sho-ben, and Rips (1974), share this assumption. For example, tothe extent cardinal shares featureswiththe category clergyman,it is likely to be deemed a member of that category If cardinalshares fewer features with the category bird than the categoryclergyman, it is deemed likely to belong to the latter categoryeven though thereare many more birds than clergymenin theworld. This feature-matching process ignores base rates of thetwo categories; hence, it is prone to error.

    Judgments of similarity follow one of the most fundamentalprinciples of cognition: stimulus generalization. It is highlyadaptive that we associate items to other itemswith whichtheyare related. Even a task as basic as classic conditioning requiresthis. The fact that cardinal ismore closely associated with cler-gyman than bird maybode verypoorlyfor our consideration ofbase rates. However, I consider this cost to be an overgeneral-ization of a process that serves us very well in other contexts.Thus, Isuggest that themanifestation ofthe representativenessheuristic is another example of a cost of an otherwise highlyadaptiveassociationisticsystem.

    Psychophysicatty Based ErrorsFrompsychophysical power functions (Stevens,1957)topros-

    pect theory's S-shaped curve (Kahneman & Tverskx 1979),from the original Weber-Fechner log function to economists'law of diminishing returns, many theorists postulated anasymptotic curve relating external stimuli (e.g,mass, cash, lightintensity) and the psychological responses to those stimuli. Fig-ure2depictsthe valuefunction ofprospect theory (Kahneman&Tversky, 1979), which represents one such nonlinear curve.

    Asystemthat translated physical intensityin a linear manneronto psychological response would impose an immensecostonany transduction system. Extreme stimuli, which occur rela-tively infrequently, wouldhave to be coded with asgreatalevelof discriminability as the more frequent middle-range stimuli.Any nonlinearsystemwith an asymptoteat the extremeend (orends) would have the benefit of eliminatingthe structures andprocesses needed to discriminate small changes in rare events,such as very intense sounds or extremely heavy weights. Ofcourse, sacrificing discriminability at the ends of the contin-uum has a cost.

    An experiment by Dinnerstein (1965) illustrates this point.Dinnerstein presented subjects with a series of weights andfound that subjects' ability to discriminate was finest at thecenter of the range of stimuli. Then a weight was introducedthat waseither above or below all the others. This caused theregion of maximal discriminability to either rise or drop de-pending on whether the new weight was heavy or light. Thisresult occurredeven though the newweight was not included inthe range of stimuli to be rated. This study demonstrates theadaptation of the nervous system to the available stimulusarray,an adaptation that allows theperceiver to extract the opti-mal amount of useful information from each situation. Of

    Figure 2. The value function of prospect theory (Kahneman &Tversky, 1979). (Seetext for discussion. From "Prospect Theory: AnAnalysis of Decision Under Risk" by D. Kahneman and A. Tversky,1979, Econametrica, 47, p. 279.Copyright 1979 byBasil Blackwell Ltd.Adapted by permission.)

    course, extractingthe optimal amount of useful information ishighly adaptive even though diminished sensitivity at the ex-tremes is a cost.

    Severaljudgment errorsmaybeamanifestationofthispartic-ular cost. I briefly enumerate several.

    Sunk cost effect. Economic decisions should be made basedon the anticipated costs and benefits that will result from thechoice of each of the alternative courses of action. Note thatfuture costsand benefitsare relevant;prior(sunk)costsare not.Ajudgment error occurswhen sunkcostsare used as abasisfordecision making.

    The sunk cost effect is manifested in a willingness to con-tinuespendingafter an investment of time, effort, or money hasalreadybeen made (Arkes& Blumer, 1985). Persons whohavealready invested substantial amountsand whohavenot yet real-izedcompensatory returns are atPointB inFigure2.Personsinthat situation are not very sensitive to further losses; a smallsubsequent expenditureof funds will therefore causenegligiblepsychological disutility Hence, such persons are willing to"throw good moneyafter bad" in a desperateattempt torecouptheir sunkcost, even though such behavior may be irrational. Ifa particular project is a poor idea, the fact that it has alreadywasted a lot of money does not make it a better idea. Yet thesunk cost effect has been shown to be powerful (Arkes &Blumer, 1985).

    Psychophysics of spending. Once a person has decided topurchase a new car, for example, he or she is located in the

  • 7/28/2019 Judgmental Errors

    6/13

    COSTS AND BENEFITS OF JUDGMENT ERRORS 491asymptotic region of a curve describing the psychophysics ofspending. Persons in this situation would be more willing topay $235 for a car radio compared with their willingness to buya radio for$235 if they had not purchased a car. Wehave gooddiscriminability (Le, the curve is steep) in the region of a fewhundred dollars on either side of our current state. Once we areseveral thousand dollars away from our current state, discrim-inability drops(i., the curve flattens), and we no longer objectto extravagant additional expenditures (Christensen, 1989).

    Reflection effect. TverskyandKahneman (1981,p. 45 3)dem-onstrated how framing the outcomes of the same gamble aslosses or as gains can lead to different decisions. They referredtothis phenomenon as the reflection effect, whichis illustratedin their following well-known example (p. 453):

    Imagine that the U.S. ispreparing for the outbreak of an unusualAsian disease, which isexpected tokill 600people. Twoalterna-tive programs tocombat thediseasehave been proposed.Assumethat the exact scientific estimatesof the consequences of the pro-grams are as follows:If Program X isadopted, 200 people will be saved.If Program Y isadopted, there is aone-third probability that 600people will be saved and a two-thirds probability that no peoplewill be saved.

    The benefit of saving200 lives is located at Point X inFigure2.The benefit of saving 600 lives is located at Point Y.ProgramY represents a relatively small gain in value over Program X.Two hundred lives savedis sogreatabenefit that the additionallives that might be saved under Program Y are too small anadditional benefit towarrant the risk of savingnoone. Hence,about three fourths of the subjects chose Program X.

    Other subjects were asked to consider two other programs:

    If Program B is adopted, 400 people will die.If Program A isadopted, there is a one-third probabilitythat no-body will die and atwo-thirdsprobability that 600people will die.

    The lossof 400 livesislocated at Point B inFigure2. The lossof 600 lives is located at Point A. Becausethe loss of 400 lives isso terrible, the loss of 200 additional lives represents only asmall additional loss invalue. Hence, about three fourthsof thesubjects chose Program A. The potential extra loss in valuebychoosing that program was more than offset by the chance ofsaving everyone.

    It iseasy to see that Program X,which isgenerally endorsed,is the same as Program B, which is generally shunned. If thevalue function were strictly linear, thisinconsistency would notoccur. By asking some subjects to consider the problem interms of lives gained, Tverskyand Kahneman (1981) exploitedthe small superiority of Y over X, a superiority which is notsufficient towarrant additional risk. Byasking other subjects toconsider the problem in termsof lives lost, Tversky and Kahne-man (1981)exploited the small inferiorityof AoverB, an inferi-ority small enough to warrant additional risk. The wording orframing of the problem directs subjects to different portions ofthe nonlinear curve.

    Anchoring The termsanchoringand anchoringand adjust-ment were popularized in the judgment and decision-making

    literature byTversky and Kahneman (1974), although they werediscussed earlier (e.g., Slovic &Lichtenstein, 1971).

    First, let us consider the psychophysical research pertainingto "induction illusions" or "context effects." We know frommany experiments on adaptation level theory (Helson, 1964)that when a medium-size circle is placed in a groupof muchlarger ones, the medium one is perceived as small. Thisconstel-lation of circles induces an adaptation level that is approxi-mately at the mean valueof the circles' areas, and the mediumcircle has an area that is below this mean. When this samemedium-sized circle is placed in a context of much smallercircles, it is perceived as large. Now its area is above the adapta-tion level.Theshift in the adaptation level isconsistent with theprinciple discussed previously by Dinnerstein(1965): It is besttohave the adaptation level change location to locatemaximaldiscriminability near the center of the stimulus continuum.

    Note that this adaptation is congruent with the relation be-tween the physical and psychological dimensions depicted inFigure 2. Point O is the current state, which is in the area ofmaximal discriminability. The asymptotes are in areas of di-minished discriminability.Consider an analogous judgment experiment (Sutherland,Dunn, & Boyd, 1983). Sixty-four hospitalized patients rated fivedifferent health states using three different methods. In the firstmethod, subjects assigned valuestothesefivehealth states on ascale anchored by perfect health and death. In the secondmethod,perfect health wasreplacedon the highend of the scaleby a health state each rater had rated less desirable. Thus, thehigh end of the scale was no longer quite as high. In the thirdmethod, death was replaced on the low end of the scale by ahealth state each rater had rated more desirable. Thus, the lowend of the scalewas no longer quiteas low. Relativeto thevaluesassigned to the various health statesusing the first method, thevalues assigned to the very same health states using the secondmethod were lower, and those assigned using the third methodwerehigher. This study showed that a patient's rating of possiblehealth states was strongly influenced by the context in whichsuch stimuli were considered. Such context-dependent effectsappear to induce inconsistent ratings just as the medium-sizedcircle was rated differently depending on the size of the circleswith which itcould be compared. However, this is a small costof an otherwise beneficial adaptation designed to extract theoptimal amount of useful information out of each situation.

    Anothergroupof judgment studies ismore closely related toacontext phenomenon demonstrated byRestle(1971).Subjectswere presented with a drawing like that in Figure 3. Restle

    E EFigure3. Stimulus used in study by Restle (1971).

  • 7/28/2019 Judgmental Errors

    7/13

    492 HAL R. ARKESvaried the length of the horizontal test line (H), the length ofthe vertical center line (C), whichcrossed the test line, and thelength of the identical vertical ends lines (E). It was expectedthat judgments of the length of Hshould decrease as C or Eincreased. It waseasyfor Restletodeterminethe influence of Eand C on H by ascertaining the slope of the function relating Eto H and C to H.

    Restle (1971) presented subjectswith one of twopossible setsof instructions. Onegroupwastold, "Pay attention to the verti-cal lines at the ends of the test line, and use them as a frame ofreference to help you in your judgments." These subjects werealso warned, "Try todisregard the center vertical line." Othersubjects were told just theopposite: They wereto use thecentervertical line as aframe ofreferenceand to ignore the end lines."The resultwasthat the line that subjects weretold toattend towas much more influential on the subjects' judgment of the teststimulus than was the line they were told to ignore.

    This study differs from the study involvingthe circles in thatinstructions are used to direct the subject's attention to the refer-ence point, which serves as the context for the ensuing judg-ment. Flexibility of frames of reference, which introductorypsychology students first appreciatewhen they viewa Neckercube, is essential to recognize the same object in different con-texts. However, this immense benefit has a cost, and manyan-choringand adjustment studies illustrate this cost.

    In the most famous such study, Tversky and Kahneman(1974) asked subjects to estimate the percentage of Africancountries in theUnited Nations. Subjects spuna"rigged"spin-ner, which landed on either 10% or 65% as a starting point.Subjects were then were asked toadjust the starting number tothe level they thought wasappropriate toanswer the questioncorrectly The median estimate for those whostarted with10%was 25%, whereas the median estimate for those who startedwith 65% was 45%.

    By directingthe subject's attention to a starting point or an-chor, Tversky and Kahneman (1974) did something analogousto what Restle (1971) asked his subjects to do. When 10% ispresented as the anchor, a context is induced that contains lownumbers. Adjustments upward move toward the area of maxi-mal discriminability in the central region of the spectrum. Be-cause such adjustments in this region are perceived to be quitesignificant, subjectsoften refrain from making themas largeaswould be warranted. This results in the insufficient adjustmentobserved by Tversky and Kahneman (1974). Of course, the op-posite result occurs when the subject's attention is drawn to thelarge anchor at the beginningof the experiment.

    Many other studies illustrate the influence of the anchor inanalogous judgment situations (e.g., Northcraft & Neale, 1987).It is true that different anchors and the subsequent insufficientadjustment result in different final estimates given differentanchors. I suggest that this "irrationality" is a worthwhile costtoachieve context-dependent judgment behavior.

    Multiple CausesTothis point I have identified various biases, forexample, the

    hindsight bias, as belonging to one of the three categories ofjudgment errors. However, some phenomena we term biases

    may have more than onecause. Hence, it wouldnot beappro-priate tocategorize thebiasasbelonging to acategory ofjudg-ment error. Instead, the various causesof thebiasmay becate-gorized according to the taxonomy presented previously.

    Perhaps the best exampleof this situation is theconjunctionfallacy. TverskyandKahneman (1983)suggested that the repre-sentativeness heuristic is one basis for this fallacy. However,they also pointed out in an earlier article (Tversky & Kahne-man, 1974) that the anchoring and adjustment heuristic mayplaya rolein themanifestationofthis fallacyinsome instances(e.g., Bar-Hillel, 1973). Tverskyand Kahneman (1983,p. 312)contended that speakers' conformity withGricean (1975) con-versational rules could hinder appreciation of theprobabilisticlaw relevant to the consideration of conjunctions. Worse yet,YatesandCarlson(1986)suggested that individual subjectsmayusemultiple procedures inarrivingat their answerson differentconjunction problems depending on the presence of variousfactors. Thus, it would be a mistake to place theconjunctionfallacy itself into onlyone of the categories of judgment errors.Itwouldbe proper, however, to place each of the various causesof the fallacy into one of the categories. Thus, the taxonomydoes not divide the judgment errors into mutually exclusivecategories. Isuggest that the causescan be sodivided. Whetherthe categories areexhaustive with regard to thecauses of judg-ment errors remains to be determined.

    To return to the example of the conjunction fallacy, anchor-ingand adjustment is a psychophysicallybasederror. Represen-tativeness is an association-based error as is the overgeneraliza-tion ofGricean principles toprobability estimates. If theenvi-ronment contains cues that foster one ofthese "incriminating"behaviors, then the fallacy will occur.

    DebiasingStrategy-Based Errors

    Biasmaynotbeanappropriate termto use todescribe subop-timal behaviors in this category, and thusdebiasing would notbeanappropriate termto use todescribetheadoption ofstrate-gies that result in higher accuracy levels.Suboptimalbehaviorsoccur in this category because the effort orcost of amoredili-gent judgment performanceisgreater than theanticipatedben-efit. The way to improve judgment within this category is toraise thecost ofusing the suboptimal judgment strategy.Typi-cally this results in the judge's utilization of the currently avail-able data in a much more thorough way, an obviously superiorstrategy.

    Consider first the study by Harknesset al.(1985) alluded toearlier and depicted in Figure 1. The investigators identifiedfour covariation strategies. The first, the Cell A strategy, con-sistsof noting how many times Tom liked a woman who had agood senseof humor.Thecovariation between Toms likingforthe woman and her sense of humor is based on the number oftimes he wanted to date such a person. The second strategy, AminusB, consists of comparing CellsA and B. To the extentAexceeds B, Tom isjudged to like women witha senseof humor.Note that these two strategies do not use Cells C and D. Thethird strategy, sum of diagonals, compares the sum of A and D

  • 7/28/2019 Judgmental Errors

    8/13

    COSTS AND BENEFITS OF JUDGMENT ERRORS 493with that of B and C. To the extent the former sumexceeds thelatter, Tom isjudged to likewomen with a sense of humor. Thefinal strategy, conditional probability is the normativeassess-ment of covariation. These final two strategies use the data inall four cells. Harkness etal. found that 6 of the 11 womenwhowere given information about Tom but who would not be dat-ing him usedone of the twoprimitive strategies. Noneof the 11women who thought they would be going out with Tomusedthese simple strategies; they all used one of the two sophisti-cated covariation estimation strategies. If a subject in a low-stakes judgment environment were using only a subset of theavailable data, itwouldbe obvioushow to improve one's judg-ment should the stakes increase: Use more data.

    An analogous finding is exemplified in a study by Petty andCacioppo(1984). Undergraduateswereexposedtothreeor ninearguments that were all of either high or lowquality.Theargu-ments related to an issue that would be of importance to agroupofundergraduates: The institutionof a newpolicy in oneyear under which all students had to pass a comprehensiveexam in their major field tograduate from the university. Need-less to say, this was the "high-involvement" group. The"low-in-volvement"subjects also evaluated either three or nine strong orweak arguments, but the issue concerned the adoption of thisnew policy at a time long after this group of students wouldgraduate. Petty and Cacioppo found that thelow-involvementsubjects weremore persuaded by nine arguments than by threearguments. The strength of the argument was not a significantfactor. For the high-involvement subjects, the strength of thearguments rather than their mere number was significant. Ifsubjects are not concerned with a proposition, merely countingthe arguments in support of it might be sufficient. If the stakesare raised, an obvious strategy isavailablethe benefits ofwhicharesubstantial: Consider the merits of the arguments.

    Inboth the Harknesset al. (1985) and the Pettyand Cacioppo(1984) studies, the presence of higher stakes resulted in lesssuperficial treatment of the data available as the basis of a deci-sion. Tetlock and Kim (1987) found that the same end could beaccomplished through slightly different means. All subjectswere presented with the responses of an actual test taker to thefirst 16 items of a personality test. Based on these responses,subjects were first asked to write a brief personality sketchofthe test taker. Then subjects were then askedtopredict how thetest taker would answer 16 new items. One group of subjectswas told beforehand that they would be interviewed by theexperimenter to learn how the subjects went about makingtheir predictions. These accountability subjects wrote morecomplex personality sketches, were more accurate in their pre-dictions of how the test takers would answerthe next 16 items,and expressed less overconfidence in their predictions. Know-ing that they would be held accountable for their predictionsraised the stakes for the subjects, whichcaused them to interactwith the stimulus materials in a less superficial way.

    Cursory interactions with currently available data cause strat-egy-based errors, and incentive promotes the adoption of amore thorough strategy.Association-Based Errors

    The influence of incentives in eliminatingassociation-basederrors isnegligible,as illustratedin an experimentby Fischhoff,

    Slovic, and Lichtenstein (1977). In this study, subjects assignedconfidence levels to their answers totwo-optionquestions, suchas"Adenwasoccupied in1839by the(a)Britishor (b)French."Ifthe analysis of thissituation by Koriat et al. (1980) is correct,subjects search forreasons to support their answer. This searchinstills ahigh (and inappropriate) levelofconfidence. Fischhoffet al. (1977, Experiment4)wantedto find out how intransigentthis overconfidence was. Subjects were asked to wager actualmoney based on the confidence levels they had assigned totheir answers. About 93% of the subjects agreed to wager in agame that would have been biased in their favor if their confi-dence levels had been appropriate for their level of accuracy Iassume that the prospect of winning or losing substantialamounts of cash based on their stated confidence levels wouldcause subjectstoscrutinize thebasisofthese stated levels. Thisadditional, highly motivated scrutiny apparently led the vastmajority of subjects to conclude that their stated confidencelevels werejustified. Nevertheless, 36 of the 39 subjects wouldhave lost money in this game, because theirhigh levelsof confi-dence were not justified.

    Incentives are not effective in debiasing association-basederrors because motivated subjectswill merelyperform the sub-optimal behaviorwith more enthusiasm. Aneven more assidu-ous search for confirmatory evidence will not loweronelsover-confidence to an appropriate confidencelevel.3

    Fischhoff (1975) andothers tried a direct approach todebiasthe hindsight effect: Tell the subjects about the bias and thenwarn them not to succumb to it. Ifthe mechanism responsiblefor the hindsight bias is the memorial primingofan outcomebyits actual occurrence, then exhortations to prevent this primingwill generallynot be effective because the priming of associa-tions between related items probably occurs automatically(Neely, in press; Ratcliff& McKoon, 1981). That is, priming isunconscious and occurs with negligible capacity usage (Posner& Snyder, 1975). It would be difficult for subjects to abort acognitiveprocess that occurs outside of their awareness."Pleaseprevent associated items from influencing your thinking"would be a curious entreaty unlikely to accomplish much de-biasing.

    There is a long history of research in cognitive psychologythat demonstrates that the occurrence of automatic processescan be maladaptive. The most commonly cited example is theStroopeffect (Stroop, 1935). Subjects are shown words and areasked to name as quickly as possible the color of the ink inwhich the word isprinted. When the word itself is the nameofsome color, suchasred, and the ink is adifferent color, subjectsexperience difficulty in suppressingthe tendency to announcered rather than the color of the ink. Theactivation of aword insemantic memory is"too automatic" for the subject toperformthe Stroop task with facility.

    In an analogous way, association-based judgment errors are a

    3 Thaler (1986), amongothers, also noted that increasing the incen-tive forrational behaviordoesnot always result in heightened rational-ity This presentsaproblem foreconomistswhohope that the irration-alities documented bypsychologists inquestionnaire studieswill dis-appearwhen financial incentives for rational behavior are introduced.

  • 7/28/2019 Judgmental Errors

    9/13

    494 HAL R. ARKESsmall cost of an otherwise adaptive association-based semanticmemory system. These errors occur when items semanticallyrelated to the judgment influence iteven when their influenceisnot conducive to increased accuracy.Todiminish an associa-tion-based judgment error, neither the introduction of incen-tives nor entreatiesto perform well will necessarily cause sub-jects to shift to a new judgment behavior. Instead, it will bemore helpful to instruct the subjects in the use of a behaviorthat will add or alter associations.

    Instructions toperform adebiasing behavior. On the basisofearlier research by Slovicand Fischhoff (1977) and by Koriat etal. (1980), Arkes, Faust, Guilmette, and Hart (1988) presentedneuropsychologists with a small case history and then askedthem to state the probability that each of three possiblediag-noses was correct. The estimates of these subjects comprisedthe foresight estimates. Other neuropsychologists were told thatone of the diagnoses wascorrect and that they should estimatethe probability they would haveassigned to the three diagnosesif they did not know which one correct. These hindsight sub-jects exhibited a bias by assigning a higher probability level tothe "correct" diagnosis than did the foresight subjects. How-ever, hindsight subjects who had to state one reason supportingeach of the diagnosesbefore making their probability estimatesmanifested nohindsight bias.

    The behaviorof considering evidence supportive of an out-come that did not occur is unlikely to be performed bysubjectswhatever their motivationunless they areasked todo so. The consequenceof performing this behavior isloweringthe inappropriateconfidence one has in the accuracy of one'sresponses and reducingthe magnitude of the hindsight effect.Koriat et al. (1980) found that this technique was effective inreducing the overconfidence people generally have in their an-swers to general knowledgequestions, and Hoch (1985) foundthe same techniquewas able to lower overconfidence in fore-casts made by business students.Notethat this "consider the opposite"strategy (Lord, Lepper,& Preston, 1984) attempts to debias by priming stimuli otherthan the ones that would normally beaccessed.Once this prim-ingoccurs,newcausal skidsaregreased. Theconsequent influ-enceof these newfactors willoccur according tothe same mech-anisms that led to the bias (e.g., hindsight, confirmation, over-confidence) in the first place. If the occurring event cued itsown causal chains, then considering the nonoccurring eventought toaccomplish theanalogous result, thereby reducingthebias.

    Another typeofdebiasinghasbeeneffective against overcon-fidence, a bias that I have postulated is a consequence ofcuingmainly supportive evidence. Murphyand Winkler (1974) foundthat weather forecasters have outstanding accuracy-confidencecalibration. For example, there is rain on 90% of the days onwhich meteorologists say there is a 90%chance of rain. How-ever, Wagenaarand Keren (1986) showed that meteorologistswere very overconfident in their answers to general knowledgequestions. This suggests that these professionals have notlearned some general debiasing strategy like "consider the op-posite," which theycan then applytodomains outside theirareaof expertise. Instead, the absence of overconfidence for this anda very fewother select groups of professionals in theirareaof

    expertise isdue to the fact that theygetrapid feedback on averylarge number of predictions the confidence level of which iscarefully recorded. Of course, daily feedback on the appropri-ateness of one's confidence is a debiasing technique almostnever available to most people.

    Cuingadebiasing behavior. Rather than instructing peoplein a different judgment behavior, it is possible to merely cuesuch a behavior. An example is provided by the research pro-gram of Nisbett, Krantz, Jepson, and Kunda (1983), whowereinterested indiscovering independent variables that would fos-ter the use of an appropriate statistical inference technique bycollege students. Tversky and Kahneman (1974) showed thatmany subjects did not use such inference techniques in manyinstances; therefore, the subjects' judgments were incorrect.Hence, the imposition on subjects of any of the effective inde-pendent variables discovered by Nisbett et al. for these taskswould constitute debiasing. Forexample, subjects in their thirdstudy were presented with the story of David, a high schoolsenior who had to choose between a small liberal arts collegeand an Ivy League university. Several of David's friends whowere attending one of the two schools provided informationthat seemed tofavor quite strongly the liberal arts college. How-ever, avisitby Davidtoeach school providedhimwith contraryinformation. Should David rely on the advice of his manyfriends (alarge sample)or on his own 1-day impressions of eachschool (a very small sample)? Other subjects were given thesamescenario with the addition of a paragraph that made them"explicitly aware of the role of chance in determining the im-pression one may get from asmall sample" (Nisbett et al., 1983,p. 353). Namely, David drew up a list for each school of theclasses and activities that might interest him during his visitthere, and thenheblindly dropped apencilon the list, choosingto do those things on the list where the pencil point landed.These authors found that if the chance factors that influencedDavid's personal evidence base were made salient in this way,subjects would be more likely to answer questions about thescenarioin a probabilistic manner(i.e, rely on the large sampleprovided by many friends) than if the chance factors were notmade salient. Such hints, rather than blatant instruction, canprovide routes to a debiasing behavior in some problems.

    Confidence as a second-order judgment. If I estimate thepotential of college football players, the proportion of men in alist of people, or the meritof programX incombatinganAsiandisease, I wouldbe performing a first-order judgment task. If Iam called on to express my confidence in any of those judg-ments, then I amperforming asecond-order judgment task.Bystating my confidence, I am rendering a judgment about myfirst-order judgment.

    Debiasing techniques aimed at the first-order judgment canalso have salutary effects on overconfidence. Forexample, Tet-lock and Kim (1987) found that subjects who knew that theywould be held accountable for their predictionsand thus wereina high-stakes situation wrote complex personality sketches ofthe person whose test they were reviewing. They also mademoreaccurate predictions concerning these test takersthandidsubjects in a low-stakes situation. This study was used to illus-trate the fact that incentives can improve strategy-based judg-ments. However, Tetlockand Kim found that the accountabil-

  • 7/28/2019 Judgmental Errors

    10/13

    COSTS AND BENEFITS OF JUDGMENT ERRORS 495ity group also wasless overconfident in their judgments thanthe control group. I hypothesize that this result was becauseaccountability subjects more thoroughly studied all availableevidence compared with the control group. As a consequence,the usualfindingoverconfidencewasameliorated. Subjectstypically have to be instructed toconsider evidence that iscon-trary totheir decision tobringtheirconfidencedowntoreason-able levels (Koriat etal.,1980). However, because confidence isa second-order judgment, attempts at improving the first-orderjudgment may also have a beneficial effect in debiasing over-confidence. (See alsoArkes, Christensen, Lai, &Blumen 1987,Experiment 2).

    Psychophysically Based ErrorsTechniques effective in debiasing psychophysically based

    errorsare muchdifferent than thoseeffective indebiasingasso-ciation-based errors.However, we must first consider what con-stitutes bias.

    Suppose that I am in line at the betting windowbefore thelast race of the day at a race track. The man in front of mebemoans his terrible luck on the prior 11 races and decides toput all hisremainingfunds ($50) on along shot on the last race.Becausehe lost all 11 of the prior races, weassume that he wasat point B in Figure 2. The loss of $50 would not represent asignificant decrease in utility The gain of several thousandwould represent an enormous increase. Calculations based onthe curvein Figure 2 might indicate that his behavior was"ratio-nal"; he was maximizingexpected utility. If the upper portionof Figure 2describedthe relation between objective light inten-sity and his subjective brightness judgments, it would not havebeen sensible for me to say, "Sir, I respectfully point out thatyour judgments do not increase in a linear fashion with objec-tive intensity. Therefore, your judgment is biased, and youmightdowell tocorrect your responses." For the same reason, itwouldnot havebeen sensibleforme topoint out that hisbettingbehavior was biased. Given the psychophysics of his situation,his behavior followed in an unbiased manner. Because therewasno bias, there was no warrant to debias.

    Kahneman and Tversky (1984) pointed out that when sub-jects are presented with their inconsistent answers on the twoversions of the Asian disease problem, many do not want toresolve the inconsistency by changing one of their answers.They can be made to realize that the inconsistency is present.However, they apparently do not consider their answers to benonnormative or "biased." This suggests howdifficult debias-ing may be on psychophysicallybased errors.

    Suppose a person's own psychophysical function is not usedas a basis for making a decision. Instead, a benevolent othermay wish to impose a "normative" framework, thereby chang-ing the original person's response. An example might be anaccountant who, through his or her professional training, real-izes that the manifestation of the sunk cost effect will haveadverse effects on the economic well-being of everyone in acompany If he or she wants to debias the sunk cost effect, whatavenues are possible?

    Incentives, which are effective in debiasing strategy-basederrors, are ineffective in debiasing psychophysically based

    errors. Sunk cost reasoning has been used to justify continuedfunding of multihillion dollar "lemons" (Arkes & Blumer,1985). Inaddition, it has been used to justify continued spend-ing on the exceedingly expensiveB-2bomber (Staff, 1989,p. 8).If saving billions of dollars is not a sufficient monetary incen-tive, thenwe mayconclude that thesunkcost effect is notpartic-ularly vulnerable tothis type of debiasing.

    It is not clear how changing the judgment behavior to addassociations, a technique effective in debiasing association-based errors, wouldeven beappliedhere. I am unaware of anysuch attempts.

    To debias psychophysically based errors, at least four tech-niques may be effective, however.

    First, because the curve relating objective gains and losses tosubjective gains and losses cannot be changed, debiasing mayoccur when new gains or losses are added to thosecurrentlyunder consideration. This will change the location of the possi-ble outcomes on the curve. For example,Northcraft and Neale(1986) asked subjects to consider spending more money on aproject that appeared to be doomed to failure. Subjectsin thecontrol group tended to continue to spend, therebymanifestingthe sunk cost effect. Other subjects were informed of the pres-enceof opportunity costs. This term refers to the fact that moneyspent on the doomed project is unavailable for use on muchmore promising ventures. These superior investments repre-sent lost opportunities if the funds are spent elsewhere. Reveal-ing to subjects the presence of this hugeadditionalcost madethe choice of the sunk cost option much less attractive, andfewer subjects succumbed to the sunk costeffect. This situationcan be understood by referring again to Figure 2. Consider thesubjects who had not been made aware of the opportunity costof a further small investment in a hopeless cause. Should theinvestment prove to be unsuccessful, the location of the sub-jects wouldshift from point B(their current position) to pointA, a small loss in utility However, those subjects who had beenmade awareof the opportunity cost of continued investment ina lost cause would realize that such behavior would actuallyresult in ashift from point B topoint Q. This represents a largerloss of psychological utility Presenting subjects with informa-tion concerning opportunity costsdecreased subjects' willing-ness to continue investing.

    Asecond way to modify psychophysical judgment behaviorsis to change the concatenation of related items. For example,Thaler (1985) asked subjects to decide whether Mr. A or Mr. Bwas happier. Mr. A won two lotteries, one for $50 and one for$25. Mr. B won a single lottery of $75. The large majority ofsubjects thought that Mr. A would be happier. Mr. Awouldreceive two separate winnings,and because of the concavityofthe value function in the region of gains, the sum of the valuesof the twowinnings would be greater than value of their sum,that is v(25) + v(50) > v(25 + 50). Thaler (1985) pointed out thatlate-nightmail-orderadvertisements take advantageofthisprin-ciplebytossinginatool set, knives,andother separable itemstomake the gain look particularly large. Adding more of the sameproduct would merely push the potential buyer along theasymptotewhere value increases quiteslowly.Thus, segregatingversus integrating gains can cause changes in one's willingnessto make purchases.

  • 7/28/2019 Judgmental Errors

    11/13

    496 HAL R. ARKESA third technique is to change one's reference point. When

    someone is 100,000calories inarrears on adiet, one is atpointA. Efforts tomove to the rightand upwardon the scale willnotresult in much improvement in the immediate future.However,if one begins the diet anew, then one is transposed topointO.Here improvement is easier to achieve thanks to the shape ofthe curve in the region of the origin. Maxims like "Today is thefirst day of the rest of your life" use this principle. (Arelatedeconomic analysis can be found in Loewenstein, 1988).

    Fourth, one can reframe losses as gains (or gains as losses) aswas accomplishedby Tversky and Kahneman (1981) in theirAsian disease example. (Also see McNeil, Pauker, Sox, &Tversky, 1982.)

    Note that these techniques do nothing to alter the shape ofthe psychophysical curve. Psychophysically based judgmenterrors occur because the relation between external stimuli andpsychological responses to those stimuli isnonlinear. Becausethe shape of the curve depicting this relation is agiven, debias-ingconsists ofchanging either the locationofthe options or thelocation ofone's reference point on the curve.

    TrainingAt least one type of debiasing lies outside the categorization

    scheme just presented. Its success is not due to its ability tocounteract the cognitive behaviors characteristic of strategy-based, association-based, or psychophysically based judgmenterrors.

    Thistypeofdebiasingis professional training. When examin-ingthe financial stateof acompany, anaccountant isunlikely tofall prey to the sunk cost effect. Standard accounting proce-duressimplyallownoplace for the consideration of sunkcosts.From a psychological perspective, this is not a very interestinginstance ofdebiasing. However, it is instructive that quite spe-cific professional trainingmay be necessaryfordebiasing to besuccessful. Arkes and Blumer (1985) showed that taking acourse or two in general economics did not inoculate studentsagainst the sunkcost effect.

    Another exampleof thissame typeofdebiasingismore gen-eral and therefore much more interesting. Lehman, Lempert,and Nisbett (1988) showed that graduate training can influencesubjects' statistical reasoning. For example, Lehman et al.showed that the importance ofcontrol groupsismore likelytobe apprehended by advanced psychology graduate studentsthan by advanced chemistry students. The superiority of thepsychology students may represent a result similar to the pre-sumed superiorityof the accountants in resisting the sunk costeffect. Namely, professional training in the techniquesof psy-chological research heightened their awareness of the impor-tance of control groups.

    Instruction in standard accounting proceduresor scientificmethodology represent examples of providing people withtools designed to reach a normatively appropriate answer. Ed-wardsand vonWinterfeldt (1986)pointed out that "if the prob-lem is important and the tools are available people will usethem and thus get right answers" (p. 679). Indeed, training in-volves giving certain (usually self-selected) people preciselythose tools needed to arrive at correct answers. The decision to

    be trained professionallyor to seek someone whoisso trained isa meta-strategy that will ameliorate some judgment errors.

    ConclusionIn his excellent reviewof the debiasing literature, Fischhoff

    (1982, p. 444) suggested that clarifying and exploiting the cog-nitive processes underlying debiasing are major theoretical andpractical tasks. The debiasing literature currently contains adesultory catalog of techniques that work, techniques that donot work, and techniques that work on some tasks but notothers. The purpose of this article was to divide judgment be-haviors into three broad categories based on functionalist crite-ria, namely the bases for their costs and benefits. With thistaxonomy, it is then possible to hypothesize which variables arelikely to be effective in debiasing judgment errors within eachcategory

    Strategy-based errors occur when the cost ofextraeffort out-weighsthe potential benefit ofextra accuracy.Given this prem-ise, debiasing should occur when the benefitsofaccurate judg-ment are increased.Association-based errors are costsofotherwise highlyadap-tive system ofassociationswithin semantic memory Errorsoc-curwhen semanticallyrelated but judgmentallyharmful associ-ations are brought to bear on the task. Debiasing requires theperformance of a behavior that will activate different associa-tions.

    Psychophysically based errors are due to the nonlinear rela-tion between external stimuli and the subjective responses tothose stimuli. Debiasing therefore requires changing the loca-tion of one's position on the curvedepicting this relation or theposition of one or more of the options.

    Twenty years of extremely creative research have docu-mented the presence of many judgment shortcomings. It ishoped that this taxonomywill help in the search fortechniqueswith which wewill be able todebiassuch errors.

    ReferencesAnderson, J. R. (1983). Th e architecture of cogni tion. Cambridge, MA:

    HarvardUniversity Press.Anderson, J. R., & Bower, G. H. (1973). Hitman associative memory.

    Washington,DC: Winston.Archer, J. (1988). The sociobiology of bereavement: Areply to Little-

    field and Rushton. Journal of Personality and Social Psychology, 5},272-278.

    Arkes, H. R..&Blumer, C.(1985).The psychology of sunkcost.Organi-zational Behavior and Human Decision Processes, 35,125-140.

    Arkes, H. R., Christensen, C, Lai, C, &Blumer, C. (1987). Two meth-ods of reducing overconfidence. Organisational Behavior and Hu-man Performance, 39,133-144.

    Arkes, H. R., Faust, D.,Guihnette,T.J.,&Hart,K. (1988). Eliminatingthe hindsight bias. Journal of Applied Psychology, 73,305-307.

    Arkes, H. R,&Freedman,M. R. (1984). Ademonstration of thecostsand benefits of expertise in recognition memory. Memory &Cogni-tion, 12.84-89.

    Arkes,H. R., & Harkness, A. R. (1980). Theeffect of making adiagno-sis on the subsequent recognition of symptoms. Journal of Experi-mental Psychology: Human Learning and Memory, 6, 568-575.

    Arkes, H. R, & Harkness, A. R. (1983). Estimates of contingency be-

  • 7/28/2019 Judgmental Errors

    12/13

    COSTS AND BENEFITS OF JUDGMENT ERRORS 497

    tween two dichotomousvariables. Journal of Experimental Psychol-ogy: General, 112,117-135.

    Bar-Hillel, M. (1973). On the subjective probability of compoundevents. Organizational Behavior and Human Performance, 9, 396-406.

    Beach, L. R, & Mitchell, T. R. (1978). A contingency model for theselection of decision strategies. Academy of Management Review, 3,439-449.Berkeley, D., & Humphreys, P. (1982). Structuring decision problemsand the "biasheuristic."Acta Psychologica, SO, 201-252.

    Billings, R. S, &Marcus, S. A. (1983). Measures ofcompensatory andnoncompensatory models ofdecision behavior: Processtracing ver-sus policy capturing. Organizational Behavior and Human Perfor-mance. 31, 331-352.

    Chapman, L., &Chapman, J. (1967).Genesisofpopular but erroneouspsychodiagnostic observations. Journal of AbnormalPsychology, 72,193-204.

    Christensen, C. (1989). The psychophysics of spending. Journal of Be-havioral Decision Making, 2, 69-80.

    Christensen-Szalanski, J. J. J. (1980). A further examination of theselection of problem-solving strategies:Theeffects of deadlines andanalytic aptitudes. Organizational Behavior and Human Perfor-mance, 25,107-122.

    Dinnerstein. D. (1965). Intermanual effects of anchors on zones ofmaximal sensitivity in weight-discrimination. AmericanJournal ofPsychology, 78, 66-74.

    Doherty, M E.,Mynatt, C.R.,Tweney, R. D,&Schiavo, M D(1979).Pseudodiagnosticity. Acta Psychologica, 43,111-121.

    Edwards, W(1983). Human cognitivecapabilities, representativeness,and ground rules for research. In P. C. Humphreys,O.Svenson,& A.Vari (Eds.), Analysing and aiding decision processes (pp. 507-513).Amsterdam: North-Holland.

    Edwards, W, & von Winterfeldt,D. (1986). On cognitive illusions andtheir implications. In H. R. Arkes & K. R. Hammond (Eds.), Judg-ment anddecision making: An interdisciplinary reader(pp. 642-679).Cambridge, England: Cambridge UniversityPress.

    Einhorn, H. J, &Hogarth, R. M. (1981). Behavioral decision theory:Processes of judgment andchoice. Annual Reviewof Psychology, 32,53-88.

    Fischhoff, B. (1975). Hindsight *foresight: The effect of outcomeknowledge on judgment under uncertainty.Journal of ExperimentalPsychology: Human Perception and Performance, 1, 288-299.

    Fischhoff, B. (1982). Debiasing. In D. Kahneman, P. Slovic, & A.Tversky, (Eds.), Judgment tinder uncertainty: Heuristics and biases(pp. 422-444). Cambridge, England: Cambridge University Press.

    Fischhoff, B., &Beyth-Marom, R. (1983). Hypothesis evaluation froma Bayesian perspective. Psychological Review, 90,239-260.

    Fischhoff, B, Slovic, P, & Lichtenstein, S. (1977). Knowing with cer-tainty: The appropriatenessof extremeconfidence.Journal of Exper-imental Psychology: Human Perception and Performance, 3, 552-564.

    Funder, D. C. (1987). Errors and mistakes: Evaluating the accuracy ofsocial judgment. Psychological Bulletin, 101,75-90.Gilovich,T.(1981). Seeingthe past in the present: Theeffect ofassocia-

    tions to familiar events on judgmentsand decisions. Journal of Per-sonality and Social Psychology, 40, 797-808.

    Gregory, WL.,Cialdini, R. B, &Carpenter, K. M.(1982). Self-relevantscenarios as mediators of likelihood estimates and compliance:Does imaginingmakeitso?Journal ojPersonality'andSocial'Psychol-ogy , 43, 88-99.

    Grice, H. P. (1975). Logic andconversation. In D.Davidson& G. Har-man (Eds.), Thelogic of grammar (pp. 64-75). Encino, CA:Dickin-son.

    Harkness,A.R., DeBono, K. G., &Borgida,E. (1985).Personal involve-ment and strategies for making contingency judgments: Astake inthe dating game makesadifference. Journal of Personalityand SocialPsychology, 49, 22-32.

    Helson,H.(l964).Adaptation-leveltheory:Anexperimentalandsystem-atic approach to behavior. New York: Harper.

    Hoch,S. J.(1985).Counterfactual reasoning and accuracy inpredictingpersonal events. Journal of Experimental Psychology: Learning,Memory, andCognition, 11,719-731.

    Hogarth, R. M. (1981). Beyond discrete biases: Functional and dys-functional aspects of judgmental heuristics. Psychological Bulletin,90,197-217.

    Hull, C. L. (1920). Quantitative aspectsof the evolution ofconcepts.Psychological Monographs, 28, (1,Whole No. 123).

    Johnson, E. J., & Payne, J. W (1985). Effort and accuracy in choice.Management Science, 31, 395-414.

    Kahneman, Q, Slovic, P., &Tversky, A. (Eds). (1982).Judgment underuncertainly: Heuristics andbiases.Cambridge, England: CambridgeUniversity Press.

    Kahneman, D, &Tversky, A. (1973). On the psychology of prediction.Psychological Review, 80, 237-251.

    Kahneman, D., &Tversky, A. (1979). Prospect theory: An analysisofdecision under risk. Econometrica, 47,263-291.

    Kahneman, D., & Tversky, A. (1984). Choices, values, and frames.American Psychologist, 39, 341-350.

    Koriat, A., Lichtenstein, S., &Fischhoff, B. (1980). Reasons for confi-dence. Journal of Experimental Psychology: Human Learning andMemory, 6,107-118.

    Kubovy,M. (1977).Response availability and the apparent spontaneityof numerical choices. Journal of Experimental Psychology: HumanPerception and Performance, 3, 359-364.

    Lehman, D.R., Lempert, R. O, &Nisbett, R. E. (1988). The effects ofgraduate training on reasoning: Formal discipline and thinkingabout everyday-lifeevents. American Psychologist, 43, 431-442.

    Lichtenstein, S., Fischhoff, B., &Phillips, L. D. (1982). Calibration ofprobabilities: Thestateof the art to1980. InKahneman, D, Slovic,R, & Tversky, A. (Eds.), Judgment under uncertainly: Heuristics andbiases (pp. 306-354). Cambridge, England: Cambridge UniversityPress.

    Loewenstein, G. F. (1988). Frames of mind in intertemporal choice.Management Science, 34, 200-214.

    Lord, C. G., Lepper, M. R., & Preston, E. (1984). Considering theopposite:Acorrective strategy forsocialjudgment.Journal of Person-ality and Social Psychology, 47,1231-1243.

    McNeil, B. J, Pauker, S.J, Sox, H C, Jr.,&Tversky, A.(1982). On theelicitation of preferences for alternative therapies. New EnglandJournal of Medicine, 306,1259-1262.

    Murphy, A. H, & Winkler, R. L. (1974). Subjective probability fore-casting experiments inmeteorology: Some preliminaryresults.Bul-letin of theAmerican Meteorological Society, 55,1206-1216.

    Neely, J. H. (in press). Semantic primingeffects invisual word recogni-tion: Aselective reviewofcurrent findings and theories. In D. Besner&G. Humphreys(Eds), Basicprocessesinreading.-Visual word recog-nition. Hillsdale, NJ:Erlbaum.

    Nisbett, R. E, Krantz,D. H.,Jepson, C, &Kunda, Z.(1983).Theuseofstatistical heuristics in everyday inductive reasoning. PsychologicalReview, 90, 339-363.

    Northcraft, G. B., & Neale, M. A. (1986). Opportunity costs and theframing of resource allocation decisions. Organizational Behaviorand Human Decision Processes, 37, 348-356.

    Northcraft, G. B, & Neale, M. A. (1987). Experts, amateurs, and realestate: An anchoring-and-adjustment perspective on property pric-

  • 7/28/2019 Judgmental Errors

    13/13

    498 1 HAL R. ARKESing decisions. Organizational Behavior and Human Decision Pro-cesses, 39, 84-97.

    Paquette, L., & Kida, T. (1988). Theeffect of decision strategy and taskcomplexity on decision performance. Organizational BehaviorandHuman DecisionProcesses, 41,128-142.

    Payne, J. W (1982).Contingent decision behavior. Psychological Bulle-tin, 92, 3&2-402.

    Payne,J. W, Bettman, J. R, &Johnson, E. J. (1988).Adaptive strategyselection in decision making. Journal of Experimental Psychology:Learning, Memory, and Cognition, 14,534-552.

    Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement onresponses toargument quantityandquality: Central and peripheralroutes to persuasion. Journal of Personality and Social Psychology,46,69-81.

    Phillips, L. D. (1983). A theoretical perspective on heuristics andbiases in probabilistic thinking. In P. C. Humphreys, O. Svenson, &A. Vari (Eds.), Analysing and aiding decision processes (pp. 525-543). Amsterdam: North-Holland.

    Posner, M. I., &Snyder, C. R. R. (1975). Attentionand cognitive con-trol. In R. L. Solso (Ed.), Information processingandcognition: TheLoyola symposium (pp. 55-85). Hillsdale, NJ:Erlbaum.

    Powell, J. L. (1988). A test of the knew-it-all-along effect in the 1984presidential and statewide elections. Journal of Applied Social Psy-chology, 18, 760-773.

    Ratcliff, R., &McKoon, G. (1981). Automaticand strategic priminginrecognition. Journal ojVerbal LearningandVerbal Behavior, 20,204-215.

    Restle, F. (1971). Instructions and the magnitudeofan illusion: Cogni-tive factors in the frame of reference. Perception and Psychophysics,9, 31-32.

    Ross, L., Lepper, M. R., Strack, F, & Steinmetz, J. L. (1977). Socialexplanation andsocial expectation: Effects of real and hypotheticalexplanationsof subjective likelihood.Journal o f Personality and So-cial Psychology, 35, 817-829.

    Shaklee, H., &Tucker, D. (1980). Arule analysisof judgments of co-variation between events. Memory andCognition, 8,459-467.

    Slobin, D. I. (1971).Psycholinguistics. Glenview, IL: Scott, Foresman.Slovic, P., &Fischhoff, B. (1977). On the psychology of experimentalsurprises. Journal of Experimental Psychology: Human PerceptionandPerformance, 3, 544-551.

    Slovic, P., &Lichtenstein, S. (1971). Comparison of Bayesian and re-gressionapproaches to the studyof information processinginjudg-

    ment. Organizational Behavior and Human Performance, 6, 649-744.

    Smith, E.E., Shoben, E J.,&Rips, L. J.(1974).Structureandprocess insemantic memory: A feature model for semantic decisions. Psycho-logical Review, 81, 214-241.

    Staff. (1989, September 4). Dont B-2 sure. TheNewRepublic, pp. 7-8.Stevens, S. S. (1957).On the psychophysical law.Psychological Review,

    64,153-181.Stroop, J. R. (1935). Studies of interference in serial verbal reactions.Journal of Experimental Psychology, IS, 643-662.

    Sutherland, H. J., Dunn, Y, & Boyd, N. F.(1983).The measurement ofvalues forstates of health with linear analog scales.Medical DecisionMaking, 3,477-487.

    Tetlock, P E., & Kim, J. I. (1987). Accountability and judgment pro-cesses in a personality prediction task. Journal of Personality andSocial Psychology, 52, 700-709.

    Thaler, R. (1985). Mental accounting and consumer choice. MarketingScience, .199-214.

    Thaler, R. (1986). The psychology and economics conference hand-book: Comments on Simon, on Einhorn and Hogarth, and onTversky and Kahneman. Journal of Business, 59, S279-S284.

    Thorngate, W(1980). Efficient decision heuristics. Behavioral Science,25, 219-225.

    Tversky, A, & Kahneman, D. (1973). Availability: A heuristic for judg-ing frequency and probability. CognitivePsychology, 5, 207-232.

    Tversky, A, & Kahneman, D. (1974). Judgment under uncertainty:Heuristics and biases. Science, 185,1124-1131.

    Tversky, A., & Kahneman, D. (1981). The framing of decisions and therationality of choice. Science, 211, 453-458.

    Tversky, A., &Kahneman, D. (1983). Extensions!versus intuitive rea-soning: Theconjunction fallacy in probability judgment. Psychologi-cal Review 90, 293-315.

    Wagenaar, W, & Keren, G. B. (1986). Doesthe expert know? The reli-abilityof predictions andconfidence ratingsof experts. In E.Hollna-gel, G. Mancini, & D. Woods (Eds.), Intelligent decision support inprocess environments (pp. 87-103). Berlin:Springer-Verlag.

    Yates, J. F, &Carlson, B. W(1986). Conjunction errors: Evidenceformultiplejudgment procedures, includingwsignedsummation.1'Orga-nizational Behavior and Human Decision Processes, 37, 230-253.

    Received December 6,1989Revision received July 10,1990

    Accepted February 12,1991


Recommended