peod wr steacenctters onctore so
models of human motivation. 2009 Published by Elsevier Inc.
ine deed to aed onees closdiencecomes struce fall t
We suspect that the story of Alypius rings true for most peo-ple. Who has not at one time or another insisted that the bearerof bad tidingsa doctor, a family member or an uninteresteddategive it to us straight without sugar-coating it, only tond that the truth can be a bitter pill to swallow? To be sure,such pursuits are occasionally premised on the hope that theinformation will turn out to be less dire than it appears (Shani,Tykocinski, & Zeelenberg, 2008). A suitor may insist on knowing
they know they are better off without.And yet, for all its intuitive appeal among psychologists, philos-
ophers and laypeople alike, there are no data that actually supportthis notion. Quite the contrary: individuals are notorious for theirtendency to seek information that they want to hear more thaninformation they do not (Brickman & Bulman, 1977; Greenwald,1980; Kunda, 1990; Pyszczynski, Greenberg, & LaPrelle, 1985; Tay-lor & Lobel, 1989; Wood, 1989). The notion that people voluntarilyexpose themselves to information that they believe will causemore harm than good seems to contradict these ndings, as wellas the motivational processes thought to underlie them.
* Corresponding author.
Journal of Experimental Social Psychology 45 (2009) 11731179
Contents lists availab
.eE-mail address: firstname.lastname@example.org (J. Kruger).What is paradoxical about the story of Alypius is not merelythat he was harmed by his actions. After all, knowledge typicallyhelps more than it hurts and individuals can hardly be expectedto perfectly predict the exceptions. Instead, what is paradoxicalabout Alypius fateand, for that matter, that of Pandora, Eveand (in case Christians did not get the hint) Lots wifewas thatit was voluntary. In his writings, Saint Augustine makes it clearthat Alypius opened his eyes despite believing that he would bebetter off not doing so.
be willing to suffer the short-term devastation associated withlearning a love interests true feelings (or lack thereof) if itmeans learning a valuable life lesson.
But there are also other occasions, we offer, in which individualsseek knowledge that they believe has no foreseeable benetknowledge that, as in the case of Alypius, they believe will causethemmore harm than good. Indeed, this is precisely the contentionof Weiner (1986) and Loewenstein (1994), both of whom have ar-gued that curiosity can cause individuals to seek information thatVisceral factors
In his confessions, Saint AugustAlypius who, though utterly opposrial shows, was nevertheless persuadAlthough determined to keep his eytest, upon hearing the cries of the auone of the gladiators he was overopened his eyes, and in so doing wain his soul than the other. . . on whosraised (St. Augustine, 1943).0022-1031/$ - see front matter 2009 Published bydoi:10.1016/j.jesp.2009.06.009scribed a friend namednd detesting gladiato-day to visit a coliseum.ed throughout the con-at the brutal killing of
with curiosity. Alypiusk with a deeper woundhat mighty clamor was
a seemingly uninterested lovers true feelings, for instance, inpart because he or she hopes that they will turn out to be morefavorable than otherwise suggested. As well, there are clearlyother occasions in which individuals expose themselves tounpleasant information because the perceived benets outweighthe perceived costs. Few individuals want to believe that theyhave cancer or HIV, for instance, but most are willing to livewith that knowledge if it grants the opportunity to extend theirlife (Dawson, Savitsky, & Dunning, 2006). Similarly, a suitor mayMotivationHedonism mon yet paradoxical aspect of human behavior that presents a counterpoint to traditional hedonisticThe paradox of Alypius and the pursuit o
Justin Kruger a,*, Matt Evans b
aMarketing Department, New York University, 40 W. 4th Street, New York, NY 10012, Ub Psychology Department, DePaul University, 2219 N. Kenmore Avenue, Chicago, IL 6061
a r t i c l e i n f o
Article history:Received 26 January 2009Revised 28 May 2009Available online 21 June 2009
Keywords:CuriosityJudgment and decision makingHeuristics and biases
a b s t r a c t
Prior work has found thatThe present work examineharmful information. In foutionable personal value. Inunderweighted its consequbelieved they would be becontrol: Visceral inuence292] analysis of visceral fadeciding whether to expos
Journal of Experime
journal homepage: wwwElsevier Inc.unwanted information
d Statesnited States
ple occasionally seek useless information, a violation of strict rationality.hether and why curiosity can also cause individuals to seek predictablyudies, participants were given the opportunity to gain knowledge of ques-h case, participants focused on their curiosity about the information andes. As a result, participants tended to seek knowledge that they themselvesoff without. Consistent with Loewensteins [Loewenstein, G. (1996). Out ofbehavior. Organizational Behavior and Human Decision Processes, 65, 272s in decision making, these effects diminished with a time delay and whenmeone else to unpleasant information. These results shed light on a com-
le at ScienceDirect
al Social Psychology
l sevier .com/locate / jesp
the private eyes written report. Actually viewing a photograph of
talNevertheless, there are a handful of studies that on the surfaceat least appear to support our thesis. Tversky and Shar (1992; seealso Bastardi & Shar, 1998; Baron, Beattie, & Hershey, 1988), forinstance, documented a number of instances in which decisionmakers postpone decisions until they gather irrelevant informa-tion. For instance, an individual might postpone purchasing a tripto Hawaii until after learning the result of an important exam, de-spite the fact that he or she would take the trip regardless of thetest outcome. To the extent that knowledge acquisition is costly,these examples could be construed as evidence of the pursuit ofknowledge that causes more harm than good.
What is unclear from these studies, however, is whether partic-ipants knowingly pursued information that was of little or no value.Rather, as Tversky and Shar (1992) argued, the information waslikely perceived as instrumental. When deciding whether to pur-chase a ticket to Hawaii, for instance, it seems as if knowingwhether one has passed or failed a major exam would be useful.As such, this research does not necessarily show that people seekinformation that they think they are better off not pursuing.
More recent research is perhaps clearer in this matter. Shaniand colleagues (Shani et al., 2008; Shani & Zeelenberg, 2007; VanDijk & Zeelenberg, 2007) report a series of studies in which partic-ipants earned money and were then offered the opportunity tolearn how much money they would have earned had they behaveddifferently. In some cases, the counterfactual outcome was eitherprobably or certainly superior to the factual one. Learning thatone has lost out is presumably unpleasant, many yet despite thisfact, participants wanted to know how much better off they wouldhave been. However, the fact that there was uncertainty abouteither whether or by how much participants would have been bet-ter off raises the possibility that there may have been a silver liningto knowledge, namely, that although one has lost money, one hasnot lost as much money as one could have. As such, these studiesalso do not necessarily show that people seek information thatthey believe they are better off without.
The present work was designed to document and explain thisparadoxical phenomenon. We conducted three studies in whichwe contrasted participants knowledge-seeking behavior with theirown assessment of whether learning the information would causethem more harm than good. This operationalization allowed us toestablish whether people seek information that they themselvesbelieve they are better off without, as opposed to information thatmerely turns out to be unwanted. This operationalization alsomade it possible to examine this paradoxical behavior ethicallywithout actually having to expose people to harmful informationand in the process (by denition) harming them.
Along the way, we also documented one source of this phenom-enon. We reasoned that when deciding whether to seek knowl-edge, people tend to focus not on the long-term consequences ofgaining the knowledgethat is, the overall utility of learning theinformationbut rather on their short-term curiosity about theinformation. Although inconsistent with rational models of humanbehavior, this prediction follows from the tendency described byLoewenstein and his colleagues for hot visceral factors to over-whelm cold non-visceral factors in judgment and decision mak-ing (Loewenstein, Weber, Hsee, & Welch, 2001). Just as theimmediate gratication of hunger, craving or arousal can causeone to eat, drink or have sex despite ones genuine (and sometimesdesperate) desire to stay thin, sober or abstinent, so too can curios-ity cause one to seek knowledge that one sincerely believes onewould be better off without.
To test this visceral-factors account, in Study 2 we experimen-tally varied participants expected outcome of learning the infor-
1174 J. Kruger, M. Evans / Journal of Experimenmation while holding curiosity constant, and in Study 3 wevaried the extent to which learning the information would satisfyones curiosity while holding the expected outcome of learning itthe act is unlikely to convey any additional benet. The hedonicimpact of such a viewing, in contrast, is far easier to guess: anger,disgust and humiliation, the likes of which we suspect the momen-tary satisfaction of ones curiosity is unlikely to match. As such,from a purely rational standpoint, one might expect individualsin this unfortunate situation to forgo the opportunity to view thephotographs.
We suspect, however, that this is not what most people in thissituation do. As in the case of Alypius, we suspect that most peoplechoose to view the photographs, despite knowing full well thatthey might be better off forgoing that opportunity. The presentstudy was designed to test this hypothesis indirectly.
ParticipantsThirty private investigators were interviewed over the tele-
phone on a volunteer basis. They were sampled randomly from alist obtained on the Internet.
Procedureconstant. If people focus on curiosity and underweight the ex-pected outcome, then people should be more sensitive to manipu-lations of the former, despite the fact that the latter is presumablyof greater consequence. Furthermore, we measured participantscuriosity about the information, as well as what they expected tobe the outcome of learning the information, to see which was abetter predictor of behavior.
In Study 4, we examined yet another moderator that followsfrom the visceral-factors account, and in doing so, examined oneof its more peculiar properties. Whereas ones own visceral factorstend to have a large and sometimes overwhelming inuence onjudgment and decision making, the visceral factors experiencedby others tend to have a much smaller inuence (Loewenstein,1996). Whereas ones own cravings for a cigarette can cause oneto light up despite ones sincere desire to quit, the cravings of oth-ers have much less inuence on behavior (or would have, if onlysmokers were willing or able to leave the decision to smoke tosomeone else). Applied to curiosity, this suggests that individualsought to be much less likely to expose others to unwanted infor-mation than they are to expose themselves to the sameinformation.
Contrary to the popular image portrayed in novels, lms andtelevision, most private investigators neither solve murders nordrive Ferraris (although several, we have discovered, wear mus-taches). Instead, a far more common activity is the decidedlyunglamorous task of spying on individuals suspected of indelity(Midwestern Detective Agency, personal communication, February4, 2004). Once the evidence is obtained (typically a photograph ofthe suspect engaged in an incriminating act), the client (typicallythe spouse of said suspect) is given a report by the investigator.He or she then must decide whether to view the incriminatingphotograph(s).
What is the individual to do? The answer, of course, depends onthe value of the information. And it is not unreasonable to suggestthat there is value inherent in discovering a spouses indelity andperhaps even value to discovering some of the lurid details. Butnote that most, if not all, of this information can be gleaned from
Social Psychology 45 (2009) 11731179Participants were told that the researchers were conducting asurvey of private investigators and the work they do. Those whoagreed to participate were then asked two questions: (1) When
hardly paradoxical if, for instance, only a small minority (8% or
was designed in part to overcome these limitations.
sonal preferences on a number of categories (e.g., food, cars, music,sports).
Once this phase of the study was nished, the experimenter col-lected each participants completed preference questionnaire. Aftera lengthy delay, the experimenter returned, explaining that the
siderably more negative outcome if it contained ridicule (M = 0.63)
J. Kruger, M. Evans / Journal of Experimental Social Psychology 45 (2009) 11731179 1175ProcedureParticipants were recruited in groups of 35. The experimenter
began by explaining that the study would take place in two phases.Phase 1 would involve making a series of preference judgments inprivate and Phase 2 would ostensibly be a group discussion aboutthose preferences. Each participant was then escorted to a separateroom in which they completed a questionnaire asking for their per-Study 2
Participants in Study 2 were given the opportunity to eavesdropon a group of individuals ostensibly talking about them behindtheir back. Half of participants were led to believe that the conver-sation was attering and complimentary, and others, insulting andderogatory. We then measured (1) participants expected outcomeof learning the information and (2) their curiosity about it. We ex-pected that despite a considerable difference in the self-reportedutility of the information, participants would underweight this dif-ference and instead focus on their curiosity about the information.As a consequence, we expected participants in the ridicule condi-tion to choose to learn information they believed they would bebetter off not knowing.
ParticipantsFifty-nine students (33 women) enrolled in an introductory
course in psychology received partial course credit for theirparticipation.fewer, to be exact) of the clients would be worse off for seeingthe photos. This did not seem to be the case, at least according tothe private investigators. On average, they reported that 33% oftheir clients would be worse off for seeing the photos, a gure thatwas signicantly larger than the 8% one would expect if their cli-ents were behaving rationally, F(1, 29) = 15.18, p < .001, g2 = .34.1
There are, however, at least two limitations of this study. First,the accuracy with which private investigators answered the ques-tions is unclear. The choice dependent measure is limited by mem-ory and the outcome dependent measure is limited by theinvestigators intuitions. Second, even if the clients did seek infor-mation that they were better off without, it is unclear whetherthey did so knowingly. After all, although the private investigatorsmay (or may not) know whether the information causes their cli-ents more or less harm than good, their clients may not. Study 2Results and discussion
According to the private investigators surveyed, an average of92% of their clients chose to view the photographs of their spouseengaged in an act of indelity, with the remaining 8% electing notto view the photographs. Although a low gure, note that it isyou obtain incriminating photos on indelity cases, what percent-age of the time do your clients choose to see the photographs?and (2) When you obtain incriminating photos on indelity cases,what percentage of the time would your clients be worse off if theysaw the photos? The order in which these two questions wereasked was counterbalanced. Once the questions were answered,the investigators were thanked and debriefed.1 Because the data were not normally distributed, we repeated this analysis using aWilcoxon signed-rank test, with similar results.than if it contained attery (M = 2.55), F(1, 57) = 7.85, p = .007,g2 = .12. There was no difference found or predicted in participantscuriosity about the information contained in the transcript, how-ever. As expected, participants were extremely curious about thecontents of the transcript and this was equally true if it containedcompliments (M = 7.86) or criticisms (M = 7.53), F(1, 57) < 1, ns. Inother words, whereas the experimental manipulation stronglyinuenced participants expected outcome of learning the informa-tion, it had no inuence on their self-reported curiosity about theinformation.
How did this difference in the expected outcome of learning theinformation inuence participants information-seeking behavior?As Table 1 shows, 93% of participants in the attery conditionchose to receive the transcript, a gure that was statistically indis-tinguishable from the 87% who did so in the ridicule condition,X2 < 1, ns. This was true despite the fact that a larger proportionof participants in the ridicule condition than in the attery condi-tion said learning the information would cause more harm thangood, X2 = 7.39, p < .01.
Did this behavior translate into a tendency for participants inthe ridicule condition to seek unwanted information? To nd out,
Table 1Proportion of individuals who reported that the knowledge would cause more harmthan good versus the proportion who chose ignorance, by condition (Study 2).
Condition Would cause more harm than good (%) Chose ignorance (%)rest of the group had just engaged in a group discussion aboutthe participants preferences. Specically, participants were toldthat the group had been instructed to focus either on their positivecharacteristics and be attering or to focus on their negative char-acteristics and be insulting, with the entire conversation ostensiblyrecorded for future analysis. Participants were further told that be-cause of an ethics requirement, each participant had the option ofhaving a typed transcript of the conversation mailed to them. Inactuality, no conversation took place and thus the transcript wasctitious.
At this point in the experiment, all three dependent measureswere collected, each under the guise of a post-experimental ethicssurvey. In addition to deciding whether to receive the transcript ona dichotomous scale (the choice measure), participants rated theexpected outcome of receiving the information on a scale from5 (more harm than good) to +5 (more good than harm), with themidpoint of the scale (0) labeled neither. They also rated howcurious they were about the information on a scale from 0 (notat all) to 10 (very).
Once the experiment was complete, all participants wereprobed for suspicion, debriefed and thanked.
Results and discussion
No participant voiced any suspicion about whether the conver-sation actually took place or that the ethics survey was actuallypart of the study.
Our rst set of analyses focused on the between-participantsdifferences in expected outcome and curiosity. As expected, partic-ipants thought that receiving the transcript would produce a con-Flattery 4 7Ridicule 39 13
knowor not knowthis information by ingesting a magic pill.If you choose to take the magic pill, the knowledge that yoursignicant other used to be very promiscuous before datingyou will instantly become known to you (along with all of thespecic details), but if you do not take the magic pill you will
tal Social Psychology 45 (2009) 11731179we compared the proportion of individuals who refused the infor-mation with the proportion of individuals who thought it wouldcause more harm than good in a X2 test of independence. As pre-dicted, we observed a signicant interaction, indicating that partic-ipants in the ridicule condition tended to seek unwantedinformation, X2 = 4.68, p < .05. We observed no such tendencyamong individuals in the attery condition, X2 < 1, ns.
Further evidence of the differential weight placed on curiosityversus expected outcome is provided by a regression analysis inwhich we predicted choice from participants ratings of their curi-osity about, and the expected outcome of learning, the informationcontained in the transcript. As predicted, whereas curiosity was astrong and reliable predictor of choice, b = .50, p < .001, expectedoutcome was neither, b = 0.14, p = .208.
Taken together, the results of Study 2 extend the results ofStudy 1 to a different domain and in a manner immune to the pri-mary alternative interpretations of that study. Of particular impor-tance is the fact that the participants themselves rated theexpected outcome of learning the information, thus providing therst evidence of which we are aware of the tendency to seek infor-mation that one believes will cause more harm than good.
One notable aspect of Study 2 was that the tendency of individ-uals to seek unwanted information, although statistically signi-cant, was rather small. One clue as to why may be the manner inwhich participants ostensibly were to learn the information.Rather than offering participants the opportunity to receive thetranscript immediately, credulity required that the typed tran-script would not be available until after some delay. As it turnsout, this subtle detail likely had a profound (but from our vis-ceral-factors account, perfectly predictable) impact on behavior,one we examine next in Study 3.
Study 3 was designed to explore a potential moderator of thetendency to seek unwanted information, and in doing so, furthertest our visceral-factors account of it. That moderator was time.One hallmark of visceral factors is that they are eeting, as is theirinuence on judgment and decision making (Loewenstein, 1994,1996). For instance, whereas hunger or sexual desire can dramati-cally increase the desirability of immediate food or sex, it has muchless inuence on the desirability of future food or sex.
The same, we offer, is true of curiosity and its inuence on thedesirability of knowledge. If people focus more on their curiosityabout information and underweight the overall impact of learningthe information, then merely introducing a delay between onesdecision to obtain the information and the moment the informa-tion will become available ought to reduce the tendency to seekunwanted informationeven if it has no inuence on the self-re-ported expected outcome of learning the information. Study 3was designed to test this hypothesis.
ParticipantsNinety-two students (51 women) enrolled in an introductory
course in psychology received partial course credit for theirparticipation.
ProcedureParticipants completed a questionnaire in which the following
hypothetical scenario was described:
Imagine that your signicant other used to be very promiscuous
1176 J. Kruger, M. Evans / Journal of Experimenbefore dating you, but that you dont know this. Imagine furtherthat you have the unusual option of choosing whether toremain forever unaware of this fact.
Participants then indicated on a dichotomous scale whetherthey would take the pill and rated the expected outcome of doingso along the same scale used to measure expected outcome inStudy 2. Participants also indicated howmuch their curiosity aboutthe information would be satised by taking the pill on a scalefrom 0 (not at all) to 10 (very much).
Participants then answered each of these questions again, onlythis time, instead of the effects of the pill being felt immediately,participants were asked to imagine that the effects would not befelt for exactly 10 months. The actual order in which the two ver-sions of the scenario (instantly versus 10-month delay) were pre-sented was counterbalanced across participants.
Results and discussion
Ourrst setof analyses concerned thebetween-conditionsdiffer-ences in curiosity and expected outcome. As predicted, participantsthought that taking the pill would satisfy their curiosity more if itseffects were immediate than if they were delayed,Ms = 7.29 versus6.47, respectively, F(1, 91) = 13.13, p < .001, g2 = .13. There was nodifference found (or predicted) in participants expected outcomeof learning the information, however. The outcome of learning theinformation was expected to be equally negative regardless ofwhether it would be known immediately (M = 2.25) or after a 10-month delay (M = 2.32), F(1, 91) < 1, ns.
How did these differences (and non-differences) translate intobehavior? As predicted, a greater proportion of participants re-ported that they would take the pill when its effects were immedi-ate (49%) than when they were delayed (33%), X2 = 5.06, p < .05,despite the fact that participants expected outcome of learningthe information was constant across conditions. Comparing thesepatterns of choice with the proportion of individuals who felt thatlearning the information would cause more harm than good re-vealed that participants in the immediate condition showed a reli-able tendency to seek out what they believed they were better offnot knowing, X2 = 12.29, p < .001. As Table 2 shows, however, thistendency disappeared in the delay condition, X2 = 3.37, ns.
Follow-up analyses revealed that this difference in choice wasmediated by curiosity (but not expected outcome). Because ofour within-participants design, we used the procedure outlinedby Judd, Kenny, and McClelland (2001). For readers unfamiliar withthis procedure, the rst step is to establish that the independentvariable (the delay manipulation) is signicantly related to boththe proposed mediator(s) and the dependent variable. As alreadymentioned, this was the case for curiosity but not expected out-come, thus ruling out the latter as a mediator. The next step is toestablish that the proposed mediator (curiosity) is signicantly re-lated to the dependent variable. Here, too, this was the case,b = .20, p = .006. The nal step involves predicting the between-conditions difference in the dependent variable from the be-tween-conditions difference in the proposed mediator. A signi-
Table 2Proportion of individuals who reported that the knowledge would cause more harmthan good versus the proportion who chose ignorance, by condition (Study 3).
Condition Would cause more harm than good (%) Chose ignorance (%)Immediate 75 51Delayed 79 67
cant relationship provides evidence of mediation, which is pre-cisely what we found, b = .30, p = .004. Taken together, these datasuggest that the inuence of the delay on choice was mediatedby curiosity (but not expected outcome), consistent with our vis-ceral-factors account.
One intriguing aspect of negative information is that we seemless inclined to expose others to it (at least others of whom weare fond) than to expose ourselves to the same knowledge. Whohas not at one point or another kept a friend, colleague or lovedone in the dark about some unpleasant personal informationbeit a bad tie, bad breath or ineffective lovemaking technique (seeRosen & Tesser, 1970, 1972, for other examples)despite the factthat we ourselves might insist that we be told immediately inthe event (unlikely as it may be) that such feedback apply to us?
The present visceral-factors analysis may help explain why. As ageneral rule, visceral factors experienced by others tend to have a
versus other) 2 (predictor: expected outcome versus curiosity)mixed-model ANOVA, with the rst variable as a between-partici-pants factor and the second variable as a within-participants fac-tor.2 This analysis revealed only an interaction, F(1, 81) = 6.03,p = .016, g2 = .069. As Table 5 shows, whereas participants decidingtheir own fate focused primarily on their curiosity about the infor-
behind your back
Table 4Proportion of information thought to cause more harm than good versus theproportion avoided, by condition (Study 4).
2 Because several of these correlations were exactly 1, we were unable to performthe usual Fisher r-to-z transformations without omitting a non-trivial number of
J. Kruger, M. Evans / Journal of Experimental Social Psychology 45 (2009) 11731179 1177Your haircut looks awful 51 33 2.44Parents thought you were
mentally disabled73 44 7.24*
A friend lets you win at chess 51 24 6.27*
There is a surprise birthdayparty planned for you
93 93 0
Current signicant other was 37 15 5.18*much smaller inuence on ones decision making than do onesown visceral factors (Loewenstein, 1996). To the extent that ourvisceral analysis of curiosity is correct, this suggests that individu-als ought to be less likely to expose others to unwanted informa-tion than they are to expose themselves to the same information.Importantly, this should be true not necessarily because of any dif-ference in the expected outcome of learning the information forthe self or others, nor necessarily because of any difference inthe targets presumed curiosity about it. Rather, the differenceought to be due to differences in the extent to which curiosityand expected outcome are weighted for the self versus others.Study 4 was designed to test this prediction.
ParticipantsEighty-four students (41 women) enrolled in an introductory
course in psychology received partial course credit in exchangefor their participation.
ProcedureParticipants completed a questionnaire in which 10 undesirable
hypothetical facts were presented (see Table 3). Half of the par-ticipants were asked to imagine that the information pertained to
Table 3Proportion of individuals who reported that the knowledge would cause more harmthan good versus the proportion who chose ignorance (Study 4).
Hypothetical fact Would cause moreharm than good (%)
You have a very low IQ 76 49 6.27*
One of your best friendsfrom childhood died
51 5 21.81*
Ex-signicant other wasunfaithful to you
68 41 5.96*
A meal you just ate wasdropped on the oor
76 51 5.26*
Friends have ridiculed you 46 20 6.68*very promiscuous
* p < .05.them (e.g., your signicant other used to be very promiscuous),and the other half imagined that the information pertained to aspecic friend (e.g., your friends signicant other used to be verypromiscuous). To ensure that participants in the latter conditionhad a specic individual in mind, participants were asked to writethe individuals initials at the beginning of the questionnaire.
Participants job, they were told, was to decide whether or notto expose the target to the information and also whether doingso would cause more good than harm or more harm than goodfor the target (both on a dichotomous scale). Participants wenton to indicate how curious the target (either themselves or theirfriend) would be about each piece of information on a scale from0 (not at all) to 10 (very).
Results and discussion
The responses of participants in the self condition enable a testof whether the results of the previous studies conceptually repli-cated. They did. Although an average of 7.28 (73%) of the facts weredeemed to cause more harm than good, participants avoided only3.10 (31%) of them, F(1, 38) = 101.22, p < .001, g2 = .73. In fact, asTable 3 shows, this difference was reliable for 8 out of the 10 indi-vidual items.
Was this tendency reduced when participants were decidingthe fate of someone else? To nd out, we submitted the ratingsto a 2 (self versus other) by 2 (choice versus expected outcome)mixed-model analysis of variance (ANOVA), with the rst variableas a between-participants factor and the second variable as a with-in-participants factor. As predicted, we observed a signicantinteraction, F(1, 82) = 5.14, p = .026, g2 = .06. As Table 4 shows, par-ticipants were less likely to submit someone else to informationthat they thought would cause more harm than good than to sub-mit themselves to the same information. Importantly, as Table 4also shows, this was true not because of any difference in the ex-pected outcome of learning the information, F < 1, ns, but ratherbecause of a difference in the amount of information sought,F(1, 82) = 10.63, p = .002, g2 = .12.
What caused this selfother difference? Our thesis is that whendeciding for oneself curiosity dominates, whereas when decidingfor someone else people are more likely to consider the full impactof learning the information. We tested this hypothesis by rst com-puting two partial correlation coefcients for each participant: onepredicting choice across the 10 pieces of information from partici-pants self-rated expected outcome of learning the informationholding constant curiosity, and the other with those predictors re-versed. We then compared these correlations in a 2 (condition: self
Condition Would cause more harm than good (%) Chose ignorance (%)
Self 73 31Other 73 43participants. Thus, the analyses reported are based on the raw correlation coefcients.However, repeating these analyses with either the un-transformable correlationsomitted or rounded to .99 did not change the results.
prior work (e.g., Baron et al., 1988; Bastardi & Shar, 1998; Tversky& Shar, 1992), but sought information that they themselves be-
tallieved to be of negative value. This was true regardless of whetherthat information was hypothetical (Studies 3 and 4) or real (Studies1 and 2) and regardless ofwhether the expected outcomewas ascer-tainedby an observer (Study1) or the decisionmaker him-or herself(Studies 24). As in the case of Alypius, participants chose to learnwhat they themselves thought they were better off not knowing.
The results also shed light on one source of this paradox. Wereasoned that curiosity, like other visceral factors, focuses atten-tion on that which will satisfy the craving at the expense of high-er-order goals such as maximizing ones expected outcomes(Loewenstein, 1996). Consistent with this explanation, we foundthat whereas manipulations of curiosity had a strong inuenceon information-seeking behavior (Study 3), manipulations of ex-pected outcomes had no measurable inuence (Study 2). Thiswas true despite the fact that it is expected outcomes, not curios-ity, that is presumably of greater consequence. Nevertheless, par-ticipants were far more sensitive to curiosity than they were tomation, participants deciding on the fate of someone else instead fo-cused primarily on the consequences of learning the information.
In summary, we found that participants were less likely to ex-pose someone else to unwanted information than to expose them-selves to the same information. Importantly, this was true despitethe fact that there was no difference in either the presumed curios-ity or the expected outcome of learning the information: Whetherdeciding the fate of someone else or oneself, the information wasseen as equally tantalizingand equally devastating (all ps > .50).Where there was a difference, in contrast, was in the relativeweight placed on that titillation and devastation. Whereas partici-pants deciding the fate of someone else focused on the conse-quences of learning the information, participants deciding thefate of themselves focused more on their curiosity about the infor-mation. As a result, the tendency to seek unwanted informationwas reduced when individuals were designing the fate of someoneelse.
When do people seek information? According to a rational-choice analysis of behavior, the answer is clear: only when doingso conveys tangible benets, such as allowing one to make a moreinformed decision (Asch, Patton, & Hershey, 1990; Laffont, 1989).
The results presented here, however, seriously challenge thatassumption. In four studies, participants not only tended to seekinformation that was of questionable value, as has been shown in
Table 5Mean decision weights (partial rs) placed on expected outcome versus curiosity bycondition (Study 4).
Condition Expected outcome Curiosity
Self .17 .30Other .30 .11
1178 J. Kruger, M. Evans / Journal of Experimenexpected outcomesat least when deciding their own fate.The results also shed light on two moderators of this paradoxi-
cal behavior. First, we found that introducing a delay betweenchoice and the satisfaction of curiosity enabled a more sober con-sideration of higher-order goals, consistent with work on other vis-ceral factors (Loewenstein, 1996). As a result, participants wereless likely to subject themselves to unwanted information if theacquisition of the information would be delayed (Study 3)an ef-fect that was mediated by curiosity (but not expected outcomes).Second, we found that whereas curiosity dominated the decisionof whether to expose oneself to unwanted information, the per-ceived curiosity of someone else was much less a factor (Study4). As a result, participants were less likely to expose someone elseto unwanted information than to expose themselves to the sameinformation.
Although these results are consistent with the work of Loewen-stein and his colleagues, they appear inconsistent (at least on thesurface) with other published ndings. For instance, numeroushealth researchers have noted the occasional reluctance amongmedical patients to obtain diagnostic tests for disorders such asHIV/AIDS (Lyter, Valdiserri, Kingsley, Amoroso, & Rinaldo, 1987)or cancer (Eaker, Adami, & Sparen, 2001; Weitzman, Zapka, Esta-brook, & Goins, 2001). To the extent that knowing one has thesediseases is more useful than not, this pattern would appear to con-tradict the results presented in this manuscript. That said, neverhas the tendency to seek out (or avoid) such information been di-rectly compared with the expected outcome of learning the infor-mation. It may very well be, for example, that the proportion ofindividuals who avoid such tests is dwarfed by the proportion ofindividuals who believe that it is a good idea to do so, consistentwith our account.
On rationality and adaptiveness
Although voluntarily exposing oneself to that which one be-lieves one would be better off without would appear irrational,several caveats are in order. First, note that although participantsbelieved that learning the information they sought would have neg-ative outcomes, they may have been wrong. For instance, althoughparticipants in Study 2 reported that the litany of curses aimed atthemwould be better left unknown, it is entirely possible that suchinformation may actually have been less harmful than participantsmay have anticipated. Among other things, the pain potentiallycaused by the knowledge would likely dissipate more quickly thanparticipants would expect (Wilson & Gilbert, 2003). A similar argu-ment could be made for each of the other unpleasant pieces ofinformation examined in this research, such as those listed in Table3. Although irrelevant from the standpoint of our thesis, whichexclusively concerns the pursuit of information perceived by theindividual to be damaging, it is centrally relevant for those inter-ested in the adaptiveness and (depending on whether one prefersthe dictionarys version of the word or the one favored by decisionscientists) rationality of the behavior.
On the one hand, we believe that it is entirely possible that par-ticipants may have underestimated the long-term consequences ofsome of the information examined in this research, particularlyself-relevant information. Consider Study 2. Few punishments areof greater long-term consequence than banishment from thegroup, particularly in our not-too-distant evolutionary past inwhich ostracism meant near-certain death (Baumeister & Leary,1995; Williams, 1997). As such, the momentary pain associatedwith reading a transcript of ones faults, for instanceespecially gi-ven the counterintuitively short duration of that pain (Wilson &Gilbert, 2003)might be offset in the long term if it enables oneto change ones behavior to better t group norms and expecta-tions (see Leary, Tambor, Terdal, & Downs, 1995, for a similar argu-ment regarding the tradeoffs associated with threatened self-esteem and social status sensitivity).
On the other hand, there is reason to doubt this interpretation.First, note that the tendency to seek unwanted information wasnot unique to status-related information, nor did it even appearto covary with this dimension. The largest effect in Table 3, for in-stance, occurred for the hypothetical fact One of your best friendsfrom childhood died, which presumably is irrelevant with respect
Social Psychology 45 (2009) 11731179to ones current social status. In fact, we suspect that if anything,participants tended to underestimate the negative outcomes oflearning the information in our studies. For instance, whereas
49% of participants believed that the knowledge of a dead class-mate would cause more harm than good, we would not be sur-prised if the actual gure was much higher. In short, we agreewith Loewenstein (1994, 1996) that as a general rule, excessive
Asch, D. A., Patton, J. P., & Hershey, J. C. (1990). Knowing for the sake of knowing:The value of prognostic information. Medical Decision Making, 10, 4757.
J. Kruger, M. Evans / Journal of Experimental Social Psychology 45 (2009) 11731179 1179attention to visceral factors in general, and curiosity in particular,has a tendency to disappoint.
That said, we do not doubt that the underlying processes thatproduce the behaviors we have documented in this research maybe adaptive. Quite the contrary. The heuristic value of seekingknowledge with little regard to the consequences of doing so canhardly be faulted in light of the fact that information helps far moreoften than it hurts. In short, although the behavior examined in thisresearch may be irrational, it would be misleadingand, we sus-pect, incorrectto conclude that it is generally maladaptive.
One limitation of the present work concerns the hypotheticalnature of several of the studies. Rather than actually presentingindividualswith a real decision, ethics and practicality required thatparticipants in Studies 3 and 4 merely guessed how they wouldbehave in such a situation. What is more, note that the situationsin which participants imagined themselves in these studies werenot only hypothetical, but fantastical. After all, participants weretold to imagine that an event was true and then decide whether toknow the information. Except in theworld of TheMatrix, one cannotdecide whether to know what one already knows.
That said, we feel that if anything, this is a conservative featureof the research. Note that by soliciting choice hypothetically, thereis little accountability for a bad decision, allowing participants freerein to custom tailor their responses in the interest of social desir-ability, demand characteristics, and so on. Now consider thatobservation in light of the fact that in each of our hypotheticalstudies there was an obvious right answer. It is irrational to seekwhat one believes one would be better off without (and vice versa),a fact that informal observation suggests was readily appreciatedby participants. Given the well-documented motive of study par-ticipants to appear rational, this presumably would have dimin-ished, rather than enhanced, our results.
One implication of these results concerns the selfother differ-ence we observed in Study 4. Recall that whereas participants weremore than willing to expose themselves to unwanted information,they were less likely to expose a friend to the same information. Ifinformal observation is any indication, this selfother differencecan occasionally lead to interpersonal discord. We resent it whenwe are kept in the dark, despite the fact that when ones role isswitched, we tend to keep others in the dark (as Study 4 showed).The irony, of course, is that not only do our gatekeepers have ourbest interests at heart, but they may even be in a better positionthan ourselves to decide what we are best off not knowing.
We thank Steve Chan, Tom Gilovich, Hyun Young Park, NealRoese and Patrick Vargas for their helpful suggestions throughoutthis research.Baron, J., Beattie, J., & Hershey, J. (1988). Heuristics and biases in diagnosticreasoning: II. Congruence, information, and certainty. Organizational Behaviorand Human Decision Processes, 42, 88110.
Bastardi, A., & Shar, E. (1998). On the pursuit and misuse of useless information.Journal of Personality and Social Psychology, 75, 1932.
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonalattachments as a fundamental human motivation. Psychological Bulletin, 117,497529.
Brickman, P., & Bulman, R. J. (1977). Pleasure and pain in social comparison. In J.Suls & R. Miller (Eds.), Social comparison processes: Theoretical and empiricalperspectives (pp. 149186). Washington, DC: Hemisphere.
Dawson, D., Savitsky, K., & Dunning, D. (2006). Dont tell me, I dont want to know:Understanding peoples reluctance to obtain medical diagnostic information.Journal of Applied Social Psychology, 36, 751768.
Eaker, S., Adami, H. O., & Sparen, P. (2001). Reasons women do not attend screeningfor cervical cancer: A population-based study in Sweden. Preventive Medicine,32, 482491.
Greenwald, A. G. (1980). The totalitarian ego: Fabrication and revision of personalhistory. American Psychologist, 35, 603618.
Judd, C. M., Kenny, D. A., & McClelland, G. H. (2001). Estimating and testingmediation and moderation in within-participants designs. PsychologicalMethods, 6, 115134.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108,480498.
Laffont, J. J. (1989). The economics of uncertainty and information. Cambridge, MA:MIT Press.
Leary, M. R., Tambor, E. S., Terdal, S. K., & Downs, D. L. (1995). Self-esteem as aninterpersonal monitor: The sociometer hypothesis. Journal of Personality andSocial Psychology, 68, 518530.
Loewenstein, G. (1994). The psychology of curiosity: A review and reinterpretation.Psychological Bulletin, 116, 7598.
Loewenstein, G. (1996). Out of control: Visceral inuences on behavior.Organizational Behavior and Human Decision Processes, 65, 272292.
Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch, N. (2001). Risk as feelings.Psychological Bulletin, 127, 267286.
Lyter, D. W., Valdiserri, R. O., Kingsley, L. A., Amoroso, W. P., & Rinaldo, L. R. (1987).The HIV antibody test: Why gay and bisexual men want or do not want to knowtheir results. Public Health Reports, 102, 468474.
Pyszczynski, T., Greenberg, J., & LaPrelle, J. (1985). Social comparison after successand failure: Biased search for information consistent with a self-servingconclusion. Journal of Experimental Social Psychology, 21, 195211.
Rosen, S., & Tesser, A. (1970). Reluctance to communicate undesirable information MUM effect. Sociometry, 33, 253.
Rosen, S., & Tesser, A. (1972). Fear of negative evaluation and the reluctance totransmit bad news. Journal of Communication, 22, 124141.
Shani, Y., Tykocinski, O., & Zeelenberg, M. (2008). When ignorance is not bliss: Howdistressing uncertainty drives the search of negative information. Journal ofEconomic Psychology, 29, 643653.
Shani, Y., & Zeelenberg, M. (2007). When and why do we want to know? Howexperienced regret promotes post-decision information search. Journal ofBehavioral Decision Making, 20, 207222.
St. Augustine (1943). The confessions of St. Augustine. New York: Liveright (J. G.Pilkington, Trans.).
Taylor, S. E., & Lobel, M. (1989). Social comparison activity under threat: Downwardevaluation and upward contacts. Psychological Review, 96, 569575.
Tversky, A., & Shar, E. (1992). The disjunction effect in choice under uncertainty.Psychological Science, 3, 305309.
Van Dijk, E., & Zeelenberg, M. (2007). When curiosity killed regret: Avoiding orseeking the unknown in decision-making under uncertainty. Journal ofExperimental Social Psychology, 43, 656662.
Weiner, B. (1986). An attributional theory of motivation and emotion. New York:Springer-Verlag.
Weitzman, E. R., Zapka, J., Estabrook, B., & Goins, K. V. (2001). Risk and reluctance.Understanding impediments to colorectal cancer screening. Preventive Medicine,32, 502513.
Williams, K. D. (1997). Social ostracism. In R. M. Kowalski (Ed.), Aversiveinterpersonal behaviors (pp. 133170). New York: Plenum.
Wilson, T. D., & Gilbert, D. T. (2003). Affective forecasting. In M. Zanna (Ed.).Advances in experimental social psychology (Vol. 35, pp. 345411). New York:Elsevier.
Wood, J. V. (1989). Theory and research concerning social comparisons of personalattributes. Psychological Bulletin, 106, 231248.
The paradox of Alypius and the pursuit of unwanted informationStudy 1MethodParticipantsProcedure
Results and discussion
Results and discussion
Results and discussion
Results and discussion
General discussionOn rationality and adaptivenessLimitationsImplications