Journal of Economic Literature

Journal of Economic Literature 2017, 55(1), 96–135
https://doi.org/10.1257/jel.20151245
Information Avoidance†
Russell Golman, David Hagmann, and George Loewenstein*
We commonly think of information as a means to an end. However, a growing theo-

retical and experimental literature suggests that information may directly enter the
agent’s utility function. This can create an incentive to avoid information, even when it
is useful, free, and independent of strategic considerations. We review research docu-
menting the occurrence of information avoidance, as well as theoretical and empirical
research on reasons why people avoid information, drawing from economics, psy-
chology, and other disciplines. The review concludes with a discussion of some of the
diverse (and often costly) individual and societal consequences of information avoid-
ance. ( JEL D82, D83)
What sense had I of her stol’n hours of lust? one can, according to standard economic
I saw’t not, thought it not, it harm’d not me: assumptions, ignore it at no cost.
I slept the next night well, was free and merry; Consistent with standard theory, there are
I found not Cassio’s kisses on her lips: countless situations in which information is
He that is robb’d, not wanting what is stol’n, useful and sought after. There are, in fact, sit-
Let him not know’t, and he’s not robb’d at all.
uations in which people seek out, and are even
—Othello
willing to pay for, apparently useless infor-
mation (see Eliaz and Schotter 2007, 2010;
1. Introduction Loewenstein 1994; Powdthavee and Riyanto
2015). Our focus in this review is, however,
T he standard economic analysis of deci-

sion making holds that information is
valuable to the extent, and only to the extent,
on the opposite phenomenon—on the many
situations in which people avoid information,
even when it is free and could improve deci-
that it leads to better decisions. A straight- sion making.1 As we will d iscuss, information
forward implication is that valid information avoidance occurs not only when there is a
should never be actively avoided, except strategic rationale for it, but also when beliefs
for situations in which ignorance confers a directly enter the utility function.
strategic advantage. Even if information has Casual observation, as well as consider-
no prospect of improving decision making, able laboratory and field research that we
review, suggests that information avoidance
* All three authors: Department of Social and Decision
Sciences, Carnegie Mellon University. We thank Joël van
der Weele and three anonymous reviewers for comments. 1 Although we limit our review to information avoid-
† Go to https://doi.org/10.1257/jel.20151245 to visit the ance by humans, animals may engage in the same behavior
article page and view author disclosure statement(s). (Jenkins and Boakes 1973).
96
Golman, Hagmann, and Loewenstein: Information Avoidance 97
is common. Investors avoid looking at their they are, particularly in comparison to her
financial portfolios when the stock mar- expectations.
ket is down, an “ostrich effect” (Karlsson, Even if people know the content of infor-
Loewenstein, and Seppi 2009; Sicherman et mation, they may yet choose to avoid attend-
al. 2016). Individuals at risk for health con- ing to it (an information avoidance tactic we
ditions often eschew medical tests (e.g., for discuss below). This may seem counterintui-
serious genetic conditions or STDs) even tive, but if you know that your bank account
when the information is costless and should, is depleted, you might still choose not to log
logically, help them make better decisions on to your bank’s website and peruse your
(Ganguly and Tasoff forthcoming; Lerman balance, and if you know that your paper got
et al. 1996, 1999; Lyter et al. 1987; Oster, rejected, you may still not want to read the
Shoulson, and Dorsey 2013; Sullivan, Lansky, reviewers’ comments. In the context of infor-
and Drake 2004; Thornton 2008). Managers mation seeking, Sicherman et al. (2016) found
often avoid hearing arguments that conflict that when the stock market was up, investors
with their preliminary decisions (Deshpande were more likely to log in multiple times on
and Kohli 1989; Schulz-Hardt et al. 2000; weekends, even though logins beyond the
Zaltman 1983), even when such arguments first did not reveal new information because
could help them avoid implementing mea- the market was closed. Investors seemingly
sures that are ill-founded. These examples gain pleasure from observing gains even
only scratch the surface of a wide variety of when no new information is revealed, and
situations in which people avoid information. likewise prefer to avert attention from out-
We do not review the broad and almost comes known to be adverse. In these types
infinite range of situations in which people of cases, one is aware quite precisely of what
fail to obtain information that is in their information will be revealed by looking, but
power to secure, but focus on a narrower can still, for hedonic reasons, be motivated
range of phenomena that we term “active to not look.
information avoidance.” Although people Second, for information avoidance to be
often fail to collect or attend to potentially “active,” the individual should choose to not
helpful information, only a small fraction obtain the information even if it were cost-
of such instances qualify as active avoid- less to obtain or even costly to avoid obtain-
ance. We specify two necessary criteria for ing. In most situations, information is costly
avoidance to be classified as “active”: (1) to obtain (carrying an opportunity cost, at
the individual is aware that the information least), so it will often be difficult in practice
is available, and (2) the individual has free to tell if a particular piece of information is
access to the information or would avoid the being actively avoided. If a medical test is
information even if access were free. expensive, an individual who foregoes the
First, for information avoidance to be test may do so due to its cost, a preference
“active,” the individual has to be aware that to not find out the information, or a combi-
the information exists. It is hardly a choice nation of the two. Indeed, it is even possi-
to avoid information that one does not even ble that people could use the cost to justify,
know is available. Knowing that information to themselves or to others, a decision to
exists does not, of course, generally mean avoid getting tested that actually had other
knowing its valence, much less its specific motives. If getting tested is costless (and we
content. Thus, a course instructor can know can rule out confounding factors such as an
that teaching ratings have been collected, impact on insurance rates), or avoiding get-
but not know how favorable or unfavorable ting tested is costly, then failing to get the
98 Journal of Economic Literature, Vol. LV (March 2017)
test clearly qualifies as active information potentially, compel them to give to char-
avoidance. Even when information is costly ity (Cain, Dana, and Newman 2014; Dana,
to obtain, but the individual who chooses Cain, and Dawes 2006; McGoey 2012).
not to obtain it would continue to do so if it Avoiding information that might challenge
were free or costly to avoid, we will classify existing beliefs (one of the varieties of infor-
this situation as a case of active information mation avoidance we discuss) can contribute
avoidance. Because the remainder of the to political polarization as voters and legis-
paper focuses exclusively on active informa- lators are less likely to find common ground
tion avoidance, we drop the word “active” (Kahan et al. 2012). The political paralysis
for brevity, and, in what follows, refer simply resulting from such polarization may, in turn,
to information avoidance. prevent enactment of potentially advan-
Information avoidance should be of tageous legislation to deal with problems
interest to economists because it is person- such as climate change (Marshall 2014).
ally and economically consequential. Most Information avoidance can also promote
obviously, it deprives people of potentially media bias. If people pay attention only to
valuable inputs into decision making. A per- media outlets that provide information con-
son who could, but does not, get tested for sistent with their beliefs, those outlets will be
a transmittable disease, for example, could discouraged from airing contradictory points
not obtain treatment for their condition and of view (Gentzkow and Shapiro 2010).2 The
might transmit it to others. Both of these desire to avoid information discordant with
consequences are present for HIV/AIDS, one’s beliefs can, more generally, lead to a
where drug treatments both prolong life wide range of economically consequential
and decrease the risk of transmission. outcomes: geographic sorting by beliefs,3
Information avoidance also deprives peo- proselytizing, and even to violence toward
ple of potentially useful feedback they could those one would like to silence (Golman et
use to fine-tune their behavior. Teachers al. 2016).
who fail to peruse their teaching ratings, for Information avoidance is not, however,
example, miss out on information they could always a bad thing. People would avoid
use to improve their teaching. Executives information much less often if they did not
and leaders who don’t tolerate criticism, obtain direct and immediate utility benefits
likewise, deprive themselves of information from doing so. For example, dieters who opt
that could help them make valuable changes for dessert may enjoy it more if they avoid
to their behavior. In fact, for reasons we will caloric information; bad teachers who don’t
discuss, the people who could most benefit look at their course ratings may have higher
from feedback are often, paradoxically, those levels of utility (although their students may
most likely to eschew it. not); and persons at risk of diseases, such as
Information avoidance can also lead, or genetic disorders, may be able to lead per-
at least “license,” people to take selfish or fectly happy lives until emerging symptoms
immoral actions (or fail to take altruistic or test results force the reality of their sit-
actions), as highlighted by the research on uation upon them. The (dis)utility people
“moral wiggle room” (Dana, Weber, and
Kuang 2007). A classic example of this effect
is the propensity of prosperous people to
2 This is in addition to the incentives a media source
avoid poor neighborhoods where observing
may have to influence public opinion.
the residents’ living conditions might induce 3 “Political segregation: The Big Sort.” The Economist,
guilt about their own relative privilege and, June 19, 2008. http://www.economist.com/node/11581447.
derive from beliefs should be considered a inconsistent, or less consistent, with standard
legitimate ingredient in their welfare. theory. Most importantly, information can
Information avoidance can also confer have direct hedonic value (positive or nega-
practical benefits for decision making and tive), separate from its usefulness. Behavioral
daily functioning. The weak teacher who theories that incorporate such “belief-based
avoids his teaching ratings might even teach utility” can help explain why a person might
better, at least in the short run. A student avoid information that has material value.
who is already suffering from performance In section 4, we discuss diverse individual
anxiety might only do worse after having her and societal consequences of information
fears confirmed with low performance on avoidance, most of which are negative, but
a practice test. The ostrich effect may also some of which are potentially beneficial.
help investors not to panic-sell when markets Section 5 concludes with a few final com-
are down. Information-avoidance that con- ments about the history of, and future direc-
tributes to self-serving biases in negotiations tions for, research on information avoidance.
might help self-righteous negotiators secure
better deals, even if the same biases con-
2. Methods of Information Avoidance
tribute to higher rates of impasse. And, by
using uncertainty to excuse inaction (similar Although it is natural to think of infor-
to taking advantage of moral wiggle room), mation avoidance as a matter of simply not
a cuckolded lover may be able to maintain, obtaining information, there are, in fact,
and continue to enjoy and benefit from, a diverse tactics that people use to avoid infor-
fulfilling relationship by ignoring informa- mation. We construe information avoidance
tion that might, if obtained, make him feel broadly, to include any behavioral or cogni-
compelled to act. tive process that enables one to avoid reach-
Section 2 of the paper discusses the differ- ing the conclusions that an unbiased perusal
ent tactics that people use to avoid informa- and analysis of information would lead to.
tion. Although one might view information
2.1 Physical avoidance
avoidance as a straightforward matter of sim-
ply not looking, there are many other tactics People can choose to avoid reading spe-
that people can and do use to avoid infor- cific newspapers or magazines, listening to
mation. People may, for example, not draw specific radio or television shows, looking
obvious conclusions from data. They can also at their teaching ratings, or having conver-
divert their attention from, or conveniently sations with specific people. In some cases,
forget, information they wish they had not they can, and do, even pay to avoid being
obtained in the first place. exposed to such information. In a clever and
In section 3, we review empirical research particularly clean demonstration of such an
and theory in economics, psychology, and effect, Eil and Rao (2011) had experimen-
other disciplines, dealing with informa- tal subjects either take an IQ test or have
tion avoidance. We organize the literature their attractiveness rated by other subjects.
according to the reasons why people avoid Subjects then received private prelimi-
information. Some of these reasons are nary feedback on a subset of IQ questions
consistent with standard economic theory, or attractiveness ratings that hinted at how
broadly construed. For example, informa- their final IQ test or attractiveness rating
tion can be avoided for strategic reasons; was likely to come out. People who initially
ignorance can be a very useful commitment received unfavorable information about
device (Schelling 1960). Other reasons are their appearance or intelligence, relative to
their expectations, were less likely to choose and Dean 2015; Sallee 2014; Sims 2003).
to obtain the full information, and some Although these lines of research highlight
were even willing to pay to avoid obtaining it. the fact that people have an ability to delib-
Ganguly and Tasoff (forthcoming) pre- erately direct their own attention, rational
sented participants with the possibility of inattention for the purpose of conserving
getting tested for herpes simplex virus 1 scarce cognitive resources would not qualify
(HSV-1) and virus 2 (HSV-2). Both are incur- as “active avoidance” under our definition,
able, but the latter, which is an incurable since obtaining the information in these situ-
sexually transmitted disease, was viewed by ations does incur an opportunity cost.
most subjects as a more serious condition. Once one recognizes that information
Five percent of participants were willing to is a source of utility in its own right (apart
forgo a $10 payment to avoid obtaining the from objective outcomes), it follows that
test results for HSV-1, while 16 percent were people may also allocate attention in ways
willing to forgo the same amount for the that respond to hedonic motivations. For
HSV-2 test. Moreover, all participants who example, people may choose to pay attention
avoided the HSV-1 results also avoided the to information that is likely to be positive,
HSV-2 results, while the converse was true while remaining inattentive to unfavorable
for only two-thirds of participants. This is or threatening information (even when that
consistent with HSV-2 results being more information may be more useful). Although
threatening. inattention motivated by hedonic consid-
Getting tested for an STD may require erations has not been treated as a form of
people to return to obtain the results and “rational inattention,” we note that there is
they may avoid information by failing to do nothing inherently irrational about avoiding
so. Sullivan, Lansky, and Drake (2004) sur- information that one suspects will under-
veyed more than 2,200 people who were mine one’s well-being.
at high risk of contracting HIV and found Inattention can take a range of forms,
that of those who had been tested for HIV, some of which border on physical avoidance.
18 percent failed to return to the test center For example, if one glances at a headline,
to obtain the results. Of those, 23 percent then decides not to pay attention to (i.e.,
said they did not go back because they were read) the associated article, this could be
afraid to get the results. classified either as a case of inattention or
physical avoidance. However, if one reads
2.2 Inattention
the article but then willfully and successfully
Even when people do physically obtain chooses not to think about it, this would be
information, or have it at their fingertips, they an unambiguous case of inattention. Brock
often have the ability to not focus their atten- and Balloun (1967) presented participants
tion on it. Cognitive psychologists have long in a lab experiment with speeches that sup-
known that attention is a limited resource, ported or rejected a link between cancer and
selectively employed to facilitate informa- smoking and that were favorable or unfa-
tion processing (Broadbent 1958; Schneider vorable toward Christianity. The speeches
and Shiffrin 1977; Simon 1971). Research by were masked by noise that participants could
economists has built on this insight, propos- remove by repeatedly pressing a button. In
ing that, like any scarce resource, attention four experiments, smokers were more eager
ought to be allocated efficiently and the opti- to remove static in the speech that rejected
mal allocation may involve rational inatten- a link between smoking and cancer than
tion to some pieces of information (Caplin the speech that supported a link between
the two, and vice versa for nonsmokers. not simply arrive at self-serving beliefs at
They found a weaker link (but significant will, but become biased because they update
in three of the four experiments) between their beliefs differently when informa-
a higher frequency of self-reported pray- tion supports their preexisting (or desired)
ing and reduced removal of static from the beliefs than when it fails to support those
anti-Christianity message. beliefs. Providing even more direct evidence
for Rabin and Schrag’s model, Mobius et al.
2.3 Biased interpretation of information
(2014) find, in an experiment, that people
When information is obtained and update more when receiving a positive signal
attended to, and its implications are adverse, about their ability than when they receive a
it is still possible for an individual to avoid negative signal. Even with a positive signal,
drawing the most logical conclusions from it. however, they update less than predicted by
Psychologists have long believed that peo- Bayesian updating. Asymmetric and conser-
ple filter out negative information in order vative updating seem to work in tandem to
to maintain mental health and well-being provide a good balance between protecting
(Taylor and Brown 1988). A number of stud- one’s ego utility while avoiding a potentially
ies in both psychology (e.g., Lord, Ross, and costly overestimation of one’s ability.
Lepper 1979) and economics (e.g., Babcock Peysakhovich and Karmarkar (2015) test
et al. 1995) find that people weigh and inter- the effect of favorable and unfavorable infor-
pret evidence in a fashion that supports what mation on the valuation of an ambiguous
they are motivated to believe, and that they gamble. Participants report a willingness to
tend to denigrate the quality of evidence that pay for a gamble in which their chance of
contradicts beliefs that they hold or would winning is determined by a draw of a red ball
like to hold. In one of the strongest demon- from an urn containing one hundred balls.
strations of the effect, Babcock et al. (1995) They are further told that there are at least
had negotiators read case materials either X red balls and at most Y red balls, which
before or after they were assigned to the imposes lower and upper bounds on their
role (plaintiff or defendant) they would be likelihood of winning. After offering a price
negotiating. Then, they predicted how the to play the gamble, participants either receive
judge on the case would rule and were paid favorable information (X is higher) or unfa-
for their accuracy, and finally attempted to vorable information (Y is lower). Favorable
negotiate a settlement (in an incentive-com- information, predictably, increases their
patible design). When they were assigned willingness to pay. However, the same does
their role before reading the case materials, not hold for unfavorable information: after
they were far more biased than when they adjusting the upper bound downward, their
were assigned after and, as a result, they valuation of the gamble does not change.
were more likely to reach a costly impasse. Lord, Ross, and Lepper (1979) recruited
In a secondary analysis, they had subjects subjects who had strong views in favor of, or
rate the importance of eight arguments in opposition to, the death penalty and pre-
favoring the plaintiff and eight favoring the sented to both groups two research studies,
defendant. Plaintiffs rated arguments favor- one suggesting the death penalty is effective
ing their side as more compelling than those at deterring crime and one indicating it is
favoring the defendant, and vice versa. The not. Subjects were then asked to evaluate
study supports the idea, later embodied in the quality of both studies and how convinc-
a model of confirmation bias proposed by ing they thought they were. Research that
Rabin and Schrag (1999), that people do conflicted with their previously held beliefs
scored lower on both measures than research States by 2100, and then present them with
that supported their prior beliefs. As a result, a second estimate that conveys either good
presenting both sides with the same evi- news (the temperature increase will be less
dence increased, rather than decreased, than originally believed) or bad news (the
belief polarization. increase is greater). Participants report their
Glaeser and Sunstein (2013) examine two belief in climate change and estimate the
mechanisms that can cause balanced news temperature increase after each statement.
to give rise to increased polarization. First, Respondents who reported a high belief in
even if people process information in a climate change increased their estimates
Bayesian fashion, the same information can more upon obtaining bad news than when
have opposite effects if people begin with the news is good. Although the news is bad
different priors (see also Benoît and Dubra for the planet, the information is desirable
2011). Information recipients whose views to the extent that it reinforces their existing
are reinforced by the information will sim- belief in climate change. Conversely, those
ply accept it, whereas those whose beliefs with a low belief in climate change adjusted
conflict with the message will tend to dismiss their estimate downward when the tempera-
it and question the quality or impartiality of ture increase was not as bad as originally
the source. Second, in a process involving thought, but do not update at all when it was
somewhat more nuanced psychology, the worse than believed.
memories and convictions activated by the Families who suffer an ambiguous loss of
receipt of information are likely to depend loved ones (e.g., those missing-in-action in
on an individual’s prior convictions, which war, “disappearances” caused by authoritar-
can produce what Glaeser and Sunstein ian regimes, abductions, or outdoor sports-re-
call a “memory boomerang.” New and con- lated disappearances) often refuse to accept
flicting information may remind people of the reality of the individual’s death, even in
evidence reinforcing their beliefs, and this the face of overwhelming evidence (Boss
reminder can outweigh the new information, 1999). Such families seem to experience a
leading to an overall affirmation of their exist- kind of expectational purgatory, facing but
ing views. not adapting to their loss (Frederick and
In the IQ and attractiveness rating study Loewenstein 1999:317). Denial of reality
discussed previously, Eil and Rao (2011) in this situation can be viewed as a form of
also asked subjects who had received pre- information avoidance with great negative
liminary feedback to provide their own pre- hedonic consequences.
dictions of where the ultimate ratings were While one might expect people who are
likely to come out, incentivizing subjects for more intelligent to be less likely to misin-
accuracy. Those who received initially favor- terpret information in a motivated fashion,
able information about their appearance or there is some research that suggests quite
intelligence tended to upwardly adjust their the opposite—i.e., that people marshal their
beliefs about these attributes. However, the intelligence in the service of believing what
reception of negative information about they want to believe. Kahan et al. (2012)
these attributes led to no comparable down- found that increased scientific expertise
ward adjustments of self-perceptions. does not lead to convergence on scientific
Sunstein et al. (2016) find similarly biased issues such as the reality of climate change.
updating in an online experiment. They Indeed, the opposite is the case: the beliefs
present participants with a statement on the of people with the highest levels of scientific
expected temperature increase in the United literacy exhibited the most extreme levels of
polarization. Another paper by Gino and asked them to provide comments on two
Ariely (2012) found that research subjects of them. Owners of both new and old cars
who scored higher on a measure of creativ- preferred evaluating advertisements of their
ity (but not intelligence) were better able to own make than of different makes. However,
come up with moral justifications for dishon- less consistent with an information-avoid-
est behavior and behaved more dishonestly ance story, participants who reported other
as a result. makes they had considered buying were not
less likely to look at and comment on the
2.4 Forgetting
advertisements of those makes.
Even when information has been received Even when people are unable, or for other
and attended to, a final information- reasons fail, to engage in motivated forget-
avoidance strategy available to individuals ting, they could still expend greater effort
is to forget the information. Although often in remembering information they wanted
thought of as a passive process, people may to remember, as compared with information
deliberately and selectively fail to rehearse that they would prefer to forget. Bernheim
negative information and therefore forget and Thomadsen (2005) suggest, for exam-
it over time (Bénabou and Tirole 2002). ple, that people with unbiased but imperfect
Motivated forgetting may help people deal memory might leave themselves reminders
with unpleasant life experiences (Anderson that bring to mind pleasant memories, but
and Huddleston 2012) or reduce cognitive choose not to leave themselves reminders
dissonance (Akerlof and Dickens 1982). about events that evoke unpleasant thoughts.
Shu and Gino (2012) conducted four lab- Bénabou and Tirole (2011) suggest that
oratory experiments in which participants people make investments based on their
completed an ability-based task and were paid beliefs to remind themselves of their social
according to their performance. The task was identities.
designed to give participants in some condi-
2.5 Self-handicapping
tions the opportunity to overreport their per-
formance (and thus increase their earnings Self-handicapping is a highly specialized
by cheating). When given an opportunity form of information avoidance that is diffi-
to cheat, those who cheated recalled fewer cult to classify into one of the other, broader,
previous items from a moral code—consis- categories. Self-handicapping refers to peo-
tent with motivated forgetting. This differ- ple’s tendency to choose tasks that are poorly
ence persisted even when participants were matched to their own abilities—either too
paid to accurately remember the items. easy or too difficult—or to take actions that
Ehrlich et al. (1957) conducted a lab undermine their performance, as a strategy
experiment to test whether people avoid for avoiding information about their abilities
information that challenges the wisdom of (see Bénabou and Tirole 2002, for a theo-
a previous choice. In a lab experiment, they retical perspective; also Alaoui 2012). In a
asked participants who had purchased cars to classic study of self-handicapping, Berglas
recall their exposure to car advertisements. and Jones (1978) randomly assigned partic-
Participants who had bought a new car (but ipants to complete a test consisting of either
not owners of older cars) were more likely soluble or insoluble questions. After com-
to recall reading advertisements promoting pleting the test, they were given positive
their car than ads promoting other mod- feedback about the number of questions
els. They then presented participants with they answered correctly (irrespective of
advertisements of eight makes of cars and how they really did). Then, prior to taking a
(Possible) Choice of (Possible)

preview of whether to decision/action
information reveal information and outcome
Utility from anticipation... Utility from outcome...
Time −1 0 1
Figure 1. Timeline of Decision to Acquire Information
s econd test that they were told would be sim- in his power (at t = 0) to obtain. At t = 0,
ilar to the first, participants were offered a the individual then decides whether to
choice between a performance-enhancing or reveal information. At time t = 1, in some,
performance-debilitating drug. In the condi- although not all, situations, the individual
tion with insoluble questions (in which sub- can then make a decision or take an action
jects believed that their strong performance that could be informed by the information.
was a fluke that would unlikely be repeated), An investor, for example, might listen to the
significantly more participants preferred the evening news at t = −1, learn that the stock
drug that interfered with performance. market had fallen, but retain some uncer-
Self-handicapping may also arise in tainty about the change in his own portfolio.
principal-agent settings, in which a princi- At time t = 0, he could then decide whether
pal is assessing an agent’s performance and to log in to his brokerage account and learn
pays her accordingly. While such incentives how his portfolio had actually performed. At
should motivate the worker to exert effort, time t = 1, in part based on whether or not
they may not be high enough to overcome a he logged in and if so what he discovered, he
potentially threatening self-signal the worker might then decide to engage in trades.4
receives if she does poorly. A noisier evalu- As figure 1 shows, there are two ways that
ation process that does not rely as much on the first information-acquisition decision
an individual’s ability, for example evaluating could affect the individual’s utility. First,
team performance, may reduce the preva- deciding whether to reveal the information
lence of self-handicapping and potentially can have a direct impact on utility from
increase average effort (Ishida 2012). anticipation or realization. The investor,
for example, will feel differently depending
on whether he logs on or not. In most sce-
3. Varieties of Information Avoidance
narios, he will have expectations (and some
In this section, we review theories that uncertainty) about what choosing to “look”
predict information avoidance, as well as will reveal, and hence how he will feel if he
empirical research providing evidence for or does reveal the information. Second, the
against the specific mechanisms implicated
by the theories. We organize this section of
4 The timeline presents a simplified setting that could
the review by the cause of (i.e., the reason
be enriched with various complexities. For example, the
for) information avoidance. situation might be repeated (as in the case of an investor
A generic timeline of decision making can who every day faces the option of logging in), and the two
help to fix the discussion. decisions might be linked in some fashion (for example, the
At time t = −1, an individual receives investor cannot trade without logging in). Similarly, some-
one may have multiple opportunities to acquire a fixed
a preview of some information that will be piece of information (e.g., the results of a genetic test).
information could inform a subsequent deci- a version, anxiety, regret aversion, optimism
sion. The investor’s decision to trade, and if maintenance, attention effects, and belief
so what trade to make, could be affected by investments. We examine each of these in
the choice of whether to reveal the infor- turn.
mation and, if information is revealed, its
3.1.1 Preferences for Resolution of
content. The quality of the action—whether
Compound Lotteries
it has beneficial or adverse consequences—
could, then, also affect his utility. Imagine an individual who is coming up
The examples of information avoidance for tenure, and a series of committees each
that we have already mentioned, as well as have to approve the case for tenure to be
many others that we discuss in this section, ultimately granted. Would the individual
can be classified into two broad catego- prefer to know of the intermediate commit-
ries corresponding to this division of utility tees’ decisions, or only to be appraised of
effects: those driven by hedonic consider- the final decision? Avoidance of information
ations (typically a desire to avoid bad news about the resolution of intermediate stages
because it will make one feel bad) and those of compound lotteries, as in this example,
driven by strategic considerations. Within can derive from risk preferences violat-
strategic considerations, we draw a further ing expected utility. In general, receiving
distinction between those involving only the information about a future lottery creates a
decision maker, and those involving other two-stage compound lottery. An individual
parties. As an example of the former, an indi- would choose not to find out the results of
vidual who was about to give a significant the first stage of the lottery if he preferred
public address might choose not to view a the reduced-compound lottery to the distri-
video of himself giving a previous talk, so as to bution of second-stage lotteries that results
prevent a presentation-debilitating plunge in from realization of the first stage (see, e.g.,
confidence or a surge of self-consciousness. Grant, Kajii, and Polak 1998; Hoy, Peter, and
As an example of the latter, during collec- Richter 2014; Snow 2010).5
tive bargaining, a labor union leader could Kreps and Porteus (1978) provide a math-
choose not to poll workers about their will- ematical framework in which a lottery is
ingness to accept management’s first offer, specified by its time of resolution along with
thus maintaining credible uncertainty about its possible outcomes and probabilities, so
workers’ actual reservation wages and posi- that a wager that pays off tomorrow based on
tioning himself to bargain more effectively a coin flip today is different from an identi-
on their behalf (Schelling 1956; 1960). cal wager based on a coin flip taking place
tomorrow (just in time to determine the pay-
3.1. Hedonically Driven Information
off). Their model provides a representation
Avoidance
The category of hedonic reasons for avoid- 5 We typically conceive of avoiding information about
ing information itself encompasses a diver- the first stage of a compound lottery as reducing that
sity of different motives, and a wide range compound lottery, implicitly adopting an ex ante view
of theories incorporate these motives and in which the compound lottery will unfold in the future.
Alternatively, one could conceive of proceeding through
make it possible to examine their implica- the first stage of a two-stage lottery as exchanging a com-
tions. Here, we examine seven distinct psy- pound lottery for a randomly selected simple lottery. Given
chological mechanisms that can produce that ex post perspective, avoiding information about the
first stage would mean continued exposure to both stages
information avoidance: preferences for reso- of uncertainty, rather than exposure to just the single
lution of compound lotteries, disappointment remaining stage.
of preferences for earlier or later resolu- this information in pieces over the course of
tion of lotteries, and, if later resolution is a few days.
preferred, then an individual would choose
3.1.2 Risk, Loss, and Disappointment
to avoid information about the outcome of
Aversion
the lottery until the payoff is to be received.
Grant, Kajii, and Polak (1998) characterize If we assume that utility depends directly
preferences to obtain or avoid information on beliefs, then information avoidance can
that result from a wide range of preferences be derived from simple assumptions. Utility
regarding compound lotteries. Their frame- function concavity alone, perhaps some-
work generalizes that of Kreps and Porteus what surprisingly, implies that information
(1978). In Grant, Kajii, and Polak’s frame- should generally be avoided in the absence
work, the preference to avoid information of adverse material consequences for doing
in order to preserve a lottery’s desired pro- so. The logic is simple: the utility downside
file of uncertainty is indistinguishable from of coming out below expectations is sim-
the preference to have that simple lottery ply greater than the upside, an effect also
rather than a materially equivalent two-stage predicted by models that incorporate loss
compound lottery. Avoiding information, aversion (Kőszegi 2010; Kőszegi and Rabin
in effect, reduces the compound lottery. A 2009). Acquiring information and resolving
preference for later resolution of uncertainty uncertainty would be akin to accepting the
(i.e., temporary information avoidance) can risk of possible disappointment or elation,
also be accommodated within a restricted rather than simply maintaining one’s prior
class of recursive models of ambiguity aver- expectations. More generally, Gul (1991)
sion (Strzalecki 2013). shows that risk aversion implies disappoint-
Following a similar line of reasoning, ment aversion, and recursive disappoint-
Palacios-Huerta (1999) and Dillenberger ment aversion in a dynamic setting (such
(2010) suggest that preferences about com- that individuals are disappointment averse
pound lotteries could stem from preferences every time some uncertainty is resolved)
to have them resolved gradually or all at necessarily leads to information avoidance
once (not just from preferences for earlier until all uncertainty can be resolved at once
versus later resolution). (See also Li 2015 (Andries and Haddad 2014; Artstein-Avidan
in the domain of ambiguity.) If individuals and Dillenberger 2015; Asch, Patton, and
prefer one-shot resolution of lotteries, then Hershey 1990; Dillenberger 2010).6 The
they may avoid partial information about strength of this account of information avoid-
their prospects and instead wait to find out ance—that the prediction is so robust—is
just the eventual outcome. Similarly, if indi- also its major weakness. It predicts informa-
viduals prefer gradual resolution, then they tion avoidance that is much more widespread
may temporarily avoid full information about than what is actually observed. In Eliaz and
the resolution of the compound lottery and Schotter (2010), for example, participants
instead gather this information in pieces. were faced with a decision under uncertainty
Indeed, Zimmermann (2014) finds that peo- about the state of the world. The same deci-
ple have heterogeneous preferences about sion was optimal in any state of the world
gradual or one-shot resolution of lotteries;
about half of subjects in his experiment
6 When information is instrumental, of course, risk
addressing the issue did, in fact, choose to
aversion can be eclipsed by the usefulness of the informa-
avoid receiving early information about the tion, e.g., to treat a medical condition if it is diagnosed (see,
outcome of a lottery, but instead chose to get for example, Fels 2015).
(i.e., the state of the world would not have medical patients do find it stressful to be
given them any useful information), yet they given more information about an unpleasant
were willing to pay to find out which state of impending procedure than they absolutely
the world they were in. Such findings may require (Miller and Mangan 1983). Some
suggest an inherent curiosity that often moti- cancer patients avoid information about the
vates information acquisition (see Golman state or prognosis of the disease so as to retain
and Loewenstein 2015; Loewenstein 1994), hope of recovery (Case et al. 2005; Leydon et
even when the expected (hedonic) impact of al. 2000; Nosarti et al. 2000). These motives
receiving the information is negative (Kruger can be represented in an expected-utility
and Evans 2009). model in which people derive utility from
Disappointment aversion leading to infor- anticipation of future events (as well as from
mation avoidance may emerge naturally in eventual outcomes) (Caplin and Leahy 2001;
some situations, without assuming utility Kőszegi 2003).
function concavity, if people use informa- In Caplin and Leahy’s (2001) model,
tion to make coarse categorical judgments utility is derived from psychological states,
(Kőszegi 2006). If the balance of evidence is which encompass not only present material
just above the threshold necessary for a pos- outcomes but also beliefs about the proba-
itive judgment, additional information could bility of material outcomes that may occur
reaffirm one’s belief (a neutral outcome) or in the future. When getting incomplete
could disappoint (a negative outcome), but information runs the risk of engendering
would be unlikely to elate (a positive out- anxiety about uncertainties that cannot be
come). In these cases, information would be resolved, a person might choose to avoid
avoided to guard against disappointment. this information. For example, obtaining a
In Bell (1985) and Loomes and Sugden medical test that would diagnose a disease
(1986), decision makers are more disap- exposes one to the prospect, if the test comes
pointed if they receive the low payoff that out positive, not of just knowing one has the
had a small probability than if the probabil- disease, but of anxiety and stress about the
ity of receiving a low payoff is large, hold- course of the disease. Avoiding the diagno-
ing constant the expected value. Outcomes sis cannot help one to avoid the disease (and
further below one’s expectation give rise to may even eliminate an opportunity to treat
more disappointment. Bell (1985) applies it), but it can help one avoid the stress and
this assumption to delegated informational anxiety. Note, however, (as will be discussed
preferences. His model implies that in break- momentarily) that people’s intuitions about
ing bad news to a recipient, the delegate the effect of information on anxiety are often
should shield the recipient and disclose the wrong—i.e., knowing the worst often engen-
information in small pieces to slowly adjust ders less anxiety than suspecting the worst.
the recipient’s expectations. Good news, in Kőszegi (2003) also proposes a model in
constrast, should be revealed immediately. which individuals derive utility from their
beliefs. Learning about a bad health state
3.1.3 Anxiety
may lead to gains in utility from any subse-
Feelings of anxiety could cause an individ- quent treatment, but causes disutility from
ual to avoid potentially useful information. the change in beliefs (i.e., the patient no lon-
Maslow (1963) suggested that people some- ger thinks of himself as healthy). The model
times avoid finding out about a risk—e.g., of predicts, contrary to standard economic the-
experiencing a disease or disaster—to reduce ory, that individuals would be less likely to
anxiety about such events. Indeed, many seek a diagnosis for more serious p otential
conditions, because obtaining the diag- show that the expected affective response to
nostic information in these cases is riskier. finding out unfavorable information (being
However, conditional on getting a diagno- devastated by the results and being unable to
sis, patients should want the most accurate cope with having a high risk of a fatal disease)
information available. Sensibly, the model significantly reduces the desire to obtain the
also predicts that individuals would not want information—except when people believe
to avoid information about conditions known there are actions they can take to reduce any
to be perfectly treatable.7 risk that is uncovered.
Eliaz and Spiegler (2006) point out that, The decision to get tested for a genetic
while incorporating anticipatory feelings into condition may depend on a number of fac-
expected utility allows for information avoid- tors. For example, a test that is perfectly
ance due to anxiety, it predicts that disliking diagnostic eliminates all worry in the event
information is independent of one’s prior of a negative result, but also leaves no room
beliefs about the issue. In reality, of course, for hope if the result is positive. On the
people tend to experience more anxiety other hand, tests that result in more accurate
when they suspect that very bad outcomes expectations but do not produce certainty
are likely, and information avoidance is more may leave hope in the event of a positive
prevalent when individuals have more neg- result, but do not eliminate worry if the result
ative prior beliefs. Women with breast can- is negative. Yaniv, Benador, and Sagi (2004)
cer symptoms that are getting worse, for vary the base rate of a condition, whether it
example, wait longer to visit a physician than is treatable, and the diagnostic accuracy of
those whose symptoms are steady or disap- a test. They find that neither the base rate
pearing (Caplan 1995), despite the patent nor the diagnosticity have an effect on the
self-destructiveness of doing so. Women
proportion of people who would like to avail
who have first-hand experience with a family themselves of the test, perhaps because the
member’s breast cancer are also more likely two effects involving worry and hope operate
to delay getting tested (Meechan, Collins, in opposing directions, or perhaps because
and Petrie 2002).8 Persoskie, Ferrer, and the preference is largely driven by power-
Klein (2014) find that people over fifty are ful heterogeneity in personal preferences.
more likely to avoid visiting a doctor when However, much as in Ferrer et al. (2015),
they are more worried about getting cancer, the potential for treatment does have a big
and that there is a positive interaction with effect; the proportion of people willing to
the subjective probability of getting cancer get tested increases from 33–47 percent with
in the future. When asked whether they no treatment to 80–93 percent when a treat-
“avoid visiting their doctor even when they ment is available.
suspect they should [visit],” 40.4 percent of Information avoidance due to anxiety can
those under fifty and 29.4 percent of those clearly be counterproductive if it delays ben-
over fifty said they did so. Ferrer et al. (2015) eficial action. Avoidance can also be counter-
productive if it leaves individuals in a state of
expectational limbo that impedes their adap-
7 Consistent with this prediction, Emanuel et al. (2015)
tation. A paper titled “Happily hopeless” by
find that, against the backdrop of 39 percent of survey
respondents who reported that they would “rather not Smith et al. (2009) illustrates such a poten-
know [their] chance of getting cancer,” those who believed tial consequence. The researchers elicited
that there isn’t much one can do to prevent cancer were the happiness and life satisfaction of people
more likely to not want to obtain the information.
8 See, also, the aforementioned studies by Eil and Rao who had colostomy or ileostomy operations
(2011) and Ganguly and Tasoff (forthcoming). one week, one month, and six months after
the operation. The main focus of the study unfavorable. In the same paper, the authors
was a comparison between those who had present empirical data on the frequency of
operations that were irreversible, and those logins to retirement investment accounts
who had procedures that were potentially when the broad market shows gains versus
reversible in the future. Apart from instances losses. When markets overall go up, then it
in which the procedure was in fact reversed, is likely that individual stocks held by inves-
uncertainty in this situation turned out to be tors have appreciated as well. In that case,
a bad thing; people with irreversible proce- logging on to check one’s savings is likely to
dures became progressively happier over present good news. On the other hand, if the
time until their happiness and life satisfaction market has declined, then logging in is more
was indistinguishable from that of the gen- likely to reveal bad news. Investor behavior
eral population. People who had potentially is consistent with a desire to avoid bad infor-
reversible operations, in contrast, although mation: investors are more likely to log in on
starting off marginally happier than those days when the market went up. In doing so,
with irreversible ones, became progressively they avoid information about current losses
more miserable, ultimately reporting happi- that may be relevant to their trading strategy.
ness far below that of the irreversible group. A follow-up paper analyzing a very large
These results didn’t examine information sample of individual 401k investors over a
avoidance per se, but do show how the kind two-year period (Sicherman et al. 2016) pro-
of uncertainty that information avoidance vides additional insights. One is that, remi-
can produce can impede adaptation and niscent of Kőszegi’s (2003) prediction that
induce long-term misery. patients facing worse outcomes (who, argu-
ably, need the information most) are least
3.1.4 Attention
likely to seek a diagnosis, investors with large
New information tends to be surprising, holdings are more likely to avoid information.
which means that it attracts attention and Another is that information avoidance seems
has a disproportionate impact on utility to be the product of stable character traits;
(Golman and Loewenstein 2015, forthcom- investors who were ostriches in 2007 (log-
ing; Karlsson, Loewenstein, and Seppi 2009; ging in selectively when the market went up)
Tasoff and Madarasz 2009). If information also tended to be ostriches in 2008, despite
is expected to be adverse, therefore, people dramatically different market conditions.
may have an incentive to avoid it. To find The authors’ analysis of multiple logins on
out something bad, at least in the short run, weekends, mentioned briefly above, is also
tends to be worse than only to suspect it. In quite revealing. Weekend logins after the
the long run, however, as we have already first provide no new information because the
discussed (see Smith et al. 2009), knowing market is closed and portfolio information is
can be better if it enables an individual to not updated on the web, so the main purpose
adapt to adverse circumstances. of such logins appears to be to “savor” the
Karlsson, Loewenstein, and Seppi (2009) information by paying attention to it, much
propose a model in which news about one’s as a child might shake a coin-packed piggy
investment portfolio, relative to lagged bank. Indeed, these noninformative logins
expectations, has a greater impact on util- display an even stronger ostrich pattern than
ity when one is attentive. Given the lag in do logins during the week.
updating one’s expectations, there is a motive Golman and Loewenstein (2015) pro-
to be inattentive when the investor suspects pose a model of information acquisition and
that the news from looking would likely be avoidance in which information is surprising
to the extent that it generates more dras- Stone (2013) model anticipated regret as a
tic revisions of beliefs. Surprise produces a cause of information avoidance in the con-
short-run boost in the attention devoted to text of choice under epistemic uncertainty.
these beliefs, and, in turn, these beliefs have Additionally, there is a large empirical litera-
a greater impact on one’s utility. Thus, infor- ture in psychology documenting consumers’
mation that is expected to produce negative avoidance of information about unchosen
beliefs can lower one’s utility even though a products, or of information about the risks
prior expectation of a negative outcome was of products they have chosen (e.g., Frey and
already a source of disutility. When there is Stahlberg 1986; Jonas et al. 2001). Ehrlich
a suspicion that the news revealed by infor- et al. (1957), for example, found that new
mation might be undesirable—for example, car owners pay more attention to advertise-
teaching ratings when a course seems to ments for the model they purchased than
have gone badly, or the value of one’s own for models they had considered but did not
portfolio when one learns from the news buy. Brock and Balloun (1967) observed
that the market has fallen—people will be that smokers attend more to pro-smoking
motivated to avoid getting this information messages and nonsmokers attended more to
to avoid the increase in attention, and hence anti-smoking messages.
weight in utility, that is likely to accompany
3.1.6 Optimism Maintenance
its receipt.
Empirical studies provide support for an Both theoretical models in economics
attentional mechanism underlying avoidance (Brunnermeier and Parker 2005) and a very
of information when bad news is suspected. large literature in psychology (summarized
Miller (1987) finds that many individuals in Sharot 2011) recognize, and provide the-
(identifiable through self-reports) distract oretical foundations for, the prevalence and
themselves from threatening information, benefits of optimism. In Brunnermeier and
for example about an impending electric Parker’s (2005) model, people can choose to
shock (as in Averill and Rosenn 1972) or hold optimistic beliefs, which are a source
about warning signs of poor academic per- of anticipatory utility and thus improve
formance. Falk and Zimmermann (2016) immediate well-being (potentially at the
find that people are more inclined to avoid risk of intensifying future disappointment).
information about whether they will be Information avoidance can occur in this
receiving electric shocks (clearly bad news if context because acquiring information can
they are) when they can distract themselves interfere with the ability to maintain unwar-
by playing a quiz game. ranted optimism. To the extent that people
are motivated to maintain optimistic expec-
3.1.5 Regret Aversion
tations, they will be similarly motivated to
Regret occurs when people compare avoid information that could force a down-
the outcome of a decision to what would ward revision of those expectations.
have happened if they had made a differ- Oster, Shoulson, and Dorsey (2013) pro-
ent choice (e.g., Loomes and Sugden 1982, vide evidence of information avoidance that
1987). Regret aversion, like disappointment is consistent with optimism maintenance by
aversion, can generate information avoid- people at risk of Huntington’s disease. Many
ance, but in the case of regret aversion, the people at known risk of carrying the disease
information avoided is about what the out- fail to get tested, despite the manifest useful-
come would have been had one taken an ness of such information for decisions such as
alternative course of action. Krahmer and whether to have children. Two key pieces of
evidence that are consistent with optimism Bénabou (2013) also presents a model on
maintenance, as opposed to other explana- the phenomenon of “groupthink” that could
tions for information avoidance, are that: be interpreted in terms of optimism mainte-
(1) people who fail to get tested (but know nance. In Bénabou’s model, groups of peo-
they are at risk) make life decisions that are ple decide whether to invest in a project,
indistinguishable from those who get tested and, before the project does or does not pay
and discover they do not carry the disease, off, each group member experiences antic-
but very different from those who get tested ipatory utility based on their own degree
and discover they do carry the disease; and of optimism and the beliefs of other group
(2) people at risk for Huntington’s appear to members. Under certain adverse equilib-
be generally optimistic about the (un)likeli- ria, group members engage in collective
hood that they have the disease. denial, ignoring “red flags” (via inattention,
Also consistent with optimism mainte- misinterpretation, and forgetting) so as to
nance are findings from the experiment, maintain the collective illusion that an unfa-
discussed earlier, by Eil and Rao (2011). vorable project will, in fact, succeed.
Presenting people with a hint about the value
3.1.7 Dissonance Avoidance
of attributes they care about (their own intel-
ligence or attractiveness), the researchers Psychologists going back to Festinger
find that people who receive hints that sug- (1957) have recognized that people dislike
gest that more detailed information might being exposed to information that conflicts
fall below their expectations (which tend to with existing beliefs (see Abelson et al. 1968
be generally optimistic), avoid obtaining the for a wide-ranging volume taking stock of
information, even when it is costly to do so. research on Festinger’s t heory approximately
People appear to exhibit optimism main- a decade later). Research on managerial
tenance in decisions about the order in decisions finds that managers avoid expos-
which stages of a compound lottery will be ing themselves to arguments that would
resolved. Consider a compound lottery con- conflict with their preliminary decisions
sisting of two independent events that must (
Schulz-Hardt et al. 2000). Managers in
both obtain in order to win a prize. Suppose German firms who were recruited for a lab
one event has a high success probability, experiment were given a hypothetical case
while the other event has a low success prob- about a firm considering relocation. They
ability. Budescu and Fischer (2001) find that were presented with an equal number of
participants prefer to have a high probability arguments in favor and against the move and
of a gain followed by a low probability of
gain and, in effect, are choosing to have
an a version to celebrating events that are at all uncertain.
much of the uncertainty resolution delayed When Loewenstein (an author on this paper) came up
to the future. In the loss domain, the pattern for tenure many years ago, he was first informed that his
reverses and people prefer a low p robability department had voted in favor, at which point the proba-
bility that he would get tenure was high. He did not want
of loss followed by a high probability of loss. to celebrate the news, however, even in his own mind,
Both preferences allow a person to keep because it wasn’t certain. Subsequent stages led to similar
hope alive as long as possible.9 upward revisions of the probability of tenure, but all, for
the same reason, without celebration. Finally, there was
only one step left; the provost had to sign to make it offi-
cial. Though there were no episodes on record in which the
9 In the real world, the preference for resolution of provost had not signed, Loewenstein was still reluctant to
uncertainty is likely to be influenced by other psycholog- celebrate, but when the provost did finally sign, it felt silly
ical factors such as superstitious thinking (e.g., that “look- to celebrate: there had only been a microscopic increase in
ing” in the middle of a lottery will “jinx” the outcome) and the probability, from 99.99 to 100 percent.
asked to make a decision on their own. They of what to look at based on a small preview
were then split into groups of five, based on snip that revealed which art piece had been
their individual decision: groups consisted of commented on and whether the comment
people who either all made the same deci- was favorable. Participants were more likely
sion, or in which a minority of one or two to choose to look in detail at comments that
participants chose differently. Participants were favorable toward the artwork that they
were then given the option to receive addi- thought other students favored. Notably, this
tional arguments, up to five in favor and five preference did not differ by how much infor-
against, with the condition that everyone mation they had in their own decision: those
in the group had to read the chosen argu- who simply picked from two art pieces with-
ments. Although the information was free, out having any idea what they were engaged
the expected effort to read the information in confirmatory search just as much as those
imposed a cost that gave participants an who had full information.
excuse for being selective. All groups chose Chater and Loewenstein (2016) propose
to obtain more information that conformed that an inherent drive for sense making (see
to the majority’s view than that conflicted also Dervin 1998) may lead people to avoid
with it. Moreover, this difference was great- information that might not fit with their cur-
est in groups in which everyone held the rent understanding of the world. The almost
same view, less pronounced in groups with willful failure of mid-century western com-
a one-person minority, and was even lower munists to accept evidence of Stalin’s atroc-
in groups in which two people held the ities illustrates the concept, and is only one
minority view. These findings suggest a ten- of myriad examples from history. Discussing
dency to avoid information that might con- this example and many others, in a charac-
flict with prior decisions, even though such teristically brilliant and insightful discussion
information might well be viewed as espe- of “Systems of Belief,”10 Jonathan Glover
cially valuable since it could lead to a war- critiques René Descartes’s view that sys-
ranted reversal of the decision. Information tems of beliefs can be built up from scratch
confirming a decision, in contrast, is unlikely from objective evidence and axioms much,
to influence the decision, especially in a as Glover depicts it, as you might build
homogeneous group. a house up from its foundation. Glover,
Scherer, Windschitl, and Smith (2013), instead, argues in favor of “an alternative
in a study of biased information seeking, model, put forward by the Austrian philos-
observe confirmatory information acquisi- opher Otto Neurath in the 1930s,” accord-
tion even when the initial decision was made ing to which constructing a belief system
arbitrarily, which they argue is driven by a from conflicting pieces of evidence “is not
desire to be right in a prediction or belief. like rebuilding a house; it’s like rebuilding
In one of their studies, participants had to a boat which happened to be afloat at sea.
decide which of two pieces of art would be Maybe the whole thing needs rebuilding,
preferred by other college students. They but at any point inevitably you have to keep
made the decision based on pictures that enough afloat for you to do the rebuilding.”
were either unobscured, partially obscured, One could extend the analogy by positing
or completely obscured (in the last case, they
picked between labels A and B with no further
10 Philosophy bites: “Podcasts of top philosophers
information). They were then given a choice
interviewed on bite-sized topics.” Accessible at: http://
to look at comments by other students on philosophybites.com/2011/10/jonathan-glover-on-systems-
the two art pieces. They made their choices of-belief.html.
that keeping the boat afloat requires avoid- 3.2 Strategically Driven Information
ing the kinds of structural damage or serious Avoidance
punctures that could be suffered from the
reception of information with obvious disso- People may avoid information as a kind of
nant implications. commitment device because they anticipate
that it will influence either their own future
3.1.8 Belief Investments
behavior or that of others. We refer to this as
Many of the varieties of hedonically driven strategically driven information avoidance.
information avoidance involve a motive to Surprisingly, as we discuss in the ensuing
not form, or focus on, an unpleasant belief. subsection, some of these situations arise
While some beliefs are intrinsically unpleas- with single individuals—i.e., in the absence
ant—e.g., believing that one has contracted of interpersonal interactions.
a serious disease—others are unpleasant
3.2.1 Intrapersonal Strategic Avoidance
because an individual has committed him-
self to an opposing belief. In many situations Intrapersonal strategic information avoid-
people invest time, money or effort—often ance may arise as a person tries to bind his
very large amounts—based on beliefs. For own hands while facing an inner conflict.
example, a devout Catholic will spend large Someone who has the choice between an
amounts of time going to church, money appealing piece of cake today and better
supporting the church, and effort complying health in the future may not want to learn
with its doctrines. In such situations, infor- just how unhealthy the cake is, so that she
mation that could potentially challenge the can justify (and enjoy) the cake right now.
beliefs underlying the investments threatens We might think of such an internal con-
the investments themselves. If the down- flict as a struggle between the present and
side risk that new information could under- future selves, modeled as distinct players in
mine beliefs is greater than the up-side a strategic game. The inner conflict could
potential for beliefs to be bolstered (which stem from a variety of psychological mecha-
is likely in a wide range of situations), people nisms, such as incoherent preferences, time
may choose to avoid information. Bénabou inconsistency, projection bias, or self-serving
and Tirole (2011) propose that people use notions of fairness.
these investments to signal their own core
values and social identities to themselves. Avoiding dynamic inconsistency: Wakker
Avoiding information that would threaten a (1988) illustrates how decision makers whose
person’s conception of his own identity helps preferences do not conform to expected util-
the person continue to enjoy his sense of self ity could recognize that resolution of the
and preserve his motivation to live up to his first stage of a compound lottery would shift
ideals. Golman et al. (2016) follow Bénabou their preference for accepting or rejecting
and Tirole in adopting the basic premise another lottery in the second stage. If they
that people become attached to their belief are sophisticated about this preference
investments. They propose, however, that reversal, Wakker shows, it might lead them
people avoid information that would conflict to avoid information about the outcome of
with investments they have made based on the first lottery to commit themselves to their
their beliefs, because they experience a kind plan for behavior with respect to the second
of sunk-cost bias and do not want to consider lottery. Suppose a gambler wants to make a
that the beliefs that informed these invest- parlay bet on his favorite baseball team in
ments might be wrong. a doubleheader (a bet that pays only if his
team wins both games), but he suspects that, avoiding the information so that they could
conditional on winning a bet on his team in resist the temptation of betting against the
the first game, he may prefer not to bet his student, which many of them clearly found
winnings on the second game. To ensure that distasteful.
he goes through with the parlay bet, he may
choose to not watch the first game until bets Motivation maintenance: In some cases,
can no longer be placed on the second game. people avoid information out of fear that
In this situation, information avoidance is a some types of information they might
kind of internal commitment device to deal obtain would be demotivating. Goulas and
with risk preferences that violate dynamic Megalokonomou (2015) show that such
consistency. a fear may be well-founded. They ana-
lyze a natural experiment in which a policy
Resisting temptation: Information avoid- change in Greece led students to no longer
ance can be used as a commitment device for observe their own and their peers’ scores on
other self-control problems as well. If peo- a national exam prior to taking the exam a
ple are sophisticated about having time-in- required second time. Students in cohorts
consistent preferences (Carrillo and Mariotti that did observe their peers’ scores and who
2000), for example, they can use information learned they were low achieving compared
avoidance to prevent themselves from recon- to their peers performed 0.3 standard devia-
sidering decisions in the future, when they tions worse the next time they took the exam,
might otherwise succumb to temptation. compared to cohorts that did not learn their
Nonsmokers, for instance, who are aware relative position. Feedback was helpful for
that nicotine is addictive but who overes- those who found themselves among the high
timate the health risk it poses, might avoid achievers; they increased their performance
information that makes them reconsider by 0.2 standard deviations.
these risks—for fear that less frightening Rogers and Feller (2016) analyze data
information about risks might tempt them to from a large online class in which students
start smoking (Carrillo and Mariotti 2000). rated three randomly assigned essays writ-
In an incentivized study, Woolley and Risen ten by their peers. Grades for the class were
(2015) asked visitors at a museum if they not relative, so others’ performance had
wanted to bet on a student’s performance. no impact on a student’s likelihood of pass-
The bet would pay between $0 and $25 if ing the class. Students who rated essays of
they won, and there was no cost to accept- average quality had a 68 percent comple-
ing the bet. In one condition, they would tion rate for the course. Students who rated
win the bet if the student finished the class higher-quality essays (with an average score
with an A or A+ grade, whereas in the other 1.6 standard deviations above the mean)
condition, they would win if the student got had a completion rate of only 45 percent.
a B− or below. Before deciding whether to To benchmark this effect, the authors note
accept the bet, participants were offered the that being assigned to rate a higher-quality
option to find out exactly how much the bet essay hurt a student’s chances of passing the
would pay out if they won. Participants were course more than (hypothetically) lowering
significantly more likely to avoid the payout one’s own grade by replacing the maximum
information when offered the bet against the possible score for the essay with the mean
student than when offered the bet that was score in the class.
contingent on the student’s success (57.8 to Huck, Szech, and Wenner (2015) test
42.9 percent). Participants appeared to be the effect of information on motivation in
a laboratory study. Participants completed e.g., Ariely et al. 2009), one could imagine
tedious real-effort tasks and were told that situations in which people might avoid infor-
half of all participants would earn EUR 1 mation to prevent themselves from becom-
for each task they completed and that the ing excessively motivated. A novice teacher,
other half would earn EUR 0.10 per task. for example, might avoid looking at teaching
In the full-information condition, par- ratings not (only) out of fear that the infor-
ticipants learned what their pay rate was mation would make them miserable, but also
prior to working on the real-effort task; in out of fear that the information might lead
a no-information condition, they did not to counterproductively high levels of anxiety
obtain this information until after they about teaching.
had worked on the task. In a third condi-
tion, participants could choose whether Avoiding projection biases: In some situ-
to learn their pay rate prior to starting the ations, such as teaching or strategic interac-
task. When given a choice, 31.6 percent of tions, it is useful for an individual to make a
participants decide not to obtain informa- guess about what other individuals believe.
tion about their piece rate. In a follow-up For example, the seller of a defective prod-
questionnaire, many of those who chose not uct, in setting a fixed price, would like to
to get the information reported that they know whether potential buyers will notice
avoided it either so as to not be demotivated the product’s defects. A teacher, likewise,
by the low wage or so as to not be pressured can convey information more effectively
by the high rate. Comparing performance armed with knowledge of what students
across the groups, those who chose not to already know. In such situations, research
learn their own wage rate performed no dif- has documented a phenomenon in which
ferently from those who found out they had better-informed individuals “project” their
a high wage, and significantly outperformed superior knowledge on less well-informed
those who had a low wage. There was no individuals. Because this “curse of knowl-
difference between the group with a choice edge” (Camerer, Loewenstein, and Weber
and the group that did not have an option to 1989) distorts individual judgments and is
learn their wage. Given a choice of whether generally detrimental, people should be
to learn their wage, many appear to have (and sometimes are) motivated to avoid
strategically chosen to not obtain informa- information to avoid being “cursed.” There
tion so as to maintain their own motivation, is, however, evidence that people do not
which, the experimental results suggest, was appreciate how information can distort these
an effective strategy. types of judgments, and hence obtain and
Bénabou and Tirole (2002) propose a even pay for, information that distorts their
model of information avoidance as a strategy judgments and causes them to lose money
for dealing with problems caused by present (Loewenstein, Moore, and Weber 2006).
bias. Present bias leads people to put too Information avoidance for the purpose of
little effort into tasks with high but delayed lessening projection bias is, therefore, a
payoffs and, according to their model, peo- theoretical possibility, but there is little if any
ple avoid information about their own true evidence that people actually avoid informa-
ability so as to maintain overconfidence for tion for this reason. Information avoidance
the purpose of offsetting this motivational can be an effective antidote to any type of
deficiency. By similar logic, to the extent projective bias involving information. For
that extreme levels of motivation can in example, researchers have found that eval-
some cases undermine performance (see, uations of other people’s decisions are overly
harsh as a result of hindsight bias—the the model became the benchmark for the
mistaken view that the outcome of a deci- program’s success.
sion should have been foreseeable (Baron According to the concept of “moral wig-
and Hershey 1988). Evaluating the decision gle room,” people are not necessarily inher-
prior to learning about whether it produced ently altruistic, but like to appear so, not
a favorable or unfavorable outcome would only to others but also to themselves (Dana,
eliminate this problem. Weber, and Kuang 2007; see, also, Broberg,
Ellingsen, and Johannesson 2007; Dana,
Abdicating responsibility: In some sit- Cain, and Dawes 2006; Grossman 2014;
uations, people may not want to obtain Lazear, Malmendier, and Weber 2012).
information for purposes of “plausible deni- Even if they would not behave in a blatantly
ability”—i.e., because they fear that obtain- selfish or immoral fashion, when they can
ing it will make them feel more culpable for interpret a situation in a way that allows
negligent or ethically questionable behavior, them to behave selfishly without compro-
leading to self-condemnation. Although we mising their identity as moral people, they
have classified this in the intrapersonal, stra- will do so and behave selfishly (Shalvi et al.
tegic category, plausible deniability can also 2015). Thus, when people have the types of
help to shield one from condemnation, and complex motives inherent in the concept of
possibly apprehension and punishment, by moral wiggle room, they may similarly be
others, so this motivation could equally well motivated to avoid information.
fit into the following subsection on interper- Consider, for example, the payoff matrix
sonal, strategic considerations. An individual in table 1 (based on Dana, Weber, and
taxpayer who honestly believes he is entitled Kuang 2007), in which a “dictator” chooses
to a tax deduction might choose not to con- between A and B. In one of the experiments
sult with an accountant who could only talk reported in their paper, subjects played this
him out of taking the deduction. Indeed, game, but the recipient’s payoffs were hid-
often the law encourages information avoid- den. However, dictators had the option to
ance by holding a person criminally respon- costlessly and privately reveal the recipient’s
sible for negligence only if the person knows payoff. This would be valuable information
or could reasonably be expected to know of a for someone with other-regarding prefer-
danger to human life or safety. ences or with preferences for efficiency,
Rayner (2012) uses the example of a pol- but could make the decision more difficult
lution mitigation program in the Chesapeake for a selfish dictator who wanted to maxi-
Bay to illustrate how organizations may skirt mize his own payoff while minimizing guilt.
accountability by avoiding information about Consistent with a desire to avoid the infor-
their efficacy. As part of the program, a com- mation, only 56 percent of dictators chose to
plex computational model of the environ- reveal the recipient’s payoff, and in the con-
ment was constructed to assess the impact dition in which information-revelation was
of environmental-protection interventions. optional, more dictators chose the “selfish”
Although the model showed continued payoff than in the corresponding game in
improvement in water quality, actual water which the recipient’s payoffs were automat-
samples showed no discernable change. ically revealed. Follow-up research (Van der
Little attention and funding was, however, Weele 2014) shows that decision makers are
directed at field measurements that could sensitive to the cost of being prosocial, and
confirm the program’s ineffectiveness; are more likely to reveal the state when the
instead “virtual” improvements coming from altruistic action is cheap. However, making
TABLE 1
The Dictator Knows His Own Payoffs and Can Choose To
Costlessly Observe the State and thus Reveal His Partner’s State-Dependent Payoffs
State
I II
A 6,1 6,5
Choice
B 5,5 5,1
the selfish action costlier to the recipient c ontributions do not have the impact they
does not affect the decision to reveal. imagine (Niehaus 2013).
Similarly, people with social preferences Grossman and van der Weele (forthcom-
may not want to hear that a possible action ing) propose a theoretical model in which
that would entail a personal cost would have decision makers avoid information so as to
greater benefits to others. With this infor- weaken the signal that their choice in the
mation, they might feel compelled to take dictator game will send about their own pro-
the action, i.e., to give others these bene- sociality. Avoiding information itself sends
fits, leaving them worse-off overall than a signal, but doing so is less informative
had they not known (Andreoni, Rao, and than choosing the selfish action after hav-
Trachtman 2011; Nyborg 2011). Naturally, ing obtained the information. They make,
consumers buying products have oppor- and find evidence for, five predictions about
tunities to learn about environmental and moral wiggle room: (1) there is a smaller
labor practices of producing firms. There incentive to act pro-socially when there is
may, however, be little upside to learning initial uncertainty about others’ outcomes;
that best practices have been adhered to, (2) those who act selfishly are more likely to
whereas finding out that the firm engages in not obtain the information; (3) selfish actions
bad practices would impose disutility from are judged more harshly if actors behaved
the renunciation, or purchase and use, of selfishly knowing the other’s payoffs than if
the product. Paharia, Vohs, and Deshpande they chose to remain ignorant; (4) more peo-
(2013) find, in a hypothetical scenario ple will choose to obtain information after
study, that consumers use moral disengage- having made the decision as dictator than
ment to justify purchasing items manufac- before (as obtaining the information after
tured using sweatshop labor, suggesting the decision does not send a bad signal); and
that this information is indeed unpleasant. (5) decision makers who choose to avoid the
Consequently, people may rationally avoid information and are selfish are willing to pay
finding out this information in the first place to remain ignorant.
and may even be willing to pay to avoid it In a lab experiment on dishonesty,
(Ehrich and Irwin 2005; Grossman and van Pittarello et al. (2016) showed participants
der Weele forthcoming). The same princi- two cards from a deck and asked them to
ple appears in charitable giving: donors may indicate whether one of the cards was a
want to avoid information about the effec- Joker. Each time they reported a Joker, they
tiveness of a charity they are emotionally lost EUR 1 from their endowment of EUR
drawn to, so as to avoid learning that their 60. In this setup, there is a strong incentive
to cheat and report that no Joker is present more likely to order the cake. Regardless of
even when there is. They measured how long whether they had expressed a desire to see
participants looked at one card or the other the information or not, showing a higher
and found that, in trials in which participants number of calories discouraged participants
cheated, they spent less time looking at the from ordering the cake; the information was
Joker card and more time looking at the taken into account, whether it was wanted
other card. When they are incentivized to be or unwanted. In a second study, the authors
untruthful, the authors surmise, dishonest show that people who believe that a menu
participants engage in a self-deception strat- with caloric information is likely to discour-
egy and shift attention away from the object age them from ordering dessert are more
reminding them of their dishonesty. likely to request a menu without caloric
Just as people avoid information to behave information. This suggests that information
in a selfish fashion with respect to others, is avoided precisely because it could lead
people about to succumb to a temptation someone to change his mind.
sometimes avoid information that could
make them feel guilty about succumbing, Saving it for later: Strategic information
or even dissuade them from doing so. In a avoidance is a device to influence the action
state of craving, for example, an individual taken at time t = 1. An interesting case arises
may choose not to learn about the long-term when the decision at this later time is once
health consequences of indulging in the again whether to acquire or avoid the infor-
behavior motivated by the craving. For exam- mation. If information will be even more
ple, cigarette smokers who are not motivated pleasurable to discover at a later time, an
to quit might prefer to not know about the individual may strategically save it (e.g., ask
health risks associated with smoking. By the others not to spoil the ending of a good book
same token, a hungry diner who wants to or show). Such a motive requires contextual
enjoy a high-calorie meal might well choose reasons why the information will be more
not to obtain nutritional information before valuable in the future than in the present.
ordering (Loewenstein and O’Donoghue One possible reason could be that delayed
2006). In such a situation, inattention to the information will come along with other
information can be perfectly rational, and pieces of information, and that these pieces
imposing the information on helpless diners of information are complements. Kocher,
via prominent calorie displays may detract Krawczyk, and van Winden (2014) show that
from their welfare. lottery players enjoy spreading out draw-
Woolley and Risen (2015) offered partici- ings (not immediately finding out whether
pants a hypothetical choice of whether to eat they’ve won each drawing) so they can savor
a slice of cake and asked if they wanted to thoughts of possibly winning (though, unsur-
know how many calories it contained before prisingly, people do not enjoy spreading
deciding whether to order it. They found that out information about whether they will be
62.7 percent of p articipants stated that they forced to receive electric shocks; Falk and
would not want to obtain caloric information. Zimmermann 2016). Another reason may
All participants, irrespective of whether they be that not knowing creates suspense, thus
wanted to see the information, were then enhancing the experience of discovering
shown either a low, medium, or high num- information (Ely, Frankel, and Kamenica
ber of calories prior to making their decision 2015). Someone watching a recorded soccer
to order. Those who had initially indicated game, for example, may not want to know
they did not want to know the calories were when goals were scored, as that would make
the ultimate outcome of each play predict- consequence of wanting others to remain
able. In these cases, avoiding information at uninformed, so we do not dwell on these
time 0 increases the utility of acquiring that examples (however, see Brocas and Carrillo
information at time t. 2007, for a discussion of this situation). In
If an uncertain outcome is known to be other cases, an individual may make it known
positive (e.g., a participant has won one of two that he will not obtain private information as
desirable prizes, but does not yet know which a kind of commitment device, so that other
one), delaying the resolution may increase players cannot rely on his knowledge of this
the enjoyment derived from the anticipa- information in their own actions. This com-
tion (Loewenstein 1987). Wilson and Gilbert mitment to an ex post suboptimal course of
(2003) argue that an inability to make sense action can change the optimal strategy for
of a positive experience enhances its plea- others and lead to a better outcome for the
sure. Lingering uncertainty inhibits hedonic individual, much like the decision to remove
adaptation, preventing a person from tak- the steering wheel in a game of chicken
ing the good outcome for granted. Golman serves as a commitment device and all but
and Loewenstein (2015, forthcoming) guarantees victory (Schelling 1960; Ponssard
describe the belief-resolution effect, whereby 1976).
attention fades after uncertainty has been There is a large literature on credible com-
resolved. Avoiding information, and thus mitment in game theory that goes beyond
maintaining uncertainty, would allow a per- the scope of our review. We focus on cases
son to continue to enjoy thinking about the of commitment that directly involve infor-
set of possible good outcomes. Such behav- mation avoidance. Information avoidance is
ior may be rare, however, because a person a particular kind of commitment device that
must be sophisticated and quite patient to be does not make all commitments possible. For
willing to forego the immediate excitement example, the agent in a trust game cannot use
of realizing a good outcome so as to savor the information avoidance to commit to return-
lasting anticipation. One example that may ing the principal’s investment. Information
be familiar to some readers is that of expect- avoidance allows a player only to commit to
ant parents avoiding information about their what would be optimal given prior beliefs.
baby’s sex until the birth. For the duration of Game theorists familiar with strategic
the pregnancy, they may anticipate the plea- moves—maneuvers intended to influence
sures of having a girl as well as the pleasures other players—recognize many situations
of having a boy, and they may particularly in which players want to be seen avoiding
enjoy resolving this uncertainty when it is private information because they can bene-
most salient, at the event of childbirth. fit from others knowing that they are unin-
formed. Schelling (1960) points out that
3.2.2 Interpersonal Strategic Avoidance
strategic information avoidance can be use-
In interpersonal interactions, information ful for resisting another player’s fi rst-mover
may be avoided in public as a strategy for advantage in a variety of games. For exam-
manipulating another player’s actions. This ple, in a Stackelberg duopoly game, the
can take a variety of forms. In some cases, leader can take market share from the fol-
obtaining information may make it publicly lower by announcing (and committing to)
available to others, which could be problem- its production level first. Similarly, in some
atic (e.g., a charity allowing outside research- games with multiple equilibria (e.g., battle of
ers to assess its effectiveness). In these the sexes) players can signal an intention to
cases, information avoidance is an incidental play their preferred equilibrium strategy by
moving first and “burning money” before the to reveal was not disclosed, and hence there
game begins (van Damme 1989; Ben-Porath was no strategic motivation to avoid the
and Dekel 1992). In both of these situations, information, very few proposers decided not
if a second mover commits to not observe to reveal the outcome of the chance event.
the first player’s strategic move, he wipes out Poulsen and Roos (2010) had pairs of
its intended effect, which is, simply, to influ- subjects play a Nash bargaining game in
ence him. which two players allocate a resource pie by
Strategic avoidance of information can demanding a share of it for themselves. If the
also strengthen one’s bargaining position sum of their demands does not exceed 100
(Schelling 1956; 1960). The intuition for percent of the pie, they each get what they
real-world bargaining situations is that infor- demand. But if the total demands exceed
mation avoidance supports a kind of brink- the size of the pie, they each get nothing.
manship, which forces the other party to Poulsen and Roos allowed the second mover,
make (or accept) a higher (lower) offer. For before the bargaining begins, to make a
example, labor union leaders aware (but, public decision whether or not to observe
of course, not sure) that their rank-and-file the first mover’s demand. If he chooses to
members might be growing weary of a strike observe the demand, the first mover has a
could publicly avoid meeting with them to strategic advantage, as the game effectively
credibly convey to management that the becomes an ultimatum game. However, if
union has no intention to end the strike and the second mover chooses not to observe
that management will have to make a better the first demand, the game is effectively a
offer to resolve the dispute. Lab studies have simultaneous-move Nash bargaining game
documented that people engaged in ultima- in which the focal equilibrium calls for an
tum bargaining avoid information to induce even split.11 Thus, committing not to obtain
the other party to accept a proposal that they information about the first mover’s demand
would not have accepted otherwise (Conrads significantly increases the second mov-
and Irlenbusch 2013), and that people er’s expected earnings. After some practice
engaged in sequential Nash bargaining avoid rounds, over 80 percent of second movers
information to induce the other party to pro- figured this out and chose not to obtain the
pose a more generous offer than they other- information. However, when the decision
wise would have (Poulsen and Roos 2010). to reveal the first demand was private and
Conrads and Irlenbusch (2013) set up a thus not an effective strategic move, over
one-shot bargaining game in which a pro- 80 percent of second movers chose to reveal
poser can make one of two offers. The pro- it, showing that information avoidance in
poser always prefers one of them. For the this context is strategic, rather than driven by
respondent, which offer is more favorable hedonic motivations.
depends on a chance event. The proposer Strategic bargainers could also use a pub-
has the option to reveal the outcome of the lic commitment to avoid information to
chance event prior to choosing one of the induce another party to make an investment
offers and thus can uncover the respondent’s (Tirole 1986; Rogerson 1992; Gul 2001; Lau
payoffs. When the decision to reveal was 2008; Hermalin and Katz 2009). These mod-
disclosed to the respondent, many proposers els analyze relationship-specific investments
chose not to obtain information about the
chance event. Effectively, they created plau- 11 Asymmetric equilibria with uneven splits exist, too,
sible deniability, and respondents were less but there would be no reason to expect one player to have
likely to reject the offer. When the decision a systematic advantage.
in the context of an imbalance in bargaining to the table as long as they do not pry into
power that may create a holdup problem. their operations.
For example, a firm may want to work with Sloof, Oosterbeek, and Sonnemans (2007)
an advertising agency to lay out their vision ran a laboratory experiment in which sub-
for a possible ad campaign before agreeing jects assigned to the role of buyers chose
to terms and payments. However, once the whether to invest (i.e., pay a cost) to raise
firm has put time and money into establish- their valuation of a seller’s good before
ing this relationship, the advertising agency entering a transaction stage in which the
might come back and ask for ancillary fees, seller sets the price. (The seller’s good was
taking advantage of the firm’s sunk costs to simply a laboratory currency that only the
expropriate more of the surplus. If the adver- buyer could convert into real money, and the
tising agency cannot assuage fears that they investment increased this monetary value.)
will engage in such behavior, the firm might The investment was at risk for a holdup
be reluctant to work with the agency in the because the seller held all the bargaining
first place. One tactic that may be available power in the transaction stage. If the buyer
to the agency is to make clear to the firm invested, the seller had an incentive to set a
that they do not inquire about the firm’s sales higher price, because the rules did not allow
projections until after they reach a deal. If buyers to reject profitable transactions. The
the agency does not know the value of the experiment manipulated whether the deci-
ad campaign for the firm, it cannot demand sion to invest was public or private. When
the entire surplus without jeopardizing the the decision to invest was public, buyers less
deal. Tirole (1986) and Rogerson (1992) frequently chose to invest and, thus, sellers
suggest that allowing a vulnerable party to earned less. While sellers were not given the
keep its information private may help them opportunity to avoid information about the
work around the holdup problem. Building investments in this study, the results suggest
on this insight, Gul (2001) develops a theo- that they would have been wise to avoid this
retical model in which the imbalance in bar- information if they could have made a public
gaining power stems from one party (e.g., commitment to do so because buyers, in this
the advertising agency) having the exclusive situation, would have been more willing to
opportunity to make offers. Gul shows that invest.
by committing to avoid information about A commitment to avoid information can
the other party’s investment (and by allow- also be used to enhance or counteract mar-
ing oneself the freedom to make additional ket power. Palfrey (1982) proposes a Cournot
offers if the initial one(s) is/are not satisfac- duopoly game in which two firms have dif-
tory), the party with the bargaining power ferent levels of risk tolerance and have an
can effectively promise not to exploit it too option to obtain information about uncertain
onerously and can thus provide incentives consumer demand. If a firm obtains infor-
for optimal investment in the relationship.12 mation about consumer demand, it will set
In the context of our example above, the its production in response to that informa-
advertising agency can indeed bring the firm tion. Importantly, it would produce less if
demand is low. However, if it chooses not to
12 Lau (2008) and Hermalin and Katz (2009) show reveal consumer demand, there is a risk that
that when bargaining is constrained to a single take-it-or- it overproduces when demand is low, which
leave-it offer, the party with bargaining power can provide would cause the price to crash and impose
better incentives for the other party to invest by avoiding
some (i.e., partial) information about (the value of) the losses on both firms. The firm that is more
investment. risk tolerant hence has an incentive not to
obtain this information. By not responding to parties are sophisticated, they could publicly
actual demand, they increase the likelihood avoid acquiring asymmetric information that
of losses imposed on the other firm, which would break the other party’s trust that they
makes that firm more cautious. The more could get a fair deal. For example, a casino
risk-averse firm must scale back production, could promise to shuffle its cards in real time
and the more risk-tolerant firm then bene- rather than to use a pre-shuffled deck so that
fits from higher prices. Strategic informa- players would know that the casino could not
tion avoidance (by the less risk-averse firm) know the deal ahead of time (and thus could
is used in this case as a competitive weapon, not remove unfavorable decks), in order
to create risk that the other firm cannot to convince patrons that they can get a fair
tolerate. game at that casino.
Roesler (2015) shows that a consumer Two or more parties who engage in risk
can counter a monopolist’s pricing power by sharing might publicly avoid information
strategically avoiding information about its about the hazard to preserve their risk-shar-
product. In a world of perfect information, ing arrangement. There are several situ-
a monopolist expropriates consumer surplus ations in which weakly risk-averse agents
by charging the consumer his maximal will- would like to agree to a state-contingent
ingness-to-pay. When the consumer has pri- contract so that no single risk-averse agent
vate information about his valuation of the needs to bear all of the environmental risk;
product, the monopolist’s belief about the but, if any party were to discover ahead of
distribution of possible valuations creates a time whether and/or when the hazard would
perception of a downward-sloping demand in fact occur, that party would be unwilling
function. The monopolist maximizes to agree to contracts that would turn out to
expected revenue, but the consumer retains be ex post losses, and the entire arrange-
some surplus in expectation. In Roesler’s ment would unravel. Hirschleifer (1971)
model, the consumer chooses to not obtain provides a simple example in which risk-
some information about the monopolist’s averse individuals endowed with state-con-
product to remain somewhat uncertain tingent claims would like to share some risk.
about his valuation of it. He shapes his infor- Suppose that a farmer can hedge his risk of
mation structure (by skewing the distribu- incurring losses from a drought by contract-
tion of signals he obtains) so that he often ing with an insurer. Both the farmer and the
has an expected valuation slightly below the insurer agree to the transaction as long as
price the monopolist would have charged if they have no information about the weather
he had obtained full information (and, in bal- forecast. However, if they knew whether or
ance, so that he occasionally has a higher val- not a drought would happen, the party that
uation). The monopolist must, in response, inevitably would incur losses would back
lower the price to capture the consumer with out. While Hirschleifer presents his example
this valuation. In effect, the consumer uses in a setting of public information, its force
strategic information avoidance to commit remains even in the case of private informa-
to buying the product even when he has a tion (as long as it is common knowledge that
low ex post valuation, and this increases the information is released). Even if the farmer
consumer surplus because it drives down the does not have access to the forecast, if he
monopolist’s price. knows that the insurer has obtained a fore-
Information avoidance can also be cast and still wants to make a deal, he can
employed to deal with an adverse selection infer that no drought was predicted and call
problem (Akerlof 1970). In theory, if both off the deal. Hence, both parties would still
want to avoid the forecast to keep the deal Principal–agent problems arise when
in place.13 one party (the principal) finds it difficult to
Efficient risk sharing in a health-insurance incentivize another, more informed party
market is similarly jeopardized by adverse (the agent) to act on his behalf. In theory,
selection. Some people may be unable to avoiding information about the agent’s action
obtain insurance if they acquire private could sometimes be a wise strategy that the
information (Rothschild and Stiglitz 1997). principal can use to improve the agent’s
Suppose it would be economically efficient incentives. If a subordinate would be more
for individuals to purchase health insurance motivated with the power to make decisions
from a cooperative with many members, on his own, for example, a manager might
and suppose that the insurance cooperative avoid information about the details of the
charges people who suffer from preexisting subordinate’s work in order to make a com-
conditions higher premiums (and offers cor- mitment not to micromanage him (Aghion
responding discounts for clean bills of health) and Tirole 1997).15 In a laboratory experi-
in order to avoid the a dverse-selection prob- ment, Dominguez-Martinez, Sloof, and von
lem. If members were obligated to inform the Siemens (2014) find evidence that subjects
insurance cooperative of positive test results in a managerial role do indeed strategically
(i.e., if the information could become pub- avoid information so as to delegate deci-
lic), they might avoid the diagnosis (even if it sion-making authority to subjects assigned
is free) because they prefer to have the risk to be workers, consistent with Aghion and
covered by their insurance policy. Moreover, Tirole’s (1997) prediction.
even if they had the right to keep test results Dewatripont and Maskin (1995) and
confidential (but could not hide records indi- Cremer (1995) show that information avoid-
cating that a test had been performed), they ance can be a useful commitment device
might still prefer not to get tested because in repeated contracting in order to prevent
the insurance cooperative could reasonably unfavorable renegotiation when the initial
assume that failure to disclose a clean bill of contract opens up again. In Dewatripont and
health is evidence of a preexisting condition Maskin’s (1995) model, the principal hires a
when a test has been performed. Hence, risk-averse agent to do a job with uncertain
only a publicly observable choice not to costs. The agent will obtain private informa-
obtain the diagnostic test ensures they can tion about his costs and take two potentially
hedge their risk of being ill.14 observable actions that depend on his cost
signal. Dewatripont and Maskin show that
in some cases, when the agent can reopen
13 Schlee (2001) points out that while individuals will
contract negotiations after his first action
want to avoid information to preserve efficient risk-sharing
arrangements, a representative agent for the full economy
will not reflect this dislike for information.
14 In some market settings, parties can agree on con- to have high costs and keep all of the savings for himself.
tracts despite an adverse-selection problem. Kessler (1998) The trade-off between keeping the cost savings and fore-
considers optimal contracting between a contractor who going the additional work merely breaks even, so there is
can acquire private information about the costs of a proj- no incentive to lie. The benefit of occasionally avoiding
ect and a client who authorizes a larger project when the information about project costs is that the smaller project
contractor reports lower costs. The contractor can extract becomes more lucrative in expectation because of occa-
a strategic rent by committing to occasionally avoid infor- sional unanticipated cost savings. Taking account of this
mation about the project costs. In these cases, the client effect, the client proposes a slightly larger project with a
will assume the worst and authorize a smaller project with higher fee than he otherwise would have.
a smaller fee, but the contractor will be able to keep all 15 See also Prat (2005) for a more obscure example of
of the cost savings when realized costs are low. Of course, a principal wanting to avoid information about an agent’s
a contractor who finds he has low costs can always claim action to improve his incentives.
but the second action is even more revealing firms and to push managers to behave more
of his true costs, the principal will strategi- responsibly, governments may privatize (or
cally choose to not observe the first action choose not to nationalize) firms in order to
and monitor just the second action. If the avoid information about their operations.
principal were to observe the first action, it
would convey information about the agent’s
4. Consequences of Information Avoidance
costs. The two parties could then renegotiate
the contract on the basis of this new infor-
4.1 Beneficial and Counterproductive
mation. Because the agent is risk averse, the
Effects on Decision Making
optimal contract ex ante effectively involves
a transfer from an agent with low costs to an Perhaps the most general and obvious
agent with high costs (with the principal pro- consequence of information avoidance is
viding insurance). Renegotiation interferes that it robs people of potentially useful infor-
with such risk sharing. After discovering he mation that could be used to enhance deci-
has low costs, the agent would no longer sion making. For example, teachers who fail
want to make the ex ante optimal transfer. to read teaching evaluations will not obtain
If the principal were to observe the agent’s feedback that could, potentially, improve
first action and thus obtain evidence about their teaching. Likewise, stock-market inves-
his costs, they would renegotiate the con- tors who do not look up the value of their
tract to reduce this transfer. So, ex ante, the portfolio when the market is down deprive
risk-averse agent would be exposed to more themselves of potentially useful information.
risk. The principal would then have to pay Yet, it is possible that asymmetric look-ups
more in expectation to compensate the agent occur for sensible, intrapersonal-strategic
for bearing this additional risk. By avoiding reasons. Perhaps, for example, investors are
information about the agent’s first action, the aware of their own predilection for panick-
principal makes a credible commitment to ing and selling at the bottom of the market;
not reopen negotiations after the agent has if so, then ignoring information could actu-
acquired his private information. ally improve investor returns (see Sicherman
In a model proposed by Cremer (1995), et al. 2016, for a detailed discussion of
firms (i.e., principals) enter into a rm’s-length this issue). More generally, in evaluating
relationships with suppliers (i.e., agents) whether information avoidance produces a
in which they observe the suppliers’ pro- net welfare gain or loss, one needs to take
duction, but choose not to listen to excuses account of the fact that people often avoid
if that product is subpar. Were they to information for perfectly good reasons. The
entertain excuses, firms might find them hedonic consequences of information acqui-
reasonable and give their suppliers second sition or avoidance, in particular, should be
chances. By committing not to give suppli- part of any welfare calculation. Caplin and
ers second chances, they maintain stronger Leahy (2004), Schweizer and Szech (2013),
incentives for good effort. Firms that avoid and Lipnowski and Mathevet (2015) recog-
information about their suppliers’ operations nize this in proposing mechanisms for the
may thus be taking a lesson from the strict optimal provision of information to patients
teacher who pays no mind to any excuses with anxiety.
for late homework and thereby demands
4.2 Ethical Transgressions
accountability. Applying similar logic to a key
policy issue, Schmidt (1996) suggests that as Bazerman and Sezer (2016) note that in the
part of a commitment to not rescue failing wake of revelations about unethical behavior
on a large scale, ranging from Bernie Madoff remain ignorant, not in an effort to commit
to corruption at FIFA (the world soccer themselves to an action, but to avoid reper-
federation), it is difficult to imagine how so cussions for what they must know is a likely
many people could have failed to notice the effect of their action (Dana 2005). Heimer
unethical behavior. With the benefit of hind- (2012) suggests that organizations may delib-
sight, the instances of wrongdoing are obvi- erately divide tasks and establish bureau-
ous. Bazerman and Sezer point out, however, cratic barriers to foster such “distributed
that unethical situations are often ambiguous ignorance.” This is particularly problematic,
and, if motivated to do so, one could main- as little blame appears to be assigned to those
tain a belief of propriety. Someone observ- who did not, but should have, known about
ing the returns on Bernie Madoff’s fund, wrongdoing. Pharmaceutical companies
for example, may think it a good product to holding the patent to a drug with no substi-
invest in. However, a professional with years tutes, for example, may sell, at a high price,
of training ought to know that the fund’s the marketing rights of the drug to a small
return looked suspicious (exceeding returns firm that can then subsequently increase the
of the broader market with no additional price. If this leads to a public backlash, the
volatility). This suspicion, in turn, should ire is directed at the firm with the market-
have prompted additional due diligence, ing rights, protecting the reputation of the
especially before advising clients to invest manufacturer.16
in the fund. Yet, in such ambiguous cases, In some cases, discovering information
people may be motivated not to look closer confers a moral responsibility to inform
so as to avoid potentially unearthing uneth- others. For example, individuals with STDs
ical behavior—an illustration of what they could be seen as morally compelled to share
call “bounded ethicality.” They argue that the information with prospective sexual part-
failing to examine questionable behavior ners. An individual who is fearful that he has
more closely, and hence enabling potentially an STD might choose to not get tested so as to
unethical behavior, is itself unethical. not confront the ethical dilemma of sharing
Failing to learn about unethical behavior any bad news that the test reveals. Moreover,
(by avoiding information) may reduce how any disease with a hereditary component,
harshly others judge an action. Bazerman once diagnosed, provides information that
(2014) recounts his own failure to notice and is relevant to parents, children, and siblings.
act on corruption he observed in an ambigu- In a study with Jewish women, Lehmann et
ous situation and notes that others are quick al. (2000) find that virtually all respondents
to absolve him of guilt. Executives may, for (100 percent and 97 percent in two scenar-
example, implement policies that induce ios) believe there is a duty to inform at-risk
unethical behavior and then fail to obtain relatives if the disease is preventable, and
information about whether employees 85 percent believe it to be a duty even when
behave unethically. For example, a consulting the disease cannot be prevented. A decision
firm that charges clients for hours billed may to get tested may, in such cases, affect not
offer large bonuses to long-working employ- just one’s own level of knowledge, but that of
ees, but not check whether the number of others who were not involved in the decision
hours billed across all clients is feasible. In to get tested. If someone feels compelled to
not collecting the information, executives
maintain “plausible deniability” in the event 16 See, for example, http://www.nytimes.com/2015/09/21/
that unethical behavior is discovered (Dana business/a-huge-overnight-increase-in-a-drugs-price-raises-
2006; Simon 2005). Agents may strategically protests.html.
disclose genetic risk information to relatives in groups, they also choose to buy into the
upon obtaining it, they may take their rela- herd mentality of the market and, in doing
tives’ (perceived) information preferences so, contribute to the momentum of extreme
into account before deciding whether to get price swings.
tested themselves.
4.5 Confirmation Bias
4.3 Spread of Disease
As already noted in the section on infor-
Avoidance of medical testing about con- mation avoidance via distorted processing of
tagious diseases, such as AIDS, can contrib- information, confirmation bias (Nickerson
ute to the spread of these diseases. While an 1998) is a common pattern of distorted
individual may choose to avoid testing—per- beliefs that is associated with, and partly the
haps rationally, given hedonic considerations result of, selective exposure to (and avoid-
(Brashers, Goldsmith and Hsieh 2002)—out ance of) information (Jonas et al. 2001; Hart
of fear of getting a positive diagnosis, this et al. 2009). Confirmation bias has diverse
kind of information avoidance imposes a negative manifestations. It can, for example,
negative externality on others. An untested causes people to accept ineffective medical
individual may fail to take precautions that treatments (and then to judge them as more
prevent the spread of the disease to oth- effective than they actually are) (Nickerson
ers, and societal welfare could be reduced 1998), and can also lead to scientific atrophy
(Caplin and Eliaz 2003). when scientists fail to challenge their own
beliefs or update them in response to valid
4.4 Groupthink
challenges from others. As Max Planck noted,
When decisions are made in groups, infor- “a new scientific truth does not triumph by
mation avoidance can play an especially convincing its opponents and making them
pernicious role, not only because many of see the light, but rather because its oppo-
the most important decisions are made by nents eventually die, and a new generation
groups, but because interpersonal interac- grows up that is familiar with it.” Researchers
tions can magnify motives leading to infor- suffering from confirmation bias may ratio-
mation avoidance. Bénabou (2013) provides nalize failed experiments, attributing them
an elegant model of such effects in a paper to chance or flawed design, and repeat sim-
on “groupthink,” a phenomenon first iden- ilar experiments until they “work,” and then
tified and discussed by Janis (1972). When conclude overconfidently that their initial
collective opinion tends toward unanimity, hypotheses are correct. This is a recipe for
people often find it individually rational to producing invalid scientific conclusions (see
adopt the shared belief as their own, rather Simmons, Nelson, and Simonsohn 2011).
than to gather their own information, even In situations of joint authorship in aca-
when the shared belief is irrational. The demia, and numerous other situations in
dearth of new information, in turn, makes it which people cooperate or delegate respon-
hard for the group to correct their irrational sibility outside of academia, people rely on
belief. Bénabou (2013) also draws a link the integrity and good will of those they are
between groupthink in organizations and working with. Having delegated parts of
collective delusions in markets that can lead a joint task, however, they may then avoid
to irrational exuberance, bubbles, and p anics information that could reveal misplaced
(Kindleberger and Aliber 2005, Reinhart and trust. An academic who has engaged in pro-
Rogoff 2009, Shiller 2005). Just as people ductive collaborations with a colleague or
choose to go along with collective opinion graduate student may be blind to warning
signs of sloppy or fraudulent practices that selectively attend to media that supports
would be more apparent from a distance their preexisting opinions, but because
(see, e.g., Bazerman and Sezer 2016). they engage in a wide range of behaviors
that contribute to selective exposure to
4.6 Media Bias
information. An example of such a behav-
If people eschew information that threat- ior is relocating to a geographic region in
ens their existing beliefs and demand only which residents share one’s views. A recent
information that supports these views, it is trend in the United States provides strik-
natural for the media to supply the infor- ing evidence of such values-based residen-
mation they demand. In the competition tial segregation. As The Economist reports
for more readers and viewers, media outlets it: “Americans are increasingly choosing
have an incentive to provide biased coverage to live among like-minded neighbors. . . .
that aligns with the perspective of their tar- Some folks in Texas recently decided
get audience. In his book Republic.com 2.0, to start a new community ‘containing
Sunstein (2007, p. 16) notes that the Internet 100 percent Ron Paul supporters.’ ” 3 (Ron
holds great promise as a democratizing tech- Paul is a staunch libertarian and was a
nology by enabling the dissemination of a Republican presidential candidate in the
greater diversity of information than was 2008 race.) A recent book, The Big Sort
previously possible. Yet, he points out, the (Bishop 2008), documents that the Ron
greater diversity of information also makes Paul community is only one example of a
it possible for people to selectively expose more general trend for Americans to form
themselves to perspectives that accord with, like-minded clusters.
and fail to challenge, their existing views, Even when people cannot avoid encoun-
and he cautions against “the risks posed by tering other people they disagree with, they
any situation in which thousands or perhaps can avoid having conversations that bring out
millions or even tens of millions of people such differences. Sugden (2005, p. 67) pro-
are mainly listening to louder echoes of their vides an elegant description of the phenom-
own voices.” Supportive of Sunstein’s con- enon of “conversational minefields” whereby
cerns, Gentzkow and Shapiro (2010) find “different topics are gradually introduced
that newspapers readers are more likely into the conversation, exploiting connec-
to demand news slanted toward their own tions with what has already been said, with
political ideology, and that firms respond the general aim of finding a topic on which
strongly to readers’ preferences (for a subtly the two partners have common opinions or
different perspective, see Garrett, Carnahan, beliefs. If a topic begins to provoke disagree-
and Lynch 2013). Similarly, users of social ment, it is dropped.”
networks are more likely to be exposed While avoiding contrary information may
to (and click through to) news stories con- lead to polarization, Druckman, Peterson,
gruent with their political beliefs than and Slothuus (2013) show that polarization
cross-cutting content (Bakshy, Messing, and can lead back to information avoidance. In a
Adamic 2015). polarized environment, partisan cues drown
out reasoned arguments. People not only
4.7 Political Polarization
form opinions based on weaker evidence,
Closely related to, and both derivative of but also believe that the arguments endorsed
and contributing to, the problem of media by those holding their position are in fact
bias is that of p
olitical polarization. Such stronger, so they rationalize not listening to
polarization occurs not only because p eople opposing arguments.
4.8 Climate Change Denial In a book about climate change denial

aptly (for the theme of this review) titled
One polarizing issue of special importance, Don’t Even Think About It, Marshall (2014)
given the threat it constitutes to humanity, is seeks to make sense of why the nations of
climate change. Although the vast majority of the world are largely ignoring an unfolding
scientists believe that climate change is real, problem that threatens to immiserate even
caused by humans, and could potentially be generations alive today. Much of his expla-
mitigated by a concerted human response, nation for the riddle involves information
large numbers of individuals in the United avoidance—e.g.,
States reject these conclusions. Rejection of
the reality of climate change makes it difficult The bottom line is that we do not accept cli-
mate change because we wish to avoid the
if not impossible for the United States, which anxiety it generates and the deep changes it
was until recently the world’s largest emitter of requires. In this regard, it is not unlike any
greenhouse gases, to enact policies to reduce other major threat. However, because it carries
such emissions. Rejection of a scientific con- none of the clear markers that would normally
sensus almost by definition requires informa- lead our brains to overrule our short-term
interests, we actively conspire with each other,
tion avoidance. Kahan, Jenkins-Smith, and and mobilize our own biases to keep it perpet-
Braman (2011) find that people with differ- ually in the background (page 228).
ent political positions disagree sharply about
how serious a threat climate change is, and
both groups tend to denigrate research that 5. Final Comments
contradicts their views (see also Kahan et al.
2012). For example, presented with a PhD Since George Stigler pioneered the eco-
scientist who is a member of the US National nomic analysis of information, we have
Academy of Sciences, climate believers and become used to thinking of information
deniers disagree about whether he really is as a means to a (typically material) end.
an “expert,” depending on whether his view Information is seen as valuable to the extent,
matches their own (Kahan, Jenkins-Smith, and only to the extent, that it enhances deci-
and Braman 2011). Surprisingly for believers, sion making. Among the stylized assump-
belief in climate change bears no connection tions of economics, this may be one of the
to an individual’s level of scientific literacy or most unrealistic and most consequential.
expertise (Kahan et al. 2012). Kahan provides Considerable research in economics, psy-
a kind of rational account of such polariza- chology, and neuroscience points to the con-
tion, pointing out the tremendous costs that clusion that people derive utility not only
an individual would bear for deviating from from possessions and experiences, but also
the dominant views of the group within which from beliefs. Seminal early contributions to
they are embedded. As a result, as summa- this perspective include Thomas Schelling’s
rized in a recent Nature column on Kahan’s 1987 paper “The Mind as a Consuming
work, “social science suggests that citizens are Organ,” and a 1986 paper by the psycholo-
culturally polarized because they are, in fact, gist Robert Abelson titled “Beliefs are Like
too rational at filtering out information that Possessions.” Neuroscience research (e.g.,
would drive a wedge between themselves and Knutson and Peterson 2005) shows that
their peers.”17 people derive immediate utility—pleasure
and pain—from learning about gains and
17 http://www.nature.com/news/why-we-are-poles- losses, and also provides strong support for
apart-on-climate-change-1.11166. the reality of anticipatory utility (Berns et al.
2006; Loewenstein 1987). Research on top- produce it could have immediate and import-
ics such as ego, meaning, framing, and men- ant policy applications—e.g., in encouraging
tal accounting provides further support for at-risk individuals to test frequently for HIV,
a point that should perhaps be seen as obvi- or in overcoming resistance to confronting
ous: most of what matters happens “inside the scientific evidence on climate change.
our heads,” and, given that there are often We hope that this review will not only help
multiple ways to interpret the same piece of to introduce a relatively unknown topic to
information, how we construe information is economists, but may also inspire new lines of
often as important as the objective content of theoretical and empirical investigation.
the information. The burgeoning economic
literature dealing with belief-based utility REFERENCES
(e.g., Caplin and Leahy 2001; Geanakoplos, Abelson, Robert P. 1986. “Beliefs Are Like Posses-
Pearce, and Stacchetti 1989; Kőszegi 2010) sions.” Journal for the Theory of Social Behaviour 16
suggests that economists are beginning to (3): 223–50.
Abelson, Robert P., Elliot Aronson, William J. McGuire,
grapple with these complexities.18 Theodore M. Newcomb, Milton J. Rosenberg, Percy
Information avoidance provides a superb H. Tannenbaum, eds. 1968. Theories of Cognitive
lens into the true complexities of consump- Consistency: A Sourcebook. Chicago: Rand McNally.
Aghion, Philippe, and Jean Tirole. 1997. “Formal and
tion in the mind. On the one hand, people Real Authority in Organizations.” Journal of Political
avoid information for conventional economic Economy 105 (1): 1–29.
reasons: for strategic purposes that promote Akerlof, George A. 1970. “The Market for ‘Lemons’:
Quality Uncertainty and the Market Mechanism.”
material outcomes. But people also avoid Quarterly Journal of Economics 84 (3): 488–500.
information for reasons not well-captured Akerlof, George A., and William T. Dickens. 1982. “The
by conventional economic analysis. In some Economic Consequences of Cognitive Dissonance.”
American Economic Review 72 (3): 307–19.
cases, they avoid information to, in effect, Alaoui, Larbi. 2012. “The Value of Useless Informa-
license them to behave as they would really tion.” University of Pompeu Fabra Department of
like to behave—providing “plausible deni- Economics Working Paper 1313.
Anderson, Michael C., and Ean Huddleston. 2012.
ability” of unethical behavior not only to “Towards a Cognitive and Neurobiological Model
other people but also to themselves. Even of Motivated Forgetting.” In True and False Recov-
more tellingly, people often avoid informa- ered Memories: Toward a Reconciliation of the
Debate, edited by Robert F. Belli, 53–120. New York:
tion simply because the information would Springer.
make them feel bad—because information Andreoni, James, Justin M. Rao, and Hannah Tracht-
carries direct, and often negative, utility. man. 2011. “Avoiding the Ask: A Field Experiment
on Altruism, Empathy, and Charitable Giving.”
As we have shown, there are many streams National Bureau of Economic Research Working
of research in both economics and psychology Paper 17648.
that either directly address, or can be con- Andries, Marianne, and Valentin Haddad. 2014. “Infor-
mation Aversion.” Society for Economic Dynamics
nected to, information avoidance. Given the 2014 Meeting Paper 1091.
diverse mechanisms leading to information Ariely, Dan, Uri Gneezy, George Loewenstein, and
avoidance, literature on the topic has not Nina Mazar. 2009. “Large Stakes and Big Mistakes.”
Review of Economic Studies 76 (2): 451–69.
been structured as a coherent body; nor, we Artstein-Avidan, Shiri, and David Dillenberger. 2015.
suspect, can it or should it. Given the import- “Dynamic Disappointment Aversion.” Unpublished.
ant consequences of information avoidance, Asch, David A., James P. Patton, and John C. Hershey.
1990. “Knowing for the Sake of Knowing: The Value
however, research on the mechanisms that of Prognostic Information.” Medical Decisions Mak-
ing 10 (1): 47–57.
Averill, James R., and Miriam Rosenn. 1972. “Vigilant
18 Psychologists have a longer tradition of recogniz- and Nonvigilant Coping Strategies and Psychophys-
ing information avoidance (see Sweeny et al. 2010 and iological Stress Reactions during the Anticipation of
Hertwig and Engel 2016). Electric Shock.” Journal of Personality and Social
Psychology 23 (1): 128–41. through Ignorance.” RAND Journal of Economics 38

Babcock, Linda, George Loewenstein, Samuel Issacha- (4): 931–47.
roff, and Colin Camerer. 1995. “Biased Judgments of Brock, Timothy C., and Joe L. Balloun. 1967. “Behav-
Fairness in Bargaining.” American Economic Review ioral Receptivity to Dissonant Information.” Journal
85 (5): 1337–43. of Personality and Social Psychology 6 (4): 413–28.
Bakshy, Eytan, Solomon Messing, and Lada A. Adamic. Brunnermeier, Markus K., and Jonathan A. Parker.
2015. “Exposure to Ideologically Diverse News and 2005. “Optimal Expectations.” American Economic
Opinion on Facebook.” Science 348 (6239): 1130–32. Review 95 (4): 1092–118.
Baron, Jonathan, and John C. Hershey. 1988. “Out- Budescu, David V., and Ilan Fischer. 2001. “The Same
come Bias in Decision Evaluation.” Journal of Per- But Different: An Empirical Investigation of the
sonality and Social Psychology 54 (4): 569–79. Reducibility Principle.” Journal of Behavioral Deci-
Bazerman, Max H. 2014. The Power of Noticing: What sion Making 14 (3): 187–206.
the Best Leaders See. New York: Simon and Schuster. Cain, Daylian M., Jason Dana, and George E. New-
Bazerman, Max H., and Ovul Sezer. 2016. “Bounded man. 2014. “Giving versus Giving In.” Academy of
Awareness: Implications for Ethical Decision Mak- Management Annals 8 (1): 505–33.
ing.” Organizational Behavior and Human Decision Camerer, Colin, George Loewenstein, and Martin
Processes 136: 95–105. Weber. 1989. “The Curse of Knowledge in Economic
Bell, David E. 1985. “Disappointment in Decision Settings: An Experimental Analysis.” Journal of Polit-
Making under Uncertainty.” Operations Research 33 ical Economy 97 (5): 1232–54.
(1): 1–27. Caplan, L. S. 1995. “Patient Delay in Seeking Help for
Benabou, Roland. 2013. “Groupthink: Collective Delu- Potential Breast Cancer.” Public Health Reviews 23
sions in Organizations and Markets.” Review of Eco- (3): 263–74.
nomic Studies 80 (2): 429–62. Caplin, Andrew, and Mark Dean. 2015. “Revealed
Benabou, Roland, and Jean Tirole. 2002. “Self-Confi- Preference, Rational Inattention, and Costly Infor-
dence and Personal Motivation.” Quarterly Journal mation Acquisition.” American Economic Review
of Economics 117 (3): 871–915. 105 (7): 2183–203.
Benabou, Roland, and Jean Tirole. 2011. “Identity, Caplin, Andrew, and Kfir Eliaz. 2003. “AIDS Policy
Morals, and Taboos: Beliefs as Assets.” Quarterly and Psychology: A Mechanism-Design Approach.”
Journal of Economics 126 (2): 805–55. RAND Journal of Economics 34 (4): 631–46.
Benoît, Jean-Pierre, and Juan Dubra. 2011. “Apparent Caplin, Andrew, and John Leahy. 2001. “Psychological
Overconfidence.” Econometrica 79 (5): 1591–625. Expected Utility Theory and Anticipatory Feelings.”
Ben-Porath, Elchanan, and Eddie Dekel. 1992. “Sig- Quarterly Journal of Economics 116 (1): 55–79.
naling Future Actions and the Potential for Sacri- Caplin, Andrew, and John Leahy. 2004. “The Supply
fice.” Journal of Economic Theory 57 (1): 36–51. of Information By a Concerned Expert.” Economic
Berglas, Steven, and Edward E. Jones. 1978. “Drug Journal 114 (497): 487–505.
Choice as a Self-Handicapping Strategy in Response Carrillo, Juan D., and Thomas Mariotti. 2000. “Strate-
to Noncontingent Success.” Journal of Personality gic Ignorance as a Self-Disciplining Device.” Review
and Social Psychology 36 (4): 405–17. of Economic Studies 67 (3): 529–44.
Bernheim, B. Douglas, and Raphael Thomadsen. 2005. Case, Donald O., James E. Andrews, J. David John-
“Memory and Anticipation.” Economic Journal 115 son, and Suzanne L. Allard. 2005. “Avoiding versus
(503): 271–304. Seeking: The Relationship of Information Seeking
Berns, Gregory, Jonathan Chappelow, Milos Cekic, to Avoidance, Blunting, Coping, Dissonance, and
Caroline F. Zink, Giuseppe Pagnoni, and Megan E. Related Concepts.” Journal of the Medical Library
Martin-Skurski. 2006. “Neurobiological Substrates of Association 93 (3): 353–62.
Dread.” Science 312 (5774): 754–58. Chater, Nick, and George Loewenstein. 2016. “The
Bishop, Bill. 2008. The Big Sort: Why the Clustering Under-Appreciated Drive for Sense-Making.” Jour-
of Like-Minded America Is Tearing Us Apart. New nal of Economic Behavior and Organization 126
York: Houghton Mifflin Harcourt Publishing. (Part B): 137–54.
Boss, Pauline. 1999. Ambiguous Loss: Learning to Live Conrads, Julian, and Bernd Irlenbusch. 2013. “Strate-
with Unresolved Grief. Cambridge, MA and London: gic Ignorance in Ultimatum Bargaining.” Journal of
Harvard University Press. Economic Behavior and Organization 92: 104–15.
Brashers, Dale E., Daena J. Goldsmith, and Elaine Cremer, Jacques. 1995. “Arm’s Length Relationships.”
Hsieh. 2002. “Information Seeking and Avoiding in Quarterly Journal of Economics 110 (2): 275–95.
Health Contexts.” Human Communication Research Dana, Jason. 2005. “Conflicts of Interest and Strategic
28 (2): 258–71. Ignorance of Harm.” In Conflicts of Interest: Chal-
Broadbent, D. E. 1958. Perception and Communica- lenges and Solutions in Business, Law, Medicine, and
tion. Oxford and New York: Pergamon Press. Public Policy, edited by Don A. Moore, Daylian M.
Broberg, Tomas, Tore Ellingsen, and Magnus Johan- Cain, George Loewenstein, and Max H. Bazerman,
nesson. 2007. “Is Generosity Involuntary?” Econom- 206–23. Cambridge and New York: Cambridge Uni-
ics Letters 94 (1): 32–37. versity Press.
Brocas, Isabelle, and Juan D. Carrillo. 2007. “Influence Dana, Jason. 2006. “Strategic Ignorance and Ethical
Behavior in Organizations.” In Ethics in Groups: 2015. “Suspense and Surprise.” Journal of Political
Research on Managing Groups and Teams, Volume 8, Economy 123 (1): 215–60.
edited by Ann E. Tenbrunsel, 39–57. Bingley: Emer- Emanuel, Amber S., et al. 2015. “Avoiding Cancer
ald Group Publishing. Risk Information.” Social Science and Medicine 147:
Dana, Jason, Daylian M. Cain, and Robyn M. Dawes. 113–20.
2006. “What You Don’t Know Won’t Hurt Me: Costly Falk, Armin, and Florian Zimmermann. 2016. “Beliefs
(but Quiet) Exit in Dictator Games.” Organizational and Utility: Experimental Evidence on Preferences
Behavior and Human Decision Processes 100 (2): for Information.” CESifo Working Paper 6061.
193–201. Fels, Markus. 2015. “On the Value of Information: Why
Dana, Jason, Roberto A. Weber, and Jason Xi Kuang. People Reject Medical Tests.” Journal of Behavioral
2007. “Exploiting Moral Wiggle Room: Experiments and Experimental Economics 56: 1–12.
Demonstrating an Illusory Preference for Fairness.” Ferrer, Rebecca A., Jennifer M. Taber, William M. P.
Economic Theory 33 (1): 67–80. Klein, Peter R. Harris, Katie L. Lewis, and Leslie
Dervin, Brenda. 1998. “Sense-Making Theory and G. Biesecker. 2015. “The Role of Current Affect,
Practice: An Overview of User Interests in Knowl- Anticipated Affect and Spontaneous Self-Affirmation
edge Seeking and Use.” Journal of Knowledge Man- in Decisions to Receive Self-Threatening Genetic
agement 2 (2): 36–46. Risk Information.” Cognition and Emotion 29 (8):
Deshpande, Rohit, and Ajay K. Kohli. 1989. “Knowl- 1456–65.
edge Disavowal: Structural Determinants of Infor- Festinger, Leon. 1957. A Theory of Cognitive Disso-
mation-Processing Breakdown in Organizations.” nance. Stanford: Stanford University Press.
Science Communication 11 (2): 155–69. Frederick, Shane, and George Loewenstein. 1999.
Dewatripont, Mathias, and Eric Maskin. 1995. “Con- “Hedonic Adaptation.” In Well-Being: The Foun-
tractual Contingencies and Renegotiation.” RAND dations of Hedonic Psychology, edited by Daniel
Journal of Economics 26 (4): 704–19. Kahneman, Ed Diener, and Norbert Schwarz, 302–
Dillenberger, David. 2010. “Preferences for One- 29. New York: Russell Sage Foundation.
Shot Resolution of Uncertainty and Allais-Type Frey, Dieter, and Dagmar Stahlberg. 1986. “Selection
Behavior.” Econometrica 78 (6): 1973–2004. of Information after Receiving More or Less Reli-
Dominguez-Martinez, Silvia, Randolph Sloof, and able Self-Threatening Information.” Personality and
Ferdinand A. von Siemens. 2014. “Monitored By Social Psychology Bulletin 12 (4): 434–41.
Your Friends, Not Your Foes: Strategic Ignorance Ganguly, Ananda R., and Joshua Tasoff. Forthcoming.
and the Delegation of Real Authority.” Games and “Fantasy and Dread: The Demand for Information
Economic Behavior 85: 289–305. and the Consumption Utility of the Future.” Man-
Druckman, James N., Erik Peterson, and Rune Slo- agement Science.
thuus. 2013. “How Elite Partisan Polarization Affects Garrett, R. Kelly, Dustin Carnahan, and Emily K.
Public Opinion Formation.” American Political Sci- Lynch. 2013. “A Turn toward Avoidance? Selective
ence Review 107 (1): 57–79. Exposure to Online Political Information, 2004–
Ehrich, Kristine R., and Julie R. Irwin. 2005. “Will- 2008.” Political Behavior 35 (1): 113–34.
ful Ignorance in the Request for Product Attribute Geanakoplos, John, David Pearce, and Ennio Stac-
Information.” Journal of Marketing Research 42 (3): chetti. 1989. “Psychological Games and Sequential
266–77. Rationality.” Games and Economic Behavior 1 (1):
Ehrlich, Danuta, Isaiah Guttman, Peter Schonbach, 60–79.
and Judson Mills. 1957. “Postdecision Exposure to Gentzkow, Matthew, and Jesse M. Shapiro. 2010.
Relevant Information.” Journal of Abnormal and “What Drives Media Slant? Evidence from U.S.
Social Psychology 54 (1): 98–102. Daily Newspapers.” Econometrica 78 (1): 35–71.
Eil, David, and Justin M. Rao. 2011. “The Good Gino, Francesca, and Dan Ariely. 2012. “The Dark Side
News–Bad News Effect: Asymmetric Processing of of Creativity: Original Thinkers Can Be More Dis-
Objective Information about Yourself.” American honest.” Journal of Personality and Social Psychology
Economic Journal: Microeconomics 3 (2): 114–38. 102 (3): 445–59.
Eliaz, Kfir, and Andrew Schotter. 2007. “Experimental Glaeser, Edward L., and Cass R. Sunstein. 2013. “Why
Testing of Intrinsic Preferences for Noninstrumen- Does Balanced News Produce Unbalanced Views?”
tal Information.” American Economic Review 97 (2): National Bureau of Economic Research Working
166–69. Paper 18975.
Eliaz, Kfir, and Andrew Schotter. 2010. “Paying for Golman, Russell, and George Loewenstein. 2015.
Confidence: An Experimental Study of the Demand “The Demand For, and Avoidance of, Information.”
for Non-instrumental Information.” Games and Eco- Unpublished.
nomic Behavior 70 (2): 304–24. Golman, Russell, and George Loewenstein. Forthcom-
Eliaz, Kfir, and Ran Spiegler. 2006. “Can Anticipatory ing. “Information Gaps: A Theory of Preferences
Feelings Explain Anomalous Choices of Information Regarding the Presence and Absence of Information.
Sources?” Games and Economic Behavior 56 (1): Decision
87–104. Golman, Russell, George Loewenstein, Karl O. Moene,
Ely, Jeffrey, Alexander Frankel, and Emir Kamenica. and Luca Zarri. 2016. “The Preference for Belief
Consonance.” Journal of Economic Perspectives 30 Kahan, Dan M., et al. 2012. “The Polarizing Impact
(3): 165–88. of Science Literacy and Numeracy on Perceived
Goulas, Sofoklis, and Rigissa Megalokonomou. 2015. Climate Change Risks.” Nature Climate Change 2:
“Knowing Who You Are: The Effect of Feedback 732–35.
Information on Short and Long Term Outcomes.” Kahan, Dan M., Hank Jenkins-Smith, and Donald Bra-
Unpublished. man. 2011. “Cultural Cognition of Scientific Consen-
Grant, Simon, Atsushi Kajii, and Ben Polak. 1998. sus.” Journal of Risk Research 14 (2): 147–74.
“Intrinsic Preference for Information.” Journal of Karlsson, Niklas, George Loewenstein, and Duane
Economic Theory 83 (2): 233–59. Seppi. 2009. “The Ostrich Effect: Selective Attention
Grossman, Zachary. 2014. “Strategic Ignorance and the to Information.” Journal of Risk and Uncertainty 38
Robustness of Social Preferences.” Management Sci- (2): 95–115.
ence 60 (11): 2659–65. Kessler, Anke S. 1998. “The Value of Ignorance.”
Grossman, Zachary, and Joel J. van der Weele. Forth- RAND Journal of Economics 29 (2): 339–54.
coming. “Self-image and Willful Ignorance in Social Kindleberger, Charles P., and Robert Aliber. 2005.
Decisions.” Journal of the European Economic Manias, Panics, and Crashes: A History of Financial
Association. Crises, Fifth edition. Hoboken, NJ: John Wiley.
Gul, Faruk. 1991. “A Theory of Disappointment Aver- Knutson, Brian, and Richard Peterson. 2005. “Neurally
sion.” Econometrica 59 (3): 667–86. Reconstructing Expected Utility.” Games and Eco-
Gul, Faruk. 2001. “Unobservable Investment and the nomic Behavior 52 (2): 305–15.
Hold-Up Problem.” Econometrica 69 (2): 343–76. Kocher, Martin G., Michal Krawczyk, and Frans van
Hart, William, Dolores Albarracin, Alice H. Eagly, Inge Winden. 2014. “‘Let Me Dream On!’ Anticipatory
Brechan, Matthew J. Lindberg, and Lisa Merrill. Emotions and Preference for Timing in Lotteries.”
2009. “Feeling Validated versus Being Correct: A Journal of Economic Behavior and Organization 98:
Meta-analysis of Selective Exposure to Information.” 29–40.
Psychological Bulletin 135 (4): 555–88. Kőszegi, Botond. 2003. “Health Anxiety and Patient
Heimer, Carol A. 2012. “Inert Facts and the Illusion Behavior.” Journal of Health Economics 22 (6):
of Knowledge: Strategic Uses of Ignorance in HIV 1073–84.
Clinics.” Economy and Society 41 (1): 17–41. Kőszegi, Botond. 2006. “Ego Utility, Overconfidence,
Hermalin, Benjamin E., and Michael L. Katz. 2009. and Task Choice.” Journal of the European Economic
“Information and the Hold-Up Problem.” RAND Association 4 (4): 673–707.
Journal of Economics 40 (3): 405–23. Kőszegi, Botond. 2010. “Utility from Anticipation and
Hertwig, Ralph, and Christoph Engel. 2016. “Homo Personal Equilibrium.” Economic Theory 44 (3):
Ignorans: Deliberately Choosing Not to Know.” Psy- 415–44.
chological Science 11 (3): 359–72. Kőszegi, Botond, and Matthew Rabin. 2009. “Refer-
Hirshleifer, Jack. 1971. “The Private and Social Value ence-Dependent Consumption Plans.” American
of Information and the Reward to Inventive Activity.” Economic Review 99 (3): 909–36.
American Economic Review 61 (4): 561–74. Krähmer, Daniel, and Rebecca Stone. 2013. “Antici-
Hoy, Michael, Richard Peter, and Andreas Richter. pated Regret as an Explanation of Uncertainty Aver-
2014. “Take-Up for Genetic Tests and Ambiguity.” sion.” Economic Theory 52 (2): 709–28.
Journal of Risk and Uncertainty 48 (2): 111–33. Kreps, David M., and Evan L. Porteus. 1978. “Tempo-
Huck, Steffen, Nora Szech, and Lukas M. Wenner. ral Resolution of Uncertainty and Dynamic Choice
2015. “More Effort with Less Pay: On Information Theory.” Econometrica 46 (1): 185–200.
Avoidance, Belief Design and Performance.” Social Kruger, Justin, and Matt Evans. 2009. “The Paradox of
Science Research Center Berlin Discussion Paper Alypius and the Pursuit of Unwanted Information.”
SP II 2015-304r. Journal of Experimental Social Psychology 45 (6):
Ishida, Junichiro. 2012. “Contracting with Self-esteem 1173–79.
Concerns.” Journal of Economic Behavior and Orga- Lau, Stephanie. 2008. “Information and Bargaining in
nization 81 (2): 329–40. the Hold-Up Problem.” RAND Journal of Economics
Janis, Irving Lester. 1972. Victims of Groupthink: A 39 (1): 266–82.
Psychological Study of Foreign-Policy Decisions and Lazear, Edward P., Ulrike Malmendier, and Roberto
Fiascoes. Oxford: Houghton Mifflin. A. Weber. 2012. “Sorting in Experiments with Appli-
Jenkins, H. M., and R. A. Boakes. 1973. “Observing cation to Social Preferences.” American Economic
Stimulus Sources That Signal Food or No Food.” Journal: Applied Economics 4 (1): 136–63.
Journal of the Experimental Analysis of Behavior 20 Lehmann, Lisa Soleymani, Jane C. Weeks, Neil Klar,
(2): 197–207. Lois Biener, and Judy E. Garber. 2000. “Disclosure
Jonas, Eva, Stefan Schulz-Hardt, Dieter Frey, and Nor- of Familial Genetic Information: Perceptions of the
man Thelen. 2001. “Confirmation Bias in Sequential Duty to Inform.” American Journal of Medicine 109
Information Search after Preliminary Decisions: An (9): 705–11.
Expansion of Dissonance Theoretical Research on Lerman, Caryn, et al. 1996. “BRCA1 Testing in Fam-
Selective Exposure to Information.” Journal of Per- ilies with Hereditary Breast-Ovarian Cancer: A
sonality and Social Psychology 80 (4): 557–71. Prospective Study of Patient Decision Making and
Outcomes.” Journal of the American Medical Associ- Petrie. 2002. “Delay in Seeking Medical Care for
ation 275 (24): 1885–92. Self-Detected Breast Symptoms in New Zealand

Lerman, Caryn, et al. 1999. “Genetic Testing in Fam- Women.” New Zealand Medical Journal 115 (1166):
ilies with Hereditary Non-polyposis Colon Cancer.” U257.
Journal of the American Medical Association 281 Miller, Suzanne M. 1987. “Monitoring and Blunting:
(17): 1618–22. Validation of a Questionnaire to Assess Styles of
Leydon, Geraldine M., et al. 2000. “Cancer Patients’ Information Seeking under Threat.” Journal of Per-
Information Needs and Information Seeking sonality and Social Psychology 52 (2): 345–53.
Behaviour: In Depth Interview Study.” British Medi- Miller, Suzanne M., and Charles E. Mangan. 1983.
cal Journal 320 (7239): 909–13. “Interacting Effects of Information and Coping Style
Li, Jian. 2015. “Preferences for Partial Information and in Adapting to Gynecologic Stress: Should the
Ambiguity.” Unpublished. Doctor Tell All?” Journal of Personality and Social
Lipnowski, Elliot, and Laurent Mathevet. 2015. “Dis- Psychology 45 (1): 223–36.
closure to a Psychological Audience.” Unpublished. Mobius, Markus M., Muriel Niederle, Paul Niehaus,
Loewenstein, George. 1987. “Anticipation and the Val- and Tanya S. Rosenblat. 2014. “Managing Self-
uation of Delayed Consumption.” Economic Journal Confidence.” Unpublished.
97 (387): 666–84. Nickerson, Raymond S. 1998. “Confirmation Bias: A
Loewenstein, George. 1994. “The Psychology of Curi- Ubiquitous Phenomenon in Many Guises.” Review
osity: A Review and Reinterpretation.” Psychological of General Psychology 2 (2): 175–220.
Bulletin 116 (1): 75–98. Niehaus, Paul. 2013. “A Theory of Good Intentions.”
Loewenstein, George, and Ted O’Donoghue. 2006. Unpublished.
“‘We Can Do This the Easy Way or the Hard Way’: Nosarti, C., T. Crayford, J. V. Roberts, E. Elias, K.
Negative Emotions, Self-Regulation, and the Law.” McKenzie, and A. S. David. 2000. “Delay in Presen-
University of Chicago Law Review 73 (1): 183–206. tation of Symptomatic Referrals to a Breast Clinic:
Loewenstein, George, Don A. Moore, and Roberto Patient and System Factors.” British Journal of Can-
A. Weber. 2006. “Misperceiving the Value of Infor- cer 82 (3): 742–48.
mation in Predicting the Performance of Others.” Nyborg, Karine. 2011. “I Don’t Want to Hear About It:
Experimental Economics 9 (3): 281–95. Rational Ignorance among Duty-Oriented Consum-
Loomes, Graham, and Robert Sugden. 1982. “Regret ers.” Journal of Economic Behavior and Organiza-
Theory: An Alternative Theory of Rational Choice tion 79 (3): 263–74.
under Uncertainty.” Economic Journal 92 (368): Oster, Emily, Ira Shoulson, and E. Ray Dorsey. 2013.
805–24. “Optimal Expectations and Limited Medical Testing:
Loomes, Graham, and Robert Sugden. 1986. “Dis- Evidence from Huntington Disease.” American Eco-
appointment and Dynamic Consistency in Choice nomic Review 103 (2): 804–30.
under Uncertainty.” Review of Economic Studies 53 Paharia, Neeru, Kathleen D. Vohs, and Rohit Desh-
(2): 271–82. pande. 2013. “Sweatshop Labor Is Wrong unless the
Loomes, Graham, and Robert Sugden. 1987. “Some Shoes Are Cute: Cognition Can Both Help and Hurt
Implications of a More General Moral Motivated Reasoning.” Organizational Behav-
Form of Regret Theory.” Journal of Economic Theory ior and Human Decision Processes 121 (1): 81–88.
41 (2): 270–87. Palacios-Huerta, Ignacio. 1999. “The Aversion to the
Lord, Charles G., Lee Ross, and Mark R. Lepper. 1979. Sequential Resolution of Uncertainty.” Journal of
“Biased Assimilation and Attitude Polarization: The Risk and Uncertainty 18 (3): 249–69.
Effects of Prior Theories on Subsequently Consid- Palfrey, Thomas R. 1982. “Risk Advantages and Infor-
ered Evidence.” Journal of Personality and Social mation Acquisition.” Bell Journal of Economics 13
Psychology 37 (11): 2098–109. (1): 219–24.
Lyter, David W., Ronald O. Valdiserri, Lawrence Persoskie, Alexander, Rebecca A. Ferrer, and William
A. Kingsley, William P. Amoroso, and Charles R. M. P. Klein. 2014. “Association of Cancer Worry and
Rinaldo, Jr. 1987. “The HIV Antibody Test: Why Perceived Risk with Doctor Avoidance: An Analysis
Gay and Bisexual Men Want or Do Not Want to of Information Avoidance in a Nationally Represen-
Know Their Results.” Public Health Reports 102 (5): tative US Sample.” Journal of Behavioral Medicine
468–74. 37 (5): 977–87.
Marshall, George. 2014. Don’t Even Think About Peysakhovich, Alexander, and Uma R. Karmarkar. 2015.
It: Why Our Brains Are Wired to Ignore Climate “Asymmetric Effects of Favorable and Unfavorable
Change. New York and London: Bloomsbury. Information on Decision Making under Ambiguity.”
Maslow, A. H. 1963. “The Need to Know and the Fear Management Science 62 (8): 2163–78.
of Knowing.” Journal of General Psychology 68 (1): Pittarello, Andrea, Daphna Motro, Enrico Rubal-
111–25. telli, and Patrik Pluchino. 2016. “The Relationship
McGoey, Linsey. 2012. “Strategic Unknowns: Towards between Attention Allocation and Cheating.” Psy-
a Sociology of Ignorance.” Economy and Society 41 chonomic Bulletin and Review 23 (2): 609–16.
(1): 1–16. Ponssard, Jean-Pierre. 1976. “On the Concept of the
Meechan, Geraldine, John Collins, and Keith Value of Information in Competitive Situations.”
Management Science 22 (7): 739–47. rocessing: I. Detection, Search, and Attention.”

P
Poulsen, Anders U., and Michael W. M. Roos. 2010. Psychological Review 84 (1): 1–66.
“Do People Make Strategic Commitments? Exper- Schulz-Hardt, Stefan, Dieter Frey, Carsten Luthgens,
imental Evidence on Strategic Information Avoid- and Serge Moscovici. 2000. “Biased Information
ance.” Experimental Economics 13 (2): 206–25. Search in Group Decision Making.” Journal of Per-
Powdthavee, Nattavudh, and Yohanes E. Riyanto. sonality and Social Psychology 78 (4): 655–69.
2015. “Would You Pay for Transparently Useless Schweizer, Nikolaus, and Nora Szech. 2013. “Opti-
Advice? A Test of Boundaries of Beliefs in the Folly mal Revelation of Life-Changing Information.”
of Predictions.” Review of Economics and Statistics Unpublished.
97 (2): 257–72. Shalvi, Shaul, Francesca Gino, Rachel Barkan, and
Prat, Andrea. 2005. “The Wrong Kind of Transpar- Shahar Ayal. 2015. “Self-Serving Justifications: Doing
ency.” American Economic Review 95 (3): 862–77. Wrong and Feeling Moral.” Current Directions in
Rabin, Matthew, and Joel L. Schrag. 1999. “First Psychological Science 24 (2): 125–30.
Impressions Matter: A Model of Confirmatory Bias.” Sharot, Tali. 2011. “The Optimism Bias.” Current Biol-
Quarterly Journal of Economics 114 (1): 37–82. ogy 21 (23): R941–45.
Rayner, Steve. 2012. “Uncomfortable Knowledge: The Shiller, Robert J. 2005. Irrational Exuberance. Prince-
Social Construction of Ignorance in Science and ton and Oxford: Princeton University Press.
Environmental Policy Discourses.” Economy and Shu, Lisa L., and Francesca Gino. 2012. “Sweeping
Society 41 (1): 107–25. Dishonesty under the Rug: How Unethical Actions
Reinhart, Carmen M., and Kenneth S. Rogoff. 2009. Lead to Forgetting of Moral Rules.” Journal of Per-
This Time Is Different: Eight Centuries of Financial sonality and Social Psychology 102 (6): 1164–77.
Folly. Princeton and Oxford: Princeton University Sicherman, Nachum, George Loewenstein, Duane J.
Press. Seppi, and Stephen P. Utkus. 2016. “Financial Atten-
Roesler, Anne-Katrin. 2015. “Is Ignorance Bliss? Ratio- tion.” Review of Financial Studies 29 (4): 863–97.
nal Inattention and Optimal Pricing.” Unpublished. Simmons, Joseph P., Leif D. Nelson, and Uri Simon-
Rogers, Todd, and Avi Feller. 2016. “Discouraged By sohn. 2011. “False-Positive Psychology: Undisclosed
Peer Excellence: Exposure to Exemplary Peer Per- Flexibility in Data Collection and Analysis Allows
formance Causes Quitting.” Psychological Science 27 Presenting Anything as Significant.” Psychological
(3): 365–74. Science 22 (11): 1359–66.
Rogerson, William P. 1992. “Contractual Solutions to Simon, Herbert A. 1971. “Designing Organizations for
the Hold-Up Problem.” Review of Economic Studies an Information-Rich World.” In Computers, Com-
59 (4): 777–93. munications, and the Public Interest, edited by Mar-
Rothschild, Michael, and Joseph E. Stiglitz. 1997. tin Greenberger, 37–72. Baltimore: Johns Hopkins
“Competition and Insurance Twenty Years Later.” University Press.
Geneva Papers on Risk and Insurance Theory 22 (2): Simon, William H. 2005. “Wrongs of Ignorance and
73–79. Ambiguity: Lawyer Responsibility for Collective Mis-
Sallee, James M. 2014. “Rational Inattention and conduct.” Yale Journal on Regulation 22 (1): 1–35.
Energy Efficiency.” Journal of Law and Economics Sims, Christopher A. 2003. “Implications of Rational
57 (3): 781–820. Inattention.” Journal of Monetary Economics 50 (3):
Schelling, Thomas C. 1956. “An Essay on Bargaining.” 665–90.
American Economic Review 46 (3): 281–306. Sloof, Randolph, Hessel Oosterbeek, and Joep
Schelling, Thomas C. 1960. The Strategy of Conflict. Sonnemans. 2007. “Does Making Specific Invest-
Cambridge, MA: Harvard University Press. ments Unobservable Boost Investment Incentives?”
Schelling, Thomas C. 1987. “The Mind as a Consuming Journal of Economics and Management Strategy 16
Organ.” In The Multiple Self, edited by Jon Elster, (4): 911–42.
177–96. Cambridge and New York: Cambridge Uni- Smith, Dylan M., George Loewenstein, Aleksandra
versity Press. Jankovich, and Peter A. Ubel. 2009. “Happily Hope-
Scherer, Aaron M., Paul D. Windschitl, and Andrew R. less: Adaptation to a Permanent, but Not to a Tem-
Smith. 2013. “Hope to Be Right: Biased Information porary, Disability.” Health Psychology 28 (6): 787–91.
Seeking following Arbitrary and Informed Predic- Snow, Arthur. 2010. “Ambiguity and the Value of Infor-
tions.” Journal of Experimental Social Psychology 49 mation.” Journal of Risk and Uncertainty 40 (2):
(1): 106–12. 133–45.
Schlee, Edward E. 2001. “The Value of Information Strzalecki, Tomasz. 2013. “Temporal Resolution of
in Efficient Risk-Sharing Arrangements.” American Uncertainty and Recursive Models of Ambiguity
Economic Review 91 (3): 509–24. Aversion.” Econometrica 81 (3): 1039–74.
Schmidt, Klaus M. 1996. “The Costs and Benefits of Sugden, Robert. 2005. “Fellow-Feeling.” In Economics
Privatization: An Incomplete Contracts Approach.” and Social Interaction: Accounting for Interpersonal
Journal of Law, Economics, and Organization 12 (1): Relations, edited by Benedetto Gui and Robert Sug-
1–24. den, 52–75. Cambridge and New York: Cambridge
Schneider, Walter, and Richard M. Shiffrin. 1977. University Press.
“Controlled and Automatic Human Information Sullivan, Patrick S., Amy Lansky, and Amy Drake.
2004. “Failure to Return for HIV Test Results among van Damme, Eric. 1989. “Stable Equilibria and For-
Persons at High Risk for HIV Infection: Results ward Induction.” Journal of Economic Theory 48 (2):
from a Multistate Interview Project.” JAIDS Journal 476–96.
of Acquired Immune Deficiency Syndromes 35 (5): Van der Weele, Joel. 2014. “Inconvenient Truths:
511–18. Determinants of Strategic Ignorance in Moral
Sunstein, Cass R. 2007. Republic.com 2.0. Princeton Dilemmas.” Unpublished.
and Oxford: Princeton University Press. Wakker, Peter. 1988. “Nonexpected Utility as Aver-
Sunstein, Cass R., Sebastian Bobadilla-Suarez, sion of Information.” Journal of Behavioral Decision
Stephanie C. Lazzaro, and Tali Sharot. 2016. “How Making 1 (3): 169–75.
People Update Beliefs about Climate Change: Good Wilson, Timothy D., and Daniel T. Gilbert. 2003.
News and Bad News.” John M. Olin Center Faculty “Affective Forecasting.” In Advances in Experimen-
Discussion Paper 883. tal Social Psychology 35: 345–411.
Sweeny, Kate, Darya Melnyk, Wendi Miller, and James Woolley, Kaitlin, and Jane L. Risen. 2015. “Avoiding
A. Shepperd. 2010. “Information Avoidance: Who, Information to Protect a Strong Intuitive Prefer-
What, When, and Why.” Review of General Psychol- ence.” Unpublished.
ogy 14 (4): 340–53. Yaniv, Ilan, Deborah Benador, and Michal Sagi. 2004.
Tasoff, Joshua, and Kristof Madarasz. 2009. “A Model “On Not Wanting to Know and Not Wanting to Inform
of Attention and Anticipation.” Unpublished. Others: Choices Regarding Predictive Genetic Test-
Taylor, Shelley E., and Jonathon D. Brown. 1988. “Illu- ing.” Risk, Decision and Policy 9 (4): 317–36.
sion and Well-Being: A Social Psychological Perspec- Zaltman, Gerald. 1983. “Knowledge Disavowal in
tive on Mental Health.” Psychological Bulletin 103 Organizations.” In Producing Useful Knowledge for
(2): 193–210. Organizations, edited by Ralph H. Kilmann, Ken-
Thornton, Rebecca L. 2008. “The Demand for, and neth W. Thomas, Dennis P. Slevin, Raghu Nath, and
Impact of, Learning HIV Status.” American Eco- Lee Jerrell, 173–87. New York: Praeger.
nomic Review 98 (5): 1829–63. Zimmermann, Florian. 2014. “Clumped or Piecewise?
Tirole, Jean. 1986. “Procurement and Renegotiation.” Evidence on Preferences for Information.” Manage-
Journal of Political Economy 94 (2): 235–59. ment Science 61 (4): 740–53.
Copyright of Journal of Economic Literature is the property of American Economic
Association and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.

Journal of Economic Literature

Uploaded by

Copyright:

Available Formats

Journal of Economic Literature

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Economic Literature

Uploaded by

Copyright:

Available Formats

Journal of Economic Literature 2017, 55(1), 96–135

We commonly think of information as a means to an end. However, a growing theo-

T he standard economic analysis of deci-

(Possible) Choice of (Possible)

Figure 1. Timeline of Decision to Acquire Information

4.8 Climate Change Denial In a book about climate change denial

Psychology 23 (1): 128–41. through Ignorance.” RAND Journal of Economics 38

Management Science 22 (7): 739–47. rocessing: I. Detection, Search, and Attention.”

You might also like

Journal of Economic Literature

Uploaded by

Copyright:

Available Formats

Journal of Economic Literature

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Economic Literature

Uploaded by

Copyright:

Available Formats

Journal of Economic Literature 2017, 55(1), 96–135

We commonly think of information as a means to an end. However, a growing theo-

T he standard economic analysis of deci-

(Possible) Choice of (Possible)

Figure 1. Timeline of Decision to Acquire Information

4.8 Climate Change Denial In a book about climate change denial

Psychology 23 (1): 128–41. through Ignorance.” RAND Journal of Economics 38

Management Science 22 (7): 739–47. ­rocessing: I. Detection, Search, and Attention.”

You might also like

Management Science 22 (7): 739–47. rocessing: I. Detection, Search, and Attention.”