And indeed, there is a vast literature on decision-making under uncertainty (see recent

reviews in Gilovich, GriYn, & Kahneman, 2002) that describes (and prescribes) a variety of
strategies to this end. It appears, however, that uncertainty is not a unitary mental state
(Kahneman& Tversky, 1982). This is important because the same level of uncertainty may
activate diVerent strategies depending on how it is interpreted. For example, it matters
whether uncertainty is quantiWed or not (e.g. Tversky & Fox, 1995), is about the magnitude
of the probability or the outcome (Harel & Segal, 1999; Keren & Gerritsen, 1999), is
attributed to chance or luck (Friedland, 1998), and whether it is about an event or about ones
beliefs about the event (Teigen, 1994; Wells, 1992). The research reported here explores
another characteristic of uncertainty that has received little, if any, attentionnamely, its
source. Does it matter whether uncertainty is due to the complexity of the world or to the
complexity of human psychology? To explore this question, our research compares peoples
reactions to uncertainty when it is attributed to chance or deception. Einhorn (1986)
distinguished between two orientations to forecasting, clinical and statistical (see Grove &
Meehl, 1996, for a review of the debate about the eVectiveness of statistical versus clinical
predictions). In clinical forecasting the all-important goal is a sense of understanding that
ensues when error is eliminated. Zero error means the decision-makers account for every
detail, large and small. To do so they attempt to construct a coherent story integrating

everything unselectively. Likelihood estimates, therefore, depend on the goodness of the

story, so that a detailed scenario consisting of what appears to be a causally linked chain of
events can give rise to a higher likelihood estimate than a scenario containing a subset of
these events (Kahneman & Lovallo, 1993; Tversky & Kahneman, 1983). This kind of
thinking is reminiscent of explanation-based decision- making models (Pennington & Hastie,
1992; Wells, 1992) in the sense of emphasizing individuals attempts to understand the
situation by imposing a coherent narrative on it (Bruner, 1986). Statistical prediction, in
contrast, acknowledges the forecasters ignorance. Rather than trying to Wt everything one
knows about the event into a coherent story, a forecaster with a statistical orientation is
willing to be selective, to consider only a subset of the features of the to-be-predicted event,
treating them as essential or deWning characteristics, and thus implying that the other
features are non-essential or accidental. This allows the forecaster to treat a large number of
events as equivalent (cf., the external view in Kahneman & Lovallo, 1993). In so doing, the
forecaster is willing to tolerate error (i.e., treat non-identical events as equivalent) to reduce
error (Einhorn, 1986).
The process of choosing and the subsequent decision outcome may make us happy or sad.
The emotion most studied in this respect is regret, and a basic Wnding is that people make
regret-averse decisions that shield themselves from the possibility of regret. Consider, for
example, a decision-maker who can choose between (a) receiving $100 and (b) Xipping a
coin that will yield $200 or nothing. If the decision-maker decides to Xip the coin and loses,
she may regret not taking the $100. This possibility of regret may be avoided by opting for
the safe $100 and not Xipping the coin (for a review, see Zeelenberg, 1999). In particular,
post-decisional information search tends to be biased in favor of information that supports the
decision (Festinger, 1957, 1964; Frey, 1986). And when people are uncertain about whether
feedback will be supportive, they prefer to avoid it completely (Northcraft & Ashford, 1990).

Curiosity and regret aversion may constitute two opposing forces in decision-making under
uncertainty. Whereas regret aversion may induce decision-makers to avoid uncertain
information, curiosity essentially entails attraction to uncertain situations. Some time ago, one
of the authors of this article witnessed an intriguing scene at the local market. A merchant

was selling wrapped packages to a crowd of people who anxiously waved their Euros. What
was the content of the package? Nobody knew! As the merchant said, the only way to know
the content was to buy the package. He only gave some vague hints about the content, such as
its a useful item and its handy. Apparently, this knowledge was suYcient to induce
curiosity and a willingness to buy. This experience suggests that curiosity may induce people
to choose for uncertain options that they might regret, and thus appears to be in conXict with
the documented reluctance to avoid decisions that might yield regret (e.g., Larrick & Boles,
1995; Ritov, 1996; Zeelenberg, Beattie, van der Pligt, & de Vries, 1996).

In the typical decision-making study participants choose

between a certain and an uncertain option, knowing that if

they choose the certain option, they will never learn the

outcome of the uncertain option. Feedback on the outcome

of the uncertain option is thus conditional on choosing it.

Participants may therefore protect themselves from regret

by choosing the certain alternative. Often, however, we

eventually learn the outcomes of uncertain alternatives

even if we opt for certain alternatives. For example, even if

you do not invest your money in stocks, you are bound to

Wnd out whether or not stocks went up. Hence, you may

often experience regret even if you dont play the game.

What would happen in our sealed-package paradigm if participants

know that they will eventually learn the content of

the package, even if they dont choose it? To answer this

question, it is instructive to separately elaborate on what

eVect such unconditional feedback would have if the main

motivation would be to minimize regret, and what if the

main motivation was to satisfy curiosity.

We

noted that after having made a decision, people may block

out information that might indicate that they made the

wrong decision. Put diVerently, they selectively expose

themselves to information by not attending to information

that would not be supportive of their decision (cf.

Festinger, 1957, 1964; Frey, 1986). Blocking out information

may, however, be easier said than done. In particular,

it has been suggested that people Wnd it very diYcult to

ignore information that is already out there. Anecdotal

evidence comes from the parents increased curiosity to

know the sex of their unborn child the instant this information

is known to the doctor.