Iterated decision making can be studied in laboratory using situations, like the Iowa Gambling Task (IGT), in which participants face repeatedly the same decision problem getting feedback after each choice. In the paper we focus on a recurring
finding in experiments carried out with the IGT, the frequency
of the contingent event effect—i.e., the fact that people consistently
prefer options associated with rare losses, independently
of their attractiveness, expected value and loss magnitude—
that has not yet received a satisfactory explanation. An experiment reveals that the effect relies on simply experiencing
rewards and punishments, not being influenced by the net outcome (loss or win) to which they are associated, and a computational model, implemented in the ACT-R cognitive architecture,
corroborates the idea that punishments and losses on one
hand, and rewards and wins on the other, play the same functional
role in determining the participants’ behavior in IGT.
