|M.Sc Student||Benjamin Shinar|
|Subject||On the Impact of RareEevents on Categorization Decisions|
|Department||Department of Industrial Engineering and Management||Supervisor||Full Professor Erev Ido|
Studies of decisions from experience in abstract choice tasks reveal a bias towards underweighting of rare events. The common deviations from maximizations reflect a tendency to select the option that gives the best payoff in most cases even when this option is associated with lower expected return (because the alternative option is much better in rare cases). The current research examines if a similar pattern emerges in perceptual and categorization decisions. It starts with a review of previous studies of the effect of rare events on perceptual and categorization decisions, and then presents a new study that compares alternative explanations for the results of these studies. Each participant, in the new study, faced four distinct (numerical and perceptual) categorization tasks under two incentive structures (rare treasure and rare disaster payoff matrixes). The results reveal a strong bias toward underweighting of rare events. In contradiction to the predictions of the optimal strategy, the proportion of missing the rare but important events was higher than the false alarm rates (incorrect identification of the common event as rare) in all eight blocks. This behavior is consistent with the underweighting bias. It seems that the bias toward underweighting of rare events in the current setting is larger than the bias observed in abstract choice tasks. In addition, the results show low perceptual discrimination in all but one. These findings favor models that assume signal contingent choice rules over models that assume learning of cutoff thresholds. The theoretical and practical implications of the results are discussed.