Base rates, contingencies, and prediction behavior.

ABSTRACT

A skew in the base rate of upcoming events can often provide a better cue for accurate predictions than a contingency between signals and events. The authors study prediction behavior and test people's sensitivity to both base rate and contingency; they also examine people's ability to compare the benefits of both for prediction. They formalize these notions and propose a new measure of the regularity in the environment (ExpPA). In two experiments they test whether the notions underlying this measure capture prediction behavior. In the first experiment, they compare participants' prediction behavior, preference, and assessment of contingencies in two data sets that differ only in their base rate. In the second, in which the contribution of contingency over base rate is manipulated, they study participants' willingness to forgo a costly predictor. Results indicate a close correspondence between ExpPA and behavior.