Institute for Mathematical Behavioral Sciences

Individuals and organizations learn from experience by increasing the probability of sampling alternatives with favorable past outcomes. Such adaptive sampling is sensible but leads to an asymmetry in experiential learning. Because decision makers continue to sample alternatives they believe are good, errors of overestimating the value of an alternative are likely to be corrected. Because decision makers may avoid alternatives they believe to be poor, errors of underestimation are less likely to be corrected. In this talk, I show how a behavioral model of learning which incorporates this asymmetry can offer an alternative account of several biases in impression formation and decision making, including risk aversion, ingroup bias, and social influence.