Affect, empathy, and regressive mispredictions of others' preferences under risk

Making effective decisions under risk often requires making accurate predictions of other people's decisions under risk. We experimentally assess the accuracy of people's predictions of others' risky choices. In four studies, we find evidence of systematic inaccuracy: predictions of others' choices are too regressive. That is, people predict that others' choices will be closer to risk neutrality than those choices actually are. Where people are risk seeking, they predict that others will be risk seeking but substantially less so; likewise, where people are risk averse, they predict that others will be risk averse but substantially less so. Put differently, people predict that others' choices will reveal a more muted form of prospect theory's fourfold pattern of risk preferences than actually prevails. Two psychological concepts, the notion of risk-as-feelings and of an empathy gap, help account for regressive mispredictions. We explore several debiasing techniques suggested by these notions and also find that self-reported ratings of empathy moderate the magnitude of regressive mispredictions.