Effects of Causal and Noncausal Sequences of Information on Subjective Prediction

Abstract

44 subjects were presented win/loss records representing either different football teams (causally linked data) or persons betting on coin tosses (noncausally linked data) and were asked to predict how likely the next game or toss would be a win. The data (sequences of wins and losses) were arranged in a temporal order of occurrence. The major findings were that causal data were integrated across temporal positions more than noncausal data, and predictions from causal data were directly related to previous wins whereas predictions from noncausal data were inversely related to previous wins. These findings suggest that subjects make qualitatively different assumptions about the representativeness of causal and noncausal data and profoundly different predictions occur as a result of these assumptions.