Regressive Judgment

Abstract

In the context of research on human judgment, regression is commonly treated as an artifact or an unwanted consequence of ill-controlled research designs. We argue that this negative image is undeserved. Regression affords not only an enlightening statistical construct but also a theoretical construct that can inspire novel research. It offers alternative accounts for a variety of well-known judgment biases, and it has inspired novel predictions of previously unknown biases. While all judgments that are less than perfectly correlated with a conditional variable must be regressive, systematic biases arise when different judgment targets are subject to unequally strong regression effects. This occurs when the information given about different targets varies in statistical uncertainty or extremity, the two determinants of regression. We illustrate the explanatory power of the regression construct for judgment biases in three paradigms: frequency estimations, performance evaluations, and illusory correlations. Because regression is a universal property of the empirical world, explanations of biases in terms of cognitive or motivational biases are insufficient and incomplete if the impact of regression is not taken into account.