The Effect of Knowledge on the Calibration of Probability Assessments.

摘要

One way to assess the validity of a set of subjective probability judgments is to examine their degree of calibration. The perfectly calibrated judge assigns probabilities so that of all the propositions assigned a probability of .XX of being true, XX% are in fact true. For example, half of the propositions given a .50 chance of being true should in fact be true. A series of experiments revealed that: (1) Although people are moderately well calibrated, their probability judgments are prone to systematic biases. The most common bias is overconfidence; (2) People are differently calibrated when dealing with items of varying degrees of difficulty; (3) Calibration is unaffected by differences in intelligence, expertise, subjects' reliance on extreme probability responses, and at least some aspects of the context in which items are presented. The implications of these results for decision makers are discussed. (Author)