Underconfidence in predicting future events

Abstract

In calibration research using future event questions, the mean of the confidence ratings assigned to each of a series of predictions is compared with the percentage of correct predictions. A person is said to be well calibrated if his/her mean confidence rating matches his/her percentage correct. In this study the temporal setting of future event questions was manipulated in order to investigate its effect on subject calibration. The subjects assigned confidence ratings to answers to questions concerning events located in the near future. Half of the questions covered a period of the next 4 days (early events); the other half covered a period of the next 14 days (late events). For both types of questions, underconfidence was found; the mean confidence rating was lower than the mean percentage correct. Although no difference in calibration was found between the two types of questions, the subjects were less confident about the late events than about the early events. These results are compared with those of some other studies in this area, and discrepancies and similarities are discussed. © 1988, The Psychonomic Society, Inc.. All rights reserved.