Categorical Confidence

Abstract

People tend to be inadequately sensitive to the extent of their own knowledge. This insensitivity typically emerges as overconfidence. That is, people’s assessments of the probability of having answered questions correctly are typically too high compared to the portion of questions they get right. Few debiasing procedures have proven effective against this problem. Those that have worked seem to be directive in character. Rather than improving subjects’ feeling for how much they know, such procedures may have suggested to subjects how their probability assessments should be changed. These successful manipulations include giving feedback and requiring subjects to provide reasons contradicting their chosen answers. The present study attempted to improve the appropriateness of confidence with a nondirective method. Subjects were asked to sort items into a specified number of piles according to their confidence in the correctness of their answers. Subsequently, they assigned a number to each pile expressing the probability that each item in the pile was correct.

(Report No. 81-10)