On the ability of monitoring non-veridical perceptions and uncertain knowledge: Some calibration studies

The overall evidence suggests that people are indeed less likely to be overconfident in a perceptual task, especially for tasks in which little or no additional processing is required and performance is limited only by perceptual variables.

Abstract

There is ample evidence in the literature to suggest that people are poor probability assessors, namely they are miscalibrated and usually exhibit overconfidence (Lichtenstein, Fischhoff and Phillips 1982). Recently, Dawes (1980) proposed that overconfidence is particularly apparent in tasks that require 'intellectual' judgments (such as general knowledge questions). In contrast, Dawes argued, we have developed a remarkable perceptual encoding system, which is not vulnerable to the same limitations as our 'intellectual' judgments. Three experiments were designed to test the ability of self-monitoring visual perceptual tasks in terms of confidence ratings. The overall evidence suggests that people are indeed less likely to be overconfident in a perceptual task. This holds especially for tasks in which little or no additional processing, beyond sensory encoding, is required and performance is limited only by perceptual variables ('state limitation' tasks in terms of Garner (1970, 1974)). Perceptual tasks that require additional processing of abstraction and discrimination ('process limitation' tasks in Garner's terminology) may be vulnerable to overconfidence. Possible explanations for the differences in confidence ratings for knowledge as compared with perceptual experience, are briefly discussed.