A Lens Model Analysis of Confidence Judgments: Beyond Calibration Measures

Abstract

To better understand the judgment processes leading to an unjustifiably high level of confidence, the present study tested the impact of task characteristics as determinants of overconfidence, using a Lens Model analysis. Participants were asked to make a repair judgment about a pavement crack, and then rate the confidence in their judgments. Two, 2 level between-participants conditions (i.e., high/low environmental predictability and yes/no confidence judgment training), and one 4 level within-participants condition (i.e., task familiarity) were used as independent variables. Lens Model parameters and calibration measures were used as dependent variables. Results showed that similar to results in prior studies, participants tended to be overconfident, in which their level of overconfidence was greater in a less predictable (i.e., more difficult) task environment. These effects were only somewhat mitigated by confidence judgment training, not being affected by increased task familiarity. An analysis of Lens Model parameters indicated that participants who were more overconfident, may have been reflected in their assessments of confidence, their belief that they were making judgments consistently.