Constraints on the Calibration of Performance

Abstract

We examined the effect of experimenter-controlled incentives and feedback on the calibration of performance. Subjects answered 36 reading comprehension and 8 mathematical multiple-choice questions and rated the accuracy of their responses. Perfect calibration was possible only when true and estimated test performance were approximately equal. Incentives for improved performance (i.e., doubling the credit people received for correct answers) adversely affected performance and calibration compared to the same incentives for improved calibration (i.e., doubling credit for minimizing the error between true and estimated performance). Feedback had no effect on performance or accuracy nor did it interact with the incentive variable. An examination of coefficient α suggested a strong response bias by individuals when calibrating their performance; individuals tended to rate their performance accuracy consistently regardless of item difficulty or whether they answered the item correctly. Educational implications were discussed.