Task predictability and performance as determinants of confidence in multiple-cue judgments

Two experiments were performed in order to study judges' confidence in their judgments. Experiment I showed that judges' confidence is a direct function of the cue intercorrelation. r_ij in a pure judgmental task. When judges recieved feedback, the effect of r_ij on confidence was, as predicted, reduced. Confidence was, however, systematically related to neither R ²_c nor performance in the feedback condition. Experiment II was a further study of the effect of feedback on confidence. In addition, the hypothesis that the lack of relation between performance and confidence is due to judges' poor knowledge of how they actually perform in probabilistic inference tasks was tested. The experiment showed that differences among judges' confidence is a direct function of task predictability and that judges' confidence is related to how they believe that they perform rather than to how they actually perform. The theoretical as well as the practical importance of studies of confidence in probabilistic inference tasks was discussed.