The effects of arguments on realism in confidence judgements

In this study we analyzed the effects of arguments on confidence ratings of answers to general knowledge questions. In two experiments subjects answered general knowledge questions and rated their confidence in the selected answers. Before each confidence rating the subjects were given a prepared argument(s) for and/or against the chosen answer. The four experimental conditions, each contrasted with a within-subject control condition, varied with respect to the type of arguments given to the subjects (for or against the chosen answer, with or without a further argument against the first argument). In general, arguments tended to result in improved calibration but in an increased overconfidence. However, these trends were only significant for the increase in overconfidence in one of the experimental conditions. Experiment 2, using a within-subject design, compared self-generated arguments, given arguments, and no arguments against the chosen answer alternative with respect to their influence on the subjects' confidence ratings. The results showed no difference in the realism of subjects' confidence ratings between the three conditions although subjects rated the given arguments as stronger in comparison with the arguments they had generated themselves. Our results suggest that arguments, whether given a to the subjects or subject generated, have no clear influence on the realism of subjects' confidence ratings.