Confidence and accuracy in judgements using computer displayed information

Abstract

Confidence and accuracy in decision making are often unrelated, contrary to popular belief. In practice, confidence is often relied upon as evidence of good decision making, since the quality of a decision is difficult to determine at the time the decision is made. Information systems are increasingly used to assist decision making in organizations. Researchers believe that task, information system, and human characteristics affect the relationship between accuracy and confidence. In this research, manipulation of task, system, or human characteristics that led to an increase in confidence in decision making did not lead to an increase in decision accuracy and vice versa. In this study decision accuracy was judged by a decision process measure instead of a decision outcome measure. It was observed that subjects who had higher numerical skills than spatial skills expressed significantly more confidence in their decisions in a problem solving task; however, decision accuracy scores were not significantly better for subjects with higher numerical skills. Thus, these subjects expressed overconfidence in their decisions. Subjects using graphical displays (instead of tabular displays) also expressed similar overconfidence in their decisions. On the other hand, when subjects were given a less complex task, their decision accuracy scores were significantly better than subjects who had a more complex task. No significant difference was found in confidence expressed in their decisions, and therefore subjects performing a less complex task expressed underconfidence in their decisions. These results and others in the paper suggest that confidence is a poor surrogate for accuracy in decision making. Sophisticated software interfaces, like graphical information displays, could lead to increased confidence in decision making without significant improvement in the quality of decisions made.