Abstract
Making a decision is invariably accompanied by a sense of confidence in that decision. Across subjects and tasks, there is widespread variability in the exact level of confidence, even for tasks that do not differ in objective difficulty. Such expressions of under- and overconfidence are of vital importance, as they relate to fundamental life outcomes. Yet, a computational account specifying the mechanisms underlying under- and overconfidence is currently missing. Here, we propose that prior beliefs in the ability to perform a task, based on prior experience with this or a similar task, explain why confidence can differ dramatically across subjects and tasks, despite similar performance. In two experiments, we provide evidence for this hypothesis by showing that manipulating prior beliefs about task performance in a training phase causally influences reported levels of confidence in a test phase, while leaving objective performance in the test phase unaffected. This is true both when prior beliefs are induced via manipulated comparative feedback and via manipulating task difficulty during the training phase. We account for these results within an accumulation-to-bound model by explicitly modeling prior beliefs based on earlier exposure to the task. Decision confidence is then quantified as the probability of being correct conditional on these prior beliefs, leading to under- or overconfidence depending on the task context. Our results provide a fundamental mechanistic insight into the computations underlying under- and overconfidence.
Competing Interest Statement
The authors have declared no competing interest.
