Variability in prior expectations explains biases in confidence reports

Abstract

Confidence in a decision is defined statistically as the probability of that decision being correct. Humans, however, display systematic confidence biases, as has been exposed in various experiments. Here, we show that these biases vanish when taking into account participants' prior expectations, which we measure independently of the confidence report.We use a wagering experiment to show that modeling subjects' choices allows for classifying individuals according to their prior biases, which fully explain from first principles the differences in their later confidence reports. Our parameter-free confidence model predicts two counterintuitive patterns for individuals with different prior beliefs: pessimists should report higher confidence than optimists, and, for the same task difficulty, the confidence of pessimists should increase with the generosity of the task. These findings show how systematic confidence biases can be simply understood as differences in prior expectations.A level of confidence accompanies all of our decisions [1]. This sense of confidence can be reported explicitly, or implicitly through behavioral markers such as the amount of time willing to wait to obtain a response [2], reaction times [3], or the predisposition to wage [4] or opt-out of the task for a lower but safe reward [5]. The use of such implicit measures has shown that a sense of confidence is present even in rodents and nonhuman primates (see [6] for a review). A quantitative approach to confidence helps formalize the concept and unify its different manifestations. In statistical decision theory, the normative definition of decision confidence is the probability of the choice being correct [7][8][9][10].