Towards a mechanistic understanding of confidence

Evaluating whether a decision is accurate or not is of critical importance. Consider taking extra caution when being unsure whether it is safe to cross a busy street. Hitherto, it has mostly been studied what such evaluations - often referred to as decision confidence - reflect. A growing consensus is that confidence reflects the probability of being correct. It remains unclear, however, how participants learn to make such self-evaluations. In this project, I propose that humans learn how to produce accurate confidence judgments based on feedback from the environment. The notion that confidence is learned by comparing it to external feedback puts a lot of weight on the availability of such feedback. Indeed, my proposed model predicts that under- and overconfidence can be experimentally induced by manipulating feedback. Finally, a major asset of my proposed model is that it allows generalization to different tasks and contexts. This can explain why humans are able to evaluate and compare their decisions across novel tasks and contexts. In three work packages, these predictions will be tested on healthy human participants. By collecting electroencephalographic recordings (EEG), the brain mechanisms will be unravelled that are involved in learning to self-evaluate decisions.