Towards a mechanistic understanding of decision confidence.

Every day we have to make numerous decisions, often based on imperfect information. Some decisions are very important, for example deciding whether it is safe to cross a busy street. Our decisions are typically accompanied by a sense of confidence in having made the correct choice. Making accurate self-evaluations is of utter importance, because confidence can be used to adapt behavior. When having low confidence in the decision that it is safe to cross the street, we will be extra cautious. Existing research on confidence has mostly tested “what” a confidence report reflects. Models have been proposed that can explain confidence reports, and these models have been linked to neural activity. However, what is currently lacking is a mechanistic understanding of confidence. How do humans learn to accurately self-evaluate their decisions? The big challenge for this field is therefore to unravel the “how” of confidence. To address the “how” question, I propose a model according to which humans learn to self-evaluate based on external feedback (Objective 1), and I will use this model to explain biases in confidence (Objective 2). Answering these questions is of critical importance, because understanding how humans learn to be confident also explains how suboptimal confidence arises. A mechanistic understanding of such sub optimality can aid at counteracting it. This has real-life consequences, given that confidence affects many choices, such as whether or not to gamble and what to purchase.