Prior Information Shapes Perceptual Confidence.

Abstract

Decisional confidence refers to the subjective evaluation of the accuracy of a decision based on sensory information. While these judgments are typically grounded in the strength of evidence leading to a decision, they are also subjected to influence from top-down factors such as prior expectations. Previous research has highlighted the impact of prior information on decision parameters such as reaction times and decision criteria placement. However, a comprehensive understanding of how prior information shapes confidence ratings is still lacking. In this study, we manipulate prior knowledge by inducing varying levels of target probability expectation (low: 33%, random: 50%, high: 67%) in a perceptual detection task. In each trial both type-1 (detection) and type-2 (confidence) responses were recorded. First, we replicate previous findings, demonstrating that decisional priors impact decision criteria but not task sensitivity. Secondly, we reveal the strong effect that prior expectations exert on type-2 decisions, with this influence being moderated by a congruency effect between the given prior, the actual stimulus presented, and the provided response. Moreover, we find that confidence is higher in correct compared to incorrect responses, with low-probability trials leading to higher confidence judgments in correct choices compared to random and liberal trials. Finally, we unveil that prior-dependent modulation rates in criterion and confidence were positively associated. These results underscore the intricate interplay between prior expectations, decision-making, and confidence levels, demonstrating that what we perceive is not solely a data-driven phenomenon but can be already shaped by the available information about the state of the world.