The effect of performance feedback on metacognitive processes: causal analysis-based approach using a Bayesian hierarchical Signal Detection Theory model

This study aims to answer two basic questions about metacognition: whether and how performance feedback influences metacognitive monitoring and regulation. We address these questions using a rigorous and novel approach: the study design and statistical analyses are justified by explicit causal assumptions using causal-theoretical principles. In two experiments with preregistered hypotheses, we tested the causal effects of the presence and value of feedback on decision confidence and response times in a perceptual decision task. Participants made a series of perceptual decisions and rated their confidence. Decision-accuracy feedback, when given, was presented at the end of the trial. We manipulated the presence of veridical feedback in Experiment 1 and the value of randomised feedback in Experiment 2. To estimate the tendency to report high or low confidence we derived a novel measure based on a Bayesian hierarchical Signal Detection Theory model. The results show that both the presence and value of accuracy feedback affects decision confidence and response times. Confidence is higher following veridical feedback and following positive compared to negative feedback (veridical or not). Response times are shorter after positive compared to negative feedback. However, the more unexpected the positive feedback the slower confidence ratings. We discuss these results in the context of the relation between external accuracy feedback and metacognitive processes.