The cognition/metacognition tradeoff

Integration-to-boundary is an optimal decision algorithm that takes samples of evidence until the posterior reaches a decision boundary, resulting in the fastest decisions for a target accuracy. For example, integration-to-boundary achieves faster mean-RT compared with taking a fixed number of samples that result in the same choice-accuracy. Here we show that this advantage comes at a cost in metacognitive accuracy. We show that integration-to-boundary results in less variability in evidence-integration, and is less predictive of choice accuracy. We test this in two experiments, in which all participants carried out two sessions that manipulated the response-mode protocol: free-response (evidence terminated by the subject response) vs interrogation (fixed number of evidence samples, which is the same as in the free-response session). In both sessions the participants observe a sequence of evidence frames (2/sec) and they first enter a choice and then a confidence response. As predicted, the latter protocol enhances metacognitive accuracy.