Abstract: How do we know if we performed well on a task? While metacognition - our ability to evaluate our own cognitive processes - has been extensively studied through local confidence (trial-by-trial accuracy assessments), real-world decisions often require a broader perspective: global confidence, a higher-order evaluation of overall performance. Despite its importance in controlling behavior and decision-making, the cognitive mechanisms shaping global confidence remain largely unknown. Here, we introduce a novel approach to studying how global confidence is formed in two vastly-studied cognitive domains: memory and perception. In both domains, participants completed decision-making tasks in blocks composed of two different item sets and expressed their global confidence at the end of each block through a choice between item sets. By investigating the relationship between global confidence and local accuracy, difficulty, response times (RT) and local confidence, we reveal the key drivers of global self-evaluation. We show that global confidence judgments rely on different information across cognitive domains: in memory tasks, both accuracy and local confidence contributed to global confidence, whereas in perceptual tasks, local confidence alone played a dominant role. Notably, neither fluctuations in RTs nor objective difficulty influenced global confidence beyond fluctuations in accuracy and local confidence, despite their well-established links to local confidence. By shifting the focus from trial-based monitoring to the construction of global confidence, our study provides new insights into self-evaluation, decision-making, and interindividual differences, paving the way for applications in education, mental health, and metacognitive interventions.

Location: JK 24/22