Uncertainty monitoring in reasoning: Cue consistency is more important than belief-logic conflict.

Dual process accounts of reasoning posit that the interactions between automatic Type 1 processes and reflective Type 2 processes are regulated through the mechanism of conflict detection. Conflict detection is thought to be triggered when two or more incompatible responses to a problem are produced. Conflict detection can be viewed as an aspect of metacognition, as both are concerned with monitoring the outputs from the reasoning process as assessing one's certainty in them. In this paper we apply the self-consistency model of metacognition to analyse cues that inform confidence judgements in reasoning, to investigate the potential sources of responses to a given problem. In four experiments (total N = 524), we analysed the relationship between confidence and accuracy on the base-rate task and syllogisms, and used a solution-evaluation task format to manipulate uncertainty experimentally. We found that in both tasks, confidence and accuracy were influenced by many cues, rather than belief-logic conflict exclusively. Belief-logic conflict was a significant cue to confidence in the base-rate task, but not in syllogisms. Across all experiments, items identified as eliciting the most uncertainty, based on the self-consistency model, showed greater conflict detection effects. Introducing an additional cue that was inconsistent with the most common response generally increased uncertainty and conflict detection. These findings support the uncertainty monitoring account of conflict detection and link it with the broader metacognition research.