Past research has found that interruptions change the relationship between confidence and accuracy. However, it is unclear how interruptions affect confidence-accuracy calibration. In this study, we used a rule-based procedural task called UNRAVEL and compared confidence-accuracy calibration between interrupted and uninterrupted trials. Results showed that participants were better calibrated in the interruption condition than in the no interruption condition. We interpret this novel effect as a result of changes in the validity of internal cues for confidence between conditions. Specifically, we explore response time as one potential mediating factor.

The relationship between confidence and accuracy has been modeled many times. This paper compares and contrasts three decision-making mathematical models (2DSD, Poisson, RTCON2) of confidence and investigates how each model predicts the effects of interruptions on accuracy, decision response time, confidence, and confidence response time.

Drift diffusion models of decision-making offer some of the most robust predictions of response time for a number of memory manipulations. Some drift diffusion models have been extended to explain confidence judgments. Many of these models assume that confidence judgments are independent and are not systematically related to other task items. In this paper the authors report a relationship between confidence judgments in procedural tasks and how the Memory for Goals model would explain this relationship.