Decision time is a crucial factor in the computation of confidence judgments during decision-making. We investigated the role of accumulation time in sequential sampling models of confidence, examining optimal confidence computation in a Bayesian observer model and fitting dynamical confidence models to empirical data.
The Bayesian analysis revealed that an optimal observer discounts the final available evidence by the accumulation time. In addition, if there is evidence about task difficulty, which is independent of the stimulus category, optimal confidence relies on this information as well.
We introduce the dynamical visibility, time, and evidence model (dynaViTE), which captures the significance of accumulation time in confidence determination, and compared its performance with other sequential sampling models. DynaViTE outperformed alternative sequential sampling models of confidence, providing the best fit to observed data. This suggests that human observers effectively utilize decision time as a key factor in confidence computation.
While the study emphasizes the importance of accumulation time, it is still an open question, whether the precise calculations involved in this process in human decision-making adhere to Bayesian principles or follow heuristics. This uncertainty invites future research to unravel the intricate mechanisms behind how humans integrate decision time in the computation of confidence.