Compulsivity is linked to suboptimal choice variability but unaltered reinforcement learning under uncertainty

Compulsivity has been associated with variable behavior under uncertainty. However, previous work has not distinguished between two main sources of behavioral variability: the stochastic selection of choice options that do not maximize expected reward (choice variability) and random noise in the reinforcement learning process that updates option values from choice outcomes (learning variability). Here we study the relation between dimensional compulsivity and behavioral variability using a computational model that dissociates its two sources. Across two independent datasets (137 and 123 participants), we found that compulsivity is associated with more frequent switches between options, triggered by increased choice variability, but no change in learning variability. This effect of compulsivity on the ‘trait’ component of choice variability is observed even in conditions where this source of behavioral variability yields no cognitive benefits. These findings indicate that compulsive individuals make variable and suboptimal choices under uncertainty, but do not hold degraded representations of option values. Lee et al. find that compulsivity is associated with choice variability under uncertainty, resulting in frequent switching between choice options but no alteration in the ability to learn from the positive or negative outcomes of these choices.