Cognitive distancing is a therapeutic technique commonly used in psychological treatment of various mental health disorders, but its computational mechanisms remain unknown. To determine the effects of cognitive distancing on computational learning mechanisms, we use an online reward decision-making task, combined with reinforcement learning modelling in 935 participants, 49.1% of whom were trained to regulate their emotional response to task performance feedback. Those participants practicing cognitive distancing showed heightened learning from negative events as well as an increased integration of previous choice values. These differences seemed to represent an evolving shift in strategy by the distancing participants during the task, from exploiting optimal choices earlier in the task (as indicated by greater inverse temperature parameters), to a late-stage increase in learning from negative outcomes (represented as higher loss learning rates). Our findings suggest adaptive changes in computational learning mechanisms underpin the clinical utility of cognitive distancing in psychological therapy.