Computational modeling of reinforcement learning and functional neuroimaging of probabilistic reversal dissociates compulsive behaviors in Gambling and Cocaine Use Disorders

Background Individuals with both Cocaine Use Disorder (CUD) and Gambling Disorder (GD) demonstrate impairments in cognitive flexibility (CF), the essential ability to adapt to changes in the environment. CF is commonly assessed in a laboratory setting using probabilistic reversal learning (PRL), which involves reinforcement learning (RL), the process by which feedback from the environment is used to adjust behavior. Aims It is poorly understood whether impairments in CF differ between individuals with CUD and GD, and how this is instantiated by the brain. We applied computational modelling of RL to gain a deeper mechanistic explanation of the latent processes underlying CF across two disorders of compulsivity. Methods We present a re-analysis of PRL data from individuals with either GD (n=18) or CUD (n=20), as well as control participants (n=18), using a hierarchical Bayesian RL approach. Furthermore, we relate behavioral findings to their underlying neural substrates through an analysis of task-based fMRI data. Results We observed lower ‘stimulus stickiness’ in GD. We also report differences in tracking expected values (EV) in individuals with GD compared to controls, with greater activity during reward EV tracking in the cingulate gyrus and amygdala. In CUD, we observed lower responses to positive punishment prediction errors (PPE) and greater activity following negative PPEs in the superior frontal gyrus compared to controls. Conclusions Using a computational approach to RL and CF, we show that individuals with GD and CUD differed in their perseverative tendencies and in how they tracked value neurally, which has implications for psychiatric classification