Disorders of compulsivity: a common bias towards learning habits

TLDR

Decision making involves goal‑directed and habitual systems driven by model‑based and model‑free learning, and the habitual system may underlie pathological fixedness, raising the question of why we repeat harmful choices. The authors used a decision task that quantifies model‑based versus model‑free learning to demonstrate a bias toward habit acquisition in binge eating, methamphetamine use, and OCD. The bias toward model‑free learning is linked to repetitive behaviors and reduced gray matter in caudate and medial orbitofrontal cortex, suggesting a shared neurocomputational dysfunction across compulsive disorders.

Abstract

Why do we repeat choices that we know are bad for us? Decision making is characterized by the parallel engagement of two distinct systems, goal-directed and habitual, thought to arise from two computational learning mechanisms, model-based and model-free. The habitual system is a candidate source of pathological fixedness. Using a decision task that measures the contribution to learning of either mechanism, we show a bias towards model-free (habit) acquisition in disorders involving both natural (binge eating) and artificial (methamphetamine) rewards, and obsessive-compulsive disorder. This favoring of model-free learning may underlie the repetitive behaviors that ultimately dominate in these disorders. Further, we show that the habit formation bias is associated with lower gray matter volumes in caudate and medial orbitofrontal cortex. Our findings suggest that the dysfunction in a common neurocomputational mechanism may underlie diverse disorders involving compulsion.