PROJECT 2 – Latent cause inference in compulsion A promising line of research has suggested that a transdiagnostic dimension of compulsive psychopathology arises from an imbalance between goal-directed and habitual systems for instrumental behavior. On this view, compulsive behaviors as in obsessive-compulsive disorder (OCD) and drug misuse share an automatic, inflexible character because they are produced by an over-dominant habit system that is goal-insensitive. Although this account has been extremely influential, it fails to explain important aspects of these disorders, such as their tight association with symptoms that suggest (paradoxically, in this view) a strong focus on specific goals, like taking actions to terminate obsessional fears or consume craved drugs. In Project 2, we leverage the latent cause inference framework to test a novel mechanism for seemingly goal- insensitive behavior linked to compulsivity. By our proposed account, rather than overreliance on habitual control, compulsivity can be understood as failure to generalize information about goal value between contexts owing to overly narrow inferences about latent causes underlying experiences. This leads to a persistence in goal value and behaviors aimed at achieving the goal, because information devaluing the goal is treated as belonging to a different latent cause and thus irrelevant to the current context (see Project 4 for devaluation experiments and Project 3 for consequences of excessive separating of latent causes for anxiety disorders). This new model preserves the successes of the habit-based account for explaining compulsive behaviors, but also extends to incorporate additional aspects of disorders of compulsivity, such as obsessions and craving. Putative habitual behavior is often measured experimentally via a signature insensitivity to changing reward contingencies, using a two-step decision task we developed. However, our recent theoretical work shows that latent cause over-separation offers a second explanation for these behaviors, leaving open the question to what extent latent cause separation contributed to previously reported associations of habits with compulsivity. Here we use the two-step decision task modified to differentiate the relative contributions of goal insensitivity and latent cause separation. Across three aims, we compare the differential association of these two mechanisms with symptoms and diagnoses in both general population (Aim 2.1) and clinical samples (OCD and alcohol use disorder, Aim 2.2). We also use fMRI to measure novel neural signatures of latent cause inference and test whether these processes are sensitive to goal-relevant symptoms using a symptom provocation manipulation (Aim 2.3). If successful, this project will provide the foundation for an entirely new way of thinking about compulsivity with potential for major impact on our theoretical understanding and treatment of this core transdiagnostic symptom. This revised view ma...