Computational Theories of Alcohol Use Disorder: Mapping Learning and Choice Mechanisms on Symptoms

Abstract

Alcohol use disorder (AUD) is characterized by a combination of symptoms including excessive craving, loss of control, and progressive neglect of alternative pleasures. A mechanistic understanding of what drives these symptoms is needed to improve diagnostic stratification and to develop new treatment and prevention strategies for AUD. To date, there is no consensus regarding a unifying mechanistic framework that accounts for the different symptoms of AUD. Reinforcement learning (RL) and economic choice theories may be key to elucidating the underlying processes of symptom development and maintenance in AUD. These algorithms may account for the different behavioral and physiological phenomena and are suited to dissect mechanisms linked to different symptoms of AUD. We here review different RL and economic choice models and how they map onto three symptoms of AUD: (1) cue-induced craving, (2) neglect of alternative rewards, and (3) consumption despite adverse consequences. For each symptom and theory, we describe findings from animal and human studies. In humans, we focus on empirical studies that investigated RL models in the context of treatment outcome in AUD. The review indicates important gaps to be addressed in the future by highlighting the challenges in transferring findings from RL and economic choice studies to clinical application. We also critically evaluate the potential and pitfalls of a symptom-oriented approach and highlight the importance of elucidating the role of learning and decision-making processes across diagnostic boundaries.