Pathologies of precision: A Bayesian account of goals, habits, and episodic foresight in addiction

The brain is thought to implement two decision-making systems: a goal-directed system in which decisions are made through planning on the basis of action-outcome relationships, and a habitual system in which behaviour reflects stimulus-response associations. A prominent theory of addiction sees it as arising due to an extreme dominance of habit over goal-directed action. The balance between these systems is thought to be arbitrated by the relative precision of their separate predictions of reward. In this paper, we argue that various factors in addiction create hyper-precise reward predictions in the habitual system and hypo-precise reward predictions in the goal-directed system, shifting the balance of behavioural control in favour of habit. Based on this, we offer a theoretical account of the utility of episodic future thinking in addiction, interpreting it as increasing the precision of reward estimates in the goal-directed system, thereby enhancing the control of this system over behaviour.