Abstract:Clinical and experimental observations show individual differences in the development of addiction. Increasing evidence supports the hypothesis that dopamine receptor availability in the nucleus accumbens (NAc) predisposes drug reinforcement. Here, modeling striatal-midbrain dopaminergic circuit, we propose a reinforcement learning model for addiction based on the actor-critic model of striatum. Modeling dopamine receptors in the NAc as modulators of learning rate for appetitive--but not aversive--stimuli in the critic--but not the actor--we define vulnerability to addiction as a relatively lower learning rate for the appetitive stimuli, compared to aversive stimuli, in the critic. We hypothesize that an imbalance in this learning parameter used by appetitive and aversive learning systems can result in addiction. We elucidate that the interaction between the degree of individual vulnerability and the duration of exposure to drug has two progressive consequences: deterioration of the imbalance and establishment of an abnormal habitual response in the actor. Using computational language, the proposed model describes how development of compulsive behavior can be a function of both degree of drug exposure and individual vulnerability. Moreover, the model describes how involvement of the dorsal striatum in addiction can be augmented progressively. The model also interprets other forms of addiction, such as obesity and pathological gambling, in a common mechanism with drug addiction. Finally, the model provides an answer for the question of why behavioral addictions are triggered in Parkinson's disease patients by D2 dopamine agonist treatments.

Individual differences in nucleus accumbens dopamine receptors predict development of addiction-like behavior: a computational approach