COMPUTATIONAL IMPLEMENTATION OF THE INCENTIVE -SENSITIZATION THEORY OF ADDICTION

Abstract

In the mid-1990s, the disease model of addiction replaced the moral model, emphasizing the underlying neurological mechanism of addiction. In accord with this view, the Incentive-Sensitization theory provides an account of compulsivity in addiction by emphasizing neural changes that accompany chronic drug use. This thesis presents the first steps towards a biologically grounded implementation of the Incentive-Sensitization theory of addiction. First, this project proposes an update of computational model structure with a learning perspective of addiction, focusing on the Incentive-Sensitization process. This project then presents multiple different plausible ways of mapping this theory onto a computational model, and examines the resulting behaviour to see whether it accords with standard interpretations of the theory. Results showed that, while repeated drug use can lead to a sensitized mesolimbic dopamine system, associative learning hyperreactivity to drug cues is crucial for the development of chronic drug use. Analyses of the computational models also revealed important ambiguities and limitations to the Incentive-Sensitization theory. This is the first step in a larger project to create a computationally tractable and biologically motivated computational model of addiction to help clarify and ground various terms in the theory.