Understanding addiction as a pathological state of multiple decision making processes: A neurocomputational perspective

Abstract

Theories of addiction in neuropsychology increasingly define addiction as a progressive subversion, by drugs, of the learning processes by which animals are equipped with, to adapt their behaviors to the ever-changing environment surrounding them. These normal learning processes, known as Pavlovian, habitual and goal-directed, are shown to rely on parallel and segregated cortico-striatal loops, and several computational models have been proposed in the reinforcement learning framework to explain the different and sometimes overlapping components of this network. In this chapter, we review some neurocomputational models of addiction originating from reinforcement learning theory, each of which explain addiction as a usurpation of one of the well-known models under the effect of addictive drugs. We try to show how each of these partially complete models can explain some behavioral and neurobiological aspects of addiction, and why it is necessary to integrate these models in order to have a more complete computational account for addiction.