From cognitive to habit behavior during navigation, through cortical-basal ganglia loops

Abstract

The distinction between cognitive goal-oriented and SR habitual behavior has long been classical in Neuroscience. Nevertheless, the mechanisms of the two types of behaviors as well as their interactions are poorly understood, in spite of significant advances in the knowledge of their supporting structures, the cortico-striatal loops. A neural network (NN) model of the dynamics of these systems during a goal navigation paradigm is presented within the framework of reinforcement learning. The model supposing, the parallel interactive learning of cognitive and habitual strategies, replicates key experimental results related to the transition between them. The biological inspiration of the NN architecture provides insights on the nature of their interactions, and the conditions of their respective engagement in the control of behavior.