Dynamics of reward based decision making: A computational study

Abstract

We consider a biologically plausible model of the basal ganglia that is able to learn a probabilistic two armed bandit task using reinforcement learning. This model is able to choose the best option and to reach optimal performances after only a few trials. However, we show in this study that the influence of exogenous factors such as stimuli salience and/or timing seems to prevail over optimal decision making, hence questioning the very definition of action-selection. What are the ecological conditions for optimal action selection?.