Anne Collins & Etienne Koechlin
Tuesday 9 June 2009
4 th Floor Seminar Room, Alexandra House
17 Queen Square, London, WC1N 3AR
Cognitive control, learning and exploration
Prefrontal cognitive control is based on building, maintaining and switching between multiple task-sets (TS, i.e. sensorimotor mappings) according to external cues and feedbacks. Little is known, however, about computational principles integrating learning and cognitive control for optimal adaptive behaviour in varying noisy environments. We propose a hierarchical reinforcement-learning model that learns, stores, monitors and reuses TS. Our model combines and generalizes existing models including those proposed by Doya (2002) and Yu & Dayan (2005). It accounts for the learning and control of multiple TSs as need arises and clarifies key concepts underlying cognitive control (default behaviour, task-switching, exploration vs. exploitation and contextual selection) and learning (hierarchical flexibility). The model makes specific predictions. We performed behavioural experiments that confirm the predictions and validate the model hypotheses.