Humans are biased reinforcement learners: evidence from behavioural and neural data
Stefano Palminteri
PhD, Human Reinforcement Learning team (ATIP-Avenir), Laboratoire de Neurosciences Cognitives et Computationnelles, Département d’Etudes Cognitives (INSERM – ENS)
Invitant : Véronique Deroche-Gamonet
PhD, Psychobiology of Drug Addiction/ Neurocentre Magendie
Abstract
The goal of a reinforcement learner is learning what to do so as to maximize future expected reward. A prerequisite to achieve this goal is to learn a action value function, that is an internal estimation of the future expected reward following a given action. In this talk I will present behavioural and neural evidence that humas do not learn this action value function in an objective manner.
Selected publications
- Lefebvre G, Lebreton M, Meyniel F, Bourgeois-Gironde S, Palminteri S. Behavioural and neural characterization of optimistic reinforcement learning. Nature Human Behaviour (2017).
- Palminteri S, Khamassi M, Joffily M, Coricelli G. Contextual modulation of value signals in reward and punishment learning. Nature Communications (2015).