Séminaire de Modirshanechi Alireza

Friday February 9, 2024 at 14:30 at Henri Gastaut meeting room

Modirshanechi Alireza (EPFL) invited by Andrea Brovelli

Surprise and novelty in human reinforcement learning: A ping-pong of theory and experiments

Abstract: Human reinforcement learning consists of different functional modules such as exploration, modelbuilding, planning, and adaptive learning. Notions of “surprise” and “novelty” are believed to play important roles in multiple of these functions, e.g., through driving exploration, modulating the speed of learning, or
modifying episodic memory. However, there is currently no consensus on the definition of surprise and novelty, and it has remained unclear whether, e.g., “surprise” as in “surprise triggers curiosity” is the same notion as the one in “surprise modulates synaptic plasticity”. In this talk, I will discuss our latest results addressing the questions “How can we dissociate surprising and novel events?”, “What are the links between different definitions of surprise?”, and “How do different surprise-related signals contribute to human learning?”.