Clémence Bergerot

This paper develops a computational framework for distributional multi-agent reinforcement learning, showing how optimism, reward discretization, and temporal discounting interact to shape decision-making.

This study shows that moderate confirmation bias improves collective decision-making in reinforcement-learning agents, but excessive bias can lead to suboptimal performance and polarization.

This paper extends homeostatic regulation into a reinforcement learning framework to model sodium-seeking behavior and dopamine signaling in rats.