Abstract

This talk proposes a Temporal Difference Reinforcement Learning (TDRL) Theory of Emotion, arguing that emotions are manifestations of the brain's reward processing system, specifically its assessment of temporal difference errors. I propose that emotions, ranging from basic joy and distress to more complex feelings like regret and hope, all stem from this fundamental process of evaluating changes in well-being. I supports this theory with computational modeling, psychological findings, and neurobiological evidence, exploring the implications for understanding emotion in humans, animals, and artificial intelligence. The paper also examines the role of emotions in social interactions and their evolution, emphasizing their adaptive function in promoting survival and learning. Finally, I discuss potential applications in fields like human-robot interaction.

Speaker

Associate Professor Joost Broekens study Artificial Intelligence, focusing on Affective Computing and human interaction with socially interactive agents. Formerly President of the Association for the Advancement of Affective Computing, he’s an associate professor at Leiden University. Co-founder of Interactive Robotics and Daisys. His research spans emotion modeling in AI, explainability, emotional expression in agents, and educational robots, with a focus on reinforcement learning and human-robot interaction.

Please note. The subtitle was primarily generated by AI in collaboration with human review and may contain some errors.