The increasing social capabilities of artificial intelligence systems raise fundamental questions about how machines learn, interpret, and participate in complex human interaction. While neural models have achieved remarkable progress in perception, language, and decision-making, understanding socially grounded behavior remains an open challenge. This workshop explores contextual games—particularly role-playing games (RPGs)—as structured yet flexible environments for studying neural and computational models of social intelligence. Contextual games provide rich interaction spaces combining narrative context, cooperation, competition, and emotional decision-making. These environments naturally elicit phenomena such as trust formation, negotiation, empathy, engagement, and collective problem solving, offering a unique experimental setting for investigating adaptive behavior in humans and artificial agents. Advances in neural computation, reinforcement learning, large language models, and multi-agent learning now enable agents capable of participating in these socially complex environments, making games an ideal testbed for studying emergent interaction dynamics. The third edition of Dungeons, Neurons, and Dialogues brings together researchers from neural computation, machine learning, cognitive modeling, game AI, and human–AI interaction to explore how contextual game environments can support the development and evaluation of socially intelligent systems. Topics include neural modeling of social behavior, adaptive dialogue systems, reinforcement learning in narrative environments, multi-agent coordination, affect-aware agents, and human-AI co-creation. Through invited talks, research presentations, and an interactive collaborative activity, the workshop aims to foster interdisciplinary discussion and promote new methodologies for studying social intelligence in artificial systems. By positioning games as experimental laboratories for neural and social learning, the workshop contributes to advancing computational approaches toward more adaptive, interpretable, and socially aware AI.
Workshop Organizers
- Pablo Barros
- Laura Triglia
- Nikhil Churamani
- Matthias Kerzel
- Pietro Gravino