Reproduce “Factorised Active Inference for Strategic Multi-Agent Interactions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jaime Ruiz-Serra, Patrick Sweeney, Michael S. Harré
Researchers at the University of Sydney introduce a factorised Active Inference model, enabling agents to explicitly infer the internal states of other agents during strategic interactions. This framework demonstrates how individual cognition influences collective dynamics in changing general-sum games, revealing patterns of strategic uncertainty, equilibrium selection, and a mechanism for bootstrapping cooperation.
Key insights
- Variational Free Energy (VFE) quantifies strategic uncertainty and surprise in multi-agent interactions, increasing sharply during game transitions.
- Ensemble-level Expected Free Energy (EFE) effectively characterizes collective states and the basins of attraction for different game equilibria, highlighting that individual Nash Equilibria are not always collectively optimal.
- Sequential game design can foster cooperation, demonstrating that an interim phase with easier cooperative conditions can 'bootstrap' desired outcomes in subsequent, more challenging interactions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.