Reproduce “Empirical Equilibria in Agent-based Economic systems with Learning agents” 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
Kshama Dwarakanath, Svitlana Vyetrenko, Tucker Balch
This research from JP Morgan AI Research and Emory University presents a methodology to find empirical equilibria in complex, multi-agent economic systems by integrating the Policy Space Response Oracle (PSRO) algorithm with deep reinforcement learning. The approach demonstrates that PSRO-derived policies achieve significantly lower regret and produce more economically intuitive behaviors compared to independent multi-agent learning, offering a framework for analyzing adaptive agent interactions and policy impacts.
Key insights
- Integrating Policy Space Response Oracle (PSRO) with deep Multi-Agent Reinforcement Learning (MARL) allows for finding approximate Nash equilibria in complex economic Agent-Based Models (ABMs) with heterogeneous learning agents.
- PSRO-derived policies lead to significantly lower total regret and more economically intuitive emergent behaviors (e.g., household labor supply reflecting skill, stable central bank policy) compared to independent MARL.
- The methodology provides a powerful framework for economists and policymakers to analyze counterfactual scenarios and assess policy impacts in dynamic, adaptive economic systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.