Think, Speak, Decide: Language-Augmented Multi-Agent Reinforcement Learning for Economic Decision-Making

2025-11-16

AI Summary

Multi-agent reinforcement learning (MARL) in economic decision-making struggles to effectively integrate semantically rich, context-dependent linguistic information with structured numerical signals. Method: This paper proposes LAMP—a Language-Augmented Multi-agent Reinforcement Learning framework—featuring a three-stage “Think-Speak-Decide” collaborative architecture: the Think module employs large language models (LLMs) for trend reasoning and caches linguistic reasoning trajectories; the Speak module generates dynamic, strategy-oriented semantic messages to align agents’ beliefs; and the Decide module fuses linguistic and numerical signals to produce interpretable actions. Contribution/Results: LAMP introduces the first linguistic reasoning trajectory caching mechanism, enabling long-horizon policy optimization and semantics-driven cooperation. In economic simulation experiments, it achieves 63.5% and 34.0% higher cumulative returns over pure MARL and pure LLM baselines, respectively, and improves robustness by 18.8% and 59.4%, significantly enhancing both decision effectiveness and interpretability.

Technology Category

• Multiagent Systems: Multiagent Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Economic decision-making depends not only on structured signals such as prices and taxes, but also on unstructured language, including peer dialogue and media narratives. While multi-agent reinforcement learning (MARL) has shown promise in optimizing economic decisions, it struggles with the semantic ambiguity and contextual richness of language. We propose LAMP (Language-Augmented Multi-Agent Policy), a framework that integrates language into economic decision-making and narrows the gap to real-world settings. LAMP follows a Think-Speak-Decide pipeline: (1) Think interprets numerical observations to extract short-term shocks and long-term trends, caching high-value reasoning trajectories; (2) Speak crafts and exchanges strategic messages based on reasoning, updating beliefs by parsing peer communications; and (3) Decide fuses numerical data, reasoning, and reflections into a MARL policy to optimize language-augmented decision-making. Experiments in economic simulation show that LAMP outperforms both MARL and LLM-only baselines in cumulative return (+63.5%, +34.0%), robustness (+18.8%, +59.4%), and interpretability. These results demonstrate the potential of language-augmented policies to deliver more effective and robust economic strategies.

Problem

• Integrating unstructured language into economic decision-making processes
• Addressing semantic ambiguity in multi-agent reinforcement learning systems
• Bridging the gap between numerical data and linguistic context

Innovation

• Integrates language reasoning with multi-agent reinforcement learning
• Uses Think-Speak-Decide pipeline for economic decision-making
• Combines numerical data with language communications in policy