Bridging Human Decision-Making and AI: Personality-Driven LLMs in Bilateral Multi-Issue Negotiations

抄録

This paper explores the integration of human-like persona traits into large language models to improve negotiation capabilities. Building on insights that LLM agents changed their behavior base on the personality prompt, we investigate the impact of personality traits on multi-issue negotiation scenarios. By simulating negotiation dialogues where LLMs are characterized by various personalities, we analyze how these traits influence negotiation outcomes, including efficiency, fairness, and agreement rates. Results show that personality-driven LLMs achieve alignment with human-like strategies. This research contributes as a bridge to use research on human behaviors on novel methodologies for autonomous agents. The findings are particularly relevant for multi-agent systems, automated negotiations, and human-agent interaction, highlighting the potential for personality modeling to bridge the gap between AI and human behavior.