ASTRA: A Negotiation Agent with Adaptive and Strategic Reasoning through Action in Dynamic Offer Optimization

2025-03-10

AI Summary

Existing negotiation agents struggle to balance self-utility and cooperation under human bounded rationality, low opponent adaptability, and weak strategic reasoning. Method: We propose an intelligent agent for dynamic multi-round negotiations that (i) pioneers the integration of real-time opponent behavioral modeling and Tit-for-Tat reciprocity into an end-to-end offer optimization framework; (ii) implements a three-stage interpretable decision pipeline—opponent modeling → inverse offer optimization → strategy fusion—combining linear programming, acceptance probability prediction, and rule-based reasoning. Contribution/Results: Evaluated across diverse simulated environments and real human-agent negotiations, our agent significantly improves agreement rate (+23.6%) and negotiator utility (+18.4%). Moreover, it functions as an interpretable negotiation coach, delivering real-time feedback and optimal offer recommendations.

Technology Category

• Multiagent Systems: Agreement, Argumentation & Negotiation
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Search and Optimization: Learning to Search

Application Category

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

Abstract

Negotiation requires dynamically balancing self-interest and cooperation to maximize one's own utility. Yet, existing agents struggle due to bounded rationality in human data, low adaptability to counterpart behavior, and limited strategic reasoning. To address this, we introduce principle-driven negotiation agents, powered by ASTRA, a novel framework for turn-level offer optimization grounded in two core principles: opponent modeling and Tit-for-Tat reciprocity. ASTRA operates in three stages: (1) interpreting counterpart behavior, (2) optimizing counteroffers via a linear programming (LP) solver, and (3) selecting offers based on negotiation tactics and the partner's acceptance probability. Through simulations and human evaluations, our agent effectively adapts to an opponent's shifting stance and achieves favorable outcomes through enhanced adaptability and strategic reasoning. Beyond improving negotiation performance, it also serves as a powerful coaching tool, offering interpretable strategic feedback and optimal offer recommendations.

Problem

• Dynamic balancing of self-interest and cooperation in negotiations.
• Overcoming bounded rationality and low adaptability in negotiation agents.
• Enhancing strategic reasoning and adaptability through opponent modeling.

Innovation

• Opponent modeling for dynamic negotiation adaptation
• Linear programming solver for offer optimization
• Tit-for-Tat reciprocity for strategic reasoning