Assistive Large Language Model Agents for Socially-Aware Negotiation Dialogues

TL;DR

This paper presents assistive LLM agents that facilitate socially-aware business negotiations by role-playing, remediating norm violations, and using a novel ICL selection method to improve negotiation outcomes, with extensive empirical validation.

Contribution

Introduces a tuning-free ICL method with a novel value impact criterion for selecting high-quality exemplars to enhance negotiation performance in LLM-based agents.

Findings

Effective in improving negotiation outcomes across multiple topics

Remediator agent successfully rewrites norm-violating utterances

Proposed ICL method outperforms baseline exemplar selection approaches

Abstract

We develop assistive agents based on Large Language Models (LLMs) that aid interlocutors in business negotiations. Specifically, we simulate business negotiations by letting two LLM-based agents engage in role play. A third LLM acts as a remediator agent to rewrite utterances violating norms for improving negotiation outcomes. We introduce a simple tuning-free and label-free In-Context Learning (ICL) method to identify high-quality ICL exemplars for the remediator, where we propose a novel select criteria, called value impact, to measure the quality of the negotiation outcomes. We provide rich empirical evidence to demonstrate its effectiveness in negotiations across three different negotiation topics. We have released our source code and the generated dataset at: https://github.com/tk1363704/SADAS.