Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Language of Bargaining: Linguistic Effects in LLM Negotiations

Stuti Sinha, Himanshu Kumar, Aryan Raju Mandapati, Rakshit Sakhuja, Dhruv Kumar

TL;DR

This work investigates how the language used in LLM negotiations shapes outcomes, beyond model architecture or game rules. By using a controlled, multi-game framework with English and four Indic framings, the authors show that linguistic framing can reverse proposer advantages and alter surplus distribution in distributive tasks while enhancing exploration in integrative tasks. The study offers six testable predictions about cultural scripts, pragmatic constraints, and stereotype activation, and it demonstrates that language functions as a latent prior that guides strategic choices. The findings highlight the need for multilingual evaluation to ensure fair and globally applicable deployment of negotiation-capable AI systems, as English-centric benchmarks risk masking qualitatively different behaviors across languages.

Abstract

Negotiation is a core component of social intelligence, requiring agents to balance strategic reasoning, cooperation, and social norms. Recent work shows that LLMs can engage in multi-turn negotiation, yet nearly all evaluations occur exclusively in English. Using controlled multi-agent simulations across Ultimatum, Buy-Sell, and Resource Exchange games, we systematically isolate language effects across English and four Indic framings (Hindi, Punjabi, Gujarati, Marwadi) by holding game rules, model parameters, and incentives constant across all conditions. We find that language choice can shift outcomes more strongly than changing models, reversing proposer advantages and reallocating surplus. Crucially, effects are task-contingent: Indic languages reduce stability in distributive games yet induce richer exploration in integrative settings. Our results demonstrate that evaluating LLM negotiation solely in English yields incomplete and potentially misleading conclusions. These findings caution against English-only evaluation of LLMs and suggest that culturally-aware evaluation is essential for fair deployment.

The Language of Bargaining: Linguistic Effects in LLM Negotiations

TL;DR

This work investigates how the language used in LLM negotiations shapes outcomes, beyond model architecture or game rules. By using a controlled, multi-game framework with English and four Indic framings, the authors show that linguistic framing can reverse proposer advantages and alter surplus distribution in distributive tasks while enhancing exploration in integrative tasks. The study offers six testable predictions about cultural scripts, pragmatic constraints, and stereotype activation, and it demonstrates that language functions as a latent prior that guides strategic choices. The findings highlight the need for multilingual evaluation to ensure fair and globally applicable deployment of negotiation-capable AI systems, as English-centric benchmarks risk masking qualitatively different behaviors across languages.

Abstract

Negotiation is a core component of social intelligence, requiring agents to balance strategic reasoning, cooperation, and social norms. Recent work shows that LLMs can engage in multi-turn negotiation, yet nearly all evaluations occur exclusively in English. Using controlled multi-agent simulations across Ultimatum, Buy-Sell, and Resource Exchange games, we systematically isolate language effects across English and four Indic framings (Hindi, Punjabi, Gujarati, Marwadi) by holding game rules, model parameters, and incentives constant across all conditions. We find that language choice can shift outcomes more strongly than changing models, reversing proposer advantages and reallocating surplus. Crucially, effects are task-contingent: Indic languages reduce stability in distributive games yet induce richer exploration in integrative settings. Our results demonstrate that evaluating LLM negotiation solely in English yields incomplete and potentially misleading conclusions. These findings caution against English-only evaluation of LLMs and suggest that culturally-aware evaluation is essential for fair deployment.
Paper Structure (34 sections, 6 figures, 3 tables)

This paper contains 34 sections, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Benchmarking and Evaluation Frameworks.
  4. Social Factors in LLM Negotiation
  5. Cross-lingual and Multilingual Contexts.
  6. Theoretical Framework
  7. Cultural and Linguistic Mechanisms
  8. Predictions
  9. Methodology
  10. System Prompts and Persona Design
  11. Model Settings
  12. Experimental Factors
  13. Evaluation Metrics
  14. Results and Analysis
  15. Ultimatum Game Results
  16. ...and 19 more sections

Figures (6)

  • Figure 1: Ultimatum Game Language Comparison showing average (a) acceptance rates, (b) average initial offer, (c) payoffs, (d) win rates (player 1).
  • Figure 2: Buy Sell Game Language Comparison showing average (a) acceptance rates, (b) seller advantages, (c) buyer advantages, and (d) win rates across different cultural behaviors.
  • Figure 3: Resource Exchange Game Language Comparison showing average (a) acceptance rates, (b) trade volume, (c) payoffs, and (d) win rates (player 1) across different cultural behaviors.
  • Figure 4: Heatmaps for Ultimatum Game comparing (a) win rate and (b) payoff for model combinations for all languages.
  • Figure 5: Heatmaps for Buy-Sell Game comparing (a) Seller advantage and (b) Buyer Advantage for model combinations for all languages.
  • ...and 1 more figures