Strategic tradeoffs between humans and AI in multi-agent bargaining
Crystal Qian, Kehang Zhu, John J. Horton, Benjamin Manning, Vivian Tsai, James Wexler, Nithum Thain
Crystal Qian and Kehang Zhu contributed equally.
Published: Proceedings of the 31st International Conference on Intelligent User Interfaces: 1625-1646 (2026)
Last updated: 2026-03
Summary
Markets increasingly accommodate large language models (LLMs) as autonomous decision-making agents. We present an empirical study comparing humans, multiple frontier LLMs, and customized Bayesian agents in dynamic multiplayer bargaining games under identical conditions. Bayesian agents extract the highest surplus with aggressive proposals that are frequently rejected. Humans and LLMs achieve comparable aggregate surplus but exhibit different trading strategies: LLMs favor conservative, concessionary proposals, while humans propose trades consistent with fairness norms that are more likely to be rejected. These findings show that performance parity can mask substantive procedural differences in complex multi-agent interactions.
Paper links
Cite
@inproceedings{horton2026strategictradeoffsbetweenhumansandaiinmultiagentbargaining,
title = {{Strategic Tradeoffs Between Humans and AI in Multi-Agent Bargaining}},
author = {Qian, Crystal and Zhu, Kehang and Horton, John J. and Manning, Benjamin and Tsai, Vivian and Wexler, James and Thain, Nithum},
booktitle = {Proceedings of the 31st International Conference on Intelligent User Interfaces},
year = {2026},
pages = {1625-1646},
doi = {10.1145/3742413.3789078},
url = {https://doi.org/10.1145/3742413.3789078}
}