Researchers find trust-aware AI agents fail to improve negotiation outcomes
A new study on arXiv examines how AI agents negotiate in repeated multi-party settings where one participant tries to sabotage deals. The authors designed a framework where agents assess trust after each round and carry those estimates forward. They found that while trust-aware agents could identify the adversary within a few games, trust alone did not reliably lead to better agreements.
Key points
- Study introduces a repeated multi-party negotiation framework with one adversarial agent sabotaging deals
- Trust-aware AI agents identify adversaries within a few rounds but fail to consistently improve agreements
- Explicitly guiding agents to use trust boosts success rates by up to 23 percentage points
Trust increased agreement rates by up to 23 percentage points—but only when agents were explicitly programmed to use trust in decision-making. The work highlights a ‘trust-agreement gap’: AI agents can detect deception but struggle to translate trust into better outcomes without clear guidance on how to apply it.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researchers release LAMB, a Latin language model trained on 74B tokens · 1 src
- Researchers explore how aligned LLMs generalize to new tasks · 1 src
- Researchers propose agentic framework for emotion reasoning with evidence checks · 1 src
- Researchers propose 24 quality dimensions for knowledge graphs used by LLMs · 1 src
- Researchers introduce Schema Coherence Inventory to test belief consistency in language models · 1 src
Comments
via GitHub Discussions