The Trust-Agreement Gap in Multi-Party LLM Negotiation
October 26, 2026
Investigation into multi-party LLM negotiation reveals that agents struggle to leverage pairwise trust estimates to mitigate manipulation in repeated games. The study compares trust-aware agents against trust-agnostic agents in environments where one participant attempts to covertly prevent agreement.
HOW THIS AFFECTS YOU
●
builderYou should account for agent manipulation when designing multi-party autonomous negotiation workflows.
●
researcherThis exposes the gap between forming trust estimates and successfully applying them in multi-agent systems.