AI negotiation optimal agreement 14%... low offers accepted 81%
Original headlineAI 협상 최적 합의 14%…낮은 제안은 81% 수용
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
Top-tier AI models reached the jointly optimal outcome in only 14% of peer-pressured negotiation tasks, according to a Stanford, Oxford, and MIT study benchmarking 18 closed- and open-weight models across 240 tasks in eight collaboration categories, with six models also tested at high reasoning settings. The same models scored 100% when solving tasks alone and 84% when paired with an unproblematic colleague, underscoring a sharp gap between individual capability and collaborative performance. The findings matter because agentic AI is increasingly deployed in multi-agent workflows where negotiation, information sharing, and team-goal alignment determine outcomes. Watch for follow-up work on whether prompting, role design, or fine-tuning can close the collaboration gap before enterprises scale agent teams.
Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.
Named in this story
Stanford University, MIT, Oxford University