Paper argues theory of mind modules in multi-agent systems should not be certified by accuracy alone
What happened
Will Beaumaster's paper argues that the theory-of-mind module in multi-agent teams, which infers actions from teammates, should not be certified on accuracy alone. The cost of errors and the correlation between errors change what accuracy threshold is acceptable. The author says accuracy fails in both directions in task allocation, Level-Based Foraging and Hanabi. In Hanabi, where errors compound, accuracy certification approved a harmful LLM deployment. Instead, the author proposes certifying the deployment itself. The paper introduces the Blind-Spot Gate, which estimates error costs, pairwise interactions and measured dependence without actually deploying the module. Experiments include modules with injected errors, trained and Bayesian modules, five LLMs and human judgments.
AI-written from the coverage · updated 2026-10-15
Coverage timeline
Follow the coverage to see the story from different sides.
- arXiv Board & Classic Game AIPickWhy accuracy fails to certify theory of mind in multi-agent teams
The paper argues that modules in multi-agent cooperation that infer actions from teammates should not be certified on accuracy alone, because the cost of errors and the correlation between errors change the acceptable accuracy threshold. The authors tested this in task allocation, Level-Based Foraging and Hanabi, finding that accuracy fails in both directions. In Hanabi, where errors compound, it approved a harmful LLM deployment.