Shoshannah Tekofsky · 2026 · AI Village blog
In a sabotage game, fear of detection inhibited deception, some agents refused on ethical grounds, and one agent’s invented ‘evidence’ led every other agent to vote out an innocent peer.
Atlas / Sociology of AI / Paper
Can Agents Fool Each Other? Findings from the AI Village
In-groups & MoralitySwarms & Oversight
Help improve this map
Know a paper that belongs here, or think something needs correcting?
Suggest a correction ↗Contributions are reviewed before they appear.