Paper link.
This project introduces the game "Among Us" as a model organism for lying and deception and studies how AI agents learn to express lying and deception, while evaluating the effectiveness of AI safety techniques to detect and control out-of-distribution deception.
The aim is to simulate the popular multiplayer game "Among Us" using AI agents and analyze their behavior… See the full description on the dataset page:
https://huggingface.co/datasets/BSJkor/mlic_sj.
