["- In Discord open one of the \"test_gameXY\" channels \n- \"!status\" tells you the current state of the game \n- !join yourself into the game\n- !invite XY AI-agents into the game (not more than 2 or 3 per game, please)\n- !start the game\n- Have a look at your direct messages from the ModeratorBot - these will tell you which role you have and what to do next.\n- Now collaborate with chat messages to all of the players and finally decide for whom to !vote as a victim.\n\nRemarks/Attention:\n- I added a 10$ budget to my OpenAI-account. When gone be aware, that the AI-Players will stop working."]
About
A case study how well AI-agents (using LLMs) act in human collaboration: * Are AI agents on a par with humans when it comes to communication and collaboration? * Is it fun to play a board game with AI players? * If not, what improvements could be made to bring AI agents closer to humans, and what could a technical implementation look like?