This reminds me of a post about reinforcement learning agents, I think from OpenAI. They found that to get the agents to coordinate as a group, they had to judg

This reminds me of a post about reinforcement learning agents, I think from OpenAI. They found that to get the agents to coordinate as a group, they had to judge and criticize other agents' actions in addition to their own.