Multi-agent RL bidders learned tacit collusion in electricity markets with nobody telling them to.

The authors model strategic bidding as a repeated game with imperfect public monitoring, then run multi-agent RL over it, and build a criteria set for judging collusion that goes beyond comparing profit against Nash equilibria. Agents sustained supra-competitive outcomes matching tacit-collusion indicators despite never being instructed to collude. Clearest domain-specific evidence yet that autonomous pricing agents create a regulatory problem their operators didn't choose and can't easily disclaim. (arXiv)