How to overcome reinforcement learning’s inherent selfishness.