What is cooperative inverse reinforcement learning (CIRL)?