AI practitioners can draw lessons from scientists working with animals to train reinforcement learning agents, a recent paper found.