Sharing rewards undermines coordinated hunting

Coordinated hunting is widely observed in animals, and sharing rewards is often considered a major incentive for this success. While most results on this topic are correlational, we reveal the causal roles of sharing rewards through computational modeling with a state-of-the-art Multi-agent Reinforcement Learning (MARL) algorithm. Using a coordinated hunting task with a group of predators hunting one prey, we show that sharing rewards is neither necessary nor sufficient for modeling animal coordinated hunting. Hunting coordination modeled through sharing rewards 1) suffers from the free-rider problem, 2) plateaus at a small group size, and 3) is not a Nash equilibrium. Moreover, individually rewarded predators outperform predators that share rewards, especially when the hunting is difficult, the group size is large, and the action cost is high. We conclude that animal coordinated hunting can be successfully modeled through reinforcement learning only when the agents are selfish, and not when the rewards are shared. Our results further offer computational support to the explanation of chimpanzee behavior that agents with only selfish interests can form coordinated hunting, and sharing rewards might simply be a byproduct of hunting, instead of an intelligent design to facilitate coordination.