A Principle of Targeted Intervention for Multi-Agent Reinforcement Learning