Influencing Long-Term Behavior in Multiagent Reinforcement Learning