Promoting Cooperative Behavior in AI

This chapter explores how learning agents can adopt social norms through public sanctions in decentralized environments, addressing collective action problems using reinforcement learning. It discusses the development of algorithms that evaluate actions based on collective norms and introduces a classifier for approving or disapproving behaviors. The chapter also examines the implications of these norms in real-world scenarios, such as traffic behaviors and driving practices, highlighting the potential for enhanced safety and community engagement.