Multi-Agent, Human-Agent, and Beyond: A Survey on Cooperation in Social Dilemmas

Recent advancements in AI have provided new perspectives on cooperation within social dilemmas, a fundamental topic in computer and social science. This survey explores AI’s role in such dilemmas, with a particular focus on multi-agent and human-agent cooperation. It delves into the intrinsic and external motivations driving cooperation among rational agents and the AI algorithms that enable human-agent collaboration. The paper also broaches the emergent field of using AI agents to foster cooperation among humans.

Highlights:

• Reviews motivations for agent cooperation and effective strategies against diverse opponents.
• Discusses AI algorithms for human-agent cooperation and human biases toward AI.
• Surveys AI’s role in enhancing human cooperation.

Importance:

Understanding and enhancing cooperation in AI systems is essential for real-world applications. This paper identifies future research directions such as employing LLMs, establishing theoretical frameworks, revisiting theories of human cooperation, and exploring practical applications. Find the full survey here.