Fostering Collective Action in Complex Societies Using Community-Based Agents

Fostering Collective Action in Complex Societies Using Community-Based Agents

Jonathan Skaggs, Michael Richards, Melissa Morris, Michael A. Goodrich, Jacob W. Crandall

Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence
Main Track. Pages 211-219. https://doi.org/10.24963/ijcai.2024/24

As AI integrates into human societies, its ability to engage in collective action is increasingly important. Human social systems have large and flexible strategy spaces, conflicting interests, power asymmetry, and interdependence among members, which together make it challenging for agents to learn collective action. In this paper, we explore the ability of community-based agents to learn collective action within a novel model of complex social systems. We first present this social model, called the Junior High Game (JHG). The JHG embodies key elements of human social systems that require players to act collectively. We then describe an agent, called CAB, which is based on community detection and formation algorithms. Via simulations and user studies, we evaluate the ability of CAB agents to interact in JHG societies consisting of humans and AI agents. These evaluations both identify requirements for successful collective behaviors in the JHG and identify important unsolved problems for developing AI agents capable of collective action in complex social systems.
Keywords:
Agent-based and Multi-agent Systems: MAS: Coordination and cooperation
Agent-based and Multi-agent Systems: MAS: Agent societies
Agent-based and Multi-agent Systems: MAS: Human-agent interaction
Agent-based and Multi-agent Systems: MAS: Trust and reputation