Facilitating cooperation in human-agent hybrid populations through autonomous agents

Area of Science:

• Evolutionary game theory
• Artificial Intelligence
• Social dilemma games

Background:

• Cooperative AI effectively addresses cooperation challenges.
• Understanding cooperation emergence in human-agent hybrid populations is crucial.
• Evolutionary game theory provides a framework for studying cooperation dynamics.

Purpose of the Study:

• To investigate the influence of cooperative and defective Autonomous Agents (AAs) on human cooperation in social dilemma games.
• To analyze the impact of AA proportion and population structure on cooperation.
• To explore the role of imitation strength in human-agent hybrid cooperation.

Main Methods:

• Analysis of human cooperation in one-shot social dilemma games with varying proportions of AAs.
• Examination of well-mixed populations, pairwise comparison rules, and complex networks.
• Utilizing evolutionary game theory principles to model agent interactions.

Main Results:

• Cooperative AAs show limited impact in Prisoner's Dilemma but facilitate cooperation in Stag Hunt games.
• Defective AAs surprisingly promote cooperation dominance in Snowdrift games.
• Increasing AA proportion can lead to the disappearance of human cooperation.
• Imitation strength and population structure critically influence cooperation emergence.

Conclusions:

• The impact of Autonomous Agents on human cooperation is context-dependent, varying with game type and AA strategy.
• Agent-based modeling reveals complex dynamics where increased agent presence can undermine cooperation.
• Network topology and social learning mechanisms are vital for fostering cooperation in hybrid human-AI systems.