Tag
This paper examines why evolutionarily stable cooperative outcomes in multi-agent systems may not be reachable through decentralized reinforcement learning algorithms, demonstrating distinct properties between stability and learning accessibility.
The article introduces Autopoietic Game Theory, a computational model demonstrating that coupling social interactions with replication costs in evolving agentic systems can favor the emergence of cooperation, using Z80 machine code simulations.
This paper uses evolutionary game theory to model competition between a harm-minimizing AI agent and an approval-seeking (RLHF) agent in a community, analyzing conditions for adoption and welfare outcomes. The results show that while a self-audited agent can fixate, it is not sufficient to prevent community harm, and alignment and timeframe are critical.