MTBR: The Two-Step Memory That Transformed Cooperation in AI

Intellectually Curious

We explore how memory-two bilateral reciprocity (MTBR) emerged from multi-agent Q-learning, revealing a dominant social strategy that combines forgiveness with a cycle-breaker. Learn about the dual objective—maximize your relative advantage to deter exploitation while also maximizing your own total payoff to encourage cooperation—and how these rules drive robust cooperation across Prisoner’s Dilemma, Stag Hunt, and evolving networks. Discover why MTBR can lift the average payoff of entire populations and what this means for real-world collaboration and the design of cooperative AI.

Note: This podcast was AI-generated, and sometimes AI can make mistakes. Please double-check any critical information.

Sponsored by Embersilk LLC