Cooperation Dynamics in Reasoning vs. Non-Reasoning Generative AI

Testing Claude 3.5 through GPT-5.2 in iterated prisoner's dilemmas shows that cooperation is influenced by reputation, strategy, and emotional signaling. While non-reasoning models respond to all three factors, reasoning models demonstrate a more concentrated reliance on strategic signals.