WOULD AN AI KEEP COOPERATING?

Archived study

Explores repeated prisoner's-dilemma play under an explicit win-stay, lose-shift rule. The conversation changes the opponent's timing and isolated defections and traces how cooperation patterns change.

Interpretation and limits

The decision-maker is a specified deterministic rule, not a measured language-model policy. Timing-sensitive paths illustrate the rule and payoff assumptions rather than a general claim about AI cooperation.

Package and provenance

The original .science export and matching poster are preserved byte-for-byte from the upstream case. The package contains the conversation, session metadata, and the resources inventoried in its internal manifest. This introduction summarizes the archived record; the research, external sources, and model evaluations were not rerun for this import.