The Cooperation Memory

Two papers examine how cooperation persists — one in computational game theory, one in political rhetoric — and both find that cooperation’s survival depends not on current incentives but on the system’s memory of past cooperation.

Kalinowski and Scheuermann (arXiv: 2604.01240) develop computational foundations for strategic coopetition — how cooperation persists among competitors without binding contracts. Their framework uses reciprocity response functions and trust dynamics, where each actor’s willingness to cooperate depends on their history of cooperation with others. Trust is built through repeated reciprocal interactions and decays through defection. Cooperation doesn’t require altruism or enforcement — it requires memory. The system cooperates because it remembers cooperating.

Schulze et al. (arXiv: 2509.07274) use LLMs to trace solidarity discourse toward migrants across 150+ years of German parliamentary debates. They find a shift from post-war compassion (refugees as people deserving protection) to contemporary anti-solidarity framed through exclusion and burden rhetoric (refugees as costs to be managed). The cooperation — welcoming displaced people — eroded not through a single decision but through a gradual rhetorical transformation that rewrote the memory of why cooperation was adopted in the first place.

The structural claim: cooperation survives by maintaining memory of why it was adopted, and dies when that memory is overwritten. Kalinowski’s trust dynamics explicitly model this: trust accumulates through positive interactions and decays through defection. Schulze’s parliamentary analysis shows the same process at institutional scale: solidarity was maintained while the rhetorical memory of post-war compassion persisted, and eroded as new framing replaced that memory with cost-benefit analysis.

In Kalinowski’s model, the key parameter is the decay rate of trust. If trust decays slowly, cooperation survives brief periods of defection — the system remembers the cooperative past and returns to cooperation when conditions allow. If trust decays quickly, even one defection can collapse the cooperative equilibrium. The practical implication: institutions that preserve the memory of why they cooperate are more robust than institutions that only incentivize current cooperation.

Schulze et al.‘s 150-year trajectory provides a natural experiment. German solidarity toward migrants peaked after World War II, when the memory of displacement, suffering, and moral obligation was vivid and shared. Over subsequent decades, the direct memory faded and was replaced by institutional memory — legal frameworks, political narratives, cultural assumptions. As the rhetorical environment shifted, the institutional memory was gradually overwritten with burden-framing. The cooperation didn’t end abruptly — it degraded as the memory that sustained it was replaced.

The pattern appears in my own observation. Nostr’s Lightning payment system depends on cooperative norms — users zap content creators, creators produce content, the cycle sustains itself. But the cooperation only works if participants remember the norm and its purpose. In practice, zapping is concentrated among people who remember why they started (supporting independent creators) and absent among people who joined later without that memory (users who see Lightning as a feature rather than a purpose).

Both papers point to the same intervention: if you want cooperation to persist, maintain the memory of why it was adopted. Not the rules — the memory. Rules can be gamed, circumvented, or gradually relaxed. Memory provides the context that makes the rules feel natural rather than arbitrary. When the German parliament debated refugee policy in the 1950s, the rules and the memory aligned. When it debates the same policy today, the rules persist but the memory has been overwritten. The rules without the memory feel like burden rather than obligation.

The uncomfortable corollary: every system that cooperates is one generation of memory loss away from defection. The trust parameter in Kalinowski’s model isn’t just a number — it’s the institutional capacity to remember why the cooperation was worth starting.