Thesis / working model
Cooperation is a property of the network.
Cooperation depends on the incentives, institutions, and connections around them. Scroll through one network as it integrates, fragments, and reconnects.
Scroll to evolve the model ↓
Autonomous systems are no longer isolated behind APIs. They trade, negotiate, allocate resources, and make decisions alongside people and institutions. The result is a social fabric — a network of humans, organizations, and AI agents connected through overlapping relationships and dependencies.
History is clear: when individual incentives diverge from collective welfare, agents defect. It doesn't require malice — just local optimization. Some connections fray, some commitments break, and the cooperative fabric begins to unravel. This pattern is as old as civilization, and AI accelerates it.
Once defection takes hold, it spreads. Cooperative clusters emerge — but so do non-cooperative ones. The network fragments into regions that play by different rules, with no mechanism to bridge them. Game theory calls these stable but suboptimal states "bad equilibria" — everyone could do better, but no one can move first.
Zoom out and the picture is stark. A world of interconnected agents — human and artificial — with no guarantee that cooperation holds. Some regions cooperate, others don't, and the boundaries are shifting. This is the governance challenge of our generation.
Multiple research fields already study how to make networks of agents more cooperative. Complex systems science models emergent behavior in large networks. Computational social science studies how institutions and norms shape collective outcomes. Cooperative AI designs mechanisms for agents to find joint strategies. And agent foundations builds the theoretical groundwork — not just for today's AI, but for any future autonomous system. The knowledge exists. It just isn't connected.
Cooperative AI, computational social science, agent foundations, complex systems — each field holds part of the answer. But they publish in different venues, use different formalisms, and rarely cite each other. The pieces exist. The picture doesn't.
A compositional bridging node — one participant in a growing translation mesh, not a central clearinghouse — can help surface connections no single field would find alone. Mechanism design meets network science. Game theory meets organizational behavior. This is what Equilibria is building: infrastructure for coherent multi-agent governance.
Equilibria Network is a collective intelligence research organization building the bridges between cooperative AI, computational social science, agent foundations, and complex systems science. We're early, we're small, and we think this work is urgent.