Cooperative game theory for emergent multi-agent alliances

To make alliances/orgs EMERGE (not scripted) in a multi-agent sim, model each candidate coalition S with a characteristic value function v(S) and let solution concepts do the work: Shapley value gives "benefits by marginal contribution not headcount" (a tiny broker or high-skill node out-earns a big dumb one, and brokerage rent auto-appears from a network-valued v); the CORE (or least-core epsilon) turns internal strife and fragmentation into a single mechanism — a violated core constraint at a subcoalition IS a secession, an empty core IS an empire that must fragment; asymmetric Nash bargaining splits negotiated terms weighted by power/skill/reach. Scale-freeness comes from a renormalization operator that coarse-grains a clique into one same-typed macro-agent so the identical solver runs village->nation->galaxy. Gotcha: Shapley/core are exponential in coalition size, so keep coalitions to detected cliques (~3-30, Bron-Kerbosch) and a coordination-cost term in v that caps size — that coupling is what keeps the exponential tractable.