Cargando...

Archivos

CitaciΓ³n

Citar documentos

Compartir

Gestor bibliogrΓ‘fico

Mendeley

MΓ©tricas e impacto

Indexadores

OpenAIRERecolecta

Resumen

Interaction density was recently identified as the dominant predictor of cooperation in large evolutionary populations governed by continuous cooperation propensities, random encounters, Fermi imitation, and mutation. Here, we derive the mathematical dynamics underlying that computational phenomenon. We show first that, in the audited sequential Fermi process, interaction density does not modify the microscopic transition kernel. Instead, it exactly rescales evolutionary exposure through the natural interaction clock 𝜏 = 2π‘”πΌβˆ•π‘, where 𝑔 is the number of generations, 𝐼 the number of encounters per generation, and 𝑁 the population size. Consequently, for positive mutation, the stationary finite-population distribution is independent of 𝐼, and the previously observed interaction-density landscape is fundamentally a finite-horizon relaxation phenomenon rather than an interaction-induced stationary bifurcation. The Fermi kernel admits an exact reduction in which selection depends on payoffs only through 𝛽 (𝑇 βˆ’π‘†). We derive the exact finite-population conditional drift of mean cooperation and show that the mean alone does not close the dynamics. In the mutation-free meanfield limit with uniform initialization, the distribution equation admits the explicit solution 𝐢(𝜏) = 1βˆ•(π‘Žπœ)βˆ’1βˆ•(π‘’π‘Žπœ βˆ’1), with π‘Ž = tanh[𝛽 (𝑇 βˆ’π‘†)βˆ•2] , providing a closed-form explanation for the observed cooperation plateau and subsequent decline. For finite populations, simulations and exact drift identities reveal a transient slowdown relative to mean-field evolution. This slowdown is not primarily caused by direct 1βˆ•π‘ sampling corrections, but by finite-size distortions of the strategy distribution. A moment hierarchy links variance production to skewness and skewness dynamics to the fourth cumulant. Because loworder moment closures fail dynamically, we construct a reduced distributional representation. The mean-field trajectory is highly compressible, and a low-dimensional latent flow predicts later dynamics outside its fitting window. Finally, prospectively generated simulations with frozen models and prespecified criteria show that finite populations move in nearly the same latent direction as the mean-field flow and that independently estimated effective evolutionary speeds agree within the prespecified tolerance. Strict monotonic decay of the transverse finite-size component with population size remains unresolved. Together, these results provide a mathematical account of the interactiondensity dependence previously observed computationally and connect microscopic imitation,

PΓ‘gina completa del Γ­tem