When Does a Dark Forest Emerge? An Open Agent-Based Model of Interstellar Strategic Regimes

The Dark Forest hypothesis is usually presented as a general consequence of uncertainty, technological asymmetry and catastrophic vulnerability. This paper asks a narrower question: under which combinations of beliefs, capabilities, signalling conditions and network incentives does a Dark Forest actually emerge? A Dark Forest Emergence Model is developed as an open, spatial agent-based Bayesian game. Civilisations enter and leave a bounded galaxy, observe noisy and temporally degraded signals, update pairwise hostility beliefs, choose whether to broadcast, form reciprocal trust links, invest through adaptive learning and select possible attacks. The model derives analytical thresholds for pre-emption, concealment, trust-network reproduction and conflict cascades, then connects them to a continuous Dark Forest Index and a discrete galactic-regime classification. Across 688 reproducible simulation runs, the hostility-offence experiment produces replicate-consistent signalling-commons and Dark Forest regions, together with mixed boundary cells and isolated alternative outcomes in individual runs. Global sensitivity analysis identifies prior hostility, coalition benefit and offensive advantage as the principal determinants of Dark Forest intensity. Ablations show that removing temporal opacity or lowering prior hostility can eliminate the regime, whereas reliable but obsolete signals need not prevent it. The results treat the Dark Forest as a conditional, endogenous phase rather than a universal equilibrium.