Recursive Polycentric Moderation (RPM) is a distributed governance architecture for online platforms in which moderation emerges from recursive, multi-community decision processes rather than centralized enforcement. The model separates local neighbor judgment from distant network evaluation, and introduces a structured appeal mechanism allowing contested content to propagate through progressively broader institutional layers. Using large-scale agent-based simulations on community-structured scale-free networks, we evaluate RPM under baseline conditions, varying decision thresholds, strategic manipulation, and network size up to 10,000 users. Results show a sharp moderation threshold tied to content extremity, stable exposure rates of approximately 2–3% independent of network scale, bounded degradation under adversarial behavior, and systematic recursive escalation of moderated content. These findings suggest that large-scale content moderation can be achieved as an emergent property of polycentric network structures, without centralized censorship or global ranking algorithms. This version includes the complete simulation code, figure-generation scripts, and publication-quality figures enabling full reproducibility of the reported results. Theoretical framework reference:Recursive Polycentric Governance (RPG)https://zenodo.org/records/18306005 Project website:https://sylebel.net
Sylvain Lebel (Fri,) studied this question.