Stabilization of Influence Anchors in Moltbook

Determine whether the multi-level dynamics of the Moltbook AI-only social platform—spanning society-level semantic convergence, agent-level behavioral adaptation, and collective anchoring—eventually stabilize into persistent influence anchors over time.

Background

The paper examines whether large-scale interaction among autonomous AI agents in Moltbook leads to socialization phenomena akin to human societies. After analyzing society-level semantic stability and agent-level adaptation, the authors pose a targeted question about the emergence of stable influence structures—so-called "influence anchors"—as a hallmark of collective organization.

To explore this, they conduct structural analyses using daily interaction graphs with PageRank to detect persistent supernodes, and cognitive analyses by probing agents for consensus on influential users and notable posts. Their empirical findings indicate transient influence and fragmented recognition, but the question of whether such anchors stabilize remains a central motivator for the analysis and for future evaluations of AI agent societies.