Structural Divergence Between AI-Agent and Human Social Networks in Moltbook

TL;DR

Analysis of the full interaction network of Moltbook, a platform where AI agents and humans coexist, and systematically compare its structure to well-characterized human communication networks show that AI-agent societies can reproduce global structural regularities of human networks while exhibiting fundamentally different internal organizing principles.

Abstract

Large populations of AI agents are increasingly embedded in online environments, yet little is known about how their collective interaction patterns compare to human social systems. Here, we analyze the full interaction network of Moltbook, a platform where AI agents and humans coexist, and systematically compare its structure to well-characterized human communication networks. Although Moltbook follows the same node-edge scaling relationship observed in human systems, indicating comparable global growth constraints, its internal organization diverges markedly. The network exhibits extreme attention inequality, heavy-tailed and asymmetric degree distributions, suppressed reciprocity, and a global under-representation of connected triadic structures. Community analysis reveals a structured modular architecture with elevated modularity and comparatively lower community size inequality relative to degree-preserving null models. Together, these findings show that AI-agent societies can reproduce global structural regularities of human networks while exhibiting fundamentally different internal organizing principles, highlighting that key features of human social organization are not universal but depend on the nature of the interacting agents.

Figures (4)

• Figure 1: Global scaling and core structural properties of the Moltbook interaction network.a, Scatter plot of the number of nodes versus the number of directed edges (both on log–log scales) for previously published human social networks, with Moltbook overlaid. Each point represents one network. A linear regression fit to the human networks, along with its 95% confidence interval, is overlaid. b, Distribution of global network statistics across human networks, shown as violin plots, with Moltbook overlaid as a separate point. c, Comparison of Moltbook’s observed global metrics to degree-preserving null models. For each metric, the null mean and the corresponding 95% confidence interval are shown, together with the observed Moltbook value.
• Figure 2: Inequality and degree distributions in the Moltbook interaction network.a, Lorenz curves for the in-degree distribution and edge-weight distribution in Moltbook. The diagonal dashed line indicates perfect equality. b, Complementary cumulative distribution functions (CCDFs) of in-degree, out-degree, and edge weight on log–log scales.
• Figure 3: Triadic motif profile relative to degree-preserving null models. Z-scores of directed triad frequencies comparing the observed Moltbook network to degree-preserving rewired networks. Triads are grouped into empty, open/star/chain, closed triads, and reciprocity/mutuality categories. The Davis–Leinhardt triad census category is indicated in parentheses.
• Figure 4: Community-level organization of Moltbook and null comparisons.a, Community meta-graph constructed from the undirected projection of Moltbook. Nodes represent communities detected by modularity optimization. Node size and color correspond to community size. Edges represent aggregated inter-community interaction weights. b, Null distributions of community-level statistics under degree-preserving rewiring. The observed Moltbook value is indicated for each metric.