Exploring the complex pattern of information spreading in online blog communities

Information spreading in online social communities has attracted tremendous attention due to its utmost practical values in applications. Despite that several individual-level diffusion data have been investigated, we still lack the detailed understanding of the spreading pattern of information. Here, by comparing information flows and social links in a blog community, we find that the diffusion processes are induced by three different spreading mechanisms: social spreading, self-promotion and broadcast. Although numerous previous studies have employed epidemic spreading models to simulate information diffusion, we observe that such models fail to reproduce the realistic diffusion pattern. In respect to users behaviors, strikingly, we find that most users would stick to one specific diffusion mechanism. Moreover, our observations indicate that the social spreading is not only crucial for the structure of diffusion trees, but also capable of inducing more subsequent individuals to acquire the information. Our findings suggest new directions for modeling of information diffusion in social systems and could inform design of efficient propagation strategies based on users behaviors.

💡 Research Summary

**
This paper presents a comprehensive empirical study of information diffusion in the LiveJournal (LJ) blog community, leveraging a massive dataset that includes approximately 56 million public posts generated over a 21‑month period, a friendship network of 9.57 million users, and 188 million undirected social links. By extracting hyperlinks embedded in posts and matching them to earlier posts, the authors reconstruct explicit information flow from a source user to a destination user, enabling a fine‑grained, individual‑level view of diffusion events.

The central contribution is the identification of three distinct diffusion mechanisms: (1) social spreading, where the source and destination are directly connected in the friendship network; (2) self‑promotion, where a user cites his or her own earlier post; and (3) broadcast, which covers all remaining citations that occur between users who are not friends. Quantitatively, social spreading accounts for 26.8 % of all citation links, self‑promotion 31.14 %, and broadcast 42.06 %. This composition contrasts sharply with platforms such as Twitter, where social contagion typically exceeds 70 % of diffusion, highlighting the particular structural features of LJ—especially its large number of topic‑based communities that allow information exchange beyond explicit friend ties.

To analyze the structural properties of diffusion, the authors reconstruct diffusion trees using a breadth‑first search (BFS) algorithm that follows citation links chronologically. They obtain 880,195 diffusion trees, whose size and depth follow power‑law distributions with exponents ≈ 1.86 and ≈ 2.97, respectively. Approximately 63 % of posts are cited only once, indicating that most cascades are shallow. When focusing exclusively on social‑spreading trees (363,115 trees), 85 % have depth ≤ 3, confirming that deep cascades are rare. Nevertheless, the few deep trees contribute disproportionately to the total number of infected nodes, suggesting that social spreading, though less frequent, is crucial for large‑scale reach. The average branching factor exceeds one in the first few generations but quickly converges to one for depths beyond 20, implying that the bulk of diffusion occurs early in the cascade.

Behavioral analysis reveals a strong tendency for users to specialize in a single diffusion mechanism. Most users either repeatedly promote their own content (self‑promotion) or primarily forward information through friend links (social spreading). This behavioral persistence likely reflects individual roles, preferences, and the way users perceive value in the platform.

The authors also evaluate the classic susceptible‑infectious‑recovered (SIR) epidemic model using the empirically measured infection probability (β ≈ 0.001). Simulations generate diffusion trees that are markedly different from the observed ones: SIR predicts many multi‑generation infections, whereas LJ diffusion is dominated by one‑ or two‑step cascades, and the model cannot reproduce the substantial broadcast component. This discrepancy underscores the inadequacy of pure epidemic analogies for modeling online information spread.

In summary, the study demonstrates that information diffusion in online blog communities is a composite process involving (i) network‑driven social contagion, (ii) deliberate self‑promotion, and (iii) platform‑mediated broadcast. Social spreading, despite its modest share, is the primary engine for deep cascades and for triggering subsequent diffusion events. The findings suggest that future diffusion models should integrate multiple mechanisms and incorporate user‑level behavioral heterogeneity rather than relying on single‑process epidemic frameworks. Such enriched models would better capture real‑world dynamics and could inform the design of more effective information propagation strategies, whether for marketing, public health messaging, or counter‑misinformation campaigns.