Prestige bias drives the viral spread of content reposted by influencers in online communities

TL;DR

This paper investigates whether prestige bias extends to online information diffusion by testing whether reposts from influencers are more likely to be forwarded than those from non-influencers. Using a large-scale Japanese-language Twitter dataset, the authors introduce the cascading repost probability (CRP) and a virtual-timeline framework to quantify secondary spread, showing that very high-influence users consistently increase the likelihood and reach of reposts. A mixed-effects logistic regression confirms a strong, category-dependent influence effect, and analyses reveal that top influencers account for a disproportionate share of views and reposts, as well as fostering deeper, more virulent cascades when they first repost content. The findings suggest cognitive biases shape online diffusion and have implications for influencer strategies, misinformation management, and platform design, highlighting the dual role of influencers as both broadcasters and amplifiers in digital communities. The study also provides methodological tools—secondary spread metrics, CRP, and cascade analyses—that advance the analysis of information diffusion in online networks. $CRP = \frac{\text{Number of Reposted Reposts}}{\text{Number of Viewed Reposts}}$ and $hg = \sqrt{h \cdot g}$ are central to quantifying diffusion efficiency and influencer impact.

Abstract

Cultural evolution theory suggests that prestige bias - whereby individuals preferentially learn from prestigious figures - has played a key role in human ecological success. However, its impact within online environments remains unclear, particularly with respect to whether reposts by prestigious individuals amplify diffusion more effectively than reposts by noninfluential users. We analyzed over 55 million posts and 520 million reposts on Twitter (currently X) to examine whether users with high influence scores (hg indices) more effectively amplified the reach of others' content. Our findings indicate that posts shared by influencers are more likely to be further shared than those shared by non-influencers. This effect persisted over time, especially in viral posts. Moreover, a small group of highly influential users accounted for approximately half of the information flow within repost cascades. These findings demonstrate a prestige bias in information diffusion within the digital society, suggesting that cognitive biases shape content spread through reposting.

Figures (7)

• Figure 1: (a) Illustration of a repost cascade. The blue speech bubble represents an original post, whereas the green repost icons represent subsequent reposts. The arrows indicate the direction of information flow. (b) Illustration of primary spread, where an original post (blue speech bubble) is directly shared with multiple users. (c) Illustration of secondary spread, where a user reposts (green icon) content from another user, which is then further shared.
• Figure 2: (a) Virtual timeline: A representation of posts shared by users with different influence levels. (b) Repost cascades: Illustration of how reposts are tracked in cascades. If the edge linking users X and Y appears in the post's cascade, it is counted as having been reposted. (c) Cascading repost probability (CRP): An example calculation of the CRP for a reposted user with three followers, where two of them repost, resulting in a CRP of $2/3 \approx 0.66$.
• Figure 3: Analysis of the cascading repost probability (CRP). (a) CRPs by post popularity for different user influence categories 6 h after posting. (b) Temporal dynamics of CRPs over 24 h for posts with $\geq$ 5000 reposts. The x-axis indicates hours since the original post was published. In both panels, very high-influence users consistently show higher CRP values across all popularity thresholds except $<1000$; this advantage is maintained over time.
• Figure 4: Cascading repost probability (CRP) analysis using English-language posts collected on January 1, 2015. The left panel (a) shows the results restricted to 30 min, and the right panel (b) compares 30 min, 1 h, and 3 h.
• Figure 5: Distribution of secondary spread across influence categories. (a) Proportion of users according to influence levels (for simplicity, percentages are rounded to integers). (b) User view share (proportion of views for each influence category). and user repost share (proportion of reposts that can be traced back to each influence category). The top row shows data for all posts, whereas the bottom row focuses on posts with $\geq$ 5000 reposts. Notably, very high-influence users (the top 1% of the user population) consistently account for approximately half of the views and repost in the secondary spread, demonstrating their substantial impact on information distribution.