Abstract
We investigate the diffusion of conspiracy theories related to the origin of COVID-19 on social media. By analyzing third-party content on four social media platforms, we show that: (a) In contrast to conventional wisdom, mainstream sources contribute overall more to conspiracy theories diffusion than alternative and other sources; and (b) platforms’ content moderation practices are subject to a "mainstream bias."
Introduction
The early stages of the COVID-19 pandemic were characterized by a "misinfodemic"—a rapid surge of conspiracy theories that filled the void left by scientific uncertainty. While social media platforms implemented various content moderation strategies, the quantitative dynamics of how these theories propagate across different digital ecosystems remained poorly understood. This research addresses three primary questions: the cross-platform dynamics of conspiracy theories, the relative influence of mainstream versus alternative sources, and the actual efficacy of content moderation in slowing their spread.
Methodology
The study utilized a multi-stage analytical pipeline to process a massive cross-platform corpus from January to March 2020, including 9.5 million Reddit submissions, 4.2 million Facebook posts, 83 million tweets, and 1.5 million 4chan posts.
- Manual Content Analysis: A sample of 11,023 unique URLs was manually classified into three categories: Supporting (reinforcing a conspiracy theory), Evidence (used as a basis for a theory), or Neither. Sources were further categorized as Mainstream, Alternative, or Other.
- Hawkes Processes: To quantify the "virality" of content, the researchers employed Hawkes processes. This mathematical model estimates a virality parameter ($\alpha$), representing the probability that a post on one platform triggers subsequent activity on another.
- Moderation Tracking: By crawling the platforms in both April and May 2020, the team identified which content was removed or flagged, allowing for a direct comparison of the diffusion patterns between moderated and unmoderated posts.
The Mainstream Paradox
A common assumption is that conspiracy theories are primarily the product of fringe, alternative media outlets. However, this research reveals a "mainstream paradox." While alternative sources produce a higher volume of conspiracy-related stories, mainstream sources—including the New York Post, Fox News, and even scientific preprint servers like bioRxiv—exert a significantly greater impact on diffusion.
As shown in Figure 6, mainstream URLs supporting conspiracy theories were shared four times more on average than those from alternative sources. Despite accounting for only 17% of the conspiracy-reinforcing stories, mainstream sources were responsible for 60% of Facebook shares and 55% of Twitter shares. This suggests that the perceived credibility of mainstream outlets facilitates wider dissemination, as users are more likely to trust and share content from these sources.
Cross-Platform Diffusion Dynamics
The study rejects the notion that fringe platforms like 4chan operate in isolation. Instead, 4chan often acts as a "feeder" for the broader social media ecosystem.
The analysis of Hawkes process parameters (Figure 3) demonstrates that conspiracy content moves fluidly between platforms. Specifically, content often originates on 4chan, migrates to Twitter, and eventually reaches Facebook. This chain of diffusion highlights the interconnected nature of modern social media, where fringe discussions can rapidly scale to reach mainstream audiences.
Content Moderation: Efficacy and Bias
While content moderation was found to reduce the probability of a story reappearing in the ecosystem, its effectiveness was hampered by two major factors: timeliness and bias.
- Mainstream Bias: There is a statistically significant bias in how platforms apply moderation. Conspiracy-reinforcing stories originating from mainstream sources were 10% to 30% less likely to be removed than those from alternative sources.
- The Velocity of Twitter: Moderation on Twitter was notably less effective, reducing in-platform virality by less than 10%, compared to over 90% on Facebook and Reddit. The high-velocity nature of retweets means that by the time content is flagged, it has often already achieved its peak reach.
- Containment Gaps: Platforms only managed to catch between 15% and 50% of conspiracy-related posts, often weeks after they had gone viral.
Conclusion
The findings suggest that current moderation strategies are insufficient to contain the spread of high-impact misinformation. The "mainstream bias" and the speed of diffusion on platforms like Twitter present significant challenges. Future interventions should move beyond simple content removal toward design-based solutions that sensitize users to conspiracy logic and address the disproportionate influence of reputable sources in spreading harmful narratives.
Summary updated