Misrepresenting Scientific Consensus on COVID-19: The Amplification of Dissenting Scientists on Twitter

TL;DR

This study investigates how dissenting scientists are amplified on Twitter during COVID-19, revealing a significant skew towards misinformation and false consensus, especially following popular dissent statements, with implications for social media moderation.

Contribution

It quantifies the amplification of dissenting scientists versus consensus scientists on Twitter during COVID-19, highlighting the role of engagement and viral spikes in spreading misinformation.

Findings

Dissenting scientists are amplified by a factor of 426 for vaccines and 43 for COVID-19 overall.

Higher engagement with dissent tweets drives false consensus more than the number of mentions of consensus scientists.

False consensus spikes follow highly popular dissent statements, especially in multilingual misinformation clusters.

Abstract

The COVID-19 pandemic has resulted in a slew of misinformation, often described as an "infodemic". Whereas previous research has focused on the propagation of unreliable sources as a main vehicle of misinformation, the present study focuses on exploring the role of scientists whose views oppose the scientific consensus. Using Nobelists in Physiology and Medicine as a proxy for scientific consensus, we analyze two separate datasets: 15.8K tweets by 13.1K unique users on COVID-19 vaccines specifically, and 208K tweets by 151K unique users on COVID-19 broadly which mention the Nobelist names. Our analyses reveal that dissenting scientists are amplified by a factor of 426 relative to true scientific consensus in the context of COVID-19 vaccines, and by a factor of 43 in the context of COVID-19 generally. Although more popular accounts tend to mention consensus-abiding scientists more, our results suggest that this false consensus is driven by higher engagement with dissent-mentioning tweets. Furthermore, false consensus mostly occurs due to traffic spikes following highly popularized statements of dissenting scientists. We find that dissenting voices are mainly discussed in French, English-speaking, Turkish, Brazilian, Argentine, Indian, and Japanese misinformation clusters. This research suggests that social media platforms should prioritize the exposure of consensus-abiding scientists as a vehicle of reversing false consensus and addressing misinformation stemming from seemingly credible sources.