Abstract
During the early stages of the COVID-19 pandemic, concerns about a "misinformation infodemic" were widespread. However, this study demonstrates that Twitter users significantly amplified the voices of experts and authorities. Accounts categorized under healthcare, science, and government experienced surges in engagement that far exceeded their increases in posting volume. This suggests a collective public shift toward seeking and promoting trustworthy information during the global health emergency.
Introduction
The World Health Organization (WHO) identified an "infodemic" early in the COVID-19 crisis, fearing that low-quality information would overshadow critical public health guidance. While social media is often criticized for the rapid spread of "fake news," this research investigates the "structural importance" of accounts—such as scientists and government officials—to see if their expertise translated into increased attention during the pandemic.
Methodology
The researchers analyzed Twitter activity from January to May 2020 using a multi-step pipeline:
- Taxonomy: Accounts were classified into 13 categories (e.g., Healthcare, Science, Government).
- Sampling: 14,200 accounts with popular COVID-19 tweets were sampled.
- Classification: A BERT-based machine learning model was trained to categorize millions of users across the COVID-19 tweet stream.
- Analysis: The study compared "tweet volume" (posting frequency) against "engagement volume" (retweets and likes) using Inverse Probability Weighting (IPW) to ensure representative results.
Results: The Selective Attention Boost
The study found that while posting increased across most categories, the surge in engagement was highly selective. Authoritative sources received a disproportionate "attention boost."
As shown in the figure above, categories like Government and Politics, Healthcare, and Science saw engagement peaks (+402%, +319%, and +281% respectively) that vastly outpaced their increases in posting volume. For instance, government accounts doubled their posting but quadrupled their engagement. In contrast, categories like Sports and Religion saw their "effectiveness" (engagements per tweet) decline as public attention shifted toward crisis-relevant information.
Discussion: Expertise vs. Clout
A key finding of the research is that traditional social media "clout" (follower count) was not the primary driver of engagement during the crisis.
The data indicates that Healthcare and Science accounts, despite often having lower average follower counts than News Media or Sports accounts, achieved significantly higher engagement ranks. This suggests that the relevance of expertise became the dominant factor in information sharing. Furthermore, the study distinguishes between "transient" authority (spikes in Healthcare/Science engagement that eventually regressed) and "persistent" authority (sustained high engagement for Government accounts).
Conclusion and Limitations
The macro-level trends on Twitter during the first wave of COVID-19 provide a counter-narrative to the "infodemic" fear, showing that "voices of reason" were the most widely shared. However, the study has limitations: it relies on account bios for categorization rather than individual tweet content, and it measures attention rather than the scientific accuracy of the messages. Future work should explore content-level quality and how these attention patterns evolve as a crisis transitions from a health emergency to a long-term societal issue.
Summary updated