When people argued about COVID-19 vaccine data on Twitter, opponents more often used language expressing certainty or pointing to cause and effect than supporters did.
The difference emerged from an analysis of 2,965 English-language posts. Ruijia Cheng, a human-computer interaction researcher now at Apple in the United States, and coauthors report the findings in Scientific Reports.
The team selected posts from October 2020 through April 2021 using vaccine-related hashtags and keywords, alongside terms such as “data” and “statistics.” Software searched the text for words associated with certainty and causation, as well as names of scientists, doctors and politicians.
About 34 percent of anti-vaccine posts contained certainty-related wording, compared with 21 percent of pro-vaccine posts. Language associated with cause and effect appeared in 44 percent versus 33 percent. Those differences remained after accounting for post length, account activity and follower counts.
One result surprised the researchers. They had expected vaccine opponents to mention prominent individuals more often, perhaps to challenge their credibility. Instead, pro-vaccine posts were more likely to name scientists, physicians or politicians.
The authors suggest that understandable language cues could help people think critically about health claims in their feeds. But the study compared patterns across groups; it did not establish a test for whether an individual post is true.
The sample captured a particular period of the pandemic and excluded posts without the selected keywords. Counting words is also an imperfect way to measure certainty or causal arguments. Mentions of organizations, such as the Centers for Disease Control and Prevention, were not counted as references to authority.
The researchers also caution that they “do not anticipate that the specific tactics and features we study here will necessarily remain stable or consistent over time.”
The research
Conspiracist attributes differentiate pro- and anti-vaccine online discourses about data ↗
Ruijia Cheng, Aaron Shaw and Benjamin Mako Hill
Scientific Reports · Accepted
DOI: 10.1038/s41598-026-76606-6