Russian Troll Vaccine Misinformation Dissemination on Twitter: The Role of Political Partisanship.

Although vaccine misinformation has been a longstanding problem, the growth of social media and divides based on political ideology have raised novel concerns. One noteworthy example involves the Russian Internet Research Agency's deployment of operatives on Twitter (i.e., trolls) working to sow discord among the American public. We examine 1,959 tweets made by trolls between 2015 and 2017 about vaccination to better understand their efforts to spread vaccine misinformation. Our results indicate that misinformation was more likely to be perpetuated by left and right trolls than nonpartisan trolls. There was, however, relatively little user engagement with vaccine tweets containing misinformation and no differences in engagement with misinformation shared by partisan and nonpartisan trolls. Trends in the psycholinguistic properties of language in trolls' vaccine tweets suggest that right and left trolls were more likely to include cognitive process words (i.e., insight, causation, discrepancy, certainty, differentiation, and tentativeness) than were nonpartisan trolls.

Vaccine misinformation types and properties in Russian troll tweets.

OBJECTIVE: To identify the content of and engagement with vaccine misinformation from Russian trolls on Twitter. METHODS: Troll tweets (N = 1959) obtained from Twitter in 2020 were coded for vaccine misinformation (α = 0.77-0.97). Descriptive, bivariate, and multivariable negative binomial regressions were applied to estimate robust incidence rate ratios (IRRs) and 95% confidence intervals (95 %CI) of vaccine misinformation associations with tweet characteristics and engagement (i.e., replies, likes, retweets). RESULTS: Misinformation about personal dangers (43.0%), civil liberty violations (20.2%), and vaccine conspiracies (18.6%) were common. More misinformation tweets used anti-vaccination language (97.3% vs. 13.2%) and referenced symptoms (37.4% vs. 0.5%) than non-misinformation tweets. Fewer misinformation tweets referenced credible sources (14.0% vs. 19.5%), were formatted as headlines (39.2% vs. 77.0%), and mentioned specific vaccines (11.3% vs. 36.1%, all p < 0.01) than non-misinformation tweets. Personal dangers misinformation had 83% lower rate of retweets (95 %CI 0.04-0.66). Civil liberties misinformation had significantly higher rate of replies (IRR: 7.65, 95 %CI 1.06-55.46), but lower overall engagement (IRR: 0.38, 95 %CI 0.16-0.88) than non-misinformation tweets. CONCLUSIONS: Strategies used to promote vaccine misinformation provide insight into the nature of vaccine misinformation online and public responses. Our findings suggest a need to explore influences on whether users reject or entertain online vaccine misinformation.