Retraction: The anti-vaccination infodemic on social media: A behavioral analysis (PLoS ONE (2021) 16:3 (e02476420 DOI: 10.1371/journal.pone.0247642)

Abstract

Following the publication of this article [1], concerns were raised regarding the methodology, results, and conclusions presented in this article. The editorial team and a subject expert have re-evaluated the article and determined that the concerns listed below remain unresolved. • The sampling reported in the study is inadequate; the sample size is too small, and the study does not report adequate information to assess whether the selected sample is sufficiently representative. In addition, it appears no attempt was made to match profiles by profile characteristics such as number of followers, how long the profile has been active, or the number of tweets. • The anti-vaccination and pro-vaccination search terms used in the study may not have been balanced appropriately and the study does not report on the justification for the choice of hashtags used. Similarly, the use of a random word generator to create a random hashtag to use as control is inappropriate and suggests that the study did not include an appropriate characterization of underlying Twitter behavior. • The study does not provide an adequate definition of “emotional language”, and the related results reporting on the use of emotional language include an outlier data-point in the pro-vaccine group, which could drive the effect significantly in a study with a small sample size. • The network analysis includes only a small number of profiles with an unbalanced number of neighbors. In addition, the clustering coefficient is inappropriate and the absence of confidence intervals in Fig 5C is problematic. As currently presented, these results are not sufficient to draw meaningful conclusions. • The reported conclusion “Our data demonstrate that Donald Trump, before his profile was suspended, was the main driver of vaccine misinformation on Twitter.” is not supported by the research reported in this study. Although the reported results suggests that people who tweet anti-vaccine content are likely to be in Trump’s network, the reported results are not sufficient to support the claim that Trump himself is driving vaccine misinformation. The authors commented that the retrieved Twitter webs and associated communities of pro- and anti-vaccine users and influencers present a snapshot taken at a given moment in time, which can be expected to change with changing opinion, circulating contents, and relevance of users within and outside of the community. Furthermore, they clarified that the study is a proxy to catch anti- and pro-vaccine discourse on Twitter and does not claim that their study is comprehensive of all vaccine-related-views on social media. In response to the issues raised above, the authors explained that the hashtags used in the study are the most widely adopted by the pro-vaccine and anti-vaccine community, since, according to the authors, most users within the pro- and anti-vaccination communities used the chosen hashtags. The authors agreed that a different choice of hashtags would likely have led to different results but stated that these results would present a similar underlying interpretation, meaning, and value.