The Anti-Vaccination Infodemic on Social Media: A Behavioral Analysis

"In this historical time, it is imperative to understand the reasons of vaccine hesitancy, and to find effective strategies to dismantle the rhetoric of anti-vaccination supporters."

As of this writing, there is hope that vaccination can help halt the ongoing COVID-19 pandemic. However, the anti-vaccination movement has been on the rise, spreading online misinformation about vaccine safety and causing a reduction in vaccination rates worldwide. This study sought to identify potential communication strategies for use by healthcare organisations and professionals to decrease the spread of vaccine misinformation by quantitatively analysing the online behaviour of Twitter users.

In order to understand whether the success of the anti-vaccination discourse is due to a particularly pronounced activity of anti-vaccination supporters online, between September and December 2020, the researchers measured the number of Twitter actions on average in a month for each profile belonging to the control, anti-vaccination, and pro-vaccination group. Control profiles were selected for the use of randomly chosen hashtags; anti-vaccination users were identified for their use of the #vaccineskill and #vaccinesharm hashtags; and pro-vaccination communicators were identified for their use of the #vaccineswork hashtag. "Twitter actions" were defined as the sum of tweets, replies, and retweets.

Key results:

• Anti-vaccination supporters tweet less, but engage more in discussion - Anti-vaccination profiles were the most active on Twitter, with 536 actions per month, compared with an average of 277 actions for the control group and only 144 actions for the pro-vaccination group, suggesting the latter is not engaged enough and highlighting a pitfall in the pro-vaccine communication strategy online. However, a calculation of the number of tweets per month revealed that anti-vaccination supporters actually tweeting the least (42 tweets per month), when compared with control and pro-vaccination profiles (123 and 93 tweets per month, respectively). This was largely compensated by the engagement of the anti-vaccination group in discussions, be it through replies or retweets. Anti-vaccination profiles replied 13 times more than control and pro-vaccination profiles and retweeted 7.4 times more than their pro-vaccination counterparts and 31.3 times more than control profiles. These findings highlight that the majority of anti-vaccination supporters act as an echo chamber for the pool of content generated by a small fraction of users. Data also suggest that pro-vaccination individuals and groups are more prone to generate new content and are not very engaged with a broader community with similar interests.
• Anti-vaccination support on Twitter is associated with a general belief in conspiracy theories and emotional behaviours - Both pro- and anti-vaccination profiles shared a larger number of science- and vaccines-related contents when compared with control profiles. The anti-vaccination group was the only one circulating conspiracy theories. Most such tweets were associated with fake news concerning ruling elites, masonries, and techniques of population control - often associated with public figures such as Bill Gates or to ongoing COVID-19, flat earth ideology, or paedophilia scandals. The anti-vaccination group shared a larger number of emotional contents per month (and/or content with emotional language) when compared with the pro-vaccination group and control group. For the anti-vaccine group, there was a strong and significant correlation between the number of published contents against the use of vaccination and the number of published contents concerning conspiracy theories, suggesting that anti-vaccination support can be seen as a part of a bigger problem connected to beliefs in unsubstantiated claims.
• Emotional language could aid the success of vaccination campaigns - The most relevant words in the anti-vaccination group were "President", "God", "People", and "Masks". In contrast, pro-vaccination profiles preferentially included words such as "Help", "Health", "Thanks", or "Research". Clustering words according to topics revealed that anti-vaccination profiles were the most engaged in political discussion, with nearly a 6-fold increase compared with the pro-vaccination group. Finally, the researchers analysed whether the use of emotional contents and language was associated with increased engagement, measured as the sum of likes, replies, and retweets on each individual tweet, but found no significant correlation between the two factors for the anti-vaccination group. On the contrary, the pro-vaccine group showed a significant correlation between the two aforementioned factors, suggesting that the use of emotional language could aid the success of the pro-vaccination communication strategy online.
• Pro-vaccination supporters are more interested in their own education and profession - The majority of pro-vaccine profiles declared their identity when compared with the control (64% vs 30%, respectively), and anti-vaccination supporters were particularly reluctant to do so (only 16%). Similarly, education and/or profession in the Twitter headline was declared 32% of the times in the pro-vaccination group, compared with 10% and 6% in the control and anti-vaccination group, respectively.
• The pro-vaccination group produces the most engaging content - The average engagement per tweet was 19.9 times higher in the pro-vaccination group when compared with the anti-vaccination group (and 5.5 times higher when compared with the control group). On average, pro-vaccination profiles were also those with a larger number of followers, when compared with control and anti-vaccination groups. In light of these results, the researchers hypothesise that the success of the anti-vaccination movement is likely driven by a stronger sense of community, built around common interests (besides vaccines), and based on personal beliefs and emotional language. There is, in this community, a pull of influencers producing the most engaging contents, with the vast majority of anti-vaccination profiles functioning as the recipient and echo chamber for these messages, whereas novel contents produced by these profiles receive little attention when compared with contents generated by an average pro-vaccination supporter.
• Anti-vaccination supporters are engaged in a virtual community led by former United States (US) President Donald Trump, who has claimed that vaccines cause autism, and other influencers - The building of a Twitter web with Cytoscape revealed that the average number of neighbours in the web was 1.45-fold higher in the anti-vaccination web when compared with the pro-vaccination web. The clustering coefficient was also higher in the anti-vaccination web, as well as the density of the network and the characteristic path length. Furthermore, there was only one large influencer in the pro-vaccination web (the World Health Organization, or WHO), in contrast to 14 large influencers in the anti-vaccination web, with the largest one being Donald Trump, 5.2 times more relevant than the WHO in the pro-vaccine web. Among the other influencers: Trump's family members, politicians, and public figures known to support his presidency. "For some politicians, social media and fake news, including those concerning vaccines, could therefore be instrumental to hold on power and determining the future course of our global society."

The researchers discuss the polarisation of social media feeds, where users are exposed to information, news, and views identified by algorithms as close to their interests. Sharing or reading conservative tweets could increase the chance that a hesitant person gets in touch with anti-vaccination beliefs. They therefore:

• "strongly encourage social media to change the polarized way they present information to users to halt the anti-vaccination infodemic and increase debate between communities..."
• "...welcome initiatives to suspend profiles that clearly share misinformation about scientific topics and are likely to have significant negative effects on society..."
• suggest "'shadow bans' for science-based contents - which could force a tweet's organic reach to drop (i.e. a small number of people would read the content)..."
• propose "info banners for tweets containing unverified information about medically-sensitive topics..."
• "...encourage social media and the scientific community to discuss the possible introduction of science knowledge tests, which could be required for users that intend to share contents containing medically-sensitive information..."
• encourage health organisations to:
  • "consider restructuring decisional pathways to...include involving citizens in decision-making processes, thus building a more engaged community when it comes to public health policy..."
  • "lobby 'indirect' anti-vaccination influencers to become active pro-vaccination communicators..., [such as]...celebrities..."
  • "use...people-centered, first-person narratives with emotional language..., [which] are more attractive and stimulate empathic responses more efficiently....[That is,] adopt a less sterile, technical language when communicating with the general public..."
  • develop "a pro-active long-term strategy for increasing the general public's science literacy and ability to read and understand at least basic scientific information..."

"A combination of the aforementioned approaches could transform social media from sources of misinformation to valuable tools to gather trustworthy, relevant news and knowledge."