CDE member Dr Gillian Bolsover and Laidlaw Scholar Janet Tokitsu Tizon write about their research on misinformation during the COVID-19 crisis in the US.
COVID-19 has upended and suspended many facets of social life. Universities, education and research, for whom in-person communication and knowledge exchange are a lifeblood, have been sorely affected by this pandemic. University campuses have shuttered their doors and the focus has shifted to adapting to online learning and providing additional pastoral care to students and graduates. These changes look set to continue, at least for the next year. However, in times of societal change, social science research holds a vital place. In order to contribute to knowledge and conversation on COVID and society, we have published a short data memo addressing an important issue in COVID discussions: health misinformation on social media.
We now live in a situation where more than half of adults get their news off social media and sometimes do not know the providence of the information they are exposed to online. This exposure may happen passively, without the user actively seeking information on that issue. The spread of the disease depends on personal actions, such as maintaining social distancing, hand-washing and mask-wearing, and these personal actions are underpinned by the information individuals have as well as their attitudes and beliefs and economic and personal constraints. However, in the case of a newly evolved disease, we lack accurate information on the disease. In these conditions, misinformation can thrive.
Social media has often been a source of misinformation both within and outside public health crises. In the West African Ebola epidemic in 2014, misinformation circulated widely on social media, with Twitter feeds showing posts stating that Ebola could be cured with blood transfusions or by the subtropical plant ewedu. Similarly, in the case of the 2015 Zika virus in Brazil false claims circulated on social media that GM mosquitos brought Zika to the country and that government vaccines were responsible for birth defects. It is not surprising therefore that false information about COVID’s origins, spread, severity, preventions and treatments have circulated on social media. Research has also shown that social media can play a role in fighting misinformation but that the way that the content of these refutations and retraction, as well as who is issuing them is crucial in whether individuals continue to believe mis-informative claims they are exposed to.
An important case study for misinformation dissemination occurred on 23 April, when US President Donald Trump appeared to propose scientific research into promising and powerful treatments for the disease: UV or high-powered light and disinfectant administered to the body through injection or other means. These comments received widespread outrage and condemnation, with the state and medical officials and the parent company of disinfectants Dettol and Lysol giving official statements condemning the administration of disinfectant products into the human body. Trump later claimed that these comments were made sarcastically. Nevertheless, disinfectant poisonings appeared to rise after Trump’s comments and several US localities also reported significant rises in poison control centre calls.
This case presents an ideal one for investigating the spread of and response to health misinformation on social media. In order to investigate this, the data memo investigates Twitter data from the day that the comments were made (23 April) and the three following days (24 – 26 April) to understand how these comments were received and responded to online. How prevalent was discussion of the comments? Did people appear to believe them? How were attempts to refute this misinformation handled? Did this refutation happen in a way consistent with what research suggests is best practices for misinformation refutation or were they likely to exacerbate the problem?
The research first examined the content of the top 200 Twitter trending topics from across the US during the study period. It then examined a random sample of 250 tweets from within political trending topics. The research found that Trump’s comments dominated discourse about politics and COVID on Twitter during the period: 80% of the posts about the coronavirus in the US and 61% of the posts about politics in the US either contested or supported Trump’s comments. Posts contesting Trumps comments were more than twice as prevalent as those supporting. However, these posts largely fell along partisan lines.
There was little evidence that individuals believed the comments but those that supported the comments attacked the media response to the comments, which added more fuel to the fire of the Trump narrative of being constantly and unfairly attacked by a lying and dishonest press. Posts that contested Trump’s comments much more frequently shared news or information rather than personal opinion. This included the retweeting of a CNN International post, a post by CNN anchor Jim Sciutto reporting Lysol’s statement and an NBC New York report that New York City poison control calls doubled after Trump’s comments. However, many contesting comments also used divisive language. In general, comments were polarising and divisive, with 91% appearing to be attempting to prevent others from speaking or undermine the value of their words and 16% appearing to be directing hate against a specific opposing group (that wasn’t a political party). This type of discourse does not evidence best practices for refuting misinformation in that it repeated that misinformation in its refutation and fell within existing partisan narratives and processes.
This level of polarisation and hate is cause for concern in that the dissemination of health information surrounding a global pandemic should not be a political issue but rather one in which information and responses are de-politicised. However, in an election year, it is perhaps not surprising that contestation over the truth about these comments has become highly polarised on Twitter. Although there was scant evidence of this being the case before, many see COVID-19 as an opportunity to change both civil society and politics for the better.
Based on this research, we make two recommendations for the handling of health misinformation on social media during this time in which correct and accurate understanding of the best-available information can be a matter of life or death. Firstly, that misinformation should not be repeated. In the current pandemic as well as more generally, we need to re-assess the idea that something is newsworthy because someone well-known says it. Both media coverage and Twitter discourse of Trump’s comments would have served to strengthen them through familiarity and to disseminate the information to those whose existing belief and attitudes would leave them susceptible to this misinformation. Secondly, in the current context of limited scientific information and heightened-public desire for information, misinformation will spread. In this situation, it is crucial that evidence-backed public health information about the disease is repeated frequently in a variety of formats, from non-political voices and in a scientifically-backed way to counteract the misinformation that will proliferate and circulate on social media, exacerbated by 24-hour news cycles and the public desire for certainty and answers.
Download the full data memo.