MIT Analytics Reveal How Anti-Maskers Leverage Data Visualization

You’ve probably seen them: the Facebook posts from a relative or a friend of a friend claiming that if you look at the numbers, the pandemic isn’t a big deal, masks don’t work – that kind of thing. Usually, there’s a chart or a graph accompanying them. It might be tempting to assume that those posts are relying on inherently flawed data – but, it turns out, that might not be the case. A new study from MIT is highlighting how data visualizations have become a “battleground” on social media, with differing sides of many pandemic debates leaning on the same data, but interpreting it in wildly different ways.

“A lot of people think of metrics like infection rates as objective,” says Crystal Lee, lead author of the study and a PhD candidate at MIT who works on the political dimensions of computing and data visualization, in an interview with MIT’s Daniel Ackerman. “But they’re clearly not, based on how much debate there is on how to think about the pandemic.”

Lee and her colleagues initially set out to collect data visualizations from social media, hypothesizing that the more data visualizations they could provide, the better-informed people would be. After scraping around half a million tweets mentioning “COVID-19” and “data” and analyzing the social connections that amplified various tweets, however, they found that anti-maskers were utilizing sophisticated data visualizations just as much as anyone else – and sometimes more.

A map of the social links between various prominent Twitter accounts posting visualizations. Dark blue is the “mainstream American politics and media” sphere; red is the “American politics and right-wing media” sphere; orange is the “British news media” sphere; teal is the “anti-maskers network” sphere; purple is the “World Health Organization and public health news” sphere.

“They are virtually indistinguishable from those shared by mainstream sources,” elaborated Arvind Satyanarayan, a professor in the Department of Electrical Engineering and Computer Science at MIT. “They are often just as polished as graphs you would expect to encounter in data journalism or public health dashboards.”

To validate this finding, the researchers engaged in “deep lurking,” following the day-to-day informal goings-on in the groups they were analyzing. The lurking confirmed the results: anti-maskers were talking a lot about data, but focusing on different elements of that data to draw their conclusions. Broadly, they eschewed accredited experts, instead teaching each other to perform basic data analysis.

The six-month study processed half a million tweets and 41,000 visualizations. The researchers used the resulting data to produce a number of their own visualizations, accessible here, that tell the story of those connected tweets and the wide range of expert and non-expert visualizations that have pervaded the conversation surrounding the pandemic. 

“Anti-maskers see the pandemic differently, using data that is quite similar,” Lee said. “I still think data analysis is important. But it’s certainly not the salve that I thought it was in terms of convincing people who believe that the scientific establishment is not trustworthy. … Data visualization is not objective. It’s not absolute. It is in fact an incredibly social and political endeavor. We have to be attentive to how people interpret them outside of the scientific establishment.”

“Convincing anti-maskers to support public health measures in the age of COVID-19 will require more than ‘better’ visualizations, data literacy campaigns, or increased public access to data,” the researchers concluded. “Rather, it requires a sustained engagement with the social world of visualizations and the people who make or interpret them.”