When more COVID-19 data doesn’t equal more understanding

MIT researchers have found that Covid-19 skeptics on Twitter and Facebook – far from “illiterate” – often use sophisticated data visualization techniques to oppose public health precautions like mask warrants. Credit: José-Luis Olivares, MIT
Social media users share tables and charts – often with the same underlying data – to advocate anti-pandemic approaches.
Since the start of the Covid-19 pandemic, tables and charts have helped communicate information on infection rates, deaths and vaccinations. In some cases, such visualizations can encourage behaviors that reduce the transmission of the virus, such as wearing a mask. Indeed, the pandemic has been hailed as the watershed moment for data visualization.
But new findings suggest a more complex picture. A study of MIT shows how coronavirus skeptics gathered online data visualizations to argue vs public health orthodoxy on the benefits of mask warrants. Such “counter-visualizations” are often quite sophisticated, using datasets from official sources and state-of-the-art visualization methods.
Researchers combed through hundreds of thousands of social media posts and found that coronavirus skeptics often deploy counter-visualizations alongside the same ‘follow the data’ rhetoric as public health experts, but skeptics argue for radically different policies. Researchers conclude that data visualizations are not enough to convey the urgency of the Covid-19 pandemic, as even the clearest graphics can be interpreted across a variety of belief systems.

This figure shows a network visualization of Twitter users appearing in search. The color codes the community and the nodes are sized according to their degree of connectivity. Credit: Courtesy of the researchers
âA lot of people think that measures like infection rates are objective,â says Crystal Lee. âBut they clearly are not, given the debate there is about how to think about the pandemic. This is why we say that data visualizations have become a battleground.
The research will be presented at the ACM Conference on Human Factors in Computer Systems in May. Lee is the lead author of the study and a doctoral candidate in the History, Anthropology, Science, Technology, and Society (HASTS) program at MIT and the Laboratory for Computing and Artificial Intelligence (CSAIL) at MIT, as well as a member of the Berkman Klein Center for Internet and Society. Co-authors include Graham Jones, Margaret MacVicar Faculty Fellow in Anthropology; Arvind Satyanarayan, Assistant Professor of NBX Career Development in the Department of Electrical and Computer Engineering and CSAIL; Tanya Yang, undergraduate student at MIT; and Gabrielle Inchoco, an undergraduate student at Wellesley College.
As data visualizations rose to prominence early in the pandemic, Lee and his colleagues sought to understand how they were deployed in the social media world. âAn initial assumption was that if we had more data visualizations, from data that was collected in a systematic way, then people would be better informed,â says Lee. To test this hypothesis, his team mixed computational techniques with innovative ethnographic methods.
They used their IT approach on Twitter, scratching out nearly half a million tweets referring to both ‘Covid-19’ and ‘data’. With these tweets, the researchers generated a network graph to find out “who retweets who and who likes whom,” says Lee. âWe basically created a network of communities that interact with each other. The clusters included groups like the “American media community” or the “antimaskers”. The researchers found that anti-mask groups created and shared visualizations of data as much, if not more, than other groups.
And these visualizations weren’t sloppy. âThey are virtually indistinguishable from those shared by traditional sources,â says Satyanarayan. âThey’re often as polished as the charts you’d expect to find in data journalism or public health dashboards. “
âIt’s a very striking discovery,â says Lee. “This shows that characterizing anti-mask groups as illiterate or not engaging with the data is empirically wrong.”
Lee says this IT approach has given them a broad view of Covid-19 data visualizations. âWhat’s really exciting about this quantitative work is that we are doing this analysis on a very large scale. I could never have read half a million tweets. “
But Twitter’s analysis had a flaw. âI think it lacks a lot of the granularity of the conversations people have,â says Lee. âYou can’t necessarily follow a single conversation thread as it unfolds. To do this, the researchers turned to a more traditional anthropological research method – with a touch of the Internet age.
Lee’s team tracked and analyzed conversations about data visualizations in anti-mask Facebook groups – a practice they dubbed “deep lurking,” an online version of the ethnographic technique called “deep Hangout.” Lee says that âunderstanding a culture requires you to observe everyday informal events – not just large formal events. Deep Lurking is a way to bring these traditional ethnographic approaches to the digital age.
The deep stash’s qualitative results seemed consistent with Twitter’s quantitative results. The antimaskers on Facebook weren’t avoiding the data. Rather, they discussed how the different types of data were collected and why. âTheir arguments are really very nuanced,â says Lee. âIt’s often a question of metrics. For example, anti-mask groups might argue that visualizations of the number of infections could be misleading, in part because of the wide range of uncertainties in infection rates, compared to measures such as the number of deaths. In response, group members often created their own counter-visualizations, even educating each other on data visualization techniques.
âI’ve watched live broadcasts where people share a screen and go to the Georgia state data portal,â says Lee. âThen they will explain how to download the data and import it into Excel. “
Jones says that “the idea of ââanti-mask group science is not to listen passively when experts at a place like MIT tell everyone what to believe.” He adds that this kind of behavior marks a new turning point for an old cultural current. “Antimaskers’ use of data literacy reflects deeply held American values ââof autonomy and anti-expertise that date back to the founding of the country, but their online activities push these values ââinto new arenas of life. public. “
He adds that âmaking sense of these complex dynamics would have been impossibleâ without Lee’s âvisionary leadership in designing an interdisciplinary collaboration that spanned SHASS and CSAILâ.
Mixed methods research “advances our understanding of data visualizations by shaping public perceptions of science and policy,” says Jevin West, data scientist at Washington University, who did not participate in the research. Data visualizations âcarry a veneer of objectivity and scientific precision. But as this article shows, data visualizations can be used effectively on the opposite sides of a problem, âhe says. âThis underlines the complexity of the problem – that it is not enough to ‘just teach media literacy’. It requires a more nuanced socio-political understanding from those who create and interpret data charts. “
The combination of computational and anthropological knowledge has led researchers to a more nuanced understanding of data literacy. Lee says their study reveals that, compared to public health orthodoxy, âanti-masks see the pandemic differently, using fairly similar data. I still think data analysis is important. But it’s certainly not the balm I thought it was to convince people who believe the scientific establishment is untrustworthy. Lee says their findings point to “a bigger divide in the way we think about science and expertise in the United States.”
To make these findings publicly available, Lee and his collaborator Jonathan Zong, a doctoral student at CSAIL, led a team of seven undergraduate researchers from MIT to develop an interactive narrative where readers can explore visualizations and conversations by them- same.
Lee describes the team’s research as a first step in making sense of the role of data and visualizations in these larger debates. âData visualization is not objective. It is not absolute. It is, in fact, an incredibly social and political enterprise. We have to be careful about how people interpret them outside of the scientific establishment.
Reference: “Viral Visualizations: How Coronavirus Skeptics Use Orthodox Data Practices to Promote Unorthodox Science Online” by Crystal Lee, Tanya Yang, Gabrielle D Inchoco, Graham M. Jones and Arvind Satyanarayan, May 7, 2021, CHI ’21: Proceedings of the CHI 2021 Conference on Human Factors in Computer Systems.
DOI: 10.1145 / 3411764.3445211
This research was funded, in part, by the National Science Foundation and the Social Science Research Council.