Since the beginning of the COVID-19 pandemic, charts and graphs have helped convey information about infection rates, deaths, and vaccinations. In some cases, this display can encourage behaviors that reduce the spread of the virus, such as wearing a mask. The pandemic has been hailed as a breakthrough in data visualization.

But the new findings reveal a more complicated situation. A Massachusetts Institute of Technology (MIT) study showed how coronavirus skeptics use online data visualization to oppose the benefits of orthodox public health mask regulations. This type of “anti-visualization” is usually quite complex, since it uses official data sets and the most advanced visualization methods.

The researchers reviewed hundreds of thousands of social media posts and found that COVID-19 skeptics often use anti-visualization while using the same “data tracking” statements as public health experts. However, the policies advocated by skeptics are not entirely different. The researchers concluded that data visualization is not enough to convey the urgency of the COVID-19 pandemic, because even the clearest diagrams can be explained by various belief systems.

“A lot of people think that indicators like infection rates are objective,” Crystal Lee said, “but obviously they’re not, based on how big the debate is about how to think about pandemics. That’s why we say visualization of data has become a battlefield. “

This research will be presented at the ACM Computer Systems Human Factors Conference in May. Lee is the first author of this study and a doctoral student in the MIT History, Anthropology, Science, Technology and Society (HASTS) project and the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), and also a researcher. from Harvard Berkman at the Klein Center for Internet and Society. ZE co-authors include Margaret-McVega College researcher Graham Jones in the field of anthropology, department of electrical engineering and computer science, and CSAIL NBX assistant professor of professional development Arvind Satyanarayan, MIT undergraduate student Tanya. Yang and Wellesley College student Give birth to Gabrielle Inchoco.

With the rise of data visualization in the early stages of this epidemic, Lee and his colleagues began to understand how it was implemented in the realm of social media. “An initial assumption is that if we have more data visualization and we collect data in a systematic way, people will get better information,” Lee said. To test this hypothesis, his team combined computer technology with innovative ethnographic methods.

They used their calculation method to collect close to 500,000 tweets mentioning “COVID-19” and “data” on Twitter. From these tweets, the researchers generated a network map to find out who retweets whom and who likes whom. Lee said, “We basically created a community network for mutual exchange.” The groups include groups such as the “American media community” or the “anti-masks.” Researchers have found that there are as many anti-mask groups as other groups in creating and sharing data visualization, or even more than them.

However, these displays are not neglected. Satyanarayan noted: “They are almost the same as the data shared by major sources. They are often as intricate as the graphs you would expect to find in the data news or on public health dashboards.”

“This is a very compelling finding. It shows that it is empirically incorrect to characterize the anti-mask group as data illiterate or not participating in the data,” Lee said. Furthermore, he also said that this calculation method has given them a broad understanding of COVID-19 data visualization. “What’s really exciting about this quantitative work is that we do this analysis on a large scale. It’s impossible for me to read 500,000 tweets.”

But the Twitter analysis has a shortcoming. “I think you lose a lot of the granularity of people’s ongoing conversations. You don’t necessarily follow the development of a conversation line,” Lee said. For this reason, the researchers turned to a more traditional anthropological research method, with the distortion of the Internet age.

Lee’s team tracked and analyzed data visualization conversations on the anti-mask Facebook group; they called this approach “deep latency.” “Understanding a culture requires that you look at everyday informal activities, not just large-scale formal events. Deep latency is a way to transplant these traditional ethnographic methods into the digital age,” Lee said.

The qualitative deep latency findings appear to be consistent with the quantitative findings from Twitter. Anti-submarine agents on Facebook did not shy away from the data. Instead, they discussed how to collect different types of data and why. Lee said: “Their argument is really subtle. This is often a question of measurement standards. For example, the anti-mask team may argue that the display of infection numbers can be misleading, in part due to measurement standards. such as the number of deaths. In contrast, the range of uncertainty in the infection rate is wide. In response, team members often create their own anti-visualizations, and even guide each other in data visualization techniques. ” .

Jones pointed out that the scientific philosophy of anti-mask groups is not to passively listen to experts in places like MIT telling others what to believe. He believes that this behavior marks a new inflection point in an old cultural trend. “The use of data literacy by anti-masks reflects the deeply ingrained values ​​of self-reliance and anti-experts in the United States, dating back to the beginning of the nation’s founding, but their online activities they have brought these values ​​to new areas of public life. “

In addition, he added, “Without Lee’s visionary leadership, planning an interdisciplinary collaboration spanning SHASS and CSAIL, it would be impossible to make sense of these complex dynamics.”

Jevin West, a data scientist at the University of Washington, said this mixed-method study has advanced his understanding of the role of data visualization in shaping public perceptions of science and politics. West was not involved in this investigation. West said: ‘Data visualization has the cloak of objectivity and scientific precision. But, as this article shows, data visualization can be used effectively on the opposite of a problem. It emphasizes the complexity of the problem:’ simply teaching media literacy “is not enough. It requires a more detailed socio-political understanding of those who create and interpret data graphs.”

By combining knowledge of computer science and anthropology, researchers have a more detailed understanding of data literacy. Lee said his research shows that maskers view the pandemic in a different way compared to public health orthodoxy, but the data used is quite similar. Lee said his findings suggest that “in the United States, there is a larger gap in our thinking about science and experience.” This gap also cuts across topics like climate change and vaccination, and similar dynamics often appear in social media discussions.

To help the public understand these results, Lee and his collaborator, CSAIL doctoral student Jonathan Zong, led a team of 7 undergraduate researchers from MIT. They developed an interactive narrative so that readers can explore the visualization and dialogue on their own.

Lee described the team’s research as the first step in understanding the role of data and visualization in these larger discussions. “The visualization of data is not objective. It is not absolute. In fact, it is an incredible social and political endeavor. We have to pay attention to how people interpret it outside of scientific institutions,” Lee said.