How antimaskers weaponize techniques of scientific analysis to attack mask mandates

After more than two decades dealing with antivaxxers, quacks, pseudoscience advocates, contrarians, and conspiracy theorists, I’ve noticed some things about how such people operate and view themselves. For example, virtually all of them believe themselves to be “brave mavericks” of some sort, unlike the run-of-the-mill “sheeple” who accept the narrative of mainstream science and, in their mind, cower before the authority of physicians and scientists on such matters. One other thing that I’ve noticed about them is that they truly believe that they are the scientific ones, the ones “following the science,” the “true” science, at least in their minds. Obviously, my observations are just that, personal observations. I don’t claim that they are scientific. That’s why I’m always interested to find studies that try to look at the characteristics of the arguments of antivaxxers, quacks, pseudoscience advocates, contrarians, and conspiracy theorists and one reason why I’m starting by listing these characteristics that I’ve noticed having observed the rhetoric of such people online going back to the days of Usenet. (Yes, I’m dating myself.) That’s why a new narrative going bubbling up among antimaskers, COVID-19 deniers, contrarians, and conspiracy theorists caught my attention over the weekend.

Here’s a representative series of Tweets:

That last Tweetstorm goes on a while longer, but those of you familiar with science denial were likely cringing at several of the quotes from the paper referenced. (More on that in a moment.) The reason, of course, is that they seems to be saying that antimaskers are more “scientific” than those supporting the primary scientific position.

Elsewhere, Michael Levitt, a professor of biophysics at Stanford and 2013 Nobel Laureate in Chemistry who’s decided that he’s an epidemiology and virology expert, leading him to become a COVID contrarian after the pandemic started, loves the article:

Then, on Friday Mike Adams’ drone Ethan Huff over at Natural News wrote about this study under the title, “Vaccine propagandists infiltrated vaccine skeptics group, found that the skeptics are MORE scientific and rigorous in their thinking compared to obedient, dumbed-down mask wearers“. That’s why I was cringing as I read those quotes; it’s almost as though they were custom-made for antimaskers and COVID-19 contrarians to take the study and use it to promote exactly the message that Huff was promoting on Adams’ site.

The quotes I’ve included (because antivaxxers and antimaskers love them so much and are widely quoting them) are, admittedly, consistent with the self-view of antivaxxers and COVID-19 cranks that I’ve seen over the years. That’s why I decided that this week’s topic for me had to be this paper, a paper from the conference proceedings of CHI 2021 under Human-Computer Interaction (cs.HC); Computers and Society (cs.CY). It is currently available on the preprint server arXiv. The article is from MIT, credited to Crystal Lee, Tanya Yang, Gabrielle Inchoco, Graham M. Jones, Arvind Satyanarayan, and titled “Viral Visualizations: How Coronavirus Skeptics Use Orthodox Data Practices to Promote Unorthodox Science Online“, which was well-received:

I found this helpful to learn, given that most conference proceedings in medicine are actually not as rigorously peer-reviewed as journal articles.

I had to see if this paper really concludes what antivaxxers and COVID-19 contrarians are claiming that it concludes. As you might imagine, the answer is: Not really. Unfortunately, the paper is written in a manner that makes it very easy for antimaskers and antivaxxers to portray it as concluding that they are more “scientific,” understand better that science is a process, and are thus more “dedicated to science” than those trying to combat COVID-19 disinformation. It’s another example of how I wish that researchers would consider how their words might be weaponized by cranks when they write papers like this, especially computer scientists without a lot of experience in techniques of disinformation. Had the authors simply asked a couple of skeptics in the trenches to read over their manuscript, they could likely have ameliorated this issue (but that’s probably just me with an overinflated view of my own importance). On the other hand, this take on the study is true:

I’m probably being unfair. It was probably impossible to write such a paper in such a manner that parts of it couldn’t be selectively quoted. On the other hand, the authors didn’t have to make it just so easy.

Still, let’s dig in. Before I do, note that this paper is about what the authors describe as “antimaskers.” Appropriately (to me, at least), they use “animasker” as a “synecdoche for a broad spectrum of beliefs: that the pandemic is exaggerated, schools should be reopening, etc.,” further noting:

While groups who hold these beliefs are certainly heterogeneous, the mask is a common flashpoint throughout the ethnographic data, and they use the term “maskers” to describe people who are driven by fear. They are “anti-mask” by juxtaposition. This study therefore takes an emic (i.e. “insider”) approach to analyzing how members of these groups think, talk, and interact with one another, which starts by using terms that these community members would use to describe themselves.

As an aside, even though the study doesn’t really address the question, in my experience the Venn diagram between antimaskers and antivaxxers has at least 80% overlap, which is another reason this study interested me. It likely could apply to antivaxxers as well.

Using the same data to come to different conclusions

It will come as no surprise to scientists how scientists can examine the same data and come to different conclusions. The differences can come from emphasizing one dataset over another, doing a different kind of analysis, and a variety of other factors. It is thus not always obvious when a different interpretation of data drifts into contrarianism or even outright science denial, although frequently it is quite obvious (for example, the claim that vaccines cause autism). The abstract of this paper (sort of) acknowledges this point:

Controversial understandings of the coronavirus pandemic have turned data visualizations into a battleground. Defying public health officials, coronavirus skeptics on US social media spent much of 2020 creating data visualizations showing that the government’s pandemic response was excessive and that the crisis was over. This paper investigates how pandemic visualizations circulated on social media, and shows that people who mistrust the scientific establishment often deploy the same rhetorics of data-driven decision-making used by experts, but to advocate for radical policy changes. Using a quantitative analysis of how visualizations spread on Twitter and an ethnographic approach to analyzing conversations about COVID data on Facebook, we document an epistemological gap that leads pro- and anti-mask groups to draw drastically different inferences from similar data. Ultimately, we argue that the deployment of COVID data visualizations reflect a deeper sociopolitical rift regarding the place of science in public life.

That last sentence is a bit of a “Well, duh!” conclusion in my book, but that doesn’t mean it’s not worthwhile to explore it. In the introduction, the authors note:

Almost every US state now hosts a data dashboard on their health department website to show how the pandemic is unfolding. However, despite a preponderance of evidence that masks are crucial to reducing viral transmission [25, 29, 105], protestors across the United States have argued for local governments to overturn their mask mandates and begin reopening schools and businesses. A pandemic that affects a few, they reason, should not impinge on the liberties of a majority to go about life as usual. To support their arguments, these protestors and activists have created thousands of their own visualizations, often using the same datasets as health officials.

This paper investigates how these activist networks use rhetorics of scientific rigor to oppose these public health measures. Far from ignoring scientific evidence to argue for individual freedom, antimaskers often engage deeply with public datasets and make what we call “counter-visualizations”—visualizations using orthodox methods to make unorthodox arguments—to challenge mainstream narratives that the pandemic is urgent and ongoing. By asking community members to “follow the data,” these groups mobilize data visualizations to support significant local changes.

So far, so good. What the authors discuss in the introduction is undeniably true. Antimaskers and those who minimize the severity of the pandemic have become very skilled at making slick-looking figures to support their claims, often using the same datasets used by public health officials. Seeing how they do this on social media is, therefore, of interest. So how did the authors examine this question?

First, they examined the circulation of COVID-related data visualizations through quantitative and qualitative methods. They started with a quantitative analysis of nearly half a million Tweets that used data visualization to talk about the pandemic, while using network analysis to identify user communities who retweet the same content or engage with each other (such as antimaskers and those supporting mask mandates). They used a computer vision model to extract feature embeddings and identify clusters in visualization designs, noting that they found that “anti-mask groups on Twitter often create polished counter-visualizations that would not be out of place in scientific papers, health department reports, and publications like the Financial Times.”

In addition, the authors undertook a six month long observational study of antimask groups on Facebook over the period from March to September 2020, a time frame during which these groups formed and consolidated, justifying it thusly:

Quantitative analysis gives us an overview of what online discourse about data and its visual representation looks like on Twitter both within and outside anti-mask communities. Qualitative analysis of anti-mask groups gives us an interactional view of how these groups leverage the language of scientific rigor—being critical about data sources, explicitly stating analytical limitations of specific models, and more—in order to support ending public health restrictions despite the consensus of the scientific establishment.

I don’t know enough about the computer algorithms and methodology to comment extensively on the nitty-gritty of the visualization analysis performed, but I can comment a bit on the observational study, whose methods the authors describe thusly:

While qualitative research can involve clinical protocols like interviews or surveys, Clifford Geertz [45] argues that the most substantial ethnographic insights into the cultural life of a community come from “deep hanging out,” i.e., long-term, participant observation alongside its members. Using “lurking,” a mode of participating by observing specific to digital platforms, we propose “deep lurking” as a way of systematically documenting the cultural practices of online communities. Our methods here rely on robust methodological literature in digital ethnography [30, 69], and we employ a case study approach [92] to analyze these Facebook groups. To that end, we followed five Facebook groups (each with a wide range of followers, 10K-300K) over the first six months of the coronavirus pandemic, and we collected posts throughout the platform that included terms for “coronavirus” and “visualization” with Facebook’s CrowdTangle tool [33]. In our deep lurking, we archived web pages and took field notes on the following: posts (regardless of whether or not they included “coronavirus” and “data”), subsequent comments, Facebook Live streams, and photos of in-person events. We collected and analyzed posts from these groups from their earliest date to September 2020.

One aspect of this paper bothers me. Nowhere do the authors identify which antimask groups in which they “deep lurked,” other than to note that, since September, Facebook has “banned some of the groups we have studied, who have since moved to more unregulated platforms (Parler and MeWe).” I don’t recall having read a study of this type that didn’t explicitly list the Facebook pages examined.

As a preview of the findings, the authors write:

While previous literature in visualization and science communication has emphasized the need for data and media literacy as a way to combat misinformation [43, 47, 89], this study finds that anti-mask groups practice a form of data literacy in spades. Within this constituency, unorthodox viewpoints do not result from a deficiency of data literacy; sophisticated practices of data literacy are a means of consolidating and promulgating views that fly in the face of scientific orthodoxy. Not only are these groups prolific in their creation of counter-visualizations, but they leverage data and their visual representations to advocate for and enact policy changes on the city, county, and state levels.

This preview rings true to those of us who have studied antivaxxers and other groups that oppose the scientific consensus. If anything, science denialists tend to delve very deeply into the data. The problem, of course, is that their preconceptions are so strong that, to echo the old saying about drunks looking for their keys and a lamppost, they use data for support rather than illumination. While it is true that many, if not most, scientists approach data trying to test a hypothesis hoping to find support for it, meaning that they have some degree of bias, science denialists take natural human bias to an extreme in which they will parse the data into finer and finer bits until they find a way to support what they want to believe. Alternatively, as this paper suggests, they create data visualizations that support their point of view.

The authors visualized the Twitter networks using the Louvain method of community detection, a commonly used method of identifying communities in large networks, and found six main networks sharing COVID-related visualizations:

1. American politics and media.
2. American politics and right-wing media (red).
3. British news media.
4. Anti-mask network.
5. New York Times-centric network.
6. World Health Organization and health-related news organizations.

Of interest, the antimask network was found to be anchored by former New York Times reporter Alex Berenson, the man who has been called the “pandemic’s wrongest man,” as well as “blogger @EthicalSkeptic, and @justin_hart.” The authors also note that a key target of this network was The Atlantic‘s @Covid19Tracking project (which collates COVID-19 testing rates and patient outcomes across the United States). The antimask network also had the highest fraction of retweets of in-network Tweets.

Here is Figure 2, which includes a sampling of some antimask COVID-19 minimizing/denying countervisualizations:

The authors comment thusly on this figure:

While there are certainly visualizations that tend to use a meme-based approach to make their point (e.g., “Hey Fauci…childproof chart!” with the heads of governors used to show the rate of COVID fatalities), many of the visualizations shared by anti-mask Twitter users employ visual forms that are relatively similar to charts that one might encounter at a scientific conference. Many of these tweets use area and line charts to show the discrepancy between the number of projected deaths in previous epidemiological and the numbers of actual fatalities. Others use unit visualizations, tables, and bar charts to compare the severity of coronavirus to the flu. In total, this figure shows the breadth of visualization types that anti-mask users employ to illustrate that the pandemic is exaggerated.

And it is true. Antivaxxers, conspiracy theorists, and antimaskers have become very adept at producing memes and, as this study shows, data visualizations with deceptive messages. Overall, the authors identified eight major clusters of types of visualizations used by antimaskers: line charts (8,908 visualizations, 21% of the corpus), area charts (2,212, 5%), bar charts (3,939, 9%), pie charts (1,120, 3%), tables (4,496, 11%), maps (5,182, 13%), dashboards (2,472, 6%), and images (7,128, 17%), also noting that The remaining 6,248 media (15% of the corpus) “did not cluster in thematically coherent ways.”

Antimask discourse

Overall, the MIT investigators found seven main themes in antimask discourse:

1. Emphasis on original content.
2. Critically assessing data sources.
3. Critically assessing data representations.
4. Identifying bias and politics in data.
5. Appeals to scientific authority.
6. Developing expertise and processes of critical engagement.
7. Applying data to real-world situations.

To be honest, I was more interested in the qualitative analysis of the content of antimask Facebook groups than the quantitative analysis of the types of data visualizations they promote on Twitter, for the simple reason that it produced potentially more actionable information. I will do this in part by examining some of the quotes cherry picked by antimaskers and then putting them into context to show (1) how cherry picked they are and (2) that they do not mean what antimaskers think they mean. However, I will also discuss key points above. I’ll start with one favorite quote by antimaskers, namely that the paper supposedly concludes that they are more sophisticated in their understanding of science. Here’s the quote in context near the beginning of the article, the cited quote in bold:

As science and technology studies (STS) scholars have shown, data is not a neutral substrate that can be used for good or for ill [14, 46, 84]. Indeed, anti-maskers often reveal themselves to be more sophisticated in their understanding of how scientific knowledge is socially constructed than their ideological adversaries, who espouse naive realism about the “objective” truth of public health data. Quantitative data is culturally and historically situated; the manner in which it is collected, analyzed, and interpreted reflects a deeper narrative that is bolstered by the collective effervescence found within social media communities. Put differently, there is no such thing as dispassionate or objective data analysis. Instead, there are stories: stories shaped by cultural logics, animated by personal experience, and entrenched by collective action. This story is about how a public health crisis—refracted through seemingly objective numbers and data visualizations—is part of a broader battleground about scientific epistemology and democracy in modern American life.

Basically, this is the final paragraph of the introduction, in which the authors set up their story, which is how antimaskers use the fact that there is no such thing as a totally objective analysis of data. Of course, scientists do their best to eliminate as much bias as they reasonably can from their analysis. Antimaskers do not and dismiss criticism that their analyses are biased with observations that there is a socially constructed aspect to all scientific knowledge. It’s the same technique that antivaxxers use. Let’s just say that, in context, this quote does not support the contention that antimaskers are more sophisticated about science. Not really. I will, however, say that this sentence quoted by antimaskers is perhaps the most grating sentence in the paper, as it really doesn’t qualify the assertion made in the way that it needs to be qualified. For one thing, it utterly ignores the role of disinformation merchants in stoking antimask narratives, seemingly assuming honest motives in the case of all antimaskers.

Then there’s the quote about how antimaskers value “unmediated access” to information and privilege. Let’s see the statement in its context, which is that this desire for “unmediated access to information” is so that antimaskers can use that information to promote their message, which is what leads to an “emphasis on original content”:

Many anti-mask users express mistrust for academic and journalistic accounts of the pandemic, proposing to rectify alleged bias by “following the data” and creating their own data visualizations. Indeed, one Facebook group within this study has very strict moderation guidelines that prohibit the sharing of non-original content so that discussions can be “guided solely by the data.” Some group administrators even impose news consumption bans on themselves so that “mainstream” models do not “cloud their analysis.” In other words, anti-maskers value unmediated access to information and privilege personal research and direct reading over “expert” interpretations. While outside content is generally prohibited, Facebook group moderators encourage followers to make their own graphs, which are often shared by prominent members of the group to larger audiences (e.g., on their personal timelines or on other public facing Pages). Particularly in cases where a group or page is led by a few prominent users, follower-generated graphs tend to be highly popular because they often encourage other followers to begin their own data analysis projects, and comments on these posts often deal directly with how to reverse-engineer (or otherwise adjust) the visualization for another locality.

Of course, data transparency is a good thing. Scientists themselves argue about data transparency, and there is a growing movement to make raw data used in analyses available to other scientists for analysis. This has led to the creation of, for instance, databanks for genomic data. The problem with the data visualizations produced by antimaskers is that the vast majority of them do not know what they are doing and don’t even try to control for their own biases.

Consistent with their bias, the discussion of data sources and reliability by antimaskers tends to focus on exaggerated concerns about how data can be “manipulated” for nefarious purposes

Many of the users believe that the most important metrics are missing from government-released data. They express their concerns in four major ways. First, there is an ongoing animated debate within these groups about which metrics matter. Some users contend that deaths, not cases, should be the ultimate arbiter in policy decisions, since case rates are easily “manipulated” (e.g., with increased testing) and do not necessarily signal severe health problems (people can be asymptomatic). The shift in focus is important, as these groups believe that the emphasis on cases and testing often means that rates of COVID deaths by county or township are not reported to the same extent or seriously used for policy making.

But, later in the article, the authors point out how this is a “sleight of hand” (their very words):

For instance, they argue that there is an outsized emphasis on deaths versus cases: if the current datasets are fundamentally subjective and prone to manipulation (e.g., increased levels of faulty testing, asymptomatic vs. symptomatic cases), then deaths are the only reliable markers of the pandemic’s severity. Even then, these groups believe that deaths are an additionally problematic category because doctors are using a COVID diagnosis as the main cause of death (i.e., people who die because of COVID) when in reality there are other factors at play (i.e., dying with but not because of COVID). Since these categories are fundamentally subject to human interpretation, especially by those who have a vested interest in reporting as many COVID deaths as possible, these numbers are vastly over-reported, unreliable, and no more significant than the flu.

Although the authors didn’t state it this way, I will. Antimaskers, because of their exaggerated concern over “manipulation” of the data, either go to great lengths to find ways to manipulate the data themselves to show what they want it to show or to place a naive faith in the “raw data,” which brings us to #3 and their criticism of data representations. The idea is conspiratorial at its core, namely that somehow the “raw data” are more pure and “can’t be spun”:

An ongoing topic of discussion is whether to visualize absolute death counts as opposed to deaths per capita, and it is illustrative of a broader mistrust of mediation. For some, “raw data” (e.g., counts) provides more accurate information than any data transformation (e.g., death rate per capita, or even visualizations themselves). For others, screenshots of tables are the most faithful way to represent the data, so that people can see and interpret it for themselves. “No official graphs,” said one user. “Raw data only. Why give them an opportunity to spin?” (June 14, 2020). These users want to understand and analyze the information for themselves, free from biased, external intervention.

This feeds into #4 and the identification of bias and politics. The authors cite various posts in which those creating data visualizations admit their bias but argue that, by sticking as close as possible to the “raw data,” they can “keep the effect of bias to a minimum.” But where does this bias that they fear so much come from? A lot of them point to “specific profit motives that come from releasing (or suppressing) specific kinds of information.” In addition, consistent with my aforementioned guesstimate of the high overlap between antimaskers and antivaxxers, the authors note that many are suspicious of the benefits of a coronavirus vaccine, predictably pointing out “how the tobacco industry has historically manipulated science to mislead consumers.” (Yes, as I’ve frequently pointed out, antivaxxers love to point to tobacco companies and how they manipulated and denied science, all while remaining conveniently ignorant that they are using exactly the same sorts of techniques to do the same thing.) Unsurprisingly, the authors found evidence that antimaskers believe that pharmaceutical companies have “similarly villainous profit motives, which leads the industry to inflate data about the pandemic in order to stoke demand for a vaccine.”

One of the more ironic, but not entirely unexpected findings of this paper is how antimaskers crave scientific respectability and, to try to gain it, have used appeals to scientific authority, which brings me to the quote about how this paper supposedly found that antimaskers are more, not less scientifically rigorous. I think you’ll see how blatantly that line (in bold) was cherry picked to convey a different meaning than what the authors intended:

Paradoxically, these groups also seek ways to validate their findings through the scientific establishment. Many users prominently display their scientific credentials (e.g., referring to their doctoral degrees or prominent publications in venues like Nature) which uniquely qualify them as insiders who are most well-equipped to criticize the scientific community. Members who perform this kind of expertise often point to 2013 Nobel Laureate Michael Levitt’s assertion that lockdowns do nothing to save lives [67] as another indicator of scientific legitimacy. Both Levitt and these anti-mask groups identify the dangerous convergence of science and politics as one of the main barriers to a more reasonable and successful pandemic response, and they construct their own data visualizations as a way to combat what they see as health misinformation. “To be clear. I am not downplaying the COVID epidemic,” said one user. “I have never denied it was real. Instead, I’ve been modeling it since it began in Wuhan, then in Europe, etc. […] What I have done is follow the data. I’ve learned that governments, that work for us, are too often deliberately less than transparent when it comes to reporting about the epidemic” (July 17, 2020). For these anti-mask users, their approach to the pandemic is grounded in a more scientific rigor, not less.

Notice how in no way are the authors claiming that antimaskers’ approach to the pandemic is more scientifically rigorous. They’re simply making a statement about how antimaskers themselves believe that their approach to the pandemic is more scientific. Also note how the authors mention Michael Levitt, whose copious pronouncements on COVID-19 have been referred to as “lethal nonsense” (based on his having declared the pandemic “over” in August). More recently, Levitt has been downplaying the number of deaths from COVID-19 in Brazil. Unfortunately, his status as a Nobel Laureate has given his pronouncements on COVID-19 far more heft in the media than they observe.

Of course, it is no surprise to any skeptic who’s followed, for example, the antivaccine movement or other science denialist movements that the cranks in those movements, even as they castigate science as “corrupt,” crave the legitimacy that science brings, which is why antivaxxers latch on to the pronouncements of physicians and scientists with no special expertise in the relevant disciplines to back up their claims. Examples of such scientists abound, and include people like James Lyons-Weiler, a bioinformaticist who’s recast himself as a vaccine expert and now—surprise! surprise!—a COVID-19 expert; Paul Thomas, a pediatrician with no expertise in clinical trials or vaccines who claims to be doing a clinical study of vaccines; Christopher Exley, a chemist who fancies himself a vaccine expert; Christopher Shaw, a neuroscientist who also fancies himself a vaccine expert; and the father-son duo of Mark and David Geier, neither of whom has any specific expertise in autism or fields relevant to vaccines. The same thing is happening with COVID-19, as the case of Levitt demonstrates. He’s a structural biologist whose Nobel Prize was in chemistry. Before the pandemic, he has no special expertise in epidemiology, infectious disease, pandemic modeling, or anything else relevant to COVID-19. Sadly, since the pandemic, he hasn’t developed any such expertise.

The list of such scientists with no special expertise who have become the darlings of antimaskers is a long and dishonorable one. Worse, occasionally, there are even scientists who really should know better, some of whom actually have relevant expertise, who have fallen down the rabbit hole of COVID-19 minimization. Great Barrington Declaration, anyone?

Then there’s the part about “individual initiative”:

Anti-maskers have deftly used social media to constitute a cultural and discursive arena devoted to addressing the pandemic and its fallout through practices of data literacy. Data literacy is a quintessential criterion for membership within the community they have created. The prestige of both individual anti-maskers and the larger Facebook groups to which they belong is tied to displays of skill in accessing, interpreting, critiquing, and visualizing data, as well as the pro-social willingness to share those skills with other interested parties. This is a community of practice [63, 102] focused on acquiring and transmitting expertise, and on translating that expertise into concrete political action. Moreover, this is a subculture shaped by mistrust of established authorities and orthodox scientific viewpoints. Its members value individual initiative and ingenuity, trusting scientific analysis only insofar as they can replicate it themselves by accessing and manipulating the data firsthand. They are highly reflexive about the inherently biased nature of any analysis, and resent what they view as the arrogant self-righteousness of scientific elites.

As a subculture, anti-masking amplifies anti-establishment currents pervasive in U.S. political culture. Data literacy, for antimaskers, exemplifies distinctly American ideals of intellectual selfreliance, which historically takes the form of rejecting experts and other elites [53]. The counter-visualizations that they produce and circulate not only challenge scientific consensus, but they also assert the value of independence in a society that they believe promotes an overall de-skilling and dumbing-down of the population for the sake of more effective social control [39, 52, 98]. As they see it, to counter-visualize is to engage in an act of resistance against the stifling influence of central government, big business, and liberal academia. Moreover, their simultaneous appropriation of scientific rhetoric and rejection of scientific authority also reflects longstanding strategies of Christian fundamentalists seeking to challenge the secularist threat of evolutionary biology [11].

This is, of course, an apt comparison, one that the antimaskers quoting this paper so approvingly seem to have…missed…in their praise of the study. I wonder why. Of course, creationism (a.k.a. evolution denial) in all its forms, particularly “intelligent design” creationism (which does not dispute that evolution takes place but co-opts scientific arguments to argue that there must be a “designer” (a.k.a. God) guiding evolution) does indeed use similar techniques. Indeed, whole organizations have been set up for this purpose. One could even say that the most prominent of these anti-evolution groups, the Discovery Institute, is mimicking science by having been set up as an “institute” to imitate legitimate scientific societies and “institutes.” It’s tempting to note how heavily evangelical Christians are represented in antimask activism and wonder how much overlap there is in a Venn diagram of creationists or evolution deniers and antimaskers. (Oh, wait. I just did.)

Two things that antimaskers get right and wrong simultaneously

Interestingly, antimaskers don’t get everything wrong. They are, however, an example of how they can get something right in the wrong way, such that it leads them astray. For instance, let’s look in context at another quote that antimaskers have been cherry picking (bolded):

Most fundamentally, the groups we studied believe that science is a process, and not an institution. As we have outlined in the case study, these groups mistrust the scientific establishment (“Science”) because they believe that the institution has been corrupted by profit motives and politics. The knowledge that the CDC and academics have created cannot be trusted because they need to be subject to increased doubt, and not accepted as consensus. In the same way that climate change skeptics have appealed to Karl Popper’s theory of falsification to show why climate science needs to be subjected to continuous scrutiny in order to be valid [42], we have found that anti-mask groups point to Thomas Kuhn’s The Structure of Scientific Revolutions to show how their anomalous evidence—once dismissed by the scientific establishment—will pave the way to a new paradigm (“As I’ve recently described, I’m no stranger to presenting data that are inconsistent with the narrative. It can get ugly. People do not give up their paradigms easily. […] Thomas Kuhn wrote about this phenomenon, which occurs repeatedly throughout history. Now is the time to hunker down. Stand with the data,” August 5, 2020). For anti-maskers, valid science must be a process they can critically engage for themselves in an unmediated way. Increased doubt, not consensus, is the marker of scientific certitude.

Doubt is the product of antimaskers, just as it was the product of tobacco companies and is the product of, for example, antivaxxers and climate science denialists. This paper highlights how this is true, pointing out how very skilled antimaskers have become at sifting through data and finding areas of uncertainty to highlight as a means of casting doubt on specific scientific consensuses.

As I’ve discussed before, climate science deniers love to quote Michael Crichton’s infamous statements, such as, “Let’s be clear: the work of science has nothing whatever to do with consensus” and “There is no such thing as consensus science. If it’s consensus, it isn’t science. If it’s science, it isn’t consensus. Period.” I’ve responded to this nonsense before in detail, which is a fundamental misunderstanding of the nature of science. (After all, as I asked, what is a scientific theory like the theory of evolution or Einstein’s theory of relativity but a statement of the current scientific consensus regarding a major scientific topic?) Also, explanatory power is all; if your model has no explanatory or predictive power, it’s useless. So, basically, antimaskers (from my perspective) have the correct understanding that science is a process not an institution and not just a body of knowledge, but they use it to draw the wrong conclusions about scientific consensus and how to approach data.

The authors also have a point about data transparency, as I mentioned above, and consistency of coding:

This plays into a third problem that users identify with the existing data: that datasets are constructed in fundamentally subjective ways. They are coded, cleaned, and aggregated either by government data analysts with nefarious intentions or by organizations who may not have the resources to provide extensive documentation. “Researchers can define their data set anyhow [sic] they like in absence of generally accepted (preferably specified) definitions,” one user wrote on June 23, 2020. “Coding data is a big deal—and those definitions should be offered transparently by every state. Without a national guideline—we are left with this mess.” The lack of transparency within these data collection systems—which many of these users infer as a lack of honesty—erodes these users’ trust within both government institutions and the datasets they release.

These statements about data collection and standardization weren’t wrong, particularly early in the pandemic. The problem is that they use this as an excuse to “analyze” the data they can access any way they want.

Implications

There are a number of implications of this study, if its findings hold up. Perhaps the implication that’s most counterintuitive to skeptics and science communicators is that increasing data literacy in the public will not, in and of itself, mitigate the effects of misinformation and disinformation. As I quoted the authors above, they have a form of data literacy “in spades.” Similarly, the authors note:

Powerful research and media organizations paid for by the tobacco or fossil fuel industries [79, 86] have historically capitalized on the skeptical impulse that the “science simply isn’t settled,” prompting people to simply “think for themselves” to horrifying ends.

Horrifying ends indeed. Also:

As David Buckingham [17] has noted, calls for increased literacy have often become a form of wrong-headed solutionism that posits education as the fix to all media-related problems. danah boyd [16] has documented, too, that calling for increased media literacy can often backfire: the instruction to “question more” can lead to a weaponization of critical thinking and increased distrust of media and government institutions. She argues that calls for media literacy can often frame problems like fake news as ones of personal responsibility rather than a crisis of collective action.

And:

While previous literature in visualization and science communication has emphasized the need for data and media literacy as a way to combat misinformation [43, 47, 89], this study finds that anti-mask groups practice a form of data literacy in spades. Within this constituency, unorthodox viewpoints do not result from a deficiency of data literacy; sophisticated practices of data literacy are a means of consolidating and promulgating views that fly in the face of scientific orthodoxy. Not only are these groups prolific in their creation of counter-visualizations, but they leverage data and their visual representations to advocate for and enact policy changes on the city, county, and state levels.

The problem isn’t a deficit in information, data, or data literacy. It’s how all that is weaponized. As the authors note, the problem isn’t that antimaskers need more scientific literacy, which characterizes their approach as uninformed. Quite the opposite. They are not uninformed, or, as the authors write, “This study shows the opposite: users in these communities are deeply invested in forms of critique and knowledge production that they recognize as markers of scientific expertise.” Nor will creating “better” data visualizations ameliorate this problem. The problem is that it’s a form of critique and knowledge production that is by design in opposition of “consensus” in a way that has a hard time even considering that the consensus might have a point, scientifically speaking. All this critique is primarily done as an act of “resistance,” rather than for purposes of scientific illumination. This makes it prone to its own set of extreme biases, biases that antimaskers don’t really admit, if they even recognize them. Meanwhile, hordes of “citizen scientists” now think themselves better than epidemiologists at analyzing data.

One huge disappointment I had about this paper is that, disappointingly, it hardly touched on the role of ideologically motivated astroturf groups in encouraging antimaskers, providing them with ideas and support, and helping them publicize their message. “Astroturf” refers to campaigns that are orchestrated by ideological groups but designed to appear to be “grassroots,” hence the term “astroturf,” or fake grass roots. Remember the Great Barrington Declaration, the document signed by three scientists that advocated, in essence, letting COVID-19 rip through the healthy population in order to achieve herd immunity, all while using “focused protection” to keep safe the elderly and those with comorbidities that put them at high risk from the disease. It was, as I put it, eugenics-adjacent, if not straight up eugenics. It was also the product of a right wing think tank that conceived and promoted the declaration, using it to lobby governments to lift “lockdowns” and “open up” again. It’s a group that likens itself and fellow antimaskers to “abolitionists“.

The authors seem to assume that the bulk of antimask activity on social media producing data “countervisualizations” is organic in nature, but we know that it’s not. Tobacco companies were real and tried to counter and deny the science showing the harmful health effects of smoking. Fossil fuel companies are real and, through their influence and cash, assisted by right wing think tanks, promote denial of climate science. The same phenomenon has occurred with COVID-19. Disinformation campaigns about COVID-19 mitigation measures promoting antimask and anti-“lockdown” demonstrations have been shown to be astroturf fueled by right wing donors, such as the Koch brothers, with support and behind-the-scenes coordination from The Convention of States project, which has been funded by Republican megadonor Robert Mercer’s family foundation, and two members of Trump’s White House, Ken Cuccinelli, Acting Director of U.S. Citizenship and Immigration Services, and Ben Carson, Secretary of Housing And Urban Development. Indeed, a number of these campaigns have been generated by well-funded organizations. One example was the promotion of hydroxychloroquine as a cure for COVID-19. (If you have a cheap cure for COVID-19, you don’t need lockdowns.) So how much of this activity described by Lee et al is truly organic, and how much is a result of astroturf? That is the question, isn’t it? What disappointing to me is that the authors hardly even touched on this question, even as an area for future research.

My one disappointment aside, I do think this article is a major contribution in that it shows that the problem is most definitely not an information deficit problem, nor is it a science literacy problem. As the authors conclude:

In other words, our paper introduces new ways of thinking about “democratizing” data analysis and visualization. Instead of treating increased adoption of data-driven storytelling as an unqualified good, we show that data visualizations are not simply tools that people use to understand the epidemiological events around them. They are a battleground that highlight the contested role of expertise in modern American life.

Science has always been a battleground. COVID-19 is just the latest front, and it’s being waged by people like this:

They can even sometimes give the appearance of being justified in their boasts, but in reality:

I couldn’t have said it better myself.