- The paper shows that COVID-19 skeptics engage with orthodox data practices to create counter-visualizations that challenge mainstream public health narratives.
- It combines quantitative Twitter analysis of 500K tweets and qualitative ethnography of Facebook anti-mask groups to map digital politicization of data.
- The research highlights that sophisticated data literacy among skeptics drives their political mobilization and critical views of expert-driven science.
This paper, "Viral Visualizations: How Coronavirus Skeptics Use Orthodox Data Practices to Promote Unorthodox Science Online" (Viral Visualizations: How Coronavirus Skeptics Use Orthodox Data Practices to Promote Unorthodox Science Online, 2021), investigates how groups skeptical of mainstream public health narratives during the COVID-19 pandemic used data visualizations to support their views and organize political action. The research combines quantitative analysis of Twitter data with qualitative digital ethnography of anti-mask Facebook groups.
Core Argument:
The central argument is that coronavirus skeptics ("anti-maskers") did not simply ignore scientific data. Instead, they actively engaged with public datasets and employed standard data visualization techniques ("orthodox practices") to create "counter-visualizations." These visualizations were used to argue against public health measures like mask mandates and lockdowns, promoting the "unorthodox" view that the pandemic's severity was exaggerated and the crisis was over. This practice highlights a deep epistemological and sociopolitical divide regarding the role and interpretation of science in public life, rather than a simple deficit in data literacy among skeptics.
Methodology:
- Quantitative Twitter Analysis:
- Analyzed nearly 500,000 tweets related to COVID-19 and data visualization (keywords like "chart," "plot," "dashboard," etc.).
- Used computer vision to classify over 41,000 images, identifying common visualization types (line charts, bar charts, maps, etc.). UMAP was used for visualizing image feature embeddings.
- Constructed a network graph of nearly 400,000 users based on interactions (mentions, replies, retweets) and identified distinct communities using the Louvain method.
- Qualitative Facebook Analysis (Digital Ethnography):
- Conducted a six-month observational paper ("deep lurking") of five anti-mask Facebook groups (10K-300K followers each) from March to September 2020.
- Collected and analyzed posts, comments, livestreams, and discussions using grounded theory.
- Focused on understanding the internal discourses, data practices, and knowledge-making processes within these groups.
Key Findings:
- Orthodox Practices for Unorthodox Claims: Anti-mask communities on Twitter produced and shared visualizations (line charts, area charts, bar charts, maps) that were often visually polished and technically similar to those used by public health officials and mainstream media. They were proficient users of visualization tools and techniques.
- Prolific Visualization Use: The anti-mask network identified on Twitter was highly active, sharing a significant number of visualizations (second highest among the top six communities studied) and exhibiting high levels of in-network amplification (retweeting within the group).
- Sophisticated Data Literacy: Anti-mask groups demonstrated sophisticated data literacy practices. They actively sought out raw data, critiqued data collection methodologies (e.g., testing protocols, definitions of COVID deaths), debated the appropriateness of different metrics (cases vs. deaths, absolute counts vs. per capita rates), and discussed visualization design choices (e.g., normalization, chart types).
- Emphasis on Unmediated Access & Personal Analysis: These groups distrusted "expert" interpretations and mainstream media, emphasizing the need to access primary data sources directly ("follow the data") and conduct personal analysis. They created tutorials and shared datasets amongst themselves.
- Critique of Bias and Politics: They were highly attuned to potential biases in data collection, analysis, and presentation, often attributing nefarious motives (political control, profit for pharmaceutical companies) to government agencies and researchers. They viewed mainstream data narratives as politically motivated and subjective.
- Community of Practice: Facebook groups served as communities where members learned data analysis skills, shared interpretations, received feedback on their visualizations, and collectively constructed their understanding of the pandemic. Data literacy became a means of group socialization and ideological reinforcement.
- Real-World Political Action: The creation and sharing of counter-visualizations were directly linked to political mobilization. Groups used their data analyses to challenge local officials, support lawsuits against health mandates, and organize protests and community events.
Implications for Practice:
- Limitations of Traditional Solutions: The paper argues that simply providing more data, creating "better" or more intuitive visualizations, or running data literacy campaigns is unlikely to bridge the gap. Anti-maskers are already data-literate and use standard visualizations; the issue is a fundamental difference in epistemology, trust, and underlying beliefs ("deep story").
- Understanding the Social Context: Effectively communicating scientific findings requires understanding the social, cultural, and political contexts in which data is interpreted. Data visualization is not a neutral tool but a "battleground" influenced by identity, values, and trust in institutions.
- Communicating Uncertainty: The tendency of official sources to present data with absolute certainty, while scientific processes inherently involve uncertainty, can erode trust. Anti-mask groups often exploited the uncertainties discussed in scientific literature but omitted in public communications. Communicating uncertainty more effectively might be crucial, though challenging.
- Rethinking HCI Assumptions: Human-Computer Interaction (HCI) research often designs visualization tools assuming users (scientists, analysts, general public) primarily seek objective understanding. This paper highlights users who leverage these tools for explicitly political and counter-establishment goals, challenging assumptions about democratization and user intent.
- Addressing the Epistemological Rift: Bridging the divide requires engaging with the underlying reasons for mistrust and the different ways groups conceptualize "science" (as an institution vs. a personal process of critical inquiry). It's a problem rooted in social dynamics and trust, not just technical presentation or individual cognition.
In conclusion, the paper demonstrates that data visualization practices during the pandemic became deeply entangled with political identity and mistrust. Anti-mask groups effectively weaponized the tools and rhetoric of data analysis to build communities, challenge scientific consensus, and mobilize political action, highlighting the complex social life of data and the limitations of purely technocratic or educational approaches to combating misinformation.