Methodology Note: Why Data-Driven Visualization Constitutes an Intervention Against Epistemic Injustice

This note is part of the methodology series of the Social Design Lab (ISVD-LAB-003). It maps the theoretical foundations for positioning data visualization as an epistemological intervention.

Data is not neutral. The very choice of what to count and what not to count determines which parts of society are rendered visible and which are left unseen. As Mimi Onuoha showed in her Library of Missing Datasets (2016), "missing datasets" — comprehensive records of police killings, nationwide eviction statistics, systematic data on violence against transgender people — are not accidental omissions but consequences of social and political dynamics. The absence of data is itself an expression of power.

This recognition fundamentally transforms the meaning of data visualization. Visualization is not merely a technical operation that makes existing data easier to see. The choice of what to visualize, the design of what visual language to employ, and the distribution channels determining who can access the visualization — all of these are implicated in either intervening against or reproducing epistemic injustice.

This note examines why and how data-driven visualization can serve as an intervention against epistemic injustice, drawing on both historical lineage and contemporary theoretical frameworks. Its purpose is to clarify the grounds on which ISVD positions its statistical dashboards — built using the e-Stat API — as a practice of "social design."

The Epistemological Lineage of Visualization — Seeing as Knowing

The recognition that data visualization can function as an instrument of social transformation is not a modern one. This lineage can be traced back at least to the mid-nineteenth century.

The polar area diagram (coxcomb chart) created by Florence Nightingale in 1858 visualized the causes of death among British soldiers in the Crimean War. What Nightingale demonstrated was the fact that poor sanitary conditions, not combat, accounted for the overwhelming majority of deaths. The crucial point is that this fact could have been read from the statistical tables alone. But tabulated numbers failed to change the perceptions of policymakers. Nightingale's coxcomb chart, by converting statistical fact into visual impact, made sanitary reform — a policy intervention — possible. Visualization functioned here as an epistemological intervention: making visible what should have been seen but was not.

At the 1900 Paris Exposition, W. E. B. Du Bois exhibited a series of data portraits visualizing the demographics, education, and economic conditions of Black Americans. Du Bois's data portraits used "objective" government statistics while reconfiguring them as counter-evidence against the dominant narrative — the racist discourse that Black people were inferior. This was a pioneering case demonstrating that the same data, through different modes of presentation, can become a counter-narrative.

The Isotype (International System of Typographic Picture Education) developed by Otto Neurath from the 1920s through the 1940s was a system for representing statistical data through pictographic symbols. Neurath's purpose was explicit: to deliver statistical knowledge to the working class, who had previously been unable to access it. Isotype was a practice of "democratic statistics" — an attempt to rectify the class-based maldistribution of data literacy.

In The Visual Display of Quantitative Information (1983), Edward Tufte systematized the aesthetics and ethics of data visualization. Tufte's principle of the "data-ink ratio" — the majority of ink should be devoted to representing data itself — appears to be a technical norm, but behind it lies an epistemological stance: "seeing is a strategy for knowing." By stripping away unnecessary decoration, the structure of data is made directly perceivable. For Tufte, good visualization was not a technique for "displaying accurately" but a device for "knowing accurately."

What the lineage of Nightingale, Du Bois, Neurath, and Tufte shares is the stance of treating data visualization not as mere display technology but as an intervention that restructures perception.

Contemporary Data Justice Theory — Power, Participation, Anti-Discrimination

Connecting this historical lineage to contemporary theoretical frameworks are Catherine D'Ignazio & Lauren F. Klein's Data Feminism (2020, MIT Press) and Linnet Taylor's data justice theory (2017).

D'Ignazio & Klein presented seven principles. Three are particularly relevant to this note. First, Examine Power: recognize that power relations are embedded at every stage of data collection, analysis, and visualization, and critically examine them. Second, Challenge Power: pursue the possibility of countering existing power structures through data practice. Third, Make Labor Visible / Elevate Emotion and Embodiment: reintegrate experiences, emotions, and embodiment that data science has excluded under the banner of "objectivity."

Taylor, in "What is data justice? The case for connecting digital rights and freedoms globally" (2017), defined data justice through three pillars. Visibility — who is rendered visible and who is rendered invisible. Engagement — who can participate in the collection and use of data. Anti-discrimination — whether data practices reproduce discrimination. These three pillars theoretically ground the proposition that data visualization is not merely a technical task but a practice that pertains to justice.

Counter-Mapping and Community GIS

The practice of treating data visualization as an epistemic intervention is also found in the tradition of counter-mapping. Indigenous peoples mapping their own lands, community-led participatory GIS (Geographic Information Systems), residents of impoverished urban areas mapping their own living conditions — these practices, through the question of "who draws the map," have challenged the power relations inherent in the production of spatial knowledge.

What does not appear on official maps does not become the subject of policy. Indigenous customary land use is not recorded in state survey maps. Informal settlements are excluded from urban planning documents. Counter-mapping demonstrates that this "exclusion from the map" is a spatial expression of epistemic injustice, and counters it through the practice of the excluded themselves generating and visualizing data.

Connecting Fricker's Hermeneutical Injustice to Data Visualization

The central argument of this note is that data-driven visualization can function as an intervention against what Fricker calls hermeneutical injustice.

Hermeneutical injustice refers to a state in which the conceptual resources (hermeneutical resources) needed to understand a given experience are socially absent. What is crucial is that this "absence" is not accidental but the result of specific groups being structurally excluded from society's interpretive practices.

Data visualization can intervene against this hermeneutical injustice through at least three pathways.

First, the construction of conceptual resources. Structurally visualizing dispersed statistical data is an act of providing individual experiences with a conceptual framework called "social patterns." To recognize that individual hardship is not "personal failure" but a structural problem, patterns beyond individual cases must become visible. Statistical dashboards function as conceptual resources that enable this pattern recognition.

Second, the reinforcement of testimonial authority. When experiential knowledge of those affected is dismissed as "not evidence" (testimonial injustice), statistical backing adds the authority of being "worth hearing" to their testimony. This is not an ideal state — testimony of affected individuals should be respected on its own terms — but within existing power structures, data-backed evidence can serve as a means to partially rectify epistemic injustice.

Third, the visualization of absence. As Onuoha's "missing datasets" demonstrated, the absence of data can itself be the result of political choices. Pointing out data absence and asking "why does this data not exist?" is itself an intervention against structural invisibility.

The Japanese Context — e-Stat, RESAS, and the Limits of Freedom of Information

Practicing data-driven visualization as an epistemological intervention in Japan requires understanding the country's specific institutional context.

e-Stat is the portal site for government statistics, providing data access through APIs. RESAS is a visualization tool specialized for regional economic analysis. While these democratize access to public data to a certain degree, they have structural limitations.

First, the problem of aggregation units. Most government statistics are aggregated at the prefectural or municipal level, making it structurally difficult to see the realities of smaller communities — for example, specific housing estates, non-regular workers in particular industries, or foreign residents by immigration status. The granularity of aggregation itself determines what can be made visible.

Second, the limits of the Freedom of Information Act (Act on Access to Information Held by Administrative Organs, enacted 1999). The law provides for information disclosure based on requests, but has the structural constraint that one cannot request what one does not know exists. This constraint directly relates to what Taylor (2017) calls the "visibility" dimension of data justice — without knowing something exists, access is impossible.

Third, the data literacy gap. Using the e-Stat API is limited to those with programming skills and data analysis knowledge. The problem Neurath confronted in the 1920s — the class-based maldistribution of statistical knowledge — persists in digitized form.

ISVD's Practice — Statistical Dashboards as Structural Revelation

ISVD's statistical dashboards, built using the e-Stat API, can be positioned within the theoretical framework outlined above as follows.

ISVD's dashboards are an implementation of D'Ignazio & Klein's "Examine Power" principle. When visualizing government statistics such as age-specific unemployment rates or occupation-specific average wages, ISVD does not merely redisplay data but reconfigures it through the design of comparison axes — for instance, disparities between age groups, wage gaps by gender, trends over time — into a form that makes structural inequality legible. This "reconfiguration" is the core of data visualization in social design.

At the same time, by juxtaposing ISVD's columns and guides alongside statistical dashboards, Tufte's "seeing" and Fricker's "construction of conceptual resources" are connected. Rather than merely showing numbers, articles that articulate what those numbers "mean" — what structural problems they reflect — are placed adjacent to the data, giving it the power to transform perception. Just as Nightingale's coxcomb charts made sanitary reform possible, visualization that lays bare structures can serve as a starting point for policy intervention.

However, caution is warranted here. As D'Ignazio & Klein warn, data visualization does not by itself achieve justice. The possibility that visualization designers unconsciously reproduce biases embedded in data, the possibility that visualization becomes a tool of surveillance, and the possibility that "having been visualized" creates the illusion that a problem has been solved — these risks always exist. For ISVD's practice to remain an intervention against epistemic injustice, the disposition to critically examine the design of visualization itself must be sustained.

→ Related: The Intellectual Coordinates of Social Design | Literature Map: From Agnotology to Structural Invisibility

Additional References