Abstract
In this paper, we describe how critical data designers have created projects that ‘push back’ against the eclipse of environmental problems by dominant orders: the pioneering pollution database Scorecard, released by the US NGO Environmental Defense Fund in 1997; the US Environmental Protection Agency’s EnviroAtlas that brings together numerous data sets and provides tools for valuing ecosystem services; and the Houston Clean Air Network’s maps of real-time ozone levels in Houston. Drawing on ethnographic observations and interviews, we analyse how critical data designers turn scientific data and findings into claims and visualisations that are meaningful in contemporary political terms. The skills of critical data designers cross scales and domains; they must identify problems calling for public consideration, and then locate, access, link, and create visualisations of data relevant to the problem. We conclude by describing hazards ahead in work to leverage Big Data to understand and address environmental problems. Critical data designers need to understand what counts as a societal problem in a particular context, what doesn’t, what is seen as connected and not, what is seen as ethically charged, and what is exonerated and discounted. Such recognition is produced through interpretive, ‘close reading’ of the historical moment in which they operate.
For many in the contemporary United States, ‘regulation’ is a dirty word, signalling excessive government and the end of liberty. The Center for Effective Government (CEG) has pushed back, insisting that regulations are ‘public protections’ and should be referred to in those terms (Center for Effective Government, 2015a). To do this, the CEG has also pushed back against data gaps that undercut recognition of the need for public protections. Using data from the US Environmental Protection Agency (EPA), for example, CEG has mapped US schools in ‘danger zones’ around industrial facilities, helping people visualise the magnitude of the problem, and the need for laws requiring emergency planning and risk reduction (see Figure 1; Center for Effective Government, 2015b). CEG’s data visualisations exemplify the kind of ‘pushback’ by critical data designers we describe in this essay, highlighting how expansive and adept interpretive practice is integral to critical data design. 1 Critical data designers draw on interpretive skill in finding, linking, visualising, and circulating available data; they pushback against entrenched ways of thinking about public problems through politically strategic and creative data configurations. Our focus here is on critical data practice in environmental pollution politics. In concluding, we zoom back out to the general challenge of critical data practice and to possibilities for supporting it through governance and education.
Work with pollution data provides particularly rich examples of critical data practice. Pollution data is remarkably heterogeneous, including data about a huge array of substances, from many kinds of collection devices, in many units of analysis, collected by many different organisations, for different purposes – and it needs to be linked or ‘networked’ (boyd and Crawford, 2012) to be meaningful and actionable. Pollution data can be overwhelmingly big in quantity, or frustratingly – and often politically – scarce. Recently (particularly since the financial crisis of 2008), in many settings, there have been notable reductions in pollution data collection; too often, insistence on austerity and small government has legitimised closure of monitoring stations and dismissal of technical staff. 2 Without data, there can be no critical data design. Critical data design thus depends on continual advocacy for data collection.
Critical data design also involves management of time out of joint. Pollution data is especially complex temporally. Often, there are long lags between the time of monitoring and the time when data is available to researchers, regulators, and the public, complicated further by lags between exposure to pollution and health effects. Sometimes exposure effects are immediate; sometimes they emerge over days, sometimes over decades, sometimes across generations. Figuring out how to characterise the temporal dimension of pollution data is also a critical dimension of critical data design.
Perhaps most basic are the challenges connecting pollution data to (human, ecosystem, and atmospheric) health data. Historically, pollution and health sciences have developed in largely separate domains; government agencies also tend to be organised in ways that make it difficult to interconnect health and pollution problems. As illustrated in the story of the Scorecard project (elaborated below), bringing health and pollution together in an accessible way relevant to local contexts in the United States took notable ingenuity and proved too expensive to sustain.
The Scorecard example also points to another challenge of pollution data – ways it is often both noisy and subject to commercial interests. One of the innovations of the Scorecard project is the way it leveraged existing data known to be imperfect. Scorecard was built around pollution data from the US Toxic Release Inventory, which includes self-reported and largely unaudited emissions data from large industrial facilities. The data thus contains many errors (some argued to be intentional) but could still be used for environmental sense making with appropriate caveats. Identifying and articulating these kinds of caveats are an important part of critical data design. 3
In this article, we discuss three projects that illustrate the challenges of pollution data and critical data design. We start with the pioneering pollution database Scorecard, released by the US NGO Environmental Defense Fund (EDF) in 1997. Described early on as an ‘Internet Bomb’ and as the ‘new gold standard’ in environmental information systems, Scorecard linked local pollution data to health data, providing users with risk profiles that helped them prioritise complaints. The other two projects we describe were also designed to be game changers, working to bring different kinds of data and people together in new ways, making environmental problems more visible and actionable. The US EPA’s EnviroAtlas brings together data sets produced and owned by various US government agencies, configured to make it easier for decision-makers (especially at the local level) to value ecosystem services (ESs) (recognising what would be lost if a road disrupted a wetland, for example, or what would be gained through urban greening). The Houston Clean Air Network (HCAN) publishes a map of real-time ozone levels in Houston, working with air quality monitoring data obtained (with considerable effort) from the Texas Commission on Environmental Quality (TCEQ) – putting air pollution on the map in a city known for its investment in both its cars and freeways, and the petrochemical industry. 4
The designers of these projects leveraged knowledge drawn from toxicology, ecology, air chemistry, and other scientific fields. They also leveraged the capacity to read and interpret the social, political, historical, and cultural context in which they worked, recognising and designing against problems with ways people habitually think and talk about problems, possibilities, and responsibilities. They thus linked scientific and technical expertise with hermeneutic expertise, taking into account what things mean, to whom, why, and to what end – becoming what we call critical data designers. 5
Before moving to our three examples, we briefly describe the methods and theoretical frames through which we have developed our conception of critical data design – as a process that can be followed ethnographically and cultivated pedagogically. We encourage both further study and cultivation of critical data design in action, building a comparative body of work that can orient and inspire teaching and students.
Configuring critical data design
The pollution data projects we describe in this essay all push back against entrenched ways of thinking about the environment and its problems. Our reference to ‘the environment and its problems’ echoes John Dewey’s important 1927 work The Public and Its Problems. The book was an extended response to journalist and social critic Walter Lippmann, who contended that the publics on which democracy depended were often eclipsed by powerful forces (corporate capital or advertising, for example) that worked against publics recognising themselves and articulating their needs and criticisms. In Dewey’s formulation, publics could be, and needed to be, provoked into existence through collective recognition of the negative externalities of state, market, or other social action. Central to that radical democratic project for Dewey were new modes and tools of communication: Only when there exists signs or symbols of activities and of their outcomes can the flux be viewed as from without, be arrested for consideration and esteem, and be regulated… As symbols are related to one another, the important relations of a course of events are recorded and are preserved as meanings. Recollection and foresight are possible; the new medium facilitates calculation, planning, and a new kind of action which intervenes in what happens to direct its course in the interest of what is foreseen and desired… The work of conversion of the physical and organic phase of associated behavior into a community of action saturated and regulated by mutual interest in shared meanings, consequences which are translated into ideas and desired objects by means of symbols, does not occur all at once nor completely. At any given time, it sets a problem rather than marks a settled achievement. (Dewey 1984 [1927]: 330–331)
We came to this way of thinking about critical data designs through sustained work – in theory and practice – oriented by radical education thinkers such as Paulo Freire (1968), Gregory Bateson (1972), Gayatri Spivak (2012), and Shoshana Felman (1982). 6 Dewey’s conception of ways ‘public problems’ take shape is just one touchstone, helping us crystallise the process of critical data design. Our understanding of this process also comes from our own effort to support critical data design among ethnographers 7 and from long-running ethnographic study of data practices in different settings. One thread of our ethnographic work has focused on ways environment and health data has been used in governance of industrial disaster, both fast (as in Bhopal and Fukushima) and slow (in chronic air pollution at sites around the world, for example) (Fortun, 2004; Fortun and Morgan, 2015). Another thread of ethnographic work has focused on how practitioners in a range of communities (from genomics to data science to demography) are leveraging data infrastructures to produce knowledge in new ways (Poirier, 2015).
In our three examples, we briefly (and far from thoroughly) draw out the process of critical data design. For each, we both interviewed the lead data designer (or designers) and became users of the systems they built.
Scorecard.goodguide.com
‘Informational strategies’ for dealing with environmental risk became law in the United States in 1986 with passage of the ‘Community Right-to-Know Act’, Title III of the Superfund Amendments and Reauthorization Act (SARA). 8 Widely regarded as the US legislative response to the 1984 chemical plant disaster in Bhopal, India (regarded as ‘the world’s worst industrial disaster’; Taylor, 2014), SARA Title III mandated a range of initiatives to support emergency planning and public access to information, including the Toxic Release Inventory (TRI), the first federal database that Congress required to be released to the public in a computer-readable format (Bass and MacLean, 1993; Young, 1994). The goal of the TRI was to allow the EPA as well as citizens to track and evaluate routine, largely legal emissions from large industrial facilities. 9
Approximately 16,000 facilities across the United States became TRI reporting facilities, reporting millions of tons of releases of a range of chemicals, including many considered carcinogenic, developmental hazards, or of ‘special concern’ because persistent and bioaccumulative. 10 Led by Monsanto, prominent industrial leaders promised significant and immediate emissions reductions (Graedel and Howard-Grenville, 2005: 31; Hamilton, 2005: 225–226). Critics began to note ‘phantom reductions’ resulting from creative emissions accounting, and the way ‘delisting’ could be an emissions reductions strategy; if a chemical was taken off the list of reportable chemicals, the overall quantity of overall emissions reported would decrease; de-listing thus became subject to well-funded corporate lobbying. In the mid-2000s, efforts to weaken TRI reporting again provoked strong criticism, making clear that environmental politics are also data politics. 11
TRI data mobilised both grassroots and national toxics activism. It also, however, provoked investment in more data – data pointing to the health consequences of exposure to TRI-reported chemicals. Bill Pease was at the centre of the storm. In the mid-1990s, Pease was lead toxicologist at the EDF, a leading US environmental organisation already at the forefront of work on toxics. And he couldn’t handle the number of calls he was getting to help interpret TRI data. People in communities around the country – many with full-time jobs – were spending hours and days in libraries, knocking on the doors of government health agencies, and sometimes in the ‘reading rooms’ of corporations to try to figure out what their exposures to TRI chemicals meant (Fortun, 2012: 319–320). Pease recognised a need for a shared resource. At the outset, Pease imagined a CD-ROM that he could send out on request by mail. A meeting with MIT computer scientist Phillip Greenspun turned the vision towards an online platform that would link TRI data to health information available through the US National Institute of Health (NIH), Center for Disease Control (CDC), and other government agencies. The vision was not to produce new data, but to link existing data, configured in ways that enabled interpretation and directed action.
Greenspun had prior experience working to make environmental data meaningful. In 1986, the State of California has passed Proposition 65, which required industry to report both what they emitted and whether the substances emitted were carcinogens or reproductive toxicants. The result was that California cut emissions covered by Proposition 65 by 25% – twice as much as the TRI at that point. What Greenspun learned from this is that ‘disclosure plus interpretation is more powerful than disclosure alone’ (Fortun, 2012: 320). Scorecard – http://scorecard.goodguide.com/ – was designed to support this kind of interpretation.
EDF launched Scorecard in 1998, saying that its purpose was ‘to make the local environment as easy to check on as the local weather’ (Krupp, 1999). Chemical Week described the website as the ‘Internet Bomb’ because of its potential effect on the reputations of chemical companies (Foster et al., 1998). Greenpeace referred to Scorecard as the ‘gold standard’ of environmental information systems because it facilitated movement from information to collaborative action, and because it was partly built on open-source software, which in Greenpeace’s view operated according to the same tenets as radical environmentalism (Fitzgerald and Hickie, 2002).
Scorecard connects TRI emissions data to chemical toxicity data drawn from over 400 US government databases. Users could type in their zip code and pull up lists of specific chemicals emitted by a specific facility, indicating whether the chemical was carcinogenic or a developmental toxin. Users can also evaluate probable risks based on a hazard ranking system based on proxy data that related all chemicals to the risk of benzene, a known carcinogen (to indicate ‘cancer potential’) or to toluene, a developmental toxin (to indicate ‘non-cancer risk’). Users could also use Scorecard to communicate with the US EPA and with polluting companies.
Scorecard built in recognition of the limits of both the data it made available, and the risk profiles it enabled users to generate – noting, for example, that its maps do not cover non-TRI reporting pollution sources, and the TRI only accounts for approximately 650 chemicals and chemical categories. Importantly, these caveats were not presented in a way that paralysed data use. Instead, they produce a savvy data literacy that positions users to see data as an important but imperfect societal resource. Data is not cast as the simple truth of the matter. Responding in the early 2000s to debates about the character of good risk communication, Kim Fortun (2004) argued that: [t]he experience of Scorecard can be dizzying. But Scorecard takes on some of the most recalcitrant problems within environmental politics - the need to deal with too little, as well as too much, information; the need to deal with contested scientific findings and intractable uncertainty about long-term effects; the need to think locally, as well as comparatively and globally. The high level of information literacy required by Scorecard can be cause for criticism. It can also be argued that the way Scorecard requires and supports high levels of information literacy makes it an appropriate technology for contemporary environmentalism. (60–61)
Scorecard.com leverages existing data by linking it. Scorecard also built in acknowledgement that the data and its meaning were far from straightforward. The goal was to advance insight, even without the promise of total accuracy. The genius of Scorecard was thus technical, as well as conceptual. It radically revamped both the conception and practice of ‘risk communication’.
https://www.epa.gov/enviroatlas
Ecologist Anne Neale was like a circuit rider, going from one U.S. government agency to another to spread her message. But there was a critical difference: Neale had to be communicatively persuasive because she wanted something from each agency she visited – their data. And she got it: water use data, crop yield data, carbon storage data, average daily potential kilowatt hours of solar energy that could be harvested per square metre within a particular subwatershed, etc. She pulled it together to create the U.S. EPA’s EnviroAtlas (www.epa.gov/enviroatlas), a web platform where users can visualise and evaluate ESs for both research and practical decision making.
Released in May 2014 after years of development, the EnviroAtlas includes an open source GIS-based mapping application, an ‘eco-health relationship browser’ that enables users to access relevant peer-reviewed research publications, and a suite of downloadable analytic tools that, in the words of its designers, ‘enable information integration across the (bio-)geophysical spectrum, in concert with anthropogenic data such as demographics, suburbanisation, and changing policies, in order to fully explore the relationships among ES and human activities’ (Pickard et al., 2015: 45). The most basic goal is to demonstrate the value of ESs, pushing back against deeply entrenched tendencies to ignore how healthy ecosystems support human health and well-being – counting as zero, as Neale explains. ‘There is going to be a segment of the population that isn’t interested in nature for nature’s sake – the ducks and the bats that we environmentalists are concerned with’, says Neale, ‘so in documenting the ESs that may be lost we ask, “can it be reframed as mosquito reduction services and quantify that into dollars or disease incidence?”’
Another goal of the EnviroAtlas to support systems decision making, pushing back against problems being seen in isolation and decisions made without regard for context or distributed impacts. ‘Taken in isolation’, the designers point out, each disciplinary field (e.g., economic, social, or ecological) can address only a limited range of management and policy related questions. Yet, when multiple disciplinary fields are linked together through an easy-to-use interface, the result is a novel tool that has the potential to enable better decision making across multiple sectors. (Pickard et al., 2015: 45)
The EnviroAtlas now draws in data from numerous federal and state agencies, as well as universities and NGOs, enabling mapping and analysis at many scales. For example, users can layer in demographic data sets from the ‘People and Built Spaces’ section (population under one year old, per cent population other than white, population with income below twice the poverty level, per cent linguistically isolated households, etc.), perhaps adding National or Community-Scale metrics on annual health costs avoided from pollutants being removed by tree cover, or the amount of carbon stored in tree biomass – supporting both local decisions (where to plant trees along a 1 mile stretch of roadway, for example) and national-scale deliberations (whether to approve a gas pipeline, for example). The EnviroAtlas also makes data sets available for download, streamlining what used to be a massively time-consuming task of gathering data produced and owned by different agencies and researchers.
The EnviroAtlas’s ‘Example Uses’ section illustrates how it can be used. In one example, the value of planting trees in Durham, North Carolina (USA) is demonstrated, drawing in data locating homes, day care centres, and schools, showing how children move through the city and different air sheds at different times of the day (see Figure 3). Text explains that homes account for only about half of children’s whereabouts during the week, and that ‘the location of daycare centres is of particular importance because of the extra vulnerability of the youngest children to unhealthful environments’. Links describe how trees and green roofs provide important filtration services and can reduce building energy consumption, and thus polluting emissions (elsewhere, if the energy used is electrical). Users are also pointed to EnviroAtlas’ Eco-Health Relationship Browser, where they can access peer-reviewed, EPA-vetted scientific publications. The goal is to be able to see where and why planting trees could make a difference – supporting both city planners and community advocates (see Figure 4).
The EnviroAtlas turns data collected by an array of sources into a new societal resource, leveraging long-running data collection efforts, most funded with public monies. The interfaces and tools it provides make innovative use of Big Data, enabling users to see the world from different angles, problematised in different ways. But it doesn’t claim to be a stand-alone solution, alerting users as they enter the system that ‘EnviroAtlas data will not replace “boots-on-the-ground measurements” or local knowledge’.
When asked what she hoped for with EnviroAtlas, Neale shared a story from about 15 years ago – a story that she would like to see replayed many times, with strong infrastructural support. Effluent from a sewage treatment plant had been warming a stream and significantly degrading fish habitat. The county was faced with huge costs for refrigeration units to cool the water before it went into the stream. Planners with training in ecology became involved and managed to convince the sewage authority to pay farmers along the stream to leave the portion of their land bordering the stream fallow. Trees were planted along the stream that filtered runoff and the temperature of the water dropped. The farmers were getting paid more than what the land was worth for agriculture, the county was saved from much higher refrigeration costs and the amount of phosphorous running into the stream from the agricultural runoff decreased significantly, improving the habitat for fish, birds, and the functioning of the ecosystem as a whole. This is the kind of success story Anne Neale and designers want to replay, supported by the EnviroAtlas.
The word ‘environment’ can evoke images of bucolic landscapes far from or even defined by the absence of humans and their urban spaces. Alternatively, reference to ‘the environment’ can evoke images of polluted waterways, frogs with five legs, and coughing children. The EnviroAtlas works at the interstices of these alternatives, providing a way to think about and approach ‘the environment’ that is more practical than sublime, offering possibilities for human activity that are protective and regenerative rather than destructive. Environmental protection becomes a proactive and positive venture.
http://houstoncleanairnetwork.com/
Houston has long had difficulty governing its air, repeatedly falling out of compliance with US federal air quality standards, particularly for ground-level ozone. And the difficulties are far from over. The United States EPA recently strengthened ozone standards, pushing Houston ever farther from consistent ‘attainment’. Across the United States, the implications of the new standards are recognised as requiring radical transformation of the transportation sector in particular. Stanford University civil and environmental engineering Mark Jacobson argues that the only possible solution for California is zero tailpipe emissions (Jacobson, 2015). Houston will have it even harder, needing to contend with pollution from its enormous industrial as well as transportation sector. The State of Texas is unhappy about these developments, leading efforts to discredit the science supporting stricter ozone standards, disputing claims that there is clear epidemiological evidence linking smog and asthma (Grant et al., 2007). The arguments are about what counts as good science. They also pit industry against regulation, economic opportunity against public health, offering residents a devil’s bargain: if they want wealth, they must sacrifice health.
Philosopher Dan Price has pushed back, working to make air pollution and its health impacts in Houston more visible and actionable. Working with atmospheric scientist Barry Lefer, Price, and a new alliance of organisations, the HCAN, envisioned a way to turn routinely collected ozone data into a map showing different and ever-changing ozone levels across the City of Houston in almost real time, in fine-grain detail. Previously, ozone data was only available as 1 h averages, updated every half hour after the reading period (levels between 2:00 and 3:00 p.m. would only be available at 3:30, for example), rendering it largely irrelevant to users making decisions about children playing outside, or whether and where to go for a run, or to hold football practice outdoors. Like Environmental Defense in the 1990s with Scorecard, the goal of the HCAN (http://houstoncleanairnetwork.com/) was to make pollution data as easy to check as the weather – to provide accurate and usable information, in a way that provoked cultural change in how people think and move about the environments in which they live (see Figure 5).
The data in the Houston Clear Air Network Ozone Map comes from 45 monitors across the City of Houston (about 1500 km2), operated by the TCEQ, Harris County, the City of Houston, and the University of Houston. Atmospheric scientist Barry Lefer had access to the data used in the Ozone Map because he operated some monitors in the network and had access to data from other monitors for purposes of comparison and validation. The data was fed to the US EPA in real time, but public release was subject to a 1 h delay. To gain access to the data in real time, for public viewing, Price and Lefer had to get the TCEQ on board, and this required involvement of someone who could ‘write a letter to the top’. The diplomacy required has been ongoing and not entirely successful; data access has decreased over time (since the project started in 2012), as TCEQ has cited security concerns and load on its servers. Price has also had ongoing negotiations with other stakeholders, including the Texas chapter of the American Lung Association, which had different ideas about the kind of information that should be available and envisioned a more traditional public education website, with little space for user engagement.
Designing the visualisation (and a mobile app) for the data was a key component of HCAN’s pushback. Similar sites may map the numerical value for ozone levels reported by a monitor at a given time (but 1 h later), but the visualisation Lefer and Price developed for HCAN takes the monitoring data, interpolates it to generate estimates of ozone levels for the immediate area around each monitor, then puts that data in time to generate almost real-time images of ozone clouds travelling across the region. By spatialising and temporalising the data in a unique way, the HCAN visualisation made ‘checking the ozone’ just like one might check the radar for an approaching rainstorm. The ozone clouds were coloured to match the codes of the Air Quality Index, the now globally standard way to community air pollution hazards (even if what counts as hazardous is different in different places): a good air quality day is green, for example, developing into clouds of yellow, orange red, purple, and maroon as air quality worsens over time, in different parts of the city.
For Price, user engagement is not only about ‘empowerment’, but about a need to change the way science is produced and operates in society. Conventionally, science is done then pushed out to users, who are then supposed to act. This linearity leaves no space to leverage the experience and perspective of people variously situated in the world science has been tasked to characterise. The HCAN site includes educational modules on the science and governance of ozone. Price envisions systems that allow users to engage their operational side, rather than simply being informed by their functional outputs; machine learning would be replaced by user-directed systems. ‘The promise of automation with correct categorisation as its endpoint, which the dominant Anglo-American tradition accepts as the role of science, has no place for decision’, Price explains, ‘The science merely performed its operations and we watched, sometimes trying to gently guide from the sidelines’ (personal communication, 26 August 2015).
Price came to his understanding of the kind of air quality knowledge and knowledge production needed through his work as both a philosopher of science (Price 2009), and on the ground in Houston, coupled with advanced, self-taught programming skills. Analytically, Price is able to parse many problems associated with conventional approaches to environmental health research and governance. In response, he’s built and envisioned alternative approaches, becoming what we’ve called a critical data designer. One initiative, for example, pushes back against the single chemical focus that has long characterised environmental health research and governance, reaching (like Anne Neale in building the EnviroAtlas) to capture complex causation and cumulative effect. For this, Price and colleagues have experimented with software originally developed to understand shopping behaviour, pushing back against deeply held assumptions about statistical validity. In a related initiative, Price advocates for much more extensive air pollution monitoring than is currently in place – to be able to support modelling with much finer granularity (see also Garnett in this theme issue for a discussion of modelling and monitoring air quality). In Price’s vision, modelling with finer granularity – at the neighbourhood level versus the 1 km map square currently supported – will provide a different, potentially transformative kind of user engagement with air quality. And even more so if connected to data on health outcomes, linking, for example, asthma-related emergency room visits to particular exposures. Price also wants to add pollen to the mix – because it is a notable asthma trigger, exacerbated by both pollution and warming conditions, but also because it has a different political charge. Houston’s (petrochemical) political economy and culture make it especially difficult to address air pollution. Pollen is less threatening. But if people could come to be interested in and concerned about pollen, they would step towards greater concern about air quality overall. Pollen literacy would almost inevitably lead to pollution (and political) literacy. Price’s envisioning of this is a critical step in the critical data design process, coupling analysis of a cultural landscape to technological possibility.
Price has learned a great deal building and sustaining Houston’s Ozone Map – about air chemistry and the health effects of particular pollutants, about the limitations of using ozone as a proxy for overall air quality, about cross-disciplinary and cross-organisational collaboration, and about the politics of data access and delivery. The problems seem endless, but – as Price insists – ‘interesting’, recognising operational challenges as cultural and conceptual challenges. This is key to critical data design. As Prices’s work illustrates, technical work itself becomes transformed when seen as cultural work and as pushback against entrenched knowledge systems.
Lessons of critical data design
New data practices are changing how problems of many kinds are recognised and addressed. But the link between data and problems is far from straightforward, and not simply a technical outcome or challenge. Problems can emerge from data – as in ‘fourth paradigm’ science involving various non-hypothesis-driven techniques. Such approaches depend, however, on the availability of very large, carefully curated data sets, which in turn depends on entwined scientific and political vision and will – and a collective capacity to make data investment make sense to multiple stakeholders, with different ways of thinking about what is valuable and credible. Technical capacity alone is not sufficient for this.
Linking data and problems can also emerge from nascent awareness of problems that aren’t yet public problems (in Dewey’s sense), coupled with awareness of data sets that could help with the translation of problems into public problems. Here, too, combined technical and interpretive capacity is required, pushing back against entrenched ways of thinking (or not thinking) about problems by locating and linking relevant data, then creating and circulating compelling data visualisations – knowing that what counts as compelling always depends on context. It is this mode of linking data and problems that we have called ‘critical data design’ (see Figure 6).
The greater Houston area has more than 270 schools in (often nested) vulnerability zones around industrial facilities. In the wake of a massive fertiliser plant explosion in 2013 that destroyed a nearby middle school, the Center for Effective Government (CEG) generated maps like these for schools across the United States. Data used in these maps is available for most large industrial facilities because of legislation passed after the 1984 Union Carbide chemical plant disaster in Bhopal, India. The resulting surge of publicly accessible pollution data enabled and accelerated critical data practise in the environmental domain. What would become CEG started in this period, as the Right-to-Know Network (Image courtesy of Center for Effective Goverment, http://tesla.foreffectivegov.org/KidsAndToxins/bin-release/; last accessed 31 August 2016). The Manchester community in Houston, Texas, which faces extraordinary industrial risk yet remains largely invisible politically. Organisations like TEJAS – Texas Environmental Justice Advocacy Services – push back against this invisibility. Critical data products like the maps made by the Center for Effective Government animate their work. Photograph by Kim Fortun. Prioritising tree planting in Durham, North Carolina (USA) to maximise health benefits for children. Image courtesy of US EPA Enviroatlas. The Eco-Health Relationship Browser. Image courtesy of US EPA Enviroatlas, https://www.epa.gov/enviroatlas. Houston Clean Air Network’s Ozone Map, which displays moving, colour-coded clouds that show different concentrations of ozone in different places in almost real time. Image courtesy of Houston Clean Air Network. Dynamics of critical data design.
Critical data design depends on capacity to read and interpret historical, social, political, and economic landscapes, attuned to how habitual representations and established programs miss or discount many problems. It also requires knowledge of data landscapes and the kind of diplomatic skill needed to gain access to many data sets. And it depends on the technical skill and creativity needed not only to make visualisations, but to craft them so that they speak to particular contexts and audiences.
In this article, we’ve begun to draw out the process of critical data design, arguing for greater recognition of how scientific and technical expertise are supplemented by hermeneutic expertise (see Figure 6). Delineating the process of critical data design, and the different types of expertise involved, advances understanding of contemporary, ‘late industrial’ (Fortun, 2012) knowledge formation in comparison to knowledge formations in other historical and cultural contexts (see, e.g., Galison and Daston, 2007; Foucault, 1970; Hacking, 2012). Figuring out the processes and types of expertise involved in critical data design also advances capacity to teach it – beyond what is often taught as ‘data analytics’, beyond claims that the hermeneutic, interpretive dimension of knowledge production is ‘subjective’ or a matter of inspiration or even creativity. Interpretation is creative, but it is also a skill that can be cultivated, validated, and evaluated, drawing on a rich history and body of work in the humanities.
Interpretive or hermeneutic capacity is capacity to understand what things mean, to differently situated actors, why, and with what effect on what will follow. Such capacity is cultivated in philosophy, literature, cultural anthropology, and other fields often omitted from the way data analytics is usually conceived. But there is growing recognition of the need for greater inclusion, to address the types of problems that are difficult if not impossible to address through established approaches. 12
Pollution problems are exemplary. Many people, variously situated, see pollution (of various kinds) as an increasingly urgent problem, with dramatically uneven and unjust impacts. But both culture and politics in many settings work against recognising this, and addressing environmental pollution thus requires creative pushback against entrenched ways of thinking and acting. As Dewey has argued, democracy as a political formation depends on this, and even more so in an era characterised by ‘wicked’ problems, made even more wicked by the ploys of actors like Exxon and Phillip Morris that actively muddy the waters, so to speak, making it more difficult than ever to make environmental sense (Oreskes and Conway, 2011). Critical data designers are thus crucial figures, deserving greater attention and analysis. Often eclipsed by the visualisations they produce, critical data designers deserve to be pulled to the foreground, observed, analysed, and supported, now and in the next generation. A growing need for critical data design as both a technical and interpretive process needs to be recognised as a public problem.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
