Redistributing the Means of Data Production, Seizing the Means of Data Circulation

by Kathryne Metcalf, Lilly Irani, and Verónica Uribe del Águila

University of California San Diego

[email protected]

Introduction

Inspired by the call to “defund big tech and refund community” (Barendregt et al. 2021), we seek to question and reimagine basic assumptions about the production and circulation of health data in the wake of the COVID-19 crisis. Through a study of a university workplace COVID-19 mitigation program, we show how the provision of data production resources and the management of resultant data is often a tool by which employers draw boundaries and reinforce managerial relations in ways which instantiate vulnerability or harm, particularly for those from marginalized groups. However, we also find that health data systems can be used to empower people to produce necessary data necessary for their health, and to nourish valuable community formations outside of strict workplace hierarchies. By attending to how data flows make particular relationships and community formations im/possible, we argue that communities should be empowered (and provided resources) to produce the kinds of data they deem useful for themselves, and to control where and how that data circulates.

Throughout the pandemic, employers have relied on a variety of tools to collect health data in order to produce “safe” shared working environments. Since March 2020, temperature checking and digital symptom screening have become commonplace, and various forms of testing and contact tracing underpin both private and public responses to rising case numbers. However, access to these tools in work settings has often been controlled by employers with unclear regard for equity or concern for the distribution of viral risk. Moreover, by reproducing rhetorics about the necessity of big data’s aggregation, many workplaces have invisibilized the possibility of other configurations, and have absorbed health data as yet another form of managerial accounting (and, indeed, one that can serve the production of economic value).

We examined these processes intimately in a larger project examining issues around health data and privacy in the context of a reopening program developed by the University of California San Diego, our employer. In the acute crisis of Spring 2020, UCSD was one of the first public universities to announce a return to in-person instruction. Through the implementation of a variety of tools, including digital symptom screening, workplace location registration, contact tracing, and wastewater testing for viral particles, UCSD promised the collaborative production of safe educational environments and the skillful management of workplace risk through the collection and deployment of health data and testing processes. Drawing on a year of ethnographic observation, 38 interviews with campus students, faculty, and staff, and our own experiences as employees and organizers, we examine how this promise was negotiated and contested, and consider how it might have been otherwise conceived.

We argue that workers should be given tools to share data accountably with the people and communities they live and work in relation to. In practice, this means that health data should not be assumed to simply flow “up” to managers made responsible for decisions about aggregate or subordinate health. Instead, we call for the redistribution of the means of health data production and control over health data circulation to autonomous actors in community. This may require large scale networks of health technologies that produce data about our bodies, but we should be given the ability to use those resources to produce and circulate data in economies outside of profit accumulation. This is not quite data about our bodies in commons, but data about our bodies in making care relations.1

Ultimately, we problematize the role “privacy” has often played in discussions about workplace data collection: complicating binary accounts about which data should be collected and what should remain “private,” we propose organizing around data’s relationality. Previous work has demonstrated how data-driven systems construct particular relations which colonize and distort our ways of knowing, doing, and connecting (Couldry and Meijas 2019). Here, however, we reach for the opposite, and ask how desirable relations and a commitment to shared human health and flourishing can be made with and through data. In doing so, we call for a proactive reimagination of data storage and circulation to begin to counter the hegemony of big tech’s data regimes.

Health Data and COVID-19

As early as March 2020, critical scholars of technology recognized that COVID-19 surveillance was suddenly both a necessary precondition of social life, and simultaneously likely to create and intensify a number of technologically-driven forms of vulnerability, risk, and marginalization (French and Monahan 2020). Since that time we’ve witnessed a huge variety of permutations in COVID data management and associated organizing, both good and bad. On one hand, citizen scientists have developed new dashboards and visualization tools designed to bridge the gap between expert and public access to data; long COVID activists and patient-advocates have used online tools to aggregate their symptoms and demand medical and research attention; and mutual aid networks have both widened and deepened in their ongoing efforts to care for their communities. On the other hand, COVID data in the US has been shared with law enforcement for policing purposes, leading to both racialized harms and popular withdrawal from a needed public health program (Molldrem et al. 2021), and abroad, facial recognition and other law enforcement technologies have been repurposed to indicate compliance with masking and other COVID policies (Kaye 2021). In these conflicts, the urgency of the public health emergency has been used to spur new community-strengthening projects while simultaneously naturalizing the oft-invasive collection of personal data for surveillance and policing purposes.

Here, though, we observe that discussions around these issues have often centered on a god’s-eye view of health surveillance–similar to what Luke Engelmann (2022) describes as “pathological omniscience”--in which either the employer or the state is solely responsible for monitoring COVID transmission, with the power to mandate or disallow the collection of health data through any number of technological tools. This framing is what we want to contest: while workplaces certainly hold much of the responsibility for maintaining a safe work environment, they should not and cannot expect to do so solely through the top-down imposition of health data collection.

We argue that by mandating the same data flows and forms of institutional visibility for all workers, employers actually damage the social and community relationships that make the daily life of the institution function. Instead, we suggest flexible modes of information sharing that allow workers access to alternative data pathways (e.g. avoiding direct supervisors) and anonymity (e.g. location-based rather than individual-based risk assessment). We will revisit this argument and its implications at the end of this essay; for now, though, we will turn to describe our case study in more detail.

Key Findings: From Privacy to Relationality

Across a year of research at UCSD, we highlight two seemingly-contradictory findings. First, nearly everyone we spoke with thought the campus COVID surveillance program was a “basically good” intervention, and all reported interacting with its health data tools (particularly digital symptom screening and free COVID test vending machines) frequently. Second, many of our informants–and particularly those from marginalized communities–simultaneously described a variety of concerns ranging from disability stigma and employment discrimination to the unconsented extraction of value through data resale. We suggest that this apparent conflict is a result of understanding particular types of sensitive data (e.g. health and medical data) as a unique property of individuals, as do many constructions of “privacy” (see Cohen 2013 for a related critique)[8]. Instead, we follow Dourish and Anderson (2006) in tracing data flows as a form of “collective information practice,” through which people situated in specific social environments can enact and negotiate a variety of social and group formations. In what follows, we describe practices of data production, data circulation, and the relational tensions we observed provoked by existing practices.

COVID Screeners: Tensions in Circulation

One of several health data-circulation tools employed by UCSD was a digital symptom screening survey, required for students and staff to enter university facilities. Upon completion of the brief form, the user is presented with a timestamped graphic of a green thumbs up, marking them as safe for work or other campus activities for the current day; a red thumbs down, indicating that they must quarantine and/or undergo COVID testing before returning to campus; or a yellow thumb, which initiates a follow-up phone call from university contact tracing. For UCSD employees (including student employees), the thumb graphic is also automatically sent to their supervisor/manager. Most of our informants were not concerned about the information recorded by the screening tool: while many admitted a sense of minor imposition while completing the survey, they were frequently quick to describe its questions as “reasonable under the circumstances.” While the specific data collected was not a significant concern, however, several people expressed confusion, frustration, or anxiety about how this information was shared. Across a variety of discussions with workers occupying positions from librarian’s assistant to full tenured faculty, our informants described how the screening tool’s notifications had uncomfortably shifted the carefully choreographed social and professional communication practices that otherwise structured their workplace relationships.

For example, Meredith—a teaching assistant (TA) for a large class with multiple instructors—described feeling distressed that the screening tool only provided information to her advisor, the instructor of record for the course. Because she also came into contact with other members of the instructional team and anticipated meeting with students as courses returned to in-person modalities, she felt that her “green thumb” should be shared broadly as a sign that she was currently safe to interact with, asking:

“But what about all the other people? Like, I guess, the rest of the instruction team plus the students that I come into contact with? So why is my advisor the main point of contact once I'm like, you know, once I'm reporting some sort of symptom? Why is that person the main point, and not all the other people? I think that's really strange to me.”

The information flow patterned through the symptom screener mirrored the hierarchical flow of communication assumed by the university’s organization of labor, in which decisions about the allocation of Meredith’s labor (here including work in physical proximity to others) are assumed to be made by her supervisor. Sharing information only hierarchically through the mediation of the screening survey represented a troubling information practice not because it was insufficiently “private,” but in fact the opposite—Meredith preferred to share her health status more broadly, and was frustrated by the imposition of a system that, as she understood it, was designed to protect community health but instead functioned only as a tool of pastoral managerial power.

In contrast, Jean, a teacher with a [State University]-affiliated primary school, was particularly apprehensive about how information collected by the screening survey might shape her supervisor’s opinion of her, particularly regarding her health and fitness to work more generally. As a frequent migraine sufferer, she reflected on how:

“I don't know how to answer the question [about symptoms] because frankly, it's a migraine, it’s not COVID. But guess what? ‘I have a headache’ comes up [on the screener]... there's something about, like, you know, your supervisor is being notified—it’s just sort of like this watchdog feeling… I'm like, is it gonna be looked badly upon if they’re constantly thinking I have a headache? Again, you know, it's just, how is that going to be interpreted? How is that going to influence people's impressions of me? You know, all of that... When it's going to your supervisor, it does feel very personal, especially when the person is going to be, like, responsible to do your performance evaluations.”

Here, Jean expresses concern about being the “kind of person” who has frequent changes in health that may need to be reported, and how that perception would shape her professional reputation. Her concern is suggestive of how patterns in health data can be used to impute medical conditions beyond those it specifically describes, linking the screening survey to larger issues regarding health and workplace disability discrimination. However, Jean also repeated throughout our discussion that she was glad that the survey was in place, and sincerely hoped it would stay—in some form—through her upcoming return to in-person instruction. These conflicting desires and anxieties raised a variety of further questions about how the information was being “looked at”—metonymically linking how the data is “seen” to how she is “seen” as an employee.

Jean’s and Meredith’s concerns converge on how their health data should–and shouldn’t–circulate in order to protect themselves and their colleagues from disease while simultaneously allowing for other personal circumstances to remain unseen. When privacy theories attempt to fix certain information as inherently sensitive and belonging to an individual, they misdirect attention to data as a kind of individual property rather than one medium of enacting relationships. Confusion, frustration, and anxiety about how survey data was “being looked at” outside of situations of acute COVID risk featured across a variety of discussions, suggesting that the complexities of health data interpretation within interpersonal relationships and the entangled choreography of cooperative work are a needed supplement to accounts of data as an individual property right, with special protections for universalized categories of “sensitive” data.

Testing as Data Production: Tensions in Sharing

Perhaps the most important form of data production has been through the distribution and use of COVID-19 tests. In January 2021, UCSD added vending machines with self-administered COVID tests to several campus facilities. While university students and many employees had access to these free tests, this privilege was not extended to all on campus: contractors and construction workers, for example, were redirected to an expensive, off-campus UCSD testing site despite routinely working in physical proximity to others at the university. Only two weeks after the first machines were placed, an email from university administrators to the student body announced that “sharing is decidedly not caring,” linking to a policy that threatened disciplinary action if students’ “friends and family” utilized the free vending machine tests rather than the $65 off-campus testing site. Several weeks later, a second email announced that “sharing is NOT caring”—now with bolded, capital letters—continuing that “while it may seem like a compassionate act to provide a friend or loved one with access to testing, using another person’s ID not only constitutes actual medical fraud, it also poses a serious threat to our public health and contact tracing efforts.”

As the recurrence of this admonition suggests, some in the campus community sought to reroute university resources as a form of material care—care which UCSD administrators explicitly sought to foreclose, reframing (and punishing) these solidarity efforts as “medical fraud.” What was at stake in criminalizing the sharing of tests? As part of the campus reopening experiment, UC San Diego announced that all employees using campus testing facilities would be included in the campus health system’s electronic records management system, EPIC. This system was a key infrastructure in larger UC efforts to create a central repository of data to fuel rapid cycles of biomedical research and innovation (Johnson 2018). In this context, sharing tests polluted research results even if it might have provided a life saving means of data production in the near term. In these conflicts we observe a differential valuation of some life and labor over others, and of scientific knowledge over everyday survival. We insist that efforts to police communities into the production of “good” data–data which is individually identifiable and produced within prescribed parameters–are similarly doomed to fail. Providing broader access to these data production tools without constraining their use would allow alternative forms of care to proliferate.

While the hierarchical structure of symptom screening and test allocation repeatedly prompted concern or frustration amongst our participants, another aspect of UCSD’s COVID mitigation program was met with universally positive sentiment: the wastewater testing system. Beginning in May 2020, university researchers collected daily wastewater samples from campus buildings to be tested for COVID; on-site residents and workers were notified of any positive results in their buildings, and were referred for further testing. Not only was the system highly effective, leading to the early diagnosis of nearly 85% of COVID cases on campus (Karthikeyan et al. 2021), but it functioned as a passive and noninvasive surveillant apparatus that linked illness to specific buildings rather than individual bodies. As compared to other tools, which relied on predetermined interpersonal relationships to function (for example, between a worker and her supervisor in the case of the screening survey), the wastewater testing system appeared to be agnostic about the kinds of relationships that came to account for disease identification, transmission, and management.

Or, almost agnostic. Some of our respondents described how friends and family members had visited their campus apartments during periods in which those buildings tested positive; because the wastewater notification system only contacted residents, they then had to share that information with their guests. More recently, following the large-scale return of students and workers to campus, notification emails were replaced altogether with an online dashboard. While the information there is public, it does not necessarily reach everyone to whom it may be relevant: we have heard numerous accounts of friends, colleagues, and administrators contacting their communities to point out that particular buildings had tested positive on certain dates, and to suggest that individuals who had spent time there might consider getting tested. As these instances suggest, it’s not that the wastewater testing system doesn’t rely on social relationships to function—it frequently has, in a variety of ways. However, it has offered flexibility about how the kinds of information flows and the ties between spaces and the people that occupy them become instantiated in disease mitigation practices. Existing communities and intimate relations share information about COVID-positive buildings on campus, but the testing system itself doesn’t assume or rely on any particular set of connections for it to function.

Conclusion

Interpersonal relationships have always been a necessary social infrastructure of institutional and urban life (Elyachar 2010, Sopranzetti 2017)—that they should be necessary to ensure the function of COVID mitigation programs comes as no surprise. However, information systems can support care relationships without trying to model them. The institutional visibility of the labor and relationships that make up such social infrastructure is frequently unnecessary for these systems to function; moreover, visibility is often undesirable to workers and community members for a variety of reasons (Suchman 1995, Star and Strauss 1999). Mandating that information flows through particular sets of assumed relations, privileging some connections over others, and making certain modes of connection difficult or impossible within the confines of technological tools functions as a form of “torque” (Bowker and Star 1999). Torque remakes and refigures relations as they become (or are forced into) structural elements of sociotechnical systems. We see torquing and related modes of undesirable visibility as the underlying causes of much of the tension we’ve explored here: for example, workers concerned that they may become the subjects of disability stigma after sharing health information with their supervisors. In contrast, the wastewater system—open to invisible, shifting, and heterogeneous negotiations of relational labor—is uncontroversial because it makes no assumptions about what kinds of relationships and data sharing practices should occur.

Ultimately, across our interviews and observations, it’s clear that the specific content of data was not typically the central tension: all of our participants were forthcoming about their willingness to share their health data toward reducing COVID transmission on campus. Rather, it was the particularities of where the data was understood to flow—including to whom and through what channels—that caused concern. As with previous studies that have examined how vulnerable populations are often unwilling to share their health data when law enforcement (Molldrem et al. 2021) or clinical researchers (Ostherr et al. 2017) may have access, we find that the involvement of supervisors, managers, and others in positions of power increases anxiety within a COVID surveillance program, particularly amongst those already marginalized. When privacy theories attempt to fix certain information as inherently sensitive and belonging to an individual, they lose sight of data as a medium through which people can enact relationships and desired disconnections.

Over the course of our research, we also repeatedly witnessed the campus community attempt to produce more data than the university’s covid mitigation program required: individuals attempted to share COVID testing resources, to circulate their symptom screening and vaccination clearances beyond their supervisors, and to share both personal and aggregate health data in a variety of ways. What becomes clear is that while people actively seek resources to produce information about their and others’ well being, they do so in ways which confound institutional expectations that seek to map data to a user body for big data medical research purposes.

This research suggests that standard formations of health data privacy, which regard health data as a property intrinsic to individuals that can be protected through the appropriate technological safeguards—are deeply insufficient for addressing the complex and relational dynamics that make up information practices within COVID surveillance programs. Further, approaches that situate contextual and institutional norms (e.g. Nissenbaum 2010, see also Vitak and Zimmer 2020 on contextual integrity and COVID) as grounds for ethical data practice can leave behind those whose relationships fall “outside” the assumed boundaries of the institution. We found across interviews that concerns about health data privacy are, in practice, concerns about the kinds of relationships made through or made impossible by health data tools. At a time of growing labor militancy and social movements that question the norms of key institutions, we find Nissebaum’s concept of “context-relative informational norms” unable to account for the distribution of concerns about data expressed by our marginalized informants. Our study subjects shared a context, but inhabited different social and structural roles and had lived through varying biographies that shaped their reactions to data practices. Further, these relationships were not approximated or made predictable by institutionally recognized relationships. These findings underscore the need to organize for flexible resources through which individuals and communities can produce and share health information, but which yet allow for selective disjunctures, discontinuities, and alternative arrangements rather than mandating particular and pre-established data flows. Finally, it can be resource-intensive to produce data about the self so that people can engage in data relations with one another. In line with Marxist thinkers who seek to repurpose technologies and infrastructures for public need rather than for profit (Vgontzas 2022; Srnicek 2017), e call for approaches that resource the production of data such as free COVID testing but allow people to choreograph their data relations. These programs should actively provide resources for people to produce and share a variety of types of health data through heterogeneous social configurations toward true collaboration in the production of safe work environments.

Works Cited

Barendregt, Wolmet, Christoph Becker, EunJeong Cheon, Andrew Clement, Pedro Reynolds-Cuéllar, Douglas Schuler, and Lucy Suchman. 2021. “Defund Big Tech, Refund Community.” Tech Otherwise. doi: 10.21428/93b2c832.e0100a3f.

Bowker, Geoffrey and Susan Leigh Starr. 1999. Sorting Things Out: Classification and its Consequences. Boston: MIT Press.

Christen, Kimberly A. (2012) "Does information really want to be free? Indigenous knowledge systems and the question of openness." International Journal of Communication 6 (2012), p. 2870–2893.

Cohen, Julie E., What Privacy Is For (November 5, 2012). Harvard Law Review, Vol. 126, 2013. Couldry, Nick, and Ulises A. Mejias. 2019. The Costs of Connection: How Data Is Colonizing Human Life and Appropriating It for Capitalism. Stanford: Stanford University Press.

Dourish, Paul, and Ken Anderson. 2006. “Collective Information Practice: Exploring Privacy and Security as Social and Cultural Phenomena.” Human–Computer Interaction 21(3):319–42. doi: 10.1207/s15327051hci2103_2.

Elyachar, Julia. 2010. Phatic labor, infrastructure, and the question of empowerment in Cairo. American Ethnologist 37(3): 452-464.

Engelmann, Lukas. 2022. “Digital Epidemiology, Deep Phenotyping and the Enduring Fantasy of Pathological Omniscience.” Big Data & Society 9(1):20539517211066452. doi: 10.1177/20539517211066451.

French, Martin, and Torin Monahan. 2020. “Dis-Ease Surveillance: How Might Surveillance Studies Address COVID-19?” Surveillance & Society 18(1):1–11. doi: 10.24908/ss.v18i1.13985.

Johnson, Christina. (2018). “Big Data, Big Wins in Medicine at UC Health.” UC Health - UC San Diego. (https://health.ucsd.edu/news/features/pages/2018-07-11-big-data-big-wins-in-medicine-at-uc-health. aspx).

Karthikeyan, Smruthi, Andrew Nguyen, Daniel McDonald, Yijian Zong, Nancy Ronquillo, Junting Ren, Jingjing Zou, Sawyer Farmer, Greg Humphrey, Diana Henderson, Tara Javidi, Karen Messer, Cheryl Anderson, Robert Schooley, Natasha K. Martin, and Rob Knight. 2021. “Rapid, Large-Scale Wastewater Surveillance and Automated Reporting System Enable Early Detection of Nearly 85% of COVID-19 Cases on a University Campus.” MSystems 6(4):e00793-21. doi: 10.1128/mSystems.00793-21.

Kaye, Brian. (2021). “Australia’s Two Largest States Trial Facial Recognition Software to Police Pandemic Rules." Reuters (https://www.reuters.com/world/asia-pacific/australias-two-largest-states-trial-facial-recognition-soft ware-police-pandemic-2021-09-16/).

Molldrem, Stephen, Mustafa I. Hussain, and Alexander McClelland. 2021. “Alternatives to Sharing COVID-19 Data with Law Enforcement: Recommendations for Stakeholders.” Health Policy (Amsterdam, Netherlands) 125(2):135–40. doi: 10.1016/j.healthpol.2020.10.015.

Nissenbaum, Helen. 2010. Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford: Stanford University Press.

Ostherr, Kisten, Svetlana Borodina, Rachel Bracken, Charles Lotterman, Eliot Storer, and Brandon Williams. 2017. Trust and privacy in the context of user-generated health data. Big Data & Society. DOI: 10.1177/2053951717704673.

Vitak, Jessica and Michael Zimmer. 2020. More Than Just Privacy: Using Contextual Integrity to Evaluate the Long-Term Risks from COVID-19 Surveillance Technologies. Social Media + Society. DOI: 10.1177/2056305120948250

Sopranzetti, Claudio. 2017. Framed by Freedom: Emancipation and Oppression in Post-Fordist Thailand. Cultural Anthropology 32(1): 68–92.

Star, Susan Leigh and Anselm Strauss. 1999. Layers of Silence, Arenas of Voice: The Ecology of Visible and Invisible Work. Computer Supported Cooperative Work (CSCW) 8(1): 9–30.

Srnicek, Nick. 2017. “We Need to Nationalise Google, Facebook and Amazon. Here’s Why.” The Guardian, August 30.

Suchman, Lucy. 1995. Making work visible. Communications of the ACM 38(9): 56–64.

Vgontzas, N. (2021). Toward Degrowth: Worker Power, Surveillance Abolition, and Climate Justice at Amazon. Vgontzas, Nantina, Toward Degrowth: Worker Power, Surveillance Abolition, and Climate Justice at Amazon (December 9, 2021), Forthcoming in New Global Studies.