“But nothing,” she declared. “What’s fair ain’t necessarily right” (301).
She is Ella—a survivor of chattel slavery, a veteran of the Underground Railroad, and a key character in Toni Morrison’s novel Beloved. She is “a practical woman” (301). A center of gravity, Ella keeps both the book and its characters grounded whenever outside forces threaten to unmoor them. She trusts lightly, if at all. She understands power and its violent machinations. She refuses reduction and rejects simple binaries. She teaches us that the conditions of life escape reductive binaries of fair and unfair, problems and solutions, past and present. When confronted with epistemological and juridical frameworks grounded in such binaries, Ella refuses their terms. Instead, she shows us how to engage the world through an active and avowed uneasiness, a discomfort with “past errors taking possession of the present” (302).
Following Ella, this essay is about unease. It seeks to register a discomfort with both emerging ideals of “data ethics” and the critical responses it has garnered, especially those grounded in political economic and feminist analyses. It is an effort to resist easy binaries, especially those bound up with Western juridical ideals of fairness/unfairness, conceptual distinctions between material/cultural, or the simple liberal calculus of inclusion/exclusion. As such, the following is offered less as a critique and more as a hesitation—a moment to pause and reflect on emergent critical and feminist accounts of the ethical problematics of data science and technology.
The Limits of Data Ethics
While there is no definitive or “universal” ethics of data science and technology, there is an increasingly robust constellation of scholarly conversations oriented toward “data ethics”—that is, toward the ethics of data’s production, circulation, application, and storage. Within these conversations, “bias” has become the dominant language through which we comprehend the ways “big data,” machine learning, and artificial intelligence shape our lives and, in particular, distributions of key liberal goods like rights, opportunities, and wealth (see Hoffmann 2019). In response, liberal ideals of “fairness” and “inclusion” have emerged as key ethical responses touted by industry, academia, and government alike. For industry players, researchers, and policymakers invested in the future of powerful, pervasive applications of data science and technology, values like fairness and inclusion communicate a vision of powerful, data-driven processes like machine learning and artificial intelligence that can evolve and be deployed generically to make the world a better and less violent place. Such visions encourage our dependence on data science and technology through, in part, readily admitting that data technologies can and do facilitate harm. But the issue is framed as a kind of moral doubling down, as an imperative not to refuse extant orders of data science and technology but, rather, to improve them (Greene, Hoffmann, and Stark 2019). Together, the dual imperatives of fairness and inclusion center on a commitment to enrolling more and more “diverse” (usually shorthand for “not white”) persons into existing circuits of power and production—whether as members of the technology workforce, as inputs to policy-making, as sites of data extraction and collection, or as boxes to be checked in processes of development and design (see Hoffmann 2020).
No doubt, the appeal to such familiar liberal concepts is practical for making data-based harms legible (and auditable) within dominant juridical and Western ethical frameworks. At the same time, however, there is an increasing, robust body of work documenting the ways ideals like fairness and inclusion are insufficient for capturing a full range of emergent, experiential, and phenomenological dimensions of data science and technology globally. This isn’t to say that fairness, rights, and inclusion have no place in critical conversations about data. However—to rework a passage from Lewis Gordon’s (2017) reading of Fanon—technical fairness, like rhetorical equality, cannot sustain its legitimacy in the face of de facto inequality, widening disparities in wealth, unequal access to basic medical and other care, and the differential impacts of climate change.
Data Justice and Data Feminism as Critical Responses
Among the most visible critical responses have been political economic and feminist approaches centered on so-called “data justice.” Political economic work, in particular, has done much to “re-politicize” (Gürses, Kundnani, and Van Hoboken 2016) normative debates that might otherwise be reduced to hollow discourses of agency and empathy (especially those represented by “human-centered data science” and “data for social good” initiatives). This work brings political power and economic forces to the fore, demanding deeper engagement with “the ideological basis [and political agendas] of data-driven processes” (Dencik, Hintz, and Cable 2019). Often, this has meant centering complex webs of exploitative relations of production often sidestepped by appeals to fairness and inclusion. Such work makes clear that depoliticized approaches to data ethics hinge on abstract liberal ideals of subjects (possessing “agency”) and freedoms (like “privacy”) in ways that obscure more than they reveal when it comes to the material conditions of lives shaped by the production and application of data science and technology.
Despite the ongoing need to reveal conditions of economic exploitation, such work can—at times—risk entrapping us into a conceptually expedient, yet ultimately untenable, dichotomy between economic and political injustice on the one hand and cultural harms on the other (see Young 1997; Meyers West 2020). For example, we have already seen the emergence of political economic discourses of “data colonialism” that, as Sareeta Amrute (2019) notes, reduce the field of colonial exploitation and violence almost solely to economies of data extraction. As Said Mustafa Ali (2017) argues, this focus on the mining and exploitation of personal data by governmental and corporate actors obscures the deeper, racialized logics of coloniality by means of a “recourse to an information-centric abstraction that allows ‘colonialism’ to be reductively, and economistically, framed in terms of labour” (3).
By contrast, feminist approaches to both studying and working with data confront directly those bodies and identities that prove elusive to data’s reductive dictates and that too often slip out of political economic discussions. Whether feminist data studies, feminist data science, or “data feminism,” these moves have in common explicit engagement with communities and groups caught up in broader discourses and machinations of power reflected in and through “data”—a power that, as Catherine D’Ignazio and Lauren Klein (2020) put it, is not equally distributed (8). Central to this movement—as with Western, often white feminisms broadly—is a focus on care as a potent counterpoint to the patriarchy’s devaluing of, among other things, vulnerability (see, for example, Zegura, DiSalvo, and Meng 2018).
At the same time, however, the appeal to Western feminist ideals of care risks sidestepping more foundational questions of data, computation, and colonialism altogether. As the Laboria Cuboniks collective’s xenofeminist manifesto (2018) argues, our current moment seems to require a feminism “at ease with computation” (33). But I remain uncertain about what it means to both care and be easy with computation when, as Philip Agre (1997) described it, computing itself is “a kind of imperialism; it aims to reinvent virtually every other site of practice in its own image” (n.p.). This is forcefully illustrated by Lily Irani’s (2019) work on the ways “other-directed” projects like social good hackathons, “user-centered design,” and empathetic design thinking offer, at best, “minimum viable futures—the least common denominator people can agree on without resorting to a longer, slower mass politics” (230). In this way, our desire for justice—gender, racial, or otherwise—is often too readily contorted to accommodate hegemonic ideals of research and design that are more than merely incidental to contemporary data science and technology.
Or, to put it another way (and with apologies to Elizabeth Anderson): what’s the point of data feminism? I don’t know. But recognizing, visualizing, and caring for difference is ultimately insufficient if it leaves dominant logics and structures intact—that is, if it ultimately reproduces “domination as [a] fundamental form of relations” (Walcott 2019, 396). Here, we would do well, I think, to remember Ella and her discomfort with “past errors taking possession of the present,” a reminder of the many ways oppressive or carceral logics pervade, often implicitly, even progressive or justice-oriented conceptions of data collection, storage, and use (Sutherland 2019).
From Critical Response to Critical Refusal
Despite these hesitations, I find these critical trajectories are often sharpest when they engage practices of active resistance and refusal—especially in the political, feminist, and abolitionist senses (see, for example, Gangadharan 2020; Cifor et al. 2019; Benjamin 2016). Central to any critically informed ethical sensibility is an imperative to not take popular and professional claims about the social value of data science and technology at face value; rather, we must always resist and refuse the disjunctions and perverse incentives fostered by such claims. That is, we must recognize that the work of refusal is still work—it represents, to use Ahmed (2018) in a different way, a kind of “counter-institutional project…creating paths for others to follow” (n.p.). In this way, refusal—as Ruha Benjamin (2016) notes—is “seeded with a vision of what can and should be, and not only a critique of what is” (970).
But what are we refusing? A few things. First, and most obviously, we refuse enlightened “data ethics” projects—and even their more critical manifestations—wherever they boil down to empty calls for fairness and inclusion. Similarly, in resisting Western feminist calls to embrace pluralism, we refuse the hollow “smile of diversity” (Ahmed 2012, 163), which—for so long—has allowed dominant orders to acknowledge expressions of resistance and difference without being actually moved by them. At the same time, however, our refusals should also look inward. We should refuse critical analyses that reduce the field of colonial violence to only problems of labor and extraction, or that reduce fields of power to problems of care. We should also refuse any easy gesture toward inclusion precisely because inclusion has too often functioned to make “others” more visible and thus vulnerable to the sorts of binaristic and juridical calculations that have dominated so much mainstream work on data ethics to date. In doing the work of refusal, however, we should not—following Audra Simpson (2007)—merely genuflect toward “alternative” methods that are neither radical nor deeply rooted in the values of the communities they claim to support. Our refusals should open up new paths, not retread those paths that ultimately risk exposing those we seek to study, defend, nourish, and draw on for inspiration and strength.
I am far from the first to articulate refusal as a core value in the context of cross-cultural struggles with and against extant technocultural orders. In fact, I could be fairly described as a “refusal” bandwagoner, hopping aboard a vehicle built—and best steered—by others. But if I can contribute anything, it may be to suggest the adoption of refusal as a universal feminist value. By universal, I do not mean universalist in the sense discussed by Anita Chan (2018), as the elevation of parochial or narrow Western practices as the only relevant pathways to the future (3). Rather, I mean universal as able to cut across cultures, generalizable to the struggles for gender justice regardless of the particular details in particular contexts. As the Feminist Data Manifest-No puts it, what allows various “feminisms” to “hang together” is exactly refusal’s negative construction (Cifor et al. 2019)—that is, its commitment to declining the dominant social, political, or economic terms on offer.
Though universality is often dismissed as both untenable and undesirable, we shouldn’t be so quick to dismiss the possibility of universal values in some sense, especially—though perhaps unintuitively—in our efforts to cultivate a cross-cultural feminist data ethics. One primary reason to tentatively endorse the search for universal values is that we don’t, as Serene Khader (2018) argues, want to preclude views and normative projects originating from the colonized as having a kind of universal force—that is, from ultimately having moral purchase on others, especially would-be colonizers (29). In this way, by allowing ourselves to conflate universalism with imperialism or the West, we concede universalism’s practical and rhetorical power, ultimately doing the imperialists’ work for them. Instead, we should ungive such a concession and strive to reimagine what a universal value can—and should—be.
Critical Refusal as Universal Feminist Value
Following Khader, we can look to bell hooks’s definition of feminism as opposition to sexist oppression, attuning us to the central role of “refusal” for a cross-cultural feminist data ethics. In terms of defining refusal as a universal feminist value, we might look to the more obscure definition of refusal as a declining to accept or submit—to a rule, a norm, a demand, an expectation. Refusal means rejecting violent inheritances (Ahmed 2017), hammering away at “those walls, those physical or social barriers that stop us from residing somewhere, from being somewhere” (Ahmed 2016, 32). Refusal also means rejecting the current terms of inclusion. Returning to Ella in Beloved, refusal also means casting out past wrongs whenever their claims extend too far into our imagination.
Refusal can be uncomfortable, no doubt. Indeed, critically reflecting on recent scholarly moves I otherwise embrace (and of which I am hardly innocent) has—for me—been an unsettling exercise, one unavoidably tangled up with “complex power relations and mixed emotions” (Leurs 2017, 131). But I interpret my discomfort as integral to the “epistemic disobedience” (Arora 2019) necessary to pierce the “white noise” (Syed 2012) that threatens to “[drown] out those sensations, feelings, knowledges, and information that…disturb the smooth operations of white power in colonial modernity” (Rault 2017). In this way, refusal’s discomforts should serve as a reminder that we can and must “[surrender] the power to inflict pain and to relate to each other through violence[s]…built on neoliberal self-regulation, productivity, utility, quality of life, and presumed able-bodiedness” (Kim 2015, 315). In being discomfited, we can better “attune” (Amrute 2019) ourselves to the spaces opened up by refusal as opportunities for reparative transformation (Cowan 2014) and “stitching” (cárdenas 2016) together new worlds.
In that spirit, I want to reiterate that the intent here has been simply to register an unease—or, put another way, to refuse an easy comfort with even welcome developments in feminist understandings of data science and technology (developments of which I have, in some small way, also been a part). And I suppose this is part of what Ella had to teach us all along—that our practical orientations have, however disquieting, broader moral and conceptual implications. That many things—from bodies to identities to scholarly movements—can occupy uncomfortable positions, but that discomfort can be clarifying—as Morrison had it—“in a world where even when you are a solution you are a problem” (302).
Anna Lauren Hoffmann is an assistant professor at the Information School of the University of Washington. Her work has appeared in New Media & Society; Information, Communication & Society; The Library Quarterly; and The Los Angeles Review of Books.