Viewing technology in the longue durée foregrounds how slavery undergirds conditions of contemporary extraction, surveillance, and the insistence on positing knowledge as outside of those arrangements.1 The rhetorical move transposing race’s technical utility into easily computed quantities grafts nicely onto the kinds of Malthusian investments we see, for example, with public sector adoption of automated decision systems. Whether we’re talking about predictive fraud detection in public health insurance, policing, humanitarian cash transfers based on the likelihood of school attendance, or high school placement within competitive public school districts, the axiomatic reasoning typically goes something like this: there is an unequal relationship between some given resource and the population of people in need of it, so algorithmic decision-making can aid the determination of who will get and who will go without. In this process, resource scarcity is accepted as immutable. Racial disproportionality is an unintended (but fixable) consequence, akin to “source-reference divergence” by natural language models—more commonly articulated in the ableist nomenclature of computer science as “hallucinations.” In the statement “facial recognition technologies don’t accurately capture dark skin,” the diagnosis is an out-of-place output—not the model’s logics, nor the society which authorizes its use. The cure is to adjust the model’s weights and biases until you get the desired, more proportional output, or to throw a wrench in the machine in order to break it. In other words, the problem is litigated within or against the confines of the machine, even though it is just a node within a historical network of world-building projects.
What, historically, sets up the problems facial recognition claims to solve? The transition from an interpersonal, relational mode of recognition—where empiric knowledge gathered either via proximity/intimacy or triangulated through surname and birthplace—to mechanistic pattern-matching between image-captured face and the database resonates with cosmopolitan ambitions. New York Times ethicist Kwame Appiah’s 2007 book Cosmopolitanism lauds the displacement of “feudal” kinship relationships, once required in order to find a place to sleep in a strange city, for the “modern” cities of strangers, where the only thing you need to get a hotel room is the money to pay for the night. There is no need to negotiate with locals, to establish where you came from, who your family is, or why you are there, because the cosmopolitan is liberated from both place and genealogical misfortune.2 It’s within this context that both biometric identification3 and consumer ancestry DNA tests4 are proffered as a salve to urban and natal alienation alike. To simply reject them or enumerate the harms they catalyze does not, in and of itself, resolve the strange conditions that make them seductive.
What digital technologies make possible within this postcolonial landscape is an “imposition of separation”5 at a previously hard-to-achieve scale. For example, US-based family policing agencies, like Pennsylvania’s Allegheny County Office of Children, Youth, and Family, have implemented a predictive risk model where all birth certificates issued within their jurisdiction listing only one parent are cataloged and labeled as a risk factor for child abuse. Prior to the introduction of this tool, family policing was already intrusive and punitive, but the mechanism for targeting families was limited to the reporting process. Adoption of predictive risk modeling expands surveillance to all families who have children under eighteen in the county,6 even if no report is ever made. To cite another example, the hotel was an early site for the surveillance and discipline of perceived sexual deviancy.7 While all who enter are subject to this gaze, the hotel’s particular algorithmic ecology8 enforces higher rates of arrest and corporate discipline among entire categories of people whose behavior has a perceived correlation to sex work, homosexuality, or criminal intent. Each kind of classification to which a group is subject can result in sanctions like suspension of credit card processing, fraud alerts, and escalating algorithmic pricing, among other consequences. The digital is not an origin story for these twin impulses of containment and discipline, but it optimizes and economizes them instantaneously—while superseding and making use of the friction of state-imposed borders. It forcibly maps “people like this,” into place, weaving the sanctity of individual guilt with dividual probability and risk—indexed by and from zip codes (or by other units that lie at the intersection of valuation and place).9 And the digital appreciates all the work neoliberalism has done to cement the “natural order of things.”10
A related irony is seen in heavy-handed references to predictive automation, articulated through analogies to the body that simultaneously deny the fundamental embodiment of the knowledge acquisition process.11 According to this logic, the so-called hallucinations generated by large language models (LLMs) are grammatically correct but substantively incoherent, and thus untrue relative to the real-world phenomena they were programmed to make linguistic claims about—a kind of superficial satisfaction of the Turing test that quickly collapses under scrutiny. In other words, engineers describe hallucinations as models that come across as fluent and natural despite being nonsensical and unfaithful to the source content (or “corpus”). They explain such lack of fidelity to the training data as the flip side of deep learning technologies’ improved capacity to summarize abstracts, generate dialogue, and translate data to text. But is the problem with generated outputs of LLMs, or with any algorithmic decision-making system writ broadly, the outputs themselves? Or is it, rather, with the implicit and explicit ontological claims made by model designers that in turn produce human complacency and overreliance on (or diffidence to) “autonomous” systems? Several high-profile plane crashes have been attributed to these dynamics: namely, when pilots lose the capacity to right the plane as a result of deskilling secondary to cockpit automation. Perhaps the underlying issue is not just the displacement of “human expertise” but specifically the way the field takes the cosmopolitan aspiration to discard kinship and birthplace even further, insisting the human body itself is a vestige of a backward past and a hindrance to be circumvented.
The forcible separation of mind and body into discrete and competing parts reflects the endurance of Cartesian duality that transatlantic slavery deeply entrenched. The Black African infidel powers economic change under the discipline of civilized/enlightened white reason and intellect yet is simultaneously rendered out of place—stubbornly overfitted to a place that no longer exists. Ugly and lantern laws converge in the governance of Blackness in public space or within the corporate-owned sociotechnical systems with which it collaborates. It’s here that we can see “hallucinations” as not just political incorrectness but a synecdoche for the logics of machine learning, according to which abled is normal—if not exceptional (supreme)—and psychosis is a type of performance degradation to be eliminated. Meanwhile, aren’t overconfident articulations of stolen work as its own—even as it completely misparses, lies, and fabricates histories to justify its poorly threaded together narrative—a defining feature of whiteness? What does it reveal that when confronted with their own reflection, computer scientists think it’s the model that’s tripping?
This is not the first historical instance where race arrives as a disability classification. I am reminded of dysaesthesia aethiopica, literally an “Ethiopian bad feeling” whereby enslaved Africans are deemed lazy, refusing to work, and less likely to feel pain. The term was coined by white supremacist Samuel Cartwright, who is better known for the similarly conceived diagnosis of “drapetomania,” a sickness compelling enslaved Black people to run away. In modernity, everyone has a place, and those who refuse to abide by their imposed assignations are out of place and as a result disabled. In the late nineteenth century, clinician John Langdon Down called white children born with trisomy 21 “Mongolian idiots,” theorizing that the hallmark features represented an infiltration of Asian DNA into the descendants of European families. In 1965, the World Health Organization agreed to abandon use of the term upon the request of the Mongolian People’s Republic, but the superficiality of this reckoning with eugenics is demonstrated by the decision to rename the syndrome after the very man who had originally termed it “Mongolian idiocy.” Similarly, while the 2021 Diagnostic and Statistical Manual of Mental Disorders forsook “Asperger’s” as a clinical term, the Nazist notions of separating the wheat from the chaff—or those that belonged in the volk from those who would contaminate its purity—which animated Hans Asperger’s conception of neurodevelopmental differences, is preserved in the alternate language of “high functioning / low-support needs” or “low functioning / high-support needs.”12 Is it “Oochie Wally” or “One Mic?”: Either racism is contaminating and disabling the pure disembodied model, meaning harms are to be enumerated, then extinguished; or race is constitutive of how such models are developed and deployed, therefore necessitating their wholesale dismantlement and the offering of a counterproposal. As the contributors to this issue of Logic(s) underscore, it is the latter; and beyond this, the vulnerability to premature death that digital infrastructures enable is due to the ways those of us rendered out of place are disciplined into the flows of projected outcomes—or, if noncompliant, wholly removed from the dataset (and society).
In conversation with Sucheta Ghoshal—who is transitioning in her role at Logic(s) to academic director—Oliver Rollins draws our attention to neuroscience’s investment in race science and, particularly, the use of functional magnetic resonance imaging (fMRI) to visibilize a biological risk of psychopathy. The state classification of certain people as criminals is so tightly indexed to race that its attempts to predict the criminal risk of children as young as five become another way to refract antiblackness through (and as) scientific abstraction—even as race or socioeconomic inequality are explicitly avoided. Echoing Chris Dancy’s contribution to Logic(s) issue 21, “Breaking the Cycle: Against the Militarization of Neuroscience Research,” Rollins calls our attention to military funding of this line of academic research and the inextricable relationship between US militarism abroad, policing domestically, and the scale of capital investment for research questions with the potential to contain populations deemed unruly. An important site where these research agendas converge are programs that try to blend methods between countering violent extremism and anti-gang-violence initiatives.13
As a matter of genre convention, unhoused people arrive in tech journalism and critical tech literature typically as the objects upon which tech is enacted. They are anecdotal seasoning for dry policy discussions, rather than agents actively negotiating and contesting the surveillance to which they are subject. Following the example of Virginia Eubanks, author of Automating Inequality: How High Tech Tools Profile, Police, and Punish the Poor, we feature the direct testimonies of the San Diego Unhoused Collective, as well as a dialogue between Ralowe T. Ampu and Eric A. Stanley, both of which examine the ways tech “innovation” is linked to the dispossession and policing of people on the streets. A recurrent theme is the way neoliberal policies, beginning in the late ’70s, have naturalized the conditions of scarcity that tech bros claim to rationalize and solve in the present. Policies aimed at gutting the US social safety net and escalating the punitive measures embedded in what remains—which, as many historians have emphasized, occurred in tandem with Black women’s inclusion in Aid to Families with Dependent Children (AFDC)—inform, for example, the shifting labor conditions for platform gig drivers. Like their antecedents in Bill Clinton’s welfare-to-work program, current compensation for gig driving has plummeted below minimum wage, even as drivers provide labor indistinguishable from that of “employees.” Historically, organized labor has not substantially engaged or allied with welfare recipients or disabled people in sheltered workshops (each receiving less than minimum wage).14 However, a new generation of labor organizers have the opportunity to change that—for instance, Rideshare Drivers United, who spoke at length with editorial fellow Eliza McCullough about their strategies for organizing against California Proposition 22, and to overcome the abysmal wages and dangerous workplace conditions of which algorithmic management has become a central pillar.
There is no future perfect or, as Fanon said, “romantic past to return to”; but Indigenous futures are also here in the present, navigating the knotty contradictions wrought by corporate infrastructures and occupation. Hanna Barakat demonstrates how Palestinians appropriate WhatsApp and Facebook to communally navigate and survive the Israeli occupying forces’ routine obstruction of traffic infrastructure in Gaza and the West Bank, linking these kinds of mesh networks to Palestinian farming practices. In the settler reading, the patterns of avocado tree cultivation to which Barakat draws our attention are considered chaotic and inefficient. Meanwhile, the kinds of order that modern agricultural management has normalized typically result in generational death and soil exhaustion. Barakat’s intervention maps Palestinian freedom dreaming to the present—and, as poet Rasha Abdulhadi frequently reminds us, it is in the present, while Palestinians are still alive, that we should celebrate them. Thinking with Rasha, let us notice—a year into the genocide’s escalation—the seduction to applaud any liberal darling who dares to say “apartheid.” Looking at the current media landscape, I see predictable failures to endorse armed resistance and, more importantly, that the resistance is alive and present.
In “To Resist Annihilation: Paternity, Inheritance, and Love against Surveillance Medicine,” Zoé Samudzi links the biomedical control through which Israel seeks to asphyxiate Palestinians to the kinds of surveillance to which Black families like her own become subject in the wake of a cancer diagnosis. What surveillance watches is uncertainty and risk—of cancer, an intifada, or any unruly threat. Data is not “like oil”; data is the production of an archive in the same tradition of anthropologists working in tandem with colonial administrators to manage the local population. And if data collection is most often an intrusion meant to classify and sort us in space, as Eubanks eloquently demonstrates in Automating Inequality, should we displace the political demand for transparency with one for the right to opacity? What kind of space or possibility opens up when we’re misrecognized in the ledger?
Western colonialism doesn’t have a monopoly on the management of risk and uncertainty. In our conversation, Romi Morrison emphasizes the significance of Black computational thought, including the encapsulation of alternate number systems and modes of nonlinear algebra in African fractals, from quilts in the Underground Railroad to hair-braiding patterns. One of the featured artists in this issue, Stina Baudin, literally weaves into denim a computation otherwise, in selected pieces from Data Studies. And in her series It Came from Outer Space, white text on a Black canvas features Haiti in recursion. Kameelah Janan Rasheed explores “the materiality and legibility of language,” wrestling “with how incoherence, incompleteness, and opacity function as dirty data that jam systems of smooth and ‘efficient’ communication.” In an era of live-streamed genocide, we are reminded again and again that “seeing isn’t believing”; rather than serve as an indictment, visual evidence usually plays into conquest. In this issue’s special visual arts section, Metabolic Flesh, Morrison’s curation recognizes these constraints but highlights a series of creative practices, each of which thinks through computation and the distributed network of politics to ask: What would such engineered forms look like if enacted in ways not encumbered by commitment to a “proof”?
In conversation with our new managing editor Rezina Habtemariam, Jada Renée Allen emphasizes the importance of Black trans conjuring and the specificity of colorism within antiblack world-building. By now it should be clear that the problem with facial recognition is not just the limitations of the camera; it is the way dark-skinned Black people are made into the very material upon which analysis is mobilized.15 We hope to solicit more work pursuing the sets of questions their dialogue makes possible. I would also like to extend a personal invitation to our readers to submit letters to the editor: we want to be more in conversation with ya’ll! In September, more than one hundred people attended the launch party for Logic(s) issue 21, “Medicine and the Body,” at the Francis Kite Club in New York City. We are so grateful for this show of support and hope to keep the conversation going with future programming and letters to the editor.
Much appreciated,
J. Khadijah Abdurahman
Logic(s) Editor in Chief