AI as Prosthesis: The Politics of Governance and the Redistribution of Competence

This essay explores whether generative AI would be better understood as a cognitive and technical prosthesis. Rather than rehearse familiar debates about whether AI replaces human labor, I want to examine how it redistributes agency without redistributing responsibility, how it reorganizes the temporal and technical expectations that institutions place on writing, reading, judgment, and making, and how it opens terrain that was previously fenced by disciplinary credentialing and cognitive normativity alike. The last of these — the emergence of what might be called prosthetic competence, both technical and cognitive — has received too little attention, and it is here that the politics of the metaphor becomes most concrete.

The usual framings are not wrong so much as flat. “AI is just a tool” borrows from an older moral drama about machines and human exceptionalism. “AI is coming for your job” borrows from another. Both flatten what is happening at the level where historical change actually registers: in routines, infrastructures, and the quiet redefinition of what counts as adequate performance.

“AI as prosthesis” gets closer to the grain. A prosthesis is not a hammer. You do not merely wield it. You attach it. You learn its resistances and affordances. You adjust your posture around it, and in time your world reorganizes so that the device is no longer an external object but part of a coupled system: body, habit, environment, maintenance regime. The metaphor matters because it shifts attention from the question of replacement to that of reconfiguration — how the coupling alters what a person can do, what institutions expect of them, and where governance sits once the interface becomes habitual.

The philosophical architecture for this claim rests on work that Andy Clark and David Chalmers began in 1998. Their “extended mind” thesis argued that certain external supports function as integral components of cognition when they are reliable, readily available, and routinely used (Clark and Chalmers 1998). Once the notebook or laptop is always there, always consulted, and always trusted, it stops being “outside” in any analytically meaningful sense. It is part of the thinking apparatus. Clark developed the argument further in subsequent work, insisting that the human mind is fundamentally a “leaky organ” whose boundaries extend into the material world through repeated, trust-calibrated coupling with external systems (Clark 2003, 2008). The insight was never that humans become machines. It was that cognition has always been scaffolded by material systems that stabilize memory, calculation, and attention over time. That is the sober corrective we forget whenever we talk about “intelligence” as if it lived entirely inside the skull.

Yet disability studies adds something the extended mind frame leaves underspecified: normativity. Aimi Hamraie’s work on universal design is essential here. Hamraie demonstrated that access is never a purely technical question but an argument about which frictions count, which bodies are treated as default, and which forms of dependence are rendered acceptable or shameful (Hamraie 2017). Prosthetic devices are not simply enabling. They are shaped by social demands for productivity, legibility, and compliance, and they arrive tethered to institutional scripts about what a body or mind should do. “Assistance” can be emancipatory. It can also be coercive. The line between the two is historically mobile, and it moves in the direction that institutions find convenient.

David Mitchell and Sharon Snyder’s foundational work on what they called “narrative prosthesis” adds a further dimension. Mitchell and Snyder showed that disability functions in literary and cultural forms as a device that props up meaning: a scaffolding for plot and moral argument (Mitchell and Snyder 2000). The concept names how physical or cognitive difference is recruited to do narrative work for an ableist culture, and the warning transfers with only minor modification. “AI as prosthesis” can become a story we tell to naturalize dependence, to aestheticize the interface, to treat structural capture as personal convenience. It can render the new arrangement inevitable.

If we keep Hamraie, Mitchell, and Snyder in view alongside Clark and Chalmers, “AI as prosthesis” ceases to be a cute metaphor and becomes a diagnosis. It forces us to ask not only what the device enables, but what it normalizes.

Start with what the prosthesis does well. The system drafts. It summarizes. It translates. It converts rough notes into a coherent outline, suggests transitions, and can serve as a tireless interlocutor for brainstorming — a way to make writing less lonely, less stuck. If you do intellectual labor for a living, the appeal is not mysterious. In fact, the appeal is precisely what makes the prosthesis frame so apt: the more useful the device, the more quickly it becomes habit, and the more habit becomes expectation.

That is where politics enters, because expectations are never evenly distributed.

In workplaces, the prosthesis will not simply make workers “more productive.” It will recalibrate what counts as adequate performance. Turnaround times shorten. Output volumes rise. The baseline shifts. The new norm is not “work plus AI” but “work as if AI were already attached.” This is a familiar historical sequence: technologies that promise to reduce burden often intensify it by creating new standards of responsiveness, availability, and polish. The device becomes ordinary. Then it becomes compulsory — first informally, then formally, and finally by the quiet violence of comparison against those who have not yet coupled.

Education is an even sharper site of contradiction. Once AI becomes prosthetic for drafting and comprehension, institutions face a choice they would rather avoid: either redesign assessment around the reality of coupled cognition, or pretend the coupling is illegitimate while everyone knows it is happening. The temptation will be to police boundary violations, because boundary policing is what institutions do when they cannot or will not redesign their own metrics. Yet the prosthesis metaphor makes the policing look strange. If cognition is now, in practice, distributed across human and machine systems, why is the student treated as an isolated unit for purposes of evaluation? And if the student is still treated as an isolated unit, why is the institution surprised when the student routes their work through the most available cognitive infrastructure? This is not a moral question first. It is a design question that institutions keep trying to resolve through punishment.

The coupling extends beyond writing and thinking into terrain that has been less examined: the capacity to build. Before generative AI, the ability to create a functioning web application, construct a relational database, generate interactive data visualizations, or write a script that processed thousands of archival records required either years of technical training or sufficient institutional resources to hire someone who had that training. The boundary between those who could make computational objects and those who could not was sharp, credentialed, and largely coterminous with disciplinary affiliation. Historians, anthropologists, literary scholars, and artists occupied one side. Computer scientists and software engineers occupied the other. Digital humanities programs tried to bridge the gap, but the bridge was narrow, and the toll was high — measured in semesters of coursework that competed with disciplinary demands rather than complementing them.

Generative AI is dissolving that boundary. Not by making everyone a programmer, but by converting programming from a skill requiring years of syntactic fluency into a collaborative practice requiring domain knowledge, clear specification, and iterative judgment. The historian who can articulate what a database should organize, what an interface should display, and what an archival query should retrieve can now produce a working prototype by describing those requirements to an AI system that translates specifications into code. The prosthesis does not replace the need to think carefully about structure, logic, and purpose. It replaces the need to memorize a programming language’s grammar in order to express that thinking as executable software.

The implications are significant. A scholar studying maritime trade routes can build a geospatial application that renders shipping data as interactive corridor maps — not by learning JavaScript and Leaflet from scratch, but by specifying what the visualization should show, testing the output, identifying errors, and iterating. A researcher working with thousands of newspaper articles can construct a script that extracts date-stamped keyword frequencies across multiple publications, producing the kind of longitudinal dataset that would otherwise require either a funded research assistant or months of manual tabulation. A teacher who wants students to interact with primary sources can build a web-based interface tailored to the specific pedagogical problem — sorting, comparing, annotating — without submitting a grant application to a digital humanities center and waiting eighteen months for a developer’s time.

What is happening here is not the democratization of coding in the triumphalist sense preferred by Silicon Valley. It is something more specific and more interesting: the redistribution of technical competence along lines that privilege domain expertise over syntactic fluency. The person who knows what the archive should organize, what the dataset should reveal, what the interface should render visible, gains the capacity to build the thing that does it. In effect, the prosthesis recodes the boundary between “having an idea for a digital project” and “being able to make one” from a credentialing barrier into an iterative conversation.

That conversion has particular force in fields where computational methods have long been recognized as valuable but remained inaccessible to most practitioners. In historical research, the gap between what scholars know they could learn from large-scale textual analysis, network mapping, or spatial modeling and what they can actually build has been a persistent structural constraint. Generative AI does not close that gap entirely — the prosthesis has its own limitations, its own patterns of error, its own tendencies to produce plausible but incorrect output. Yet it narrows the gap enough to shift who can do computationally intensive work from a credentialed minority to a much wider population of domain specialists willing to learn the collaborative discipline of specifying, testing, and repairing.

The skeptic’s objection is fair and needs to be absorbed rather than dismissed: code produced through prosthetic coupling may be functional without being robust, legible without being secure, and sufficient for a prototype without being adequate for production deployment. This is accurate as far as it goes. Yet the objection draws the wrong boundary around the problem. For the historian building an interactive archive, the literary scholar constructing a textual analysis pipeline, or the teacher designing a classroom tool, the relevant standard is not whether the code meets enterprise software engineering criteria but whether it produces a reliable, functional object that would not otherwise exist. The alternative to prosthetic code is not better code. In most cases, there is no code at all. The prosthesis creates capacity where there was none, and the question of quality, as real as it is, remains secondary to the question of possibility.

The redistribution of competence extends beyond disciplinary boundaries. It reaches into the politics of cognition itself.

Institutions have always defined “competence” through specific output formats: linear prose, timed examinations, structured oral presentations, and sequential task completion. What presents as a neutral standard of intellectual performance is, in practice, a sorting mechanism calibrated to neurotypical processing. Hamraie’s analysis of how design regimes produce “the normate” — a subject position whose unmarked status renders its own contingency invisible — applies directly to cognitive assessment (Hamraie 2017, 19–22). The person with dyslexia is not less capable of historical analysis; they are penalized by an evaluation regime that treats fluent written production as a proxy for analytical thinking. The person with ADHD is not less capable of sustained intellectual work; they are penalized by institutional temporalities that treat linear task sequencing as the only legitimate workflow. The autistic researcher whose pattern recognition exceeds their neurotypical peers is nonetheless disadvantaged by assessment structures that reward a narrow repertoire of communicative conventions. In each case, the institution measures deviation from a cognitive default and records that deviation asa deficit. The deficit is institutional. The label lands on the person.

Margaret Price’s work on “mental disability” in academic culture is clarifying here. Price argued that the norms governing academic discourse — the seminar, the conference presentation, the written examination — are not neutral containers for intellectual exchange but embodied practices that systematically privilege certain kinds of minds while rendering others incoherent or invisible (Price 2011). The classroom is not merely a place where content is assessed. It is a site where cognitive normativity is performed and enforced. What counts as “participation,” “clarity,” and “rigor” are not objective descriptors but institutional conventions that carry the full weight of ableist assumptions.

AI as a cognitive prosthesis intervenes at exactly this joint. It converts between cognitive modes: translating spatial or associative thinking into linear prose, handling surface-level mechanics so that a writer with dysgraphia can concentrate on argumentation, and providing structural scaffolding for someone whose executive function processes organization differently. The system can serve as an intermediary between how a person actually thinks and the institutional format that has historically gatekept the expression of that thinking. In effect, the prosthesis separates the idea from the narrow channel through which institutions have demanded it travel.

The implications for cognitive equality are real but require careful specification. For neurodivergent users, the prosthesis does not fix a deficit. It routes around an institutional barrier — one that was never a measure of intellectual capacity but rather a measure of conformity to a particular cognitive style. The student with dyslexia who uses AI to handle sentence-level mechanics and produce polished academic prose is not cheating. They are accessing the same expressive range that neurotypical students reach through a processing pathway the institution happened to privilege. The researcher with ADHD who uses AI to organize scattered but brilliant associative insights into a structured argument is not being “helped” in the condescending sense. They are using a prosthesis that compensates for an institutional design failure: the failure to recognize that analytical capacity and linear organizational fluency are separate competencies that happen to have been bundled by convention.

Yet the critical edge matters here, and there are at least two tensions worth holding in view. The first is the question of whether prosthetic access produces genuine cognitive equality or better masking. If the institution never revises its norms — if it continues to treat linear prose, timed output, and sequential processing as the unmarked standard — then AI does not accommodate neurodivergent cognition so much as make neurodivergent people more legible to systems that never examined their own assumptions. The person still has to pass through the institution’s preferred format. The prosthesis simply makes the translation invisible. That is access of a kind, but it leaves the normative architecture intact and places the burden of adaptation entirely on the user rather than on the institution that designed the barrier. What presents as accommodation is, in practice, assimilation by other means.

The second tension is that the dependency problem strikes harder here than anywhere else in the prosthetic landscape. If the device is what makes the playing field level, then a pricing change, a model update, or an expired institutional license is not an inconvenience. It is a revocation of access. For the neurotypical user, losing the prosthesis means slower writing, more effortful organization, reduced output — a degradation of convenience. For the neurodivergent user whose equitable participation is routed entirely through the prosthetic interface, losing the prosthesis can mean losing the capacity to participate at all. Dependence, in this case, is not a market relation. It is a civil rights question dressed as a subscription fee.

At some point, the prosthesis frame encounters its own limits. A prosthesis assumes separability: you attach the device, you can detach it, the boundary between user and system remains analytically legible even when practically blurred. Yet the cases accumulating in this essay press against that assumption. The historian whose research methodology now presupposes iterative human-machine conversation, the neurodivergent student whose institutional participation is routed entirely through the interface, the scholar whose workflow has reorganized so thoroughly around AI coupling that removal would not restore a prior state but produce a different and diminished subject — these are not cases of a person using a detachable tool. The coupling has become constitutive.

The concept that names this threshold predates current technology. Manfred Clynes and Nathan Kline coined “cyborg” in 1960 to describe an organism-machine integration so thoroughgoing that the boundary itself dissolves — not as metaphor but as a functional description of a self-regulating human-machine system in which the mechanical components become as integral as the organic ones (Clynes and Kline 1960). Their original context was space travel: an astronaut whose physiological regulation was partially offloaded to implanted devices would be free to concentrate on exploration rather than survival. The insight was that coupling, once sufficiently deep and habitual, ceases to be supplementary. It becomes structural.

Donna Haraway radicalized the concept by insisting that the cyborg was already the condition of late-twentieth-century subjectivity, not a future to be anticipated but a present to be reckoned with politically (Haraway 1985). Haraway’s cyborg refused the clean boundaries between organism and machine, between nature and culture, between the self-sufficient liberal subject and the technologically enmeshed one. The “Cyborg Manifesto” argued that these boundaries were never stable to begin with and that their maintenance served ideological functions: naturalizing certain forms of autonomy while rendering others deviant. For Haraway, the political task was not to resist cyborgization but to claim it — to insist on the right to narrate one’s own coupling rather than having it narrated by the institutions and markets that profit from it.

Cyborgization — treated as a process rather than a fixed state — names what happens when the prosthesis passes the habit threshold. Clark’s later work on “natural-born cyborgs” converges here from a different direction, arguing that human beings are by nature prone to deep integration with cognitive technologies and that the boundaries we draw between “self” and “tool” are matters of convention rather than ontology (Clark 2003). The question is not whether cyborgization occurs. It always has, from writing systems to eyeglasses to smartphones. The question is what changes when the coupling involves a system that generates, reasons, and adapts, and when the infrastructure that sustains the coupling is controlled by commercial platforms operating at a planetary scale.

The transition from a prosthetic to a cyborg relation changes the governance stakes. You govern a prosthesis through regulation of the device: its safety, its availability, and its terms of use. You govern a cyborg relation by regulating the person-system composite, thereby shaping the conditions under which cognition itself is constituted. No existing institutional or legal framework is designed to do that. Employment law assumes that an employee's capacities are their own. Educational assessment assumes a student whose cognition is bounded by their skull. Intellectual property law assumes an author whose creative contribution is distinguishable from the contributions of their tools. Each of these assumptions becomes unstable once the coupling is deep enough that separating human contribution from machine contribution is not merely difficult but conceptually incoherent.

For neurodivergent users, the stakes of cyborgization are sharpest. If the AI system is not an optional enhancement but the infrastructure through which equitable cognitive participation becomes possible, then the person-system composite is not a convenience but a civil rights configuration. Regulating, restricting, or withdrawing the technological component is not analogous to revoking a workplace perk. It is analogous to removing a wheelchair ramp — except that the ramp is owned by a corporation, priced as a subscription, and subject to unilateral modification. The costs of that arrangement are absorbed entirely by the user.

The deeper problem is governance, and here the prosthesis metaphor becomes uncomfortably literal. Prostheses require maintenance, calibration, and an ecosystem that decides what the device can do, how it behaves when it is uncertain, and how it fails. With AI, that ecosystem is not your body’s physiology or a local clinic. It is a platform stack: training data, compute infrastructure, moderation policy, pricing tiers, terms of service, enterprise procurement, and update schedules that can alter the device without warning. If the prosthesis becomes habitual — and certainly if the relation becomes cyborg — then governance becomes intimate. The interface sits inside your cognitive routine, but it is administered elsewhere.

This is where the political economy stops being an optional background and becomes constitutive. Shoshana Zuboff’s account of surveillance capitalism, whatever one thinks of its emphases, remains clarifying on this specific point: platforms extract value by positioning themselves as intermediaries and converting mediated activity into predictive assets (Zuboff 2019). The more the prosthesis becomes the default route for everyday thinking-work and making-work, the more that work is routed through systems whose incentives are not aligned with epistemic care. The intermediary recodes the relationship: what appears to be assistance functions as extraction.

Nick Couldry and Ulises Mejias sharpen the stakes by framing large-scale data extraction as a colonial relation — not as an analogy but as a structural claim about appropriation. Drawing on a long tradition of dependency theory and postcolonial critique, Couldry and Mejias argued that modern capitalism extends itself by annexing human experience and interaction as raw material for value (Couldry and Mejias 2019). Within that frame, “AI as prosthesis” names a double movement: the device helps you, and the device appropriates the traces of your helping yourself. When the historian builds an application through iterative conversation with an AI system, each exchange refines the platform’s training data while producing an artifact the historian needs. Dependence is not merely psychological. It is infrastructural.

Kate Crawford’s insistence that AI is not an abstraction but an extractive industry provides the material counterweight (Crawford 2021). Crawford demonstrated that the planetary costs of AI systems are distributed along familiar lines of geopolitical asymmetry: minerals are extracted from mines in the Global South, data annotation labor is performed by low-wage workers, and the environmental costs of computation are externalized onto communities with the least political leverage to resist them. The prosthesis has a supply chain. So does the cyborg.

For prosthetic technical competence, the governance problem carries an additional dimension. The non-specialist who builds through AI coupling depends not only on the platform’s continued availability and pricing but also on its continued capability. A model update that changes how the system handles code generation, a policy revision that restricts certain outputs, a corporate decision to deprecate a feature — any of these can degrade or destroy a workflow that the user has built their practice around. The prosthesis is not yours. It is leased, and the lease terms are written by the lessor.

Prosthetic systems distribute agency. They rarely distribute liability. When AI produces an error — a fabricated citation, a flattened argument, an invented statistic, a broken function — the reputational and professional risk sticks to the human user. That is not an accident. It is a governance choice. Platforms insist the system is “assistive.” Organizations insist the human remains accountable. The gap between the two becomes the user’s burden: auditing, verifying, cross-checking, and disclaiming. Prosthetic cognition reduces some kinds of labor while creating new ones, especially supervision and repair. The work does not disappear. It is translated into a different register and shouldered by a different party.

The same asymmetry applies to prosthetic code. The scholar who builds an application through AI coupling bears full responsibility for its accuracy, its functionality, and its failures — while the platform that generated the code bears none. If the script misparses a dataset, the error is the researcher's responsibility. If the web application breaks after a library update, the repair falls to the builder. The prosthesis does not simply extend capacity. It disciplines the user into a managerial posture toward its own supports, converting the act of making into a continuous practice of inspection and maintenance. What presents as creative empowerment doubles as unpaid quality assurance.

Mitchell and Snyder’s concept of narrative prosthesis is worth returning to here. If “AI as prosthesis” becomes the dominant story — the comfortable metaphor that frames dependence as enhancement and structural capture as personal convenience — then the narrative itself does political work by foreclosing harder questions about who profits from the coupling and who bears its costs (Mitchell and Snyder 2000). Haraway would add that the political task is not to refuse the coupling but to insist on knowing its terms: to be a cyborg with a politics rather than a consumer with a subscription (Haraway 1985).

If the metaphor is to do analytical work rather than serve as atmosphere, it needs to be operationalized. Building on the frameworks outlined above — Clark and Chalmers on cognitive extension, Hamraie on normative access, Haraway on cyborg politics, Zuboff on platform capture, Couldry and Mejias on data colonialism, and Crawford on material extraction — I would propose five tests.

The first is habit. The decisive threshold is not whether you use AI but whether your workflow reorganizes around it — whether going without feels like losing capacity rather than choosing a different method. Clark’s “natural-born cyborg” argument suggests this threshold will arrive sooner and more quietly than most users expect (Clark 2003). Prostheses change history when they become ordinary. They become cyborg relations when their removal is no longer experienced as an inconvenience but as an amputation.

The second is governance. Who sets the parameters of the prosthesis, and how legible are those parameters to the user? What genres does the system compress into templates? What uncertainty signals does it provide or withhold? What does it refuse, and on whose authority? These are epistemic politics encoded as interface design — and opacity, as Zuboff’s framework reminds us, is not a bug but a revenue model (Zuboff 2019).

The third is norm formation. Once prosthetic cognition and prosthetic technical competence become common, which new expectations of speed, output, and capability become naturalized — and which people are punished for failing to meet them? The device that “helps” the historian build a digital archive today becomes the baseline expectation that renders the non-digital historian inadequate tomorrow. The prosthesis that gives the neurodivergent student equitable access today becomes the invisible infrastructure that institutions use to avoid redesigning their own assessment norms. Hamraie’s analysis of how “the normate” structures design regimes is directly applicable (Hamraie 2017). The risk is that prosthetic access substitutes for institutional transformation — praised as inclusion, when in practice, it is merely delegation.

The fourth is dependency and repair. What happens when access is revoked, when pricing changes, when institutional licenses expire, when models shift, when privacy regimes tighten or loosen? Dependence is not a feeling. It is a relation to an infrastructure, and it is most visible when the infrastructure withdraws. For users whose equitable cognitive participation depends on the prosthesis, withdrawal is not inconvenient. It is exclusion. Couldry and Mejias’s framework clarifies the structural dimension: what appears to be individual dependency is, at scale, an extractive relation between the platform and the population (Couldry and Mejias 2019). The subscription renews. The terms do not.

The fifth is responsibility. How is liability assigned when an agency is distributed? In practice, the default resolution is to make the human user responsible for the system’s failures — convenient for platforms, punitive for workers, scholars, and students. If AI is prosthetic, then the politics of audit, disclosure, and liability are central, not an afterthought. If the relation has become cyborg, then the question is not merely who is responsible but who gets to narrate the coupling and on whose terms, which is, as Haraway insisted, a question about power before it is a question about policy (Haraway 1985).

None of this requires us to treat AI as literally analogous to a prosthetic limb, nor to treat the language of cyborgization as a science-fiction prediction. Metaphors need not be anatomically faithful to be analytically sharp, and process concepts need not name a completed transformation to identify a trajectory. What “AI as prosthesis” does, at its best, is keep a contradiction in view: the system can genuinely extend capacity — including the capacity to build computational objects, and including the capacity of neurodivergent people to participate equitably in institutions that never redesigned their norms — while simultaneously tightening governance, intensifying expectations, and deepening dependence on infrastructure administered by others. What cyborgization adds is the recognition that, at some threshold of integration, the language of “using a tool” gives way to that of constitutive coupling, and the governance questions shift accordingly. That contradiction is not a temporary glitch on the road to a stable settlement. It is the terrain.

The question that matters most is not whether AI becomes more capable, but whether it becomes habitual, and for whom. The historian who now builds applications, the teacher who now constructs interactive archives, the scholar who now generates datasets from primary sources, the neurodivergent student whose analytical capacity finally finds institutional expression — each has gained something real. Each has also entered into a dependency relationship with a platform whose incentives, longevity, and governance are beyond their control. Prostheses are political when they become ordinary. They become cyborg relations when the ordinary becomes constitutive. And they are most political of all when they redistribute competence across boundaries that institutions have spent decades treating as natural.

Clark, Andy. 2003. Natural-Born Cyborgs: Minds, Technologies, and the Future of Human Intelligence. New York: Oxford University Press.

Clark, Andy. 2008. Supersizing the Mind: Embodiment, Action, and Cognitive Extension. New York: Oxford University Press.

Clark, Andy, and David J. Chalmers. 1998. “The Extended Mind.” Analysis 58, no. 1: 7–19.

Clynes, Manfred E., and Nathan S. Kline. 1960. “Cyborgs and Space.” Astronautics, September, 26–27, 74–76.

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.

Crawford, Kate. 2021. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven: Yale University Press.

Hamraie, Aimi. 2017. Building Access: Universal Design and the Politics of Disability. Minneapolis: University of Minnesota Press.

Haraway, Donna J. 1985. “A Manifesto for Cyborgs: Science, Technology, and Socialist Feminism in the 1980s.” Socialist Review, no. 80: 65–108.

Mitchell, David T., and Sharon L. Snyder. 2000. Narrative Prosthesis: Disability and the Dependencies of Discourse. Ann Arbor: University of Michigan Press.

Price, Margaret. 2011. Mad at School: Rhetorics of Mental Disability and Academic Life. Ann Arbor: University of Michigan Press.

Zuboff, Shoshana. 2019. The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. New York: PublicAffairs.