Sorry About Your Lost Privilege

24. 9. 2026 / Muriel Blaive

 
Consider language. Native English speakers possess an advantage so thoroughly naturalized that it is barely perceived as an advantage at all. The rest of us are expected to publish in the same journals and according to the same linguistic standards, except that we must either invest considerably more time in producing acceptable English or pay somebody to correct it. Professional editing of an article can easily cost €2,000; editing a book can cost €5,000 or more. Over a career of thirty years, the cumulative cost can approach the price of a small house. This is effectively a tax imposed on scholars for not having been born into the language that became academically dominant. Yet the finished article appears, the CV records another publication, and the radically different resources necessary to produce it disappear from view.

The same applies to research assistance. Prestigious universities employ graduate students and research assistants who locate sources, compile bibliographies, scan material, verify references and summarize books and articles. Earlier generations of predominantly male academics had another source of assistance: their wives, who typed manuscripts, corrected texts, maintained bibliographies and sometimes contributed considerably more substantial intellectual labour. They might receive an affectionate acknowledgment — “To my dear wife, without whom this book would never have been the same” — but certainly no co-authorship. None of this rendered the resulting scholarship inauthentic. Academic authorship was understood, quite reasonably, to mean intellectual responsibility rather than the literal performance of every operation involved in producing a book. It is curious that this distinction should suddenly become so difficult to understand when the assistant is a machine.

And then there is access to knowledge itself. This may be the greatest academic privilege because scholars who possess it can so easily mistake it for the natural state of research. I am writing these lines at the British Library. I came here because there are too many books I need that I cannot obtain in Austrian libraries, despite Austria being a rich European country. To read them, I have had to travel from Vienna to London, pay for transportation and accommodation, and devote several weeks to placing myself physically close enough to the books to consult them. Over an academic career, I have spent many thousands of euros simply gaining access to scholarly material. A scholar at Harvard, Oxford or another exceptionally well-resourced institution can click on a reference and obtain the article, request a book and have it delivered, consult databases whose subscription prices she may never even know, or ask the library to acquire material she needs. This infrastructure is subsequently converted, almost invisibly, into individual academic productivity.

Communist studies offer an even more spectacular example. After the Soviet archives opened, enormous collections of documents were microfilmed, digitized, and sold commercially. Wealthy American universities could spend millions of dollars acquiring entire collections, giving their scholars local access to bodies of Soviet documentation that researchers elsewhere could consult only by travelling to Moscow or other former Soviet archival centers, obtaining funding and visas, finding accommodation, navigating inventories, persuading archivists to produce files, waiting for documents and dealing with restrictions that could change unpredictably. One scholar could search Soviet archival material from an American university library while another had to cross a continent merely to discover whether an archive would allow her to see the relevant file. Yet when their books appeared, the vast investment in one scholar’s research infrastructure did not appear on the title page. Both books became evidence of individual scholarly achievement.

These inequalities become still more striking when one looks beyond Western Europe and North America. Scholars working in less well-resourced academic systems may have excellent training and genuinely original ideas while lacking subscriptions to major databases, comprehensive research libraries, funds for international travel, professional language editing or armies of graduate assistants. Some work in English; many do not. In either case, participation in the international academic conversation has depended not simply on intelligence or originality but on access to the infrastructure through which intelligence and originality become professionally recognizable. A brilliant argument badly expressed in the dominant academic language has never competed on equal terms with a conventional argument presented in impeccable academic English.

This is one reason why I find some of the indignation about AI remarkably unconvincing. Artificial intelligence suddenly gives scholars without this infrastructure access, at negligible cost compared with the alternatives, to forms of assistance that wealthier scholars have long obtained through money, geography or institutional affiliation. I now have a language editor, a translator, something resembling a research assistant and an intellectual interlocutor available at any hour. AI cannot give me a Soviet archival document that has never been digitized, nor can it put an unavailable book into my hands. The British Library remains the British Library. But it can reduce the disadvantages that arise once information becomes accessible, and for scholars working in poorer institutions and countries the difference may be even greater than it is for me.

Apparently this is the moment at which academic assistance has become morally troubling.

The timing deserves some attention. Assistance was compatible with scholarly authenticity when it was provided by a graduate student, a research assistant, an expensive professional editor, “my dear wife,” or the institutional infrastructure of an elite university. Now that scholars without these resources can obtain some comparable forms of assistance for the price of a monthly subscription, we suddenly hear passionate defenses of unaided intellectual production. This does not mean that everyone who objects to AI is consciously defending privilege. There are serious reasons to worry about hallucinated references, fabricated information, homogenized prose, intellectual dependency and the industrial production of academic bullshit. But there is also a sociological question here that deserves considerably more attention: which forms of academic capital are losing their scarcity value because of AI, and who previously benefited from their scarcity?

The most obvious is command of academic English. Perhaps the most interesting, however, is accomplished academic writing itself. Academia contains exceptionally intelligent people who do not write particularly well. It also contains people with the opposite and professionally very useful gift: the ability to make relatively ordinary ideas appear sophisticated. A modest proposition can be surrounded with an impressive theoretical vocabulary, embedded in 9,000 elegantly constructed words and presented with sufficient conceptual assurance to acquire the appearance of intellectual depth. Entire careers are built on this capacity. Until recently, producing such prose was a scarce skill and therefore an important comparative advantage. AI is rapidly making it less scarce.

This is usually presented as evidence against AI: the machine can produce sophisticated prose without sophisticated thought. I draw a somewhat different conclusion: perhaps we have been attributing too much intellectual significance to the ability to produce sophisticated prose. If almost everyone can now generate the surface characteristics of accomplished academic writing, those characteristics cease to function as reliable markers of intellectual distinction. And that may be one of AI’s most genuinely egalitarian effects. A scholar’s idea no longer needs to lose simply because another scholar happens to be a native English speaker, can afford a better editor, or has been socialized from the age of eighteen into the rhetorical codes of an elite Anglophone university. A researcher in Lagos, Prague, La Paz or Tbilisi can increasingly present an argument in English of the same superficial professional quality as a professor at Harvard. That does not make the arguments equally good. It does something much more interesting: it removes one reason for not judging them on their intellectual merits.

The question can therefore move backwards, from presentation to proposition. What exactly is being said? Is the question original? Does the argument explain something? Is there evidence for it? Does the conceptual apparatus illuminate the object or merely decorate it? If linguistic polish and professional academic presentation become widely available, then originality, intelligence and judgment have a chance to matter more, not less. Some of the academics most vulnerable to AI may therefore be precisely those whose comparative advantage consisted in presenting conventional ideas exceptionally well. I admit to finding this prospect rather amusing.

There is another misunderstanding about AI that I encounter frequently, particularly among people who assure me that they do not need to work with it in order to know how it works. Generative AI is, among other things, a peculiar intellectual mirror. Give it a conventional question and it will produce a conventional answer in excellent prose. Give it a weak premise and it is perfectly capable of constructing an elegant argument upon weak foundations. One can then contemplate the result and conclude that AI is banal or stupid. Sometimes it certainly is. But sometimes one has simply received one’s own banal question back in professionally edited form.

The experience becomes quite different when the interaction itself becomes intellectual work. One can reject an answer as obvious, point out that two concepts have been confused, introduce contradictory evidence, ask what would follow if one’s hypothesis were false, demand an example from an entirely different historical setting, reject the resulting analogy as superficial and try again. Used in this way, AI creates an unusually rapid iterative environment in which propositions can be tested, reformulated, contradicted and displaced. Its fluency does not eliminate the need for intellectual judgment; it makes judgment more important, because somebody still has to recognize what is interesting, what is trivial and what is simply wrong.

For me, however, the most exciting possibility lies in AI’s extraordinary indifference to disciplinary boundaries. One can begin with a problem concerning post-communist memory politics and discover an analogous problem in the sociology of expertise; from there move into epistemology, legal history, anthropology or the history of medicine; and eventually return to the original historical problem and see something that was invisible before. This is one of the capacities I have always admired in Michel Foucault: his ability to assemble materials and intellectual traditions that conventional disciplinary organization kept apart. Medicine, psychiatry, prisons, sexuality, law, architecture, administrative practices, political economy and philosophy could become components of the same inquiry because the question, rather than the discipline, determined what was relevant.

I have joked to friends that with AI, we can all be Michel Foucault. Obviously, AI cannot give everyone Foucault’s intelligence, originality, or judgment. But it can democratize one of the conditions that made his intellectual practice possible: the possibility of moving across enormous bodies of knowledge, disciplines, periods, and problems in search of unexpected connections. What once required decades of interdisciplinary reading, an unusually rich intellectual environment or the fortunate presence of exactly the right colleague at dinner can sometimes now begin with a simple question: has anybody in another field encountered a problem structured like this one? The answers still require verification. Some connections will be superficial and others wrong; analogy remains no substitute for evidence. But the territory across which curiosity can operate has expanded dramatically.

This suggests a possibility considerably more interesting than the endless lamentations about the death of academia. If polished prose becomes cheap, polished prose becomes less valuable as a marker of distinction. If competent summaries and conventional interpretations become readily available, reproducing them becomes less impressive. What may consequently become more valuable are precisely the things that remain difficult to automate: asking an unexpected question, recognizing a connection that matters, distinguishing an illuminating analogy from a superficial one, noticing that an authoritative answer is nonsense, finding the evidence that destroys one’s own hypothesis, and remaining curious after receiving the first perfectly plausible answer. In other words, AI may devalue some of the proxies through which academia has traditionally recognized intelligence and force us to pay rather more attention to intelligence itself.

This, I suspect, is one reason why some of the panic is so intense. I do not mean that every critic of AI is secretly defending his status or that objections to AI can be reduced to material interests. But technologies do not merely create new possibilities; they depreciate existing forms of capital. When a competence that has been difficult, expensive or institutionally restricted becomes widely available, the people who possessed it lose part of their comparative advantage. What one person experiences as democratization can therefore quite sincerely be experienced by another as decline. It is worth asking whether at least some of the hand-wringing about the “death of academia” is in fact mourning for particular academic distinctions. My sympathy for this particular loss is limited.

There is, however, a serious problem that I cannot gloat away: students. I have not taught regularly since 2022, just before generative AI became ubiquitous, and I therefore have little direct experience of what it has done to student work. But here the problem seems fundamentally different. An established scholar learned to construct an argument before a machine could construct one for her, learned to read before a machine could summarize a book, learned to formulate questions before a machine could propose them, and accumulated enough knowledge to recognize at least some of the occasions on which AI confidently produces nonsense. A student can now potentially bypass precisely the intellectual labor through which these capacities were traditionally acquired. What functions as liberation for the intellectually trained can function as deskilling for someone who is still being trained.

I honestly do not know the solution. Universities may have to distinguish much more carefully between intellectual operations students must first learn to perform themselves and those that can subsequently be delegated. Assessment may have to shift from evaluating finished products toward evaluating processes of reasoning and the capacity to explain and defend an argument. Oral examinations may become more important again. Entirely new pedagogical practices may emerge. This is a genuine problem, and one I would much rather see academics debating than issuing declarations that they will “never respect” anything written with AI.

For centuries, academia has represented radically unequal conditions of intellectual production as a fair competition between individual minds. Some contestants were native speakers of the dominant language; some could afford editors; some had research assistants or invisible domestic assistance; some worked in institutions that could spend fortunes acquiring databases and entire archival collections; some happened to live beside the greatest research libraries in the world. Others spent their own money, time and labor trying to approximate these conditions, while scholars in less well-resourced academic systems often had no realistic possibility of approximating them at all. In the end, the CVs were compared line by line and the infrastructure disappeared.

AI will not abolish these inequalities. Rich universities will acquire better systems, proprietary databases will remain expensive, physical archives will remain physical, and money will continue to purchase time, mobility and assistance. But some long-standing advantages are becoming less exclusive. The capacity to express an intelligent idea in internationally acceptable academic English no longer needs to depend quite so heavily on birthplace, institutional affiliation or personal wealth. If that means that a genuinely original scholar working at an underfunded university can compete more effectively with a mediocre scholar surrounded by magnificent institutional resources, I have considerable difficulty seeing this as the death of academia.