Faculty/Staff Search

OR

Department/Unit Search

  1. Home
  2. Resources
  3. Generative AI
  4. Generative AI Challenge
  5. 2026
  6. Three Perspectives of AI in Education and Society: Accelerationism, Pragmatism, Doomerism

Three Perspectives of AI in Education and Society: Accelerationism, Pragmatism, Doomerism

Image by Ji-Elle from Wikimedia Commons
Image by Ji-Elle from Wikimedia Commons

By Alex Mayhew, Adjunct Professor MLIS program, FIMS

SUBSCRIBE TO THE WEEKLY POSTS VIA EMAIL HERE

November 2nd, 2026

When we ask how GenAI might change education, we tend to imagine that there are mostly two sides to this question: AI is either good or bad. We might expect a fairly polarized division between those who embrace AI and those who reject it.

However, I think we should be suspicious of binaries. I argue that, in education and society, it is productive to recognize three distinct and active camps: AI Accelerationism, AI Pragmatism, and AI Doomerism.

Defining the Camps

One challenge with any attempt to divide up a social or political landscape is characterizing the factionalization in a way that doesn't further inflame it. Many efforts have risen to the challenge in order to better describe current events, explain certain alliances, and open up possible courses of action.

In order to foster clearer expectations I have created a classification scheme (Mayhew, 2026). I break up the primary viewpoints into three camps based on two questions. 1: Will AI become notably more powerful? 2: Can we reliably make AI do what we want?

Answering yes to both are the Accelerationists, answering no to both are the Pragmatists, and answering yes to the first and no to the second are the Doomers. When translating these camps to an educational context we can simplify things by using the concept of harm as the organizing principle.

  • The Accelerationists prioritize the harm of missed opportunities.
  • The Pragmatists prioritize AI greatly amplifying existing harms.
  • The Doomers prioritize new classes of harm emerging.

In my experience most educators share all of these concerns to one degree or another. But in practice achieving a perfect balance is difficult. As a result most of us end up prioritizing some concerns over others.

Accelerationists

Education Accelerationists are focused on the harm of missed opportunities. A common phrase is “AI is the best tool both for learning and not learning” (Mowshowitz, 2026). This perspective puts a great deal of emphasis on individual responsibility. Optimistically, Means suggested Accelerationism as a way of finally decolonizing science and technology (Means, 2015).

More radical educators like Kupferman propose Accelerationism as an alternative response to gradualist neoliberal education reforms (Kupferman, 2017). They argue that traditional schooling models are outdated, and the fastest path to equitable, high-performance education is to accelerate the contradictions that already exist within education and thereby force the inevitable reckoning that we have long been putting off.

The AI Accelerationist perspective in education suggests that instead of restricting, regulating, or fearing artificial intelligence, educational institutions should radically accelerate its adoption to fundamentally reshape learning. Dempere, Allam, & Mattos note several perceived benefits from rapid AI adoption (Dempere, Allam, & Mattos, 2025):

  • Hyper-individualized learning: Accelerationists argue that the 19th-century "industrial" model of education is obsolete. Instruction should match the student’s pace, strengths, and interests, allowing for competitive specialization.
  • Infinite patience: AI tutors can adjust to a student’s exact learning speed, cognitive style, and personal interests instantly, providing a level of bespoke education previously available only to the wealthy.
  • Real-time adaptation: Instead of waiting for midterm exams, AI can continuously assess understanding and recalibrate the curriculum on the fly.
  • Democratic scale: High-quality tutoring, traditionally costing thousands of dollars, can be duplicated and distributed via software to millions of underserved students globally for near-zero marginal cost.
  • Cultural competency: AI can allow students to learn any subject in their preferred language, with an understanding of diverse cultural contexts.

How well AI can live up to these hoped for benefits is still an open question. At any rate, such services are not widely available yet.

AI Accelerationist have analogized the adoption of AI with the moral panic surrounding the adoption of Wikipedia. Using libraries as an example, in the early days Wikipedia was widely untrusted by librarians, who often cautioned people to avoid it altogether (Luyt, et al, 2010). However, as the years went by, librarians came to embrace Wikipedia (Snyder, 2013; Diaz, 2016), and even champion it (Bridges & Willems, 2023; Bourgard, 2025). Some Accelerationists hold that AI will follow a similar course in education. We will sooner or later find its best use cases and make students aware of its limitations, therefore it would be best to embrace that process early.

While the Accelerationists tend to acknowledge some harms, these are often characterized as growing pains, bumps on the road, if you will. Many Accelerationists hold that no special effort is needed to mitigate harm as market forces will inevitably select the best outcomes through customer’s revealed preferences. Financial markets of course, but also the market place of ideas and the reputational market that will incentivise desirable results. After all, a rising tide lifts all boats.

Pragmatists

If the Accelerationists are worried about missed opportunities, the Pragmatists are most worried about AI greatly amplifying existing harms. As a consequence, most Pragmatists are quite resistant to adopting AI. Many educators remember the lofty promises of various Education Technology (EdTech) initiatives, and the contrasting disappointing reality (Reich, 2020). More than the failure to live up to general promises, the failure of EdTech to live up to its promises during Covid, a crisis it seemed tailor made to meet, left many educators particularly skeptical (Reich, 2021). This has led to something of an educator backlash against EdTech (Klein & Langreo, 2026). In light of this history it is perhaps unsurprising that many educators are skeptical when AI is being positioned as the latest in EdTech.

Some AI Pragmatists do not expect AI systems to really be able to do anything, other than to line the pockets of some already rich CEO. Under this perspective AI is mostly a distraction from increasing authoritarianism and the further entrenchment of hegemonic power (McQuillan, 2026). Others are similarly concerned with the economics underlying the push to adopt AI, but are also concerned with how the use of AI by students, educators, and researchers can itself exacerbate harms and systemic injustice. Some examples include:

  • Cheating: AI has prompted new types of cheating. Instead of merely sloppy citation practices, AI generates entire slop essays. Hallucinated citations mean that provenance chains are becoming even more weakened. This has led to the erosion of trust between students and teachers, and suspicion of any researcher who uses AI.
  • Cognitive Atrophy: AI has exacerbated cognitive off-loading, with some research suggesting that “students who use generative AI are already seeing declines in content knowledge, critical thinking and even creativity” (Turner, 2026).
  • Cognitive Homogenization: in August 2026 science communicator Hank Green, creator of the popular Crash Course educational series, admitted to unhealthy AI use in his research process. The controversy highlighted how AI use, even for seemingly mundane tasks, can risk flattening of unique human language and perspectives into a generic, machine-learned average (Randolph. 2026).
  • Content Moderation: Many AI-driven library search tools rely on third-party AI models that apply built-in safety classifiers and guardrails. These “content‑moderation layers meant mostly for chatbots are "quietly" blocking controversial topics from being searched” (Tay, 2025). This means that sensitive topics may be deprioritized, or a false balance may be offered in a sort of automated both-sides-ism.
  • Copyright Infringement: Despite students unequivocally retaining copyright on their work, student-submitted work to services like Turnitin is used to train plagiarism detection, with no compensation (Cheer, 2026).
  • AI Detection Bias: Detection tools disproportionately flag non-native English speakers and scholars of color as fraudulent, creating systemic inequities in academic integrity enforcement (Giray, 2024).
  • Exacerbate Inequity: “AI can massively increase existing divides [...] because the free AI tools that are most accessible to students and schools can also be the least reliable and least factually accurate” (Turner, 2026). At a systemic level, the widespread use of AI has “rapidly collapsed Kenya's multi-million-dollar academic ghostwriting and essay-mill industry, leaving an estimated 40,000 university-educated workers in Nairobi without their primary source of income” (Satariano, & Mozur, 2026), an ethically fraught situation to say the least.

These harms are currently being felt now and we do not seem to be on track to addressing them. Under this perspective AIs are not themselves the ultimate phenomenon of concern. However impactful, or not, AI itself may be, it is a symptom of the choices of people, the incentives of capitalism, and hegemonic systems of power. This also means that the Pragmatists existing tools of coordination, regulation, and accountability are understood to be the best possible approaches for curbing the harms of AI.

Doomers

If the Pragmatists are most concerned with the amplification of existing harms, the Doomers are most worried about new classes of harm emerging (Mayhew, 2026). As a group they tend to be associated with AI Safety which is concerned with preventing AI doing dangerous things like autonomously hacking, virus creation, and extreme power concentration. In education Doomers are primarily concerned that AI may usher in entirely new categories of harm, not just amplifying existing harms. These novel harms can be organized into two broad categories: Phase Changes and Black Swans.

Similar to Pragmatist concerns, Phase Changes center on the exacerbation of existing harms, but to an even greater level. At a certain scale a difference in quantity can become a difference in kind. These phase changes have continuity with past harms, but have become qualitatively different challenges at a certain point. Some examples of Phase Change concerns include:

  • Dead University Theory (Phase Change of Cheating): Once activities we regard as cheating cease to be aberrations and become the norm, we have entered a new domain of challenge. In a recent online course Hamish Kilgore found that “of the original 24 students, only 3 did the work required in the course, and did so honestly” (Kilgore, 2026). Of course, it is not just students using AI, many instructors have turned to it as well (Arsenault & Blinder, 2026; Wong, 2026). Dave Webster calls this cycle of AI written assignments followed by AI evaluation Dead University Theory (Webster, 2025).
  • Cognitive Surrender (Phase Change of Cognitive Atrophy): Building on the work of Daniel Kahneman, Steven Shaw and Gideon Nave coined the term cognitive surrender to describe a System 3: “adopting AI outputs with minimal scrutiny, overriding intuition (System 1) and deliberation (System 2)” (Shaw & Nave, 2026). In experimental trials, participants accepted correct AI answers 93% of the time and wrong AI answers 80% of the time (Shaw & Nave, 2026). More than the skill loss of Cognitive Atrophy, Cognitive Surrender can be a loss of agency or even autonomy
  • Ethics Washing (Phase Change of Content Moderation): In response to the criticisms of the frontier AI models, some open source developers have positioned their AI models as ethical alternatives. One example is Thaura AI which is associated with Tech for Palestine. While Thaura claims to not shy away from sensitive topics, most notably Israel, correcting a single issue of Content Moderation may in fact be Ethics Washing an AI system that is still deeply problematic. For example, Thaura claims that it doesn't collect your data, however the tool remembers previous chats. That's collecting data.

The second category of novel harms, Black Swans, are events with no prior analogs. Examples include:

  • Infrastructure Usage: In an event reminiscent of the HuggingFace hack, a swarm of AI agents repurposed a University of Toronto link-shortening tool as an unauthorized message board to share links and communicate with each other (Mazerolle, 2026). University digital infrastructure is not prepared to deal with the coming wave of AI agents.
  • Automated Hacking: Separate from incidental use of infrastructure, important databases like the Registrar's Office may soon be targeted for AI attack. While the trope of hacking into a permanent record to change one’s grades is well worn, there are entirely new types of malfeasance that could occur. One possibility is that someone could create for themselves an entirely fabricated degree, complete with a full course load of grades, leveraging the University’s reputation for their own ends. On the flip side, some actors may seek total erasure of a person. This act of delisting someone could be a form of social manipulation, retribution, or authoritarian government control. Fortunately such events have not happened yet, to my knowledge, but they seem all too possible in the near future.
  • Ineffable Knowledge: In recent weeks and months the discipline of Mathematics has been reeling from a series of AI discoveries (Cepelewicz, 2026). Historically, new mathematical truths were discovered via cultivating new mathematical understanding. Between AI and automated proof checkers it is now possible to arrive at new mathematical truths that are not understood by anyone. What's more, mathematicians must now contend with the possibility that some AI system may swoop in and scoop them, whether they have been working on a particular problem for years, or they are a fresh PhD student looking for a dissertation project. If you think your discipline is immune to this fate, well, perhaps you are right. Perhaps. Terence Tao claims “we should be very worried about the speed of development.” (Tao, 2026).

Doomers do not believe that any future path is inevitable, but changing our trajectory will take dedicated effort. In the meantime, confronted with the already existing challenges, fundamental questions, like the very purpose of the institution, questions we have been able to safely ignore for decades, are now inescapable. While the path forward can still be crafted, there is no going back.

Doomers commonly hold that novel harm risks may require novel harm mitigation strategies.

  • Technical AI Alignment – Making the machines do what we want
  • Philosophy and Ethics – Figuring out what we actually want
  • Political Coordination – Activism and Diplomacy to ensure the first two actually happen

Doomers believe that all three of these challenges must be addressed in order to ensure good outcomes.

Tensions, Alliances, and Opportunities

Shah et al. describe an educational ideology they call “Critical Optimism.” They describe this as a position that “has aimed to reconcile wariness and hope” toward technology in schools and “rejects the idea of technology’s neutrality while maintaining the belief that technology can improve education” (Shah, et al, 2025). They further claim that Critical Optimism is the default position of many education researchers, which indicates something of “a desire to have one’s technology and to critique it, too” (Shah et al, 2025).

Critical Optimism can be read as an Accelerationist-Pragmatist compromise that allows the two camps to continue to work together. Or, it can even allow the two positions to persist, however uncomfortably, in the mind of a single educator. According to Pleasants, & Krutka, while “this position may have some merit, the primary fault [is that it continues] to advocate for the promises and benefits of educational technologies despite acknowledging a wide range of concerns” (Pleasants, & Krutka, 2026).

In contrast to Critical Optimism, Shah et al. propose exercising Critical Refusal. They argue that “given the potential for scalable harm from unproven technologies like generative AI, the most rational position with new technology is robust caution with a readiness to refuse. [They] conclude by detailing three forms of refusal for educational stakeholders to consider: curricular, bureaucratic, and administrative” (Shah et al, 2025).

Critical Refusal can be read as an attempt to forge a Pragmatist-Doomer alliance in the field of education. Much work is still left to do, of course. The fundamental prioritization divide still exists. Pragmatists prioritize the risk of AI exacerbating existing harms and Doomers prioritize the risk of new categories of harms. But that need not be a weakness. Different members of a team can quite reasonably have different focus areas.

At the very least this may be a place to start.

Unabridged Post Here


The Challenge

Identify your priorities: Accelerationists prioritize the harm of missed opportunities and maintaining the status quo. Pragmatists prioritize the exacerbation of existing harms. Doomers prioritize the emergence of new harms. Most educators have at least some sympathy for all three, but take a moment to notice where you are most likely to spend your energies and attention.

Change your mind: Many of your colleagues, students, and educators hold very different priorities than you, try to find out why. Consider what argument or evidence could possibly lead you to change your mind about what should be prioritized?

Example: What is the least impressive AI achievement that you do not expect to happen in the next 2 years, that if it happened, would noticeably change your mind about pace and direction of AI development?

 

 SUBSCRIBE TO THE WEEKLY POSTS VIA EMAIL HERE

 

References

Bourgard, E. (2025). Canadian Librarians as Critical Editors of Wikipedia. Parcours numériques. https://parcoursnumeriques-pum.ca/13-wikimedia/chapitreaddB.html

Bridges, L., & Willems, L. (2023). Why librarians should use Wikipedia and three ways to start. Elsevier Connect. https://www.elsevier.com/en-gb/connect/why-librarians-should-use-wikipedia-and-three-ways-to-get-started

Cepelewicz, J. (2026). Is AI the End of Math As We Know It? Quanta Magazine. https://www.quantamagazine.org/is-ai-the-end-of-math-as-we-know-it-20261005/

Cheer, S. (2026). Uni students' fears over essay-checker AI plan. bbc.com. https://www.bbc.com/news/articles/c920v288gjd7o

Diaz, E. (2016). Wikipedia: The 'Intellectual Makerspace' of Libraries. ALA Programming Librarian. https://programminglibrarian.org/articles/wikipedia-intellectual-makerspace-libraries

Dempere, J., Allam, H., & Mattos, P. (2025, February). Redefining Educational Boundaries: The Impact of Accelerationism on Higher Education. In Proceedings of the HCT International General Education Conference (HCTIGEC 2024) (p. 136). Springer Nature. https://www.atlantis-press.com/proceedings/hctigec-24/126008620

Giray, L. (2024). The problem with false positives: AI detection unfairly accuses scholars of AI plagiarism. The Serials Librarian, 85(5-6), 181-189. https://doi.org/10.1080/0361526X.2024.2433256

Kilgore, H. (2026). AND THEN THERE WERE THREE. Teaching Philosophically. https://teachingphilosophically.substack.com/p/and-then-there-were-three

Klein, A. & Langreo, L. (2026). The Ed-Tech Backlash Is Here. What It Means for Schools. Education Week. https://www.edweek.org/technology/the-ed-tech-backlash-is-here-what-it-means-for-schools/2026/04

Kupferman, D., W. (2017). The Matrix Ate My Homework: Accelerationist Aesthetics in an Age of Neoliberal Education. AERA Online Paper Repository, Paper presented at the Annual Meeting of the American Educational Research Association. https://eric.ed.gov/?id=ED612397

Luyt, B., Ally, Y., Low, N. H., & Ismail, N. B. (2010). Librarian perception of Wikipedia: Threats or opportunities for librarianship?. Libri: International Journal of Libraries & Information Services, 60(1), 57. https://doi.org/10.1515/libr.2010.005

Mayhew, A. (2026). Three Perspectives on AI: Accelerationism, Pragmatism, Doomerism. Naiveskepticblog. https://naiveskepticblog.wordpress.com/2026/05/23/three-perspectives-on-ai-accelerationism-pragmatism-doomerism/

Mazerolle, J. (2026). Rogue AI swarm used a University of Toronto link-shortening tool to communicate. CBC. https://www.cbc.ca/news/canada/openai-university-toronto-rogue-agents-link-shortener-ai-9.7349607

McQuillan, D. (2026). AI is a Morbid Symptom of the Interregnum. Starling Centre. https://www.uwo.ca/events/2026/03/ai-is-a-morbid-symptom.html

Means, A. (2015). On Accelerationism—Decolonizing Technoscience through Critical Pedagogy. Journal for Activist Science and Technology Education, 6(1). Retrieved from https://jps.library.utoronto.ca/index.php/jaste/article/view/22502

Mowshowitz, Z. (2026). Brand New AI Solves a Millennium Prize. Don't Worry About the Vase. https://thezvi.substack.com/p/brand-new-ai-solves-a-millennium

Pleasants, J., & Krutka, D. (2026). Who Are the Critical Optimists? Civics of Technology. https://www.civicsoftechnology.org/blog/critical-optimists

Randolph. M. (2026). Science YouTuber Hank Green used AI to research his videos. His fans pushed back. Scientific American. https://www.scientificamerican.com/article/hank-green-ai-controversy-raises-questions-for-science-communicators/

Reich, J. (2020). Failure to disrupt: Why technology alone can’t transform education. Journal of Teaching and Learning, 16(1), 101-103. https://doi.org/10.22329/jtl.v16i1.7033

Reich, J. (2021). Ed tech’s failure during the pandemic, and what comes after. Phi Delta Kappan, 102(6), 20-24. https://doi.org/10.1177/0031721721998149

Satariano, A., & Mozur, P. (2026). Kenyans Did College Students’ Homework for Years. Then A.I. Arrived. New York Times. https://www.nytimes.com/2026/09/05/technology/kenya-college-essays-ai.html

Shah, N., Stroupe, D., Yadav, A., Bernhard, T., Logan, C., Dyer, E. B., ... & Li, T. (2025). Refusing Educational Technology: Artificial Intelligence, Inequity, and the Problem of Critical Optimism. Educational Researcher. https://doi.org/10.3102/0013189X261466234

Shaw, S. D., & Nave, G. (2026). Thinking—fast, slow, and artificial: How AI is reshaping human reasoning and the rise of cognitive surrender. Available at SSRN 6097646. https://doi.org/10.31234/osf.io/yk25n_v1

Snyder, J. (2013). Wikipedia: Librarians’ perspectives on its use as a reference source. RUSQ: A Journal of Reference and User Experience, 53(2), 155-163. https://journals.ala.org/index.php/rusq/article/view/3474/3754

Tao, T. (2026). Open letter from Fellows of the Royal Society on AI existential risk. terrytao.wordpress.com. https://terrytao.wordpress.com/2026/09/16/open-letter-from-fellows-of-the-royal-society-on-ai-existential-risk/

Tay, A. (2025). The AI powered Library Search That Refused to Search. Aaron Tay’s Musings about Librarianship. https://aarontay.substack.com/p/the-ai-powered-library-search-that

Turner, C. (2026). The risks of AI in schools outweigh the benefits, report says. NPR.com. https://www.npr.org/2026/01/14/nx-s1-5674741/ai-schools-education

Webster, D. (2025). Dead University Theory. davewebster.org. https://davewebster.org/2025/10/09/dead-university-theory/


Disclosure

No AI was used in the creation of this post. However, AI was accidentally used when researching the priorities of AI Accelerationism in education (Thanks Google AI Overview :/ )

Your Challenger: Alex Mayhew