A Response to Joel Mokyr
Joel Mokyr has been awarded the Nobel Memorial Prize in Economic Sciences in 2025 “for having identified the prerequisites for sustained growth through technological progress.” You can hear his interview at https://player.wbur.org/hereandnow/2025/12/01/joel-mokyr.
Listening to Joel Mokyr discuss the long arc of technological progress is always bracing. He repeatedly reminds us that technological change has been the engine of modern prosperity — that inventions like electricity, microprocessors, and now AI are “general purpose technologies” that transform everything they touch.
He also insists, correctly, that technological advances often expand human flourishing rather than contract it. Electricity benefits billionaires and people on food stamps alike. Bicycles democratized transportation. Microprocessors made computation ubiquitous and cheap.
His recent interview extends this optimism to AI. And in many respects, he’s right: AI will generate stunning new capabilities, including in education. But one claim in particular deserves closer examination, because it is becoming a central assumption in today’s discourse:
AI will revolutionize schooling by allowing every student to receive individualized instruction — a personal tutor for everyone.
This vision is as appealing as it is intuitive. It promises efficiency, equity, and tailored support. Mokyr presents it as a natural extension of AI’s power: if we can customize retail, advertising, medicine, and entertainment, why not learning?
But this is precisely where the social risks begin.
Not because personalization is bad — but because what we call “personalized AI” often rests on a deeply individualistic model of learning that threatens to erode the very things education exists to cultivate.
The Missing Variable: Learning as a Social Practice
Mokyr speaks as an economist, rightly focused on material progress and productivity. But education is not simply a service delivered to individuals. It is a practice — a social, interpretive, communal activity.
A classroom is not merely a room of separate minds absorbing content. It is:
- a community of inquiry
- a space for public reasoning
- a site where students learn to disagree, persuade, interpret, and listen
- a microcosm of democratic life
To treat learning as something best optimized through private, personalized AI tutoring is to misunderstand what learning is for.
We do not educate students simply to maximize their productivity.
We educate them to prepare them for life with others.
And this is exactly the point made — implicitly and explicitly — by Carlo Rotella and so many humanists right now:
AI threatens not the content of education, but its social form.
The Individualization Fallacy
Mokyr’s vision assumes that if each student receives perfectly optimized instruction, then learning improves. But this assumption hides several dangers.
1. The Collapse of Shared Inquiry
If students learn in individualized, AI-mediated bubbles, they lose the opportunity to:
- test ideas against others
- refine their beliefs in public
- develop interpretive courage
- understand perspectives that challenge their own
The “perfect tutor” may be pedagogically efficient, but it produces lopsided intellects — strong in private reasoning, weak in shared reasoning.
2. The Atrophy of Academic Agency
A highly personalized AI will anticipate what each student needs and smooth the path ahead.
But the essence of learning is not frictionless optimization.
It is friction.
It is confusion.
It is the “slow work” of formulating questions without knowing the answer in advance.
If AI supports the student too much — defines problems, frames interpretations, offers ready-made insights — the student risks losing the ability Rotella’s student Tyler sought:
“How do we ask the questions on our own?”
AI can scaffold. But if it scaffolds too invisibly, it replaces the very cognitive muscles education is supposed to build.
3. The Loss of Public, Democratic Competence
A society of private learners may be highly skilled but poorly socialized.
If we imagine AI tutors replacing much of classroom interaction, then we are also imagining:
- fewer opportunities to deliberate with peers
- fewer encounters with difference
- fewer occasions to build collective meaning
A democracy built on citizens who have learned largely alone with machines is at risk — not because of AI itself, but because of the weakening of the civic capacities that education sustains.
This is the hidden cost of over-individualization.
Progress Should Not Erase Community
Mokyr is correct that technological progress brings enormous material benefits.
He is also right that innovation often comes with unintended consequences — asbestos, pollution, inequity.
In education, the danger is not toxicity or climate damage.
The danger is far quieter:
We may unintentionally redesign education around the individual at the expense of the collective.
We could end up with schools where each student learns efficiently, privately, and alone — and in doing so, we degrade the communal ecology that makes learning meaningful.
This is not Luddite anxiety. It is a sober recognition that:
- material progress is not the only form of progress, and
- AI-driven personalization may inadvertently weaken the human practices we cannot afford to lose.
Toward a Better Future: AI as Shared Infrastructure, Not Private Tutor
Mokyr is right that AI has extraordinary potential.
But its deepest promise is not in isolating every learner with a bespoke virtual teacher.
Instead, AI’s promise lies in becoming infrastructure for collective reasoning — a tool used publicly, in groups, under human direction.
A tool like Topoi.
Rather than optimizing each student in isolation, Topoi imagines:
- teachers orchestrating shared dialogue
- students reasoning together with AI as a visible, accountable resource
- intellectual process made transparent, not hidden in private chats
- learning as a public act, not a private transaction
This version of AI does not threaten the social function of education — it strengthens it.
Conclusion: Progress Worth Pursuing
Joel Mokyr reminds us that technologies transform the world.
He also reminds us that progress has costs — sometimes only visible in retrospect.
The challenge for AI in education is not to reject innovation, nor to fear personalization.
It is to ensure that progress does not hollow out the human practices education is meant to cultivate.
Efficiency is not the highest good.
Individual optimization is not the whole story.
Material progress is necessary — but not sufficient.
If we are going to build AI for learning, let it be in service of the thing students most need from education in the 21st century:
the ability to think together.