The Hayekian argument for homeschooling

Why a Nobel Prize-winning economist's most famous idea points toward radically personalized education.

By Andrew Garber, founder of TheHomeschoolingCompany. Reviewed August 12, 2026.

Friedrich Hayek was not writing about homeschooling. In 1945, he was writing about one of the largest economic arguments of the twentieth century: whether a sufficiently intelligent central authority could successfully plan an economy.

His answer was no, and his reasoning was more subtle than most people realize. The problem was not that central planners were stupid or evil or that they were missing some sufficiently clever equation. The problem was that the information required to coordinate a complex economy does not exist in one place. It is scattered among millions of people, much of it local, contextual, constantly changing, and impossible to meaningfully compress into a report. The guy running a hardware store in Des Moines knows something about demand for hammers in his neighborhood that an economist in Washington does not. A factory manager knows something about his machines that the central planning office does not. A customer knows whether she would rather spend $10 on lunch or save it. Hayek's insight, the one that would earn him a share of the 1974 Nobel Memorial Prize in Economic Sciences, was that decentralized systems can make use of knowledge that no central authority could ever completely possess.

Knowledge "never exists in concentrated or integrated form."

F.A. Hayek, The Use of Knowledge in Society (1945)

I think his knowledge problem also happens to be one of the strongest arguments for homeschooling ever made.

The knowledge problem, except your kid is the economy

Imagine you are designing the perfect education for one child. You would want to know what they are interested in, how quickly they learn, what they are unusually good at, what they struggle with, whether they learn better by reading something or hearing it or seeing it or arguing about it or building something. You would want to know what makes them curious, what bores them, what frustrates them, what makes them disappear into a project for six hours without noticing the time, and what they might want to do with their life. You would want to know that they hate arithmetic worksheets but spent an entire afternoon calculating gear ratios because they are obsessed with bicycles. That they supposedly "don't like writing" but voluntarily wrote 4,000 words of lore for their Minecraft server. That they got a mediocre grade on the biology test but then asked forty questions about genetics after the lesson was over.

Parents know an enormous amount of this. Kids themselves know some of it. And critically, a lot of it is only discovered as the kid actually learns. Your kid may not know that they love genetics until they start learning about horse breeding. They may insist that they hate math while voluntarily spending their weekend calculating stats for their fantasy basketball league. A kid who looks mediocre at writing assignments might spend four hours writing an absurdly detailed history for a strategy game they play. This is all information, and it is exactly the kind of local, contextual, constantly-changing information Hayek was talking about.

What the system knows

Age

12

Grade

7

Math score

78

Reading level

9.2

State standard

Complete unit 4 by October 17

What the parent knows

Twelve years

Of direct observation of one specific person

The gear ratios

An afternoon of voluntary arithmetic, for bicycles

The 4,000 words

Written for fun, by a kid who "doesn't like writing"

The forty questions

Asked about genetics after the lesson was over

Compare that with what a centralized education system can work with. None of it is useless, but it is an incredibly lossy representation of a human being. The school is working with a compressed profile of a seventh grader. The parent is working with twelve years of direct observation of a specific person. The resolution gap between those two is enormous, and it maps directly onto Hayek's original insight: the central planner lacks the local knowledge that would be required to make good decisions, and that knowledge cannot be efficiently transmitted upward because it is too granular, too contextual, and too dependent on things that change from week to week.

What a parent brings to the decision

The obvious objection is that parents can be wrong, and of course they can. A parent can misunderstand their child, push them in the wrong direction, teach badly, mistake temporary enthusiasm for genuine aptitude, or impose their own preferences on a kid who wants something entirely different. Homeschooling does not magically convert parents into omniscient educational planners, and any honest argument for it has to acknowledge this.

But Hayek's argument was never that decentralized actors possess perfect information. Nobody does. The question is where decisions should be made when the relevant knowledge is scattered and constantly changing, and when we are deciding what an individual child should learn tomorrow morning, it seems almost absurd to locate that decision with a state bureaucracy hundreds of miles away that has never met the kid. The strongest case for homeschooling is not that parents always know best. It is that educational decisions should be made as close as possible to the child, where the highest-resolution information actually exists, and that the child themselves should have agency in that process. A good homeschool constantly discovers new information through the student's behavior: what did they choose to spend more time on, where did they get stuck, what did they ask about after the lesson was supposedly over, what project suddenly made everything click. Education, in this framing, is a discovery process, something closer to iterative experimentation than to a twelve-year conveyor belt.

So why doesn't everyone homeschool?

Because individualized education is expensive. This is the part of the homeschooling conversation that people frequently skip.

Traditional homeschooling solves a huge part of the knowledge problem, but it creates another problem: parental labor. Someone has to find the curriculum, decide what comes next, teach algebra, explain mitosis, read the essay, figure out whether the kid actually understands cell division or merely memorized enough vocabulary to pass the quiz. Someone has to keep records, plan projects, find resources, answer questions, adjust pacing, and do it all again tomorrow. Even if the curriculum costs zero dollars, the parent's time does not. For a household with two working parents, this can become brutal.

People misunderstand the economics of the traditional classroom. A class with one teacher and twenty-five kids is not some bizarre educational philosophy we accidentally stumbled into. It is an economies-of-scale technology. One teacher, one lesson, one schedule, one curriculum, many students. You dramatically reduce the amount of expert labor required per child, and the cost of that reduction is standardization. The teacher cannot give twenty-five completely different lectures simultaneously, cannot let one student spend three weeks on the Punic Wars while another races ahead in calculus and another decides to learn biology through veterinary medicine. The system has to batch students because human attention is expensive and cannot be parallelized. For most of modern educational history, that tradeoff was unavoidable. You could have personalization, but personalization required dedicated expert labor for each student. Or you could have scale, but scale required treating every student the same.

Individual tutoring remained available to families who could afford it, while a classroom spread the cost of instruction across a larger group.

AI changes the economics

The most important educational consequence of AI is not that kids can ask a chatbot to explain the quadratic formula. It is that the marginal cost of adapting an education to one specific human being is collapsing, and that collapse breaks the historical tradeoff between personalization and scale.

What a system would have to know about Sarah

• She is eleven, loves horses, and wants to become a veterinarian.
• She reads three years above grade level and is weak on fractions.
• She understands math faster when she can visualize the problem.
• She mastered yesterday's biology material immediately.
• She misunderstood one specific part of today's lesson on cell division.
• She then spent forty minutes voluntarily asking questions about genetics.

Tomorrow's curriculum should be different because of those facts. A good parent could theoretically produce that adapted curriculum. A good private tutor could do it. A conventional school cannot realistically do it at meaningful resolution for every student, because doing it requires an amount of labor per student that the economics of the classroom cannot support. Software can, and at a cost so low that it fundamentally changes what kind of education is possible.

This means we can get the information advantages of homeschooling, the proximity to local knowledge, the ability to adapt to one specific child, without requiring the parent to personally perform every piece of educational labor. AI lets us separate control over education from the labor required to produce it. Those used to be bundled together: if parents wanted genuine control, they frequently had to carry instruction, curriculum design, grading, planning, records, and educational project management themselves. They should not have to.

Let the curriculum carry the teaching

This is a huge part of what we are building at The Homeschooling Company. Parents should retain sovereignty over their child's education. What are our goals, what are our values, what does my child care about, where do I want them challenged, what kind of person are we trying to help them become? Those are strategic decisions. Parents should make them. But control does not mean the parent needs to spend Sunday night writing a custom geometry worksheet. The parent should not have to simultaneously deliver every subject, design the curriculum, grade the work, maintain records, and research every next step in order to exercise that control.

This is how we separate those jobs today. Placement sets a starting level, and we build lessons around the interests your family shares. The Always-On Mentor answers questions about any passage your child highlights and can explain the idea another way after a missed answer. Submitted work is graded as soon as it is turned in, with written feedback, so grading is never the parent's job. Your family can see what was taught, what your child understood, and where they need help. The parent keeps the authority to judge the course without having to write every lesson in it.

There is a bad version of AI education

AI does not automatically solve the knowledge problem, and it is worth being explicit about this. You could easily use AI to make education even more centralized, to build one gigantic model that decides the Correct Curriculum for every eleven-year-old in America and then generates superficially personalized versions of it. Johnny likes dinosaurs, so his algebra worksheet has dinosaurs in it. That is not personalization. That is Mad Libs.

The system has to actually respond to local information. The family's goals have to matter. The student's interests have to matter. The student's performance has to matter. And the system has to be willing to change. If a kid becomes obsessed with ancient Egypt next Tuesday, that should be educationally useful information. If they master eighth-grade math six months early, the answer should not be to sit around until ninth grade. If an educational experiment completely fails, you should be able to kill it Wednesday morning without convening a district committee or waiting for a five-year textbook adoption cycle.

For me, the consequence is that families need control over the use of the software as well as access to it. Better generation does not remove the need to question a course, change its direction, or decide that it is wrong for this child.

Any class. Any interest. Any kid.

Once you accept this framework, another assumption starts looking strange: why do schools offer the classes they offer? Not because those are the only things worth learning. Schools offer classes that are economically feasible to provide. A teacher needs to exist. Enough students need to want the class. A classroom needs to be available. The schedules have to work. So every school necessarily offers a limited menu, and the range of things a child can seriously study is determined by whether twenty-four other kids living within five miles of them happen to want the same class at the same time.

A kid wants aerospace engineering? Why not? Marine biology? Roman military logistics? Fashion design taught rigorously enough to incorporate geometry, economics, history, chemistry, and writing? The historical answer was cost: you could not hire a private expert to construct a bespoke course for every child. AI changes that too. A course no longer needs twenty-five students to justify its existence. It needs one.

A curriculum that can change with the learner

This is where the argument ultimately lands. We spend an extraordinary amount of time arguing about what the curriculum should be, which books, how much history, algebra in eighth grade or ninth, whether coding should be mandatory, what belongs in the canon. These can be useful questions, but I think they obscure the deeper one: what if there is no single correct curriculum?

Not because knowledge does not matter, and not because children should only do whatever they feel like, but because the best education depends on the child receiving it. The correct curriculum for a future engineer should look different from the correct curriculum for a future novelist. The kid who can do calculus at twelve should not have the same mathematical sequence as the kid who needs another year to master fractions. The child obsessed with Rome should be allowed to go absurdly deep into Rome. The child obsessed with airplanes should be able to learn physics through airplanes. None of those choices need to be permanent, because the kid changes, the curriculum changes, and the relevant information is local, dispersed, and constantly being discovered.

That is the Hayekian argument for homeschooling. For most of history, we could understand the argument without being able to economically act on it, because personalized education required personalized labor and personalized labor was expensive. AI changes the cost structure. Homeschooling gives us the decentralized architecture. AI gives us scalable educational labor. Put them together and you can build something that has rarely existed outside the wealthiest households in history: a genuinely individualized education. Not a better standardized curriculum, not a chatbot bolted onto school, not worksheets with your kid's favorite superhero inserted into the word problems. An educational system that actually changes around the child, where the curriculum adapts to the student instead of demanding that the student adapt to the curriculum.

Hayek did not establish that homeschooling works better for every child. His argument explains why the information a family holds should carry real weight in deciding what that child's education looks like.

At The Homeschooling Company, we build curriculum around your child, not the other way around. Any subject. Any interest. A placement diagnostic sets the starting level, the mentor offers help whenever your child seems stuck, and you see each grade land in the Parent Dashboard.

An example algebra lesson

Your parent dashboard

See today’s plan, completed lessons, grades, and feedback. Attendance is saved when your child checks in each day, and a weekly summary shows how the week went.