Refactoring American Public Education

• American students need to leave school able to read — well, not a newspaper anymore, but read relatively sophisticated text, read a whole book, maybe even read and enjoy popular fiction.

• They need to be able to write at least simple emails, write a focused paragraph, maybe even able to write a simple essay. Nothing outrageous, but say one page of prose, 250 words.

• They need to be able to perform simple calculations without needing a calculator, or at least to be able to use a calculator to do things like convert from yards or feet to miles, from Celsius to Fahrenheit, and figure miles per gallon.

In my parents' and grandparents' generation, this was understood. Admittedly, my parents' generation is now a century ago, but their expectations were formed a century before that. They were, proverbially, the "Three R's" — Reading, wRiting, and aRithmetic.

What American public education delivers today is entirely too often none of the above. Hundreds of inner-city school systems have few — if any — students who graduate able either to read or to write.

The Education Establishment says the problem is that we need to spend more money. This might be a convincing argument if only American public education hadn't been getting more and more money for the last 30, 40, 50 years. But as it stands, increasing education funding doesn't increase the effectiveness of education. Increasing education funding has a nearly perfect negative correlation with student results: the more money is spent on education, the worse students perform.

The truth is that the Education Establishment — the teachers' unions, the academic education departments at the college and university level, government agencies, and non-governmental organizations — have been pretty uniformly pernicious. In place of well-known and proven methods, like phonics, we have the whole-word approach. For arithmetic, Common Core, while not inherently flawed, is often taught by educators no more mathematically adept than their students, reducing instruction to rote memorization without understanding.

Over the past 50 years, however, there has been a lot of really promising, effective  research on the pernicious effects of current approaches:

• Breaking students up into age-segregated cohorts — "First grade, second grade" — ignores students' individual differences and puts students whose sixth birthday was the day before school opened into the same cohort as students whose birthday fell on the day after school opened in the previous year.

• Age-segregation also loses the opportunity for the more advanced students to participate in helping their less advanced peers.

• Schools rarely foster a “growth mindset” among students. As Carol Dweck notes, a growth mindset means believing you can learn, so you will learn.

The crazy part of this is it's not exactly rocket science. With no background in education as an academic discipline at all, I developed a course in 1989 called TeX for TeXnical Typists, in which I laid out a, well, philosophy around which I focused my course.

[In this approach], the students are exposed to a very few concepts at one time; these concepts are then immediately reinforced with exercises. This quick reinforcement has two effects: it helps the students learn each concept, and reduces confusion; more importantly, each successful exercise reinforces the students' confidence. With greater confidence, the students are more willing to experiment on their own, and more willing to brave the mystic realms of The TeXbook.

So, let's try something new.

I've actually been thinking about this problem for, at this point, at least 17 years, going back to an article in 2008, back when we were still Pajamas Media: A One-Room Schoolhouse for the 21st Century. I expanded that into a longer proposal that was picked up and republished by The Country Schools Association — The Cosmopolitan One Room School (CORS): A Modest Proposal. (Bootlegged, to be honest: me and Dylan.)

At the time, it was what I considered just a "modest proposal" à la Jonathan Swift.  But it has stayed in my mind, as I looked at other aspects of the American public school system, culminating in an article in 2015: Let's Start Calling Public Schools What They Are: One of the Biggest Swindles in History.

In that article, I concluded:

In the public schools of New York, we have a situation where simple arithmetic shows that hundreds of thousands of dollars of revenue per classroom aren’t making it to the classroom; where the administrators are receiving high salaries, well up into six figures, and where there are literally more “non-teaching professionals” than teachers; where the students aren’t leaving school with the basic education they need, much less the quality education they were promised and for which the taxpayers are paying.

The modest proposal was a simple one: in place of the current public schools, we establish one-room schools in commercial office space, paying commercial rates, establishing the infrastructure including desks and chairs, tables, computers for each student and for the teacher, along with whiteboards and projection for video and and high-quality audio. To pay for this, we invest exactly what public schools are paying per student, and once we pay for all this rather palatial infrastructure, we pay the teachers with the remainder.

For the details as of this, you can refer to Who Benefits From The Public Schools, but the results come out to be that teachers could make upwards of $300,000 for a 9-month school year.

One expects the competition for school teacher jobs would be much greater than it is today.

I will be expanding on this in future articles, but the basic ideas are:

One-Room Schools

Organize schools with small class sizes, mixing students from first to eighth grade. This offers younger students role models, while older students deepen their understanding by teaching peers. (Teaching a topic is when you really learn it.)

Individualized Instruction.

The "modern" school system approach of a teacher lecturing to a room full of students and periodically testing them actually dates back to the Prussian school systems instituted by Frederick the Great in 1763. They were a massive advance at the time, but they were primarily intended to educate rural Prussians to be basically literate and better suited for the factory jobs that were becoming the new standard.

The whole model of these schools was essentially industrial: the students were workpieces moving through a series of workstations until they were delivered — "graduated" — ready to take their place in the factories of industrializing Prussia. Some proportion of these workpiece students were expected to fail, and they were discarded.

To be fair, this really was an improvement over schooling at the time, and there were plenty of places for students who weren't as capable in farms and labor. But this isn't the 18th century, and there aren't nearly as many jobs for farm hands and muleskinners as there once were.

But many of the problems with this model today are the consequences of expecting today's students to be uniform raw material for the education factory, and the discard rate is unacceptably high.

Kumon Education

After the end of World War II, Japan saw a need for Japanese students to be educated to be competitive with the rest of the world. In 1954, Toru Kumon developed an approach that emphasized individual, self-paced education with mastery of a topic as the goal. In Kumon's methods, students proceed through a sequence of worksheets, working individually but with regular individual assistance from the teacher as needed.

This method has two problems of scale: first, it requires a lot of worksheets and can put a lot of demands on the teacher's time. Second, for students who are having difficulty, it's hard for a collection of predefined worksheets to adapt.

AI-Assisted Learning

Fortunately, technology has come a long way since 1954. There is a lot of research going on using an AI tutor to make new worksheets on demand, and to adapt to the issues a struggling student may have — or alternatively, to let a student who is finding a topic easier to advance more quickly. There's really nothing conceptually new to this, going back to P. L. Plessey's mechanical teaching machine in 1924, and followed by B.F. Skinner's work with teaching machines in the 1950s and 1960s.

The challenge with teaching machines is similar to the challenge with Kumon — it takes a lot of work to develop the material. Kumon and programmed instruction don’t just teach; they engineer learning step by step, creating a mountain of materials where a textbook would only need a hill.

But using large language models (LLMs)  and reasoning systems — that is, AI that solves a logical problem instead of, as with an LLM, mimicking human interaction based on a statistical model — offer the opportunity to build a system that can adaptively generate worksheets and lessons in real time.

Every Student an A Student

Letter grades, the familiar ABCDF, are inherently suited to the Prussian factory model — a workpiece meets standards or is classified as substandard, with little room for variation. But, as we've discussed above, it's not clear that discarding the inferior workpieces is a desirable outcome.

Students, I mean, not workpieces.

There are well-known approaches that foresee a student working for mastery. It's actually not a new idea. For many generations, skills like blacksmithing, candle making, or printing were taught by a master of the skill taking on an apprentice who eventually was able to perform the skill at a level to be able to work independently — a journeyman — and eventually to have mastered the skill, qualifying him to take on apprentices of his own.

The key here is mastery.

In 1984, Benjamin Bloom published “The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring” (Educational Researcher, vol. 13, no. 6, 1984, pp. 4–16) on the "Two Sigma Problem." What Bloom observed was that students working somewhat independently, with periodic one-on-one tutoring, were achieving results two standard deviations — "two sigmas" — above a control group. That is, mapping this to letter grades, students were achieving solid As instead of gentleman's Cs.

The problem, as Bloom observed, is that this also has scaling issues: how many instructors would be needed to implement a two-sigmas approach?

Again, AI offers a technology that could be a big advantage. With adaptive AI-based lessons, students would be able to work more independently, for longer.

Now, this doesn't mean that the need for skilled instructors would be eliminated. But it would mean that their time could be devoted to students who were running into real difficulties.

Growth Mindset

There is one more aspect that would change the direction of education in a very useful way.  In 2006, Carol Dweck published the book "Mindset: The New Psychology of Success." I recommend the book to anyone interested in teaching, and to adults who are helping educate children — their own or others. But the basic insight can be summarized in one sentence: A growth mindset means believing you can learn, so you will learn.

Individual Intervention

There is one more issue that needs to be addressed. Individual students can have specific challenges: specific learning disabilities, like dyslexia; perceptual issues, like poor eyesight or impaired hearing; or a language barrier, like limited English fluency.

These issues might need to be dealt with by specialists who can be added on a consulting basis, or by identifying teachers already skilled with these issues. The one thing that is needed is to not discard these students. But by eliminating the massive drain on allocated resources coming from the administration-heavy current factory model of education, there can be resources available to cope with specific challenges.

Also for our VIPs: Why Governments and Other Big Organizations Act Like Idiots

Conclusion

American public education is broken, churning out students unprepared for life while squandering billions on bloated administration and outdated methods. The solution isn’t more money — it’s a radical rethink. By reviving one-room schools, prioritizing individualized instruction, leveraging AI-driven learning, and fostering a growth mindset, we can aim for every student to master the Three R’s and beyond, not just scrape by with a C. Bloom’s Two Sigma Problem shows us what’s possible: A-level outcomes for all, not just the lucky few. The tools — Kumon’s mastery focus, AI’s adaptability, Dweck’s mindset — exist. We just need the courage to dismantle the factory model and build something better.

This is only the beginning. In future articles, I’ll explore how to implement one-room schools practically, from funding to teacher training; dive deeper into AI’s role in scaling personalized education; and address how to support students with unique challenges without breaking the bank. The goal? An education system where no student is discarded, and every student can become an A student. Join me as we refactor education for the 21st century.