CXXXV. The Teacher or The Machine?

My youngest son had a teacher in elementary school who could really read the room. Twenty-eight kids, all at different speeds, all from different homes, all having different kind of days. Five days a week. She knew who excelled, who needed more time, and who to keep the eyes in the back of her head on. She knew who had eaten breakfast, who was slow to warm up, and who was too quick to please.

None of that lived in a database, and none of it ever will.

We are about to hand teachers one of the most powerful tools ever invented. The only question that matters is whether we plan for it or just let it land and hope everything works out.

Here is what artificial intelligence can finally do for education. It can meet each student where the student actually is. It can retire the fiction of the "average student." It can give the kid who needs three more examples exactly that in a language they understand, and for the kid who is bored with a harder problem, it can gamify it until he or she gets it. This can all happen at the same time, in the same room. Done correctly, fewer kids will slip through the cracks. Done wrong, more kids will continue to not keep up and teachers will have yet another expectation in a room full of expectations.

We all know that direct individual teaching is a gold standard. In 1984, Benjamin Bloom showed that a student tutored one-to-one performed two standard deviations above a student in a normal class. The catch, in Bloom's own words, was that one-to-one tutoring is "too costly for most societies to bear." Group teaching has been the business method of managing more kids at a reduced cost for a very long time. Those that don’t cut it? That cost in lost opportunity, potential crime, and illness gets distributed throughout the community.

AI can truly revolutionize education by assisting both teacher and student.

A World Bank trial in Nigeria gave secondary students six weeks of AI tutoring guided by their teachers, and the students gained what normally takes close to two years. A recent U.S. randomized controlled trial found something similar: students using an AI tutor learned more than twice as much, in less time, than their peers in one of Harvard's best active-learning classrooms.

So that is the promise. Here is the danger.

The same tool that can tutor a child can also do the child's work for him. Researchers ran the experiment. In a study published in the Proceedings of the National Academy of Sciences, students given an ordinary chatbot blew through their practice problems, then scored worse than their peers once the chatbot was taken away. They had used it as a crutch. A second group, given a version built with guardrails to coach rather than answer, held their ground. Same technology. Opposite outcome. The difference was entirely in the design and the supervision, which is to say, the difference was the teacher.

That is the whole argument in a sentence.

Treated as a teaching assistant, AI is one of the great developments in the history of the profession. Treated as a replacement, it accelerates exactly the social disconnection and skill erosion we are already drowning in. AI can do any worksheet given to it, but it cannot notice who may be sick, who may have arrived with an empty stomach, or who is the quiet kid in the third row that stopped raising his hand because something happened at home. We keep confusing the delivery of content and quantitative outcomes with the work of actual teaching. It's much more than that.

Which makes the timing of all of this disastrously bad.

We are handing schools the most consequential instrument in a generation at the exact moment we have buffoons in charge of education and are running teachers ragged with ideology and procedures. Roughly one in eight teaching positions in this country is now either empty or filled by someone not fully certified for the job. The drumbeat against teachers, education, and public schooling has never been louder, and there will be vested interests willing to profit from less teachers in the classrooms.

So what should a new generation of teachers be doing to get ready? What should we be doing as a society?

The opposite of panicking. Stop trying to beat the machine at the things AI will simply be better at, which is the drill work, the leveling of material to different learners, and the instant feedback at late hours of the evening. AI can assist with all of that. Lean into the human side of reading the room. The teacher of the next decade will be less a lecturer and more a maestro, building the conditions in which a kid and a very powerful tool can do something neither could do alone.

And what does society owe the teacher in return? Everything we have been too cheap to provide. Pay them like leaders of the community. Train them on these tools instead of dropping them in their lap, or worse, ignoring them or trying to replace them. Fund the guardrails so the AI shows up as a tutor and not a crutch.

My son's teacher never needed a dashboard to know her kids. No system will ever replace that, and no system should have to. The classroom isn't broken. It’s the slapdash way we treat teachers and students as pieces.

Teachers want to teach. Kids want to be inspired and stimulated. Forty years ago, Bloom proved what every child deserves, and we decided we couldn't afford it. Now, the individualized tutor is here. The only thing we have left to do is give teachers the support and the training and the creative freedom to do it right.

Estonia is a small country with a long border with Russia and a clear-eyed sense of what it cannot afford to get wrong. So it is telling that this NATO ally is moving to put American AI tools directly into its classrooms, and treating how students learn as a matter of national readiness rather than a passing tech fad. Compare that to here, where most districts are still arguing about phones, let alone artificial intelligence, and you see some of the problems I’ve witnessed for an entire career in public service. People defend the procedure they are comfortable with instead of asking whether it produces the outcome they claim to want. AI in education is not going away. A country the size of one American county is treating that as obvious. We should, too.

Jeff Bezos isn’t done just with Amazon. He’s put some of his time and funding into building a chip that runs the way a mind does. Read that again. We are trying to reverse engineer the most phenomenal thing we know of, and we are close enough that it is worth a billionaire's fortune to try. The promise is staggering. Computers that reason the way we reason, and maybe one day machines we run by thought alone. I’m naturally optimistic with tech, but I’m not blind to the dangers. A tool this powerful could have both the pros and cons of being human. While we are creative creations, capable of building something resembling ourselves and how we think, we are also selfish. Can we be decent enough to deserve what we make?

The Senses

The memory does slowly fade
Yet, much will remain remembered
Persistent views from a child's past
First loves, and teenage adventures

That first job, a marriage union.
The joyful birth of a newborn.
Some snapshots forever retained.
The vision weakens, the light dims

A closer look for certainty
The eye is too easily fooled
Alert observance is the key
For an accurate perception
An eye may drift from reality
What is seen may not be what is.

Less is heard in a smaller world
Muted voices replace clear sound
Many echoes, and missed chatter
More distant sound not heard at all
Something of much importance lost
There may be a valid concern
So often, just valueless noise.

Taste, smell, and touch will diminish
No major aging problem there
Awareness, and wisdom continue
Too often cast quickly aside
A store of knowledge so needed
A life's history unheeded
The most valuable discarded.

In closing, my son told me a story that reminded me of how we let legalistic thinking creep into our ability to treat each other as humans. He was at a little league baseball game where he overheard the umpire saying the pay he had just received didn’t reflect the number of games he umpired. The response from the people giving him the pay was that he would need to take it up with someone else.

If there is a leadership lesson to be learned, give some respect to your power of discretion. If you trust someone as an umpire, you should trust his comments on his pay. If you’ve gotten an education, use it to discern the correct answer. It is better to treat people as you would want to be treated in these cases, and not use the shortcut of deflecting an issue that causes more work for someone else and for the ump. Take a leap of faith that people can be honest and have integrity.

Keep plugging along. Enjoy your summer. Get some sun. Carry your book your reading with you. Always strive to be a good human.