Blaming AI for Destroying Education Makes Me Want to Scream

Every day, another op-ed lands in my feed: AI is destroying education. AI is killing critical thinking. Kids are cheating their way through school, and they’ll never learn to think.

It makes me want to scream. Not because nothing is breaking — something is. But because of what is breaking, and our complete unwillingness to say it out loud.

Here’s what’s actually true. We built an education system around information retrieval, and then we ranked children by who could retrieve the most, fastest. That’s the game.

And then — recently, conveniently, right about when the machines got good at the game — we started insisting that what we really value, above all else, is critical thinking.

That’s nonsense. And I can prove it in two words: tests and papers.

We say critical thinking. We measure retrieval.

Look at how we actually assess kids. The two instruments that decide their futures are the standardized test and the paper.

The test, in the overwhelming majority of cases, rewards critical thinking minimally or not at all. It rewards retrieval and pattern-matching against an answer key. The best human stochastic parrots win.

The paper does support thinking to a degree. But the paper is a monologue you write alone (and you better write it alone or you’ll get sent to the Honor Council), over days, with the entire internet open, every gap quietly patched and every weak point edited out before another human ever sees it. As a vehicle for the thing we claim to worship, it is, generously, mediocre.

If critical thinking were actually our highest value, we would have built the system around the activity that produces the most of it. We didn’t. We made that activity optional and scheduled it after school, next to some other club that gets even less support.

I mean debate.

Why debate beats the paper, plainly

I’m not being romantic about an activity I’ve spent my life in. The mechanics are just better:

• You can’t retrieve your way out. No lookup, no editing, no week to hide your gaps. You think in real time against a live opponent. That’s cognition under load — the only kind that’s actually hard.

• You have to argue the side you don’t believe. Switch-side debate forces you to build the strongest version of the position you walked in hating. You can’t cherry-pick one thesis and ignore the rest of reality.

• Your reasoning gets attacked immediately, by someone trying to break it. A paper gets one reader, days later. Debate stress-tests your argument in front of you, in the moment, by a person whose entire job is finding the weakest joint and snapping it.

• You have to listen, not just produce. You find the actual point of disagreement and refute that — not the strawman you wish they’d made. Nobody talks back to a paper.

• You have to adapt. New argument, hostile judge, two minutes on the clock. That’s judgment — weighing, prioritizing, deciding what matters under pressure.

• You can’t fake understanding. Fancy prose hides a shallow grasp on the page. One cross-examination question and the bluff is over.

Now hold the two things next to each other. The paper is the single assessment most easily done by AI. Debate is the one format AI cannot simply do for you, because at some point you are standing in the room, alone, thinking. And we made the paper central and debate optional. That is not an accident of scheduling. It is a confession about what we actually value.

The absurdity is recursive

Here’s where I lose it.

We built a retrieval curriculum. We called it critical thinking. AI trivially beats retrieval. So we panic that AI is destroying critical thinking, and our proposed fix is to pay AI companies to build Socratic tutors that will debate kids through the material.

Read that again. We are begging the machine to supply the dialogue, clash, and both-sides reasoning our curriculum deliberately left out, while still trying to steer students toward tidy answers. We outsourced the cure to the thing we’re blaming for the disease.

Ironically, we already have a real Socratic method that requires students to confront hard questions. It’s called debate. We defunded it.

What’s actually breaking

So let me say the quiet part. AI does not threaten critical thinking. It threatens the continued workability of an easy sorting function.

If retrieval is free, and every kid can win the content-curriculum Olympics with a phone in their pocket, then the contest can no longer rank anybody. And ranking was the product the whole time. The panic that gets dressed up as “kids won’t be able to think” is really “our easy instrument for sorting kids just broke.”

Follow the incentives. A high school’s and university’s prestige rides on test scores — SAT, AP, the standardized numbers that feed the rankings, that drive the U.S. News & World Report placement, that buy the reputation. A university gets rewarded for how well its admits played the content game in school before they became adults, not for how much thinking they did when they got there.

And then there’s the second lever, the one nobody says out loud nicely: exclusion.

The fewer kids you let in, the more prestige your educational institution gets. The smaller the percentage you admit to what is supposedly one of the greatest educations on Earth — best professors, best facilities, best environment — the more the world admires you for it.

Sit with how ass-backwards that is.

Prestige education is a positional good: its entire value comes from other people not having it. That’s why “we admit only X%” is a brag and not an admission of failure. If the goal were actually educating people — educating the whole child, whatever phrase is in the mission statement this year — scarcity would be the bug. In a positional-good system, scarcity is the product. We built a competition to see how many kids we could keep out, and we called it the best education in the world. And we say it’s for the good of the children.

And it never had to change

There’s a third gift the retrieval system handed the people who run it — the one that never makes it into a brochure. It lets you teach the same stuff, the same way, every damn year.

Same units, same worksheets, same canned labs, same test-prep packet. Only the delivery system had to improve.

The only innovation anyone was ever asked to produce was perfecting the standardized delivery. No reinvention. No risk. No reckoning with whether any of it built a mind.

You could run the identical machine for thirty years and get rewarded for running it a little more efficiently each time. A retrieval curriculum isn’t just easy to grade and easy to rank — it’s easy to repeat. That’s the part institutions love most and admit least. And AI threatens that too, because the moment you can’t pretend the content was the point, you have to ask what you were actually doing all those identical years.

There is no way to beat AI at the retrieval game, or at any quasi-standardized curriculum built around predictable process. The question is whether we are finally willing to stop playing it.

So stop screaming at the machine

We built a system whose three products are ranking, scarcity, and the right never to change — and we called it education. AI is dissolving all three. That’s the threat. Not thinking. Thinking was never the thing being measured.

If you genuinely cared about critical thinking, AI would be the best news you’ve had in a generation — because it just made the retrieval contest worthless, and it cannot touch the one thing that builds actual minds: kids in a room, arguing with each other, under pressure, with nowhere to hide.

So blame AI for destroying education, and I’ll keep screaming, because you’re mourning something that was never primarily about thinking. You’re mourning the standardized sorting machine.

Build the thing the machine can’t replace. Spend some time thinking about how to do it. Or admit what the system was always for.

There’s hope, and I felt it this week

Earlier this week, I was at an event where an AI walked a group of us through a critical-thinking exercise, and I have to tell you about it, because it’s this whole argument in miniature.

The setup: four people, each with their own wants, resources, and constraints. Some of those wants were compatible. Some were flatly mutually exclusive. All of them were legitimate. Every move you could make to help one person carried a cost for another, and there was no answer key — your job was to advise a group that actually had to decide. So you asked questions to pull out more information. Sometimes clarifying questions, sometimes challenging ones. You worked it through with other humans, out loud, in real time. You had to draw in facts and ideas and values and ethics all at once and make the best call you could in the moment, knowing it wasn’t clean. In other words: real human life.

And in that short stretch, I’m certain I did more thinking in 20 minutes than I’ve done writing most papers.

That’s why the MIT “cognitive debt” study is useful but incomplete. If you ask people to use AI to outsource the essay, of course, you may see less cognitive engagement. But that does not tell us what happens when AI is used to create dilemmas, pressure, dialogue, challenge, and uncertainty. Hook people up to the same machines during that exercise, and you might find the opposite story: not AI replacing thought, but AI provoking it.

That’s the hope. It was never AI versus your brain. It’s what you point the AI at. Aim it at the paper, and it’ll happily do your not-thinking for you. Aim it at the dilemma, make it build the impossible scenario, ask the hard question, and create a situation where there is no answer beyond what humans can reason through together, and it lights you up.

Same machine. Opposite job. The same machine that can flatten a paper into fake fluency can also generate dilemmas, force dialogue, create clash, and pull students into real judgment. AI is not the enemy of thinking. Bad assignments are. We can’t keep the old assignments and then rely on the technology to fix the pedagogy.

We spent a generation building a system that asked kids to retrieve answers so we could rank them and never change. We finally have a tool that can help run the Socratic seminar we always claimed we couldn’t afford: hard questions, no clean answers, humans thinking out loud together.

We could have done that without technology. We should have done it without technology. But if AI is what finally forces us to bring hard thinking back, I’ll take it.