The Essay Was Never the Point

AI can now produce the work. Schools still have to build the mind that can judge it.

I've spent my career in games and media. I also help govern a school.

But the sharpest education conversations I've had lately were at my own dinner table.

My youngest daughter was voted by her classmates onto her high school's Honor Council, the peer-led panel that hears and adjudicates academic-integrity disputes. It's the kind of role you're proud of your kid for earning, right up until you watched what it required of her.

The disputes reaching her generation aren't the ones honor codes were written for. Increasingly, they involve AI. Work that may or may not be the student's own, produced with help that may or may not be against rules that may or may not exist.

A teenager is being asked to draw a line that universities, school districts and legislatures haven't managed to draw.

Last episode I wrote about the competition for our kids' attention. AI creates a different problem.

It doesn't just compete for a student's attention. Increasingly, it offers to do the thing we wanted the student to do with that attention.

It writes. It summarizes. It codes. It produces essays, lab reports, study guides and passable first drafts of almost anything, instantly, politely, and with the confidence of an intern on day three.

The usual question is how much better AI will get.

It will get better.

The more interesting question for education is what becomes valuable when producing a plausible answer is no longer particularly difficult.

The finished essay was never the whole point

Schools have always faced a measurement problem. Thinking happens inside a student's head.

So we look for evidence.

An essay. A problem set. A lab report.

But those artifacts have always done two jobs at once. The finished work showed us something about what a student understood. And producing the work was one of the ways the student developed that understanding in the first place.

Writing an argument helps form the argument. Solving the problem builds the ability to solve the next one.

AI can now separate those two things.

A polished paragraph doesn't necessarily tell us who did the thinking behind it. A correct answer doesn't necessarily tell us who reasoned toward it. A finished assignment can increasingly exist without the student having experienced much of the difficulty the assignment was designed to create.

That's more consequential than cheating.

Cheating is old.

What's new is the possibility of producing genuinely good work while bypassing the practice that was supposed to make the student better.

The kids are already there

This isn't a future policy problem. Pew reported earlier this year that 54 percent of American teens have used chatbots for schoolwork.

RAND found something even more interesting. In its December survey, 67 percent of students agreed that the more students use AI for schoolwork, the more it will harm their critical-thinking skills.

Students are adopting the tool while simultaneously worrying about what the tool may be doing to them.

We should take both halves seriously.

Trying to keep AI out of education entirely looks increasingly unrealistic.

Pretending every use of it is educational looks equally unserious.

The question isn't whether they used AI

RAND offers a useful distinction between cognitive augmentation and cognitive offloading.

Used one way, AI can make the student work harder intellectually.

Challenge this argument.

Give me the strongest counterexample.

Quiz me until I can explain this without notes.

Tell me what assumption I'm making that may not survive contact with the evidence.

Used another way, the same technology removes the difficult part.

Write this.

Solve this.

Summarize this so I don't have to read it.

Tell me what I think.

One randomized study gave the risk a memorable name, metacognitive laziness. Students supported by ChatGPT improved the essays they produced, but the researchers found no corresponding advantage in knowledge gain or transfer.

That's the distinction schools should care about. The useful question isn't “Did they use AI?” It's “What did they still have to do?”

And after they used it, can they do more without it, or less?

What stays scarce

I've spent a career in businesses where the cost of making things kept falling.

That never made it easier to know what was worth making.

AI is now doing something similar to intellectual production. It can generate a competent draft almost instantly.

But abundance creates its own problem.

Which answer is actually good?

What's missing?

Which assumptions did it smuggle in?

What would a smart critic attack?

What happens when this recommendation meets the real world?

As producing a competent answer gets easier, judging one gets more important.

And judgment isn't a substitute for knowledge.

Judgment is what knowledge becomes after enough practice.

A student with no history can't reliably recognize bad history because a chatbot delivered it fluently. A student who has never learned to write has a harder time hearing what's wrong with machine-generated prose. A student who has never worked through the problem is poorly equipped to supervise a machine solving one.

We shouldn't expect students to oversee reasoning they have never learned to perform.

The better AI gets at doing novice work, the more intentional schools will have to become about preserving the experiences that eventually create expertise.

What I'd ask

When a vendor or a teacher describes how AI fits into schoolwork, or when my daughter thought out loud about a hard case, these are the questions I find myself reaching for:

1. Where does the student still have to struggle?

2. Does AI enter before or after the student has attempted the problem?

3. Can the student explain and defend the result without the machine?

4. Are they interrogating the answer or merely accepting it?

5. What knowledge or skill is being practiced that will still be theirs when the tool is gone?

6. After a month of working this way, can they do more without AI — or less?

That last question is the same one I asked last episode, translated. What has the product been optimized to produce?

What school is still for

None of this is an argument against AI.

I use it almost every day. It can be an extraordinary tool.

It's an argument for being much clearer about what education is trying to produce.

Last episode I argued that a school's job isn't to control a student's attention forever. It's to develop a young person capable of deciding what deserves it.

AI adds another requirement.

They also have to decide which answers deserve to be believed.

That requires knowledge. Skepticism. Practice. The ability to recognize when something sounds convincing but isn't.

And, inconveniently, it requires doing some things the hard way before technology makes them easy.

That's why Papert's idea of hard fun keeps getting more useful.

He also named the trap. In 1987 he called it technocentrism, the mistake of asking what the computer does to the child instead of what the child does with the computer. Swap in AI and the warning hasn't aged a day.

The struggle was never an inefficiency we needed to engineer out of education.

Sometimes the struggle is the education.

As for the Honor Council, my daughter graduated this spring, and the dinner-table seminar went with her.

What encourages me is that the panel she served on still gets the hard cases. Real stakes, incomplete rules, no answer key. In other words, the students have to exercise judgment.

If you teach, parent, build AI tools or govern a school, I'd like to hear how this is playing out. Where is AI making students more capable, and where has the thinking quietly left the building?

Further reading: RAND, More Students Use AI for Homework, and More Believe It Harms Critical Thinking; Pew Research Center, How Teens Use and View AI; Yizhou Fan et al., “Beware of Metacognitive Laziness”; OECD, Digital Education Outlook 2026; Ted Chiang, “ChatGPT Is a Blurry JPEG of the Web”, The New Yorker.