To My Children: The Quran, Effort, and the Machine That Answers for You

My children,

The other evening, one of you came toward me, notebook under your arm and a tablet in hand, with the look of someone who had just found a treasure.

“Look, Dad: I gave it the exercise, and it did everything.”

In a few seconds, the machine had lined up the answers, neat, correct, with not a single erasure. You were waiting for me to marvel.

I stood silent for a moment, and I think you did not understand.

I owe you an explanation. Here it is.

There are two verses in the Qur’an that I would like you to carry with you all your life:

These verses speak first of man’s responsibility before God: each is answerable for what he has accomplished. They should not be read as a lesson in pedagogy about schoolwork.

But they also carry a word I would like you to retain: sa‘y, the effort, the process, the direction toward which one’s action tends.

And that reminds me of a very simple truth: in learning too, no one can do the inner work that changes you for you.

The exercise from the other evening was finished, but it had not been done by you.

The notebook was neat, but your mind had hardly worked.

You had obtained the answer without going through the path.

Whereas, when we learn, the path sometimes matters more than the answer.

Think of someone who wants to become stronger. He will not grow his muscles by watching someone lift the weights for him. He can watch to understand the gesture. He can ask for guidance. He can be corrected. But at some point, it is up to him to bear the load.

For the mind, there is something similar.

A mathematics exercise was not invented merely to fill a box. It exists to force you to search, hesitate, trip up, start again, and ultimately understand something you could not do before.

If the machine does all this work for you, the assignment advances.

But you, far less.

What an experiment conducted with nearly a thousand high school students shows

And now, we have scientific data that allow us to measure this risk.

In 2025, researchers published in the Proceedings of the National Academy of Sciences (PNAS) a large experiment conducted with nearly a thousand high school students in a school in Turkey.

During mathematics sessions, some students worked without artificial intelligence. Others had access to a tool based on GPT-4 and close to a standard conversational assistant. A third group used a different version, designed with pedagogical safeguards to provide hints, signal errors, and accompany reasoning rather than immediately provide the solution.

The results are striking.

During the exercises, students using the classic AI achieved results about 48% higher than those in the group without AI.

With the AI designed as a true tutor, results during training were even 127% higher.

At first glance, one might conclude that the machine makes students much better.

But then came the exam.

This time, no one had access to the artificial intelligence.

Students who had freely used GPT-4 then achieved results 17% lower than those who had worked without AI.

They had done better when the machine was present, but learned less to succeed when it disappeared.

The researchers observed why: many simply asked for the answer and copied it. The tool had become a crutch.

By contrast, among those who had used the tutor designed to give hints rather than solutions, this drop almost disappeared. In the exam without AI, their results were not significantly different from those of the group that had worked without artificial intelligence.

We must be precise: this second method did not make them better than the others on the exam. It mainly preserved the benefit of the aid during training without paying later the price of dependence.

That is why I do not want to teach you to fear artificial intelligence.

I want to teach you to use it.

The effort that makes you grow

Researchers who study memory have long known of a strange phenomenon.

The methods that give the impression of learning the easiest are not always the ones that help you retain best.

Re-reading again and again can give an impression of mastery. Recognizing an answer when someone shows it to you can give the impression that we knew it.

But trying to recall an answer from memory, testing yourself, producing your own solution, or doing an exercise again without looking forces the brain to work harder — and this effort can strengthen learning.

Psychologists sometimes speak of “desirable difficulties.”

The important word is not “difficulties.”

It’s desirable.

It’s not about making life unnecessarily hard for the learner. A child who is completely lost does not learn better by being left to suffer over his sheet.

The useful effort is one that is within reach: sufficiently challenging to force reflection, but not so hard as to discourage or prevent understanding.

And it is precisely here that artificial intelligence can become extraordinary.

Not when it thinks for you.

But when it helps you think.

So I am not asking you to flee these machines.

You will grow with them. You will study with them. You will probably work with them. They will allow you to learn things that my generation found far harder to access.

I use them myself.

But I would like you to remember how the first revealed verses begin, according to the Muslim tradition:

Read.

Or recite.

In either case, there is an act to be performed.

Someone can read a text in front of you. A machine can summarize it. It can explain a difficult word. It can even produce ten interpretations in a few seconds.

But no one can understand for you.

A few verses later:

I do not want to make this verse say what it does not say.

The Qur’an does not give us here a scientific theory stating that we necessarily learn better by writing by hand.

But what a magnificent image: knowledge taught, transmitted, recorded, preserved.

The calamus is a tool.

Today, the tablet, the computer, and artificial intelligence are others.

The question is perhaps not: which tool do you use?

The real question is: who learns?

You, or the tool?

That is why, in my eyes, a notebook full of scribbles can be worth more than a flawless page from which you cannot explain a single line.

Three rules for using AI without giving it your brain

I would like us to adopt three simple rules at home.

The first is the rule of the first attempt.

Before asking anything of the machine, try it yourself.

Not necessarily for an hour. A few minutes can be enough.

Write down what you understand. Look for a lead. Do a calculation. Formulate a hypothesis.

Even a mistake can become useful when it is truly yours, because you can then understand why it was wrong.

The second is the rule of the master.

When you seek the intelligence of artificial intelligence, ask it to teach you rather than to work for you.

Instead of: “Do this exercise.”

Try: “Don’t give me the answer. Help me with a first hint.”

Or: “Here’s my reasoning. Tell me where I’m wrong.”

Or: “Ask me one question at a time so I can find the solution myself.”

A good master does not seek to impress his student by showing that he can solve the problem.

He seeks to make his student able to solve it without him.

The third is the screen-off rule.

When AI has helped you, turn it off.

Then start again.

Without looking.

redo the calculation. Explain the lesson in your own words. Solve a similar exercise. Explain what you have just learned to your brother, to your sister, or even to me.

If you can explain it without the screen, something has passed from the machine to you.

Otherwise, the information was in front of your eyes.

And here is an invocation I would like you to know:

Look closely at what this invocation asks for.

Knowledge.

Not merely a correct answer.

Not a good grade obtained by any means.

Not an immaculate page.

Knowledge.

An answer can appear on a screen in one second.

Knowledge asks for more.

It requires that something enter you, link to what you already know, resist forgetting, and emerge again when you need it.

That is what remains when the screen is off.

One day, my children, you will face a problem for which no one will have written the right prompt.

That may be an exam.

Perhaps your first job.

Perhaps a difficult decision.

Perhaps a mistake you will have made.

Perhaps someone you love and whom you will have to help.

Life contains many questions that do not have a single answer at the bottom of the page.

That day, AI may perhaps advise you.

But it cannot be brave for you.

It cannot decide your values.

It cannot assume the consequences of your choices.

And it cannot replace what you will have built within yourself, effort after effort, mistake after mistake, understanding after understanding.

So use the machine.

Use it a lot, even.

But make it a helpful master, never a replacement.

Let it help you go further.

Never to stop walking.

Because the goal of school has never been to produce notebooks without mistakes.

The aim is to build human beings capable, one day, of moving forward without someone whispering the answer.

Your father.

Scientific source: Bastani et al., Generative AI without guardrails can harm learning: Evidence from high school mathematics, PNAS, 2025. See study.