This is a polished English translation of an Icelandic opinion piece originally published in Visir.

By Helgi S. Karlsson, Behavior Analyst, Psychologist, and Teacher

A teacher sits with a writing assignment that is well written. Perhaps too well. The text is clean, the arguments are neatly arranged, and the language may even be better than the teacher expected. Still, a question many teachers now recognize begins to appear: did the student write this, or did they let artificial intelligence do it?

The next day the teacher asks the student to explain the thinking behind the text. Then it may become clear that the writing was much clearer than the understanding. Perhaps the student was not trying to deceive anyone. Perhaps they simply used a tool that felt natural to use. Still, the teacher is left with a new problem: what was I actually assessing?

Parents recognize this too. They ask whether their child is really learning anything, or merely letting AI do the work. Students ask in their own way: may I use this? When may I use it? Where is the line?

This is how AI has first entered school life: not as a distant future, but as daily uncertainty.

That uncertainty is understandable. But schools cannot solve it simply by turning away from the tool. AI has already entered learning, work, and ordinary life. The task is therefore not only to set boundaries, but to teach students to use it with responsibility, criticism, and understanding.

When uncertainty grows, we often reach for a simple word. In this case the word is usually "cheating."

That is understandable. But if we see this primarily as a cheating problem, we are only seeing the surface. The deeper question is not only: who wrote the text? It is whether the student understands, reflects, and can stand behind their own thinking.

That is why the discussion about AI in schools needs to move one level higher.

The real question is this:

How do we teach children to think in a world where they always have access to a machine that can answer almost everything?

We cannot build the school of the future on the hope that children will not use a tool society itself is already using. Nor can we pretend that this is only a new version of Google. A search engine finds information. AI can shape language, arguments, examples, explanations, and even an entire path of thought.

That changes the assignment. It changes the problem.

If a school assignment mainly asks students to retell information, then it has become a weaker measure of understanding than it used to be. If a student can ask AI to write the answer without needing to understand the material, the assignment no longer reliably shows what the student understands. It may measure access, technical skill, or boldness, but not necessarily understanding.

Schools therefore do not first need more surveillance. They need clearer teaching of thought.

I do not say this to minimize cheating as a real problem. Schools need rules, boundaries, and fair assessment. Sometimes a ban may be justified, especially with younger children, sensitive assessment, or when student privacy and rights are at stake.

But a ban by itself does not teach children to think better. It only moves the problem elsewhere.

A better foundation is this question:

Can the student show how the thinking came into being?

In my view, this is the next major step in schoolwork: moving the focus from the final text to the way the student thought their way toward the final answer.

Not only: what did you submit?

But also:

Here AI stops being a shortcut away from thinking and becomes a mirror for thinking. A student who uses it well should not merely submit a polished result. They should be able to show judgment, doubt, correction, and their own position.

The same applies to this article. In writing it, I used AI as a conversation and writing tool: to discuss the idea, sharpen the argument, compare wording, look for objections, and examine which sources mattered. That does not mean the thinking disappeared from the process. On the contrary, the tool made the demand for judgment clearer. I had to choose, reject, rephrase, verify sources, take responsibility for the position, and stand behind the final text.

That is the core. AI does not automatically become cheating when it enters a working process. It becomes a problem when it releases the student from thinking. Used well, however, it can become a powerful tool for reflection, criticism, and clearer thought.

An old and good school word now becomes central again: reflection.

Teachers have long spoken about reflection, but it has too often been treated as a beautiful word in a curriculum or a final question at the end of an assignment. Now reflection must be taught as a daily skill. Students must learn to ask better questions, doubt the first answer, compare arguments, change their mind, and explain why.

AI does not make this less important. It makes it more urgent.

The role of the teacher therefore does not shrink. It becomes more important.

The teacher of the future will not primarily be a monitor checking whether a text is "real." The teacher will be a coach of questions, a guardian of understanding, and the person who helps students see the difference between an answer that sounds good and an answer that survives reflection.

This does not have to be complicated. Schools can begin with three simple rules.

First, say clearly when AI may be used and why. Students do not only need rules. They need to understand the purpose of the rules.

Second, require students to show how their own thinking came into being when AI is used. Not to shame them, but to teach them. A student who cannot show that path has probably not yet reached enough understanding.

Third, assess understanding more often in living contexts, even in simple ways. Students can explain briefly, compare answers, revise an earlier answer, or use knowledge in a new situation. Text alone is no longer enough.

This is also fairer to students. Children who learn to use AI as a tool for better reflection become stronger. Children who learn to hide behind it become weaker. The difference is not in the tool itself, but in how we teach them to use it.

International guidance points in the same direction. UNESCO has emphasized a human-centered approach, ethical use, and competence for both students and teachers. The European Commission has published guidance for teachers on critical, responsible, and ethical use of AI and data in learning. The European Union's AI Act also shows that when AI begins to influence assessment, access to education, or monitoring of students, we enter a particularly sensitive risk area.

Icelandic authorities have also placed education, competence, digital citizenship, and standards for AI use in schools into the national AI action plan. That is good. But policy must become daily teaching.

We do not need to wait for a perfect system to begin well.

We can begin with a simple rule:

AI must not become a way for children to avoid thinking. It should help them see more clearly how they think.

If we do this badly, AI will weaken learning by making submission easier than understanding. If we do it well, it can help us teach deeper questions, clearer arguments, stronger judgment, and real reflection.

The future will not first ask whether students used AI. It will ask whether they learned to think with it without losing themselves.

I thank A. for deep conversation, critical review, and an important contribution to the structure of the frame.