The question every school is asking, and why the answer is more specific than yes or no

By Chris Calder, AI Education Consultant | Meta Pedagogy

The wrong question

“Should we ban AI?” is the question most schools are asking. It is the wrong question because it treats AI as one thing. It is not. A student using AI to generate an essay they submit as their own work is doing something fundamentally different from a student using AI to challenge their reasoning after they have written an argument themselves. Banning AI treats both of those students the same way.

The useful question is: is the student using AI to think, or to avoid thinking?

Thinking vs churning

When a student uses AI to skip the cognitive work, they are churning. The output looks right. The essay reads well. The answers are plausible.  However, the student has not built the understanding that the work is supposed to represent. They have produced something without learning anything.

When a student uses AI after doing their own thinking, as a way to test their reasoning, fill gaps in their understanding or push their argument further, they are thinking. The AI is doing what a good teacher does in a one-to-one conversation: asking questions, challenging assumptions, pointing out weaknesses.

The distinction matters because the first use undermines learning and the second supports it. A blanket ban eliminates both. A blanket permission enables both. Neither policy is fit for purpose.

Why bans fail

Schools that ban AI face three practical problems.

Students are using AI anyway. A ban that cannot be enforced is a policy that teaches students to hide what they are doing rather than use AI responsibly. Every school that has tried a blanket ban has found that usage continues, it just moves underground.

A ban removes the opportunity to teach responsible use. Students will leave school and enter workplaces where AI is a standard tool. A school that bans AI entirely sends students into that environment with no framework for using it well and no understanding of where it fails.

A ban does not address the underlying assessment problem. If a piece of homework can be completed by AI without the student understanding anything, the issue is the assessment design, not the AI. Banning AI protects a broken assessment model rather than fixing it.

Why open access fails too

Allowing unrestricted AI use creates a different set of problems. Students who use AI to generate work without doing the thinking accumulate what the Cognitive Debt Framework calls cognitive debt: a cumulative deficit that builds when students repeatedly produce correct outputs without constructing the underlying understanding.

The student may not know it is happening. The work looks finished and the answers look right. The debt compounds over time until the student cannot build on what they appear to have learned. A knowledge gap is visible and a teacher can address it directly. Cognitive debt is hidden by the quality of the output.

This is the real risk of unrestricted AI access in schools. It is not plagiarism in the traditional sense. It is students who genuinely believe they understand something because they have produced work that looks like understanding, when in fact the AI did the cognitive work for them.

The Cognitive Debt Framework

The Cognitive Debt Framework is a five-stage model designed to prevent AI from replacing student thinking. At each stage the student does the cognitive work first and AI plays a specific, limited role.

At Stage 1 (Surface) the student states what they already know before any teaching. AI only asks probing questions to surface misconceptions. It does not correct or teach. The starting position stays the student’s own.

At Stage 2 (Struggle) the student attempts the problem and gets stuck. AI gives hints, not answers. It provides only enough support to keep the student thinking, within the zone of proximal development. If AI gives more than that, it has replaced the cognitive work.

At Stage 3 (Speak) the student explains their reasoning to a peer first. Verbal articulation forces thinking into the open. AI then pressure-tests the argument after peer discussion, finding weak points and challenging assumptions.

At Stage 4 (Show) the student explains how their thinking changed from Stage 1 to the present. AI reflects the student’s own thinking back to them without supplying the answer. The student accounts for the journey, not just the destination.

At Stage 5 (Spiral) the student returns to the concept weeks later at a harder level. AI retrieves earlier work and helps connect past and present understanding. If AI ends up re-teaching the concept instead of building on it, the original learning did not stick and the debt is visible.

The debt test: if a student can produce an answer but cannot do stages 1, 3 or 4 out loud, the thinking was offloaded and the debt is accumulating.

What schools should do instead of banning

Design tasks where AI use is structured, not prohibited. The protection comes from task design, not from policing tool access. A well-designed task makes it difficult to skip the thinking because the thinking is built into the structure of the task itself.

Make the distinction between thinking and churning explicit to students. Students who understand why the distinction matters are more likely to use AI in ways that support their learning. Students who are simply told “do not use AI” without understanding the reasoning will work around the rule.

Use the Cognitive Debt Framework or a similar structure to define where AI sits in the learning process. If staff know that AI is allowed at Stage 2 but not at Stage 1, and that Stage 3 requires verbal explanation before AI challenge, the policy has specificity. A policy that says “use AI responsibly” has none.

Train staff to recognise cognitive debt. The signs are students who produce good written work but cannot explain it verbally, students who struggle with follow-up questions on topics they have apparently mastered, and students whose performance drops under exam conditions compared to coursework.

The short answer

Should schools ban AI? No. Should schools allow unrestricted AI use? Also no. Schools should define where AI supports thinking and where it replaces thinking, build that distinction into task design and make it visible to staff and students. The Cognitive Debt Framework provides one structure for doing this.

If your school is working through this question, book a free consultation with Meta Pedagogy.  You can also download the Cognitive Debt Framework here.