Schools Are Still Treating AI Like a Technology Problem

Most conversations about AI in schools begin in familiar places: infrastructure, privacy, policy, assessment, professional development, platform selection, and academic integrity.

Those are legitimate concerns.

They are also downstream of something schools have spent less time examining.

AI is entering the learning environment at the level of habit.

Students are developing patterns around when they turn to it, what they hand over, how long they remain with uncertainty, what kinds of questions they ask, and which parts of the learning process still feel like their responsibility.

Those patterns begin forming long before they appear in a policy document.

So while AI clearly creates technological and operational questions, schools are also facing a developmental and cultural challenge:

What capacities should students continue developing as increasingly capable systems become available to support them?

The Questions Reveal the Frame

Listen to the questions that tend to dominate institutional conversations about AI.

Which tools should we approve?

How do we regulate use?

How do we protect student data?

How should teachers disclose AI use?

How do we detect misuse?

What should our policy say?

How should assessment change?

Each question has a place.

Together, though, they reveal how schools are interpreting the problem. AI is something entering the institution, so the institution responds by deciding how to govern it.

What receives less attention is what repeated use may be doing to the learner.

A policy can establish whether AI is permitted on an assignment.

It cannot decide how quickly a student turns to AI when confused.

A platform agreement can address privacy.

It cannot determine whether a student still practices forming an argument before asking for assistance.

An assessment redesign can make shortcutting more difficult.

It cannot, by itself, build judgment.

Those are questions of practice and development.

The Same Tool Can Produce Very Different Learners

Imagine two students using the same AI system for the same assignment.

One encounters uncertainty and immediately asks AI what to do. The response is plausible, so the student follows it.

The other begins independently, reaches a point of uncertainty, and then brings AI into the process. They question the response, compare it with what they already understand, reject part of it, refine another part, and make the final decision themselves.

The technology is identical.

The cognitive responsibility is not.

Over time, those differences accumulate.

One student may become increasingly skilled at obtaining useful output.

The other may also be developing judgment about when support is useful, when it is premature, what deserves skepticism, and what thinking still belongs to them.

Both may produce strong work.

Only the process tells us what capacities are being strengthened along the way.

That is where the educational question becomes more complicated than tool access.

Support and Substitution Can Look Similar From the Outside

AI can remove friction from learning.

Sometimes that is beneficial.

A multilingual student may use it to clarify unfamiliar phrasing. A student struggling to organize a large project may use it to compare possible approaches. A learner who cannot see why an explanation is not making sense may ask for another representation.

In each case, AI can make more cognitive space available for the work that actually deserves attention.

But the same capability can also remove the very thinking a student still needs to practice.

Planning becomes delegation.

Clarification becomes answer-seeking.

Feedback becomes rewriting.

Brainstorming becomes idea generation without prior thought.

The visible interaction may look almost identical.

The difference sits in what responsibility remains with the learner.

That makes developmental judgment more useful than simple categories of allowed and prohibited use.

The question is not only whether AI was involved.

It is what the student was still responsible for doing.

Schools Have Always Shaped More Than Academic Output

Schools have never been only places where information is transferred.

Students also learn how to persist through confusion, communicate unfinished thinking, make decisions with incomplete information, revise after feedback, collaborate across differences, organize themselves, and decide when they need help.

Much of that development happens indirectly.

A difficult assignment teaches more than its content.

A group project develops capacities beyond the final presentation.

Waiting before receiving an answer can be part of learning.

So can explanation, modeling, guided practice, conversation, experimentation, and appropriate support.

AI changes the conditions surrounding all of those experiences.

It can extend them.

It can also quietly bypass them.

That does not make AI uniquely harmful. It makes intentional design more necessary.

The availability of support has changed.

Schools now have to become more explicit about which forms of cognitive responsibility they are trying to preserve at different stages of development.

Habits Form Before Institutions Reach Consensus

Students are not waiting for schools to settle the AI question.

They are already developing routines.

Some use AI only when a teacher permits it. Others use it across homework, revision, studying, planning, research, and everyday questions. Some approach it cautiously. Others reach for it automatically.

Repeated use begins creating defaults.

When uncertainty appears, do I stay with it?

When I have an idea, do I develop it before seeking alternatives?

When AI gives me something polished, do I interrogate it?

When I am capable of doing something myself, do I still practice doing it?

These are small decisions.

Together, they become habits.

And once enough individual habits become normal across a school community, they begin shaping culture.

That sequence is easy to miss when institutional attention begins with policy.

Policy Can Establish Boundaries. Culture Determines What Becomes Normal.

A school can write clear expectations around AI and still develop unhealthy practices.

Another can permit broad experimentation and develop thoughtful ones.

The difference often sits in the signals surrounding use.

What do teachers model?

What do assignments reward?

What happens when students admit uncertainty?

Is speed consistently valued over process?

Are students expected to explain their reasoning?

Do adults distinguish between support that builds capacity and support that replaces it?

Does the school share enough language that students encounter roughly coherent expectations from one classroom to the next?

Culture forms through those repeated experiences.

AI does not create that culture by itself.

It enters one that already exists and begins interacting with it.

In environments where completion is the dominant signal of success, AI makes completion easier.

Where explanation, judgment, revision, and intellectual ownership are visible parts of learning, AI can support those practices too.

The Work Is Larger Than Technology

Schools will continue making decisions about platforms, privacy, policy, assessment, and professional learning.

They should.

But those decisions become more coherent when they sit underneath a clearer developmental question:

What should AI support here, and what still needs to remain with the learner?

The answer will not be the same for every age, subject, task, or student.

Nor should it be.

As student judgment develops, the nature of appropriate support can change with it.

That moves the institutional conversation beyond whether AI should be present and toward something schools already understand well: how to scaffold development without prematurely removing responsibility.

AI may be new.

The educational work underneath it is not.

Schools are still trying to help young people become capable of thinking, deciding, creating, questioning, and acting with increasing independence.

The challenge now is making sure increasingly capable support strengthens that development rather than quietly replacing parts of it before anyone notices.

By Carl Murray