The Classroom Caught Between Two Extremes

  

When Technology Becomes the Question, Not the Answer.

A school wants to prepare its students for an AI-driven future. Some teachers want stronger restrictions on technology. Parents want more AI exposure. Students are caught between the two. The institution now has to decide: should it teach students to use AI, teach them to work without it, or teach them when each approach is appropriate?

Read the case → 

CASE STUDY 03
Educational Management & Learning Design

The Case

A secondary school had begun discussing how it should prepare students for a world increasingly shaped by artificial intelligence.

Some teachers were already using AI tools to generate examples, create practice activities and provide students with additional explanations. Others were becoming increasingly concerned that students were using AI to complete work they should be doing themselves.

Parents were divided as well.

Some believed the school was not doing enough to prepare their children for the future. They wanted students to develop practical AI skills early.

Others were worried that children were becoming too dependent on screens and digital tools. They wanted more time spent reading, writing, discussing and solving problems without technology.

The school leadership therefore faced a difficult question:

What should the role of technology actually be in learning?

The debate quickly became polarised.

One group argued for stronger restrictions:

Students need to learn how to think without machines before they learn how to think with them.

Another argued for greater AI integration:

If AI is going to be part of their professional and everyday lives, students need to learn how to use it responsibly - not be protected from it.

Both arguments appeared reasonable.

But neither answered the deeper educational question.

• A student who can solve a problem without AI demonstrates one kind of capability.
• A student who can use AI to solve a more complex problem demonstrates another.

And a student who can recognise when NOT to trust an AI-generated answer demonstrates something different again.

The school realised that simply deciding whether to allow or ban AI would not solve the problem.

It needed to decide something much more fundamental:

What should students still be able to do independently, what should they learn to do with AI, and what should they learn about AI itself?

That question would eventually force the school to reconsider not only its technology policy, but also its curriculum, teaching practices and assessment methods.

-----------------------------------------------------------------------------------
The Problem

• Should they be restricted?
• Should teachers use it?
• Should assessments be completed without it?

It needed to separate the tool from the learning. 

The school's immediate problem appeared to be AI use. But that was only the surface.

The deeper problem was that the institution had not clearly defined 

• which capabilities students should develop independently, 
• which could be enhanced by technology, 
• and which required students to understand the technology itself.

Without that distinction, every discussion about AI became a debate about permission.

Should students be allowed to use it?

These were important questions, but they were not the most important ones.

The more fundamental question was:

What do we actually want students to learn?

If a student uses AI to produce an excellent answer, has the student demonstrated excellent learning?

If a student is prohibited from using AI and produces a weaker answer independently, has the student demonstrated greater learning?

Neither question has a simple answer. It depends on what the learning outcome is.

A student may need to demonstrate that they can reason independently, write clearly and solve fundamental problems without assistance. In another situation, the ability to use AI effectively, question its output and improve upon it may itself be an important competency.

The school therefore faced a problem much larger than technology policy:

AI could change how students learn, how teachers teach and how work is produced.

But the institution still had to decide what students should know, what they should be able to do, and how it could tell whether genuine learning had taken place.

Until those questions were answered, an AI policy alone could not solve the problem.

-----------------------------------------------------------------------------------

Three Different Kinds of Learning

One reason the AI debate becomes so difficult is that using AI is not the same thing as learning with AI. The school began to distinguish between three different capabilities.

1. Learning Without AI

There are things students should still be able to do independently.

• Read.
• Write.
• Reason.
• Calculate.
• Explain.
• Solve problems.
• Make decisions.

These capabilities provide the foundation on which other forms of learning can build.

2. Learning With AI

There are also situations where AI can extend what a learner is able to do.

Students might use AI to generate alternative explanations, receive feedback, explore ideas, practise a skill or work through a difficult problem.

Here, the important question is not simply whether AI was used. It is how the student used it.

• Did the student question the output?
• Improve it?
• Verify it?
• Make meaningful decisions about it?

3. Learning About AI

There is a third capability that is increasingly difficult to ignore.

Students need to understand what AI can and cannot do.

They need to recognise that AI-generated information can be inaccurate, biased or misleading. They need to understand the importance of verification, privacy, responsible use and human judgement.

In other words, students need to become not merely users of AI, but informed users of AI.

The Three Capabilities

The school therefore began to see AI literacy as something broader than learning how to operate an AI tool:

Learn independently → Learn with AI → Understand AI

These capabilities are not competitors. A future-ready learner may need all three.

The challenge for the school was to determine WHEN each capability should be developed, when AI should be available, and when it should deliberately be absent.

-----------------------------------------------------------------------------------

The Access-Independence Dilemma

Once the school began looking at AI through these three capabilities, another problem became apparent.

Access to AI is not equally valuable, or equally necessary, for every learner.

Consider a student who is struggling to understand a concept that most of the class has already mastered. Without additional support, the student may continue falling behind.

An AI tutor could provide another explanation, adjust the difficulty, generate examples, or allow the student to practise privately without feeling embarrassed about asking the same question repeatedly.

For that learner, restricting access to AI may not simply protect independent learning.

It may remove a useful pathway to learning.

At the same time, unrestricted access creates another risk. A student who can immediately ask AI to solve every problem may never develop the underlying ability to solve those problems independently.

The challenge, therefore, is not simply deciding whether AI should be available.

It is deciding what kind of assistance helps a learner move forward without replacing the learning that the learner needs to develop.

This creates a difficult balance:

Too little assistance
→ Some learners may struggle unnecessarily.

Too much assistance
→ Some learners may become dependent on the tool.

The educational question becomes:

When does AI function as a scaffold that helps a learner become more capable - and when does it become a substitute for the capability itself?

And that answer may depend on the learner, the learning objective, the stage of learning and the nature of the task.

There may be no single “AI allowed” or “AI prohibited” rule that works equally well for everyone.

 

-----------------------------------------------------------------------------------

Can We Teach Responsible AI Use?

The school could create rules.

• Students should verify AI-generated information.
• They should protect personal data.
• They should not submit AI-generated work as their own.
• They should understand bias and misinformation.
• They should use AI responsibly.

All of these principles are important.

But a new question emerged:

Can responsible AI use really be taught simply by giving students a set of rules?

Young learners are naturally curious. They experiment, explore and learn outside the boundaries of the classroom. 

If an AI tool can explain something more clearly than a textbook, help a student overcome a learning difficulty, translate an unfamiliar concept, or provide instant feedback, it is unrealistic to expect a curious learner to ignore it simply because a school has placed the activity in a “no AI” category.

At the same time, unrestricted use can create another problem.

A student may gradually stop asking:

“Can I solve this?”

and start asking:

“Can AI solve this for me?”

That difference is critical.

Responsible AI use therefore cannot be reduced to following rules.

It requires judgement.

A learner needs to gradually develop the ability to ask:

• Why am I using AI?
• What am I asking it to do?
• Could I do this myself?
• Is AI helping me understand, or simply giving me the answer?
• How do I know that the response is correct?
• What should I keep private?
• When should I stop using the tool and think for myself?

These are not questions that can be mastered through a single lesson on AI ethics.

They develop through experience, modelling, discussion, feedback and gradually increasing responsibility.

The challenge for the school, therefore, was becoming clearer:

The goal cannot simply be to make students obey rules about AI. The goal must be to help them develop the judgement to make increasingly responsible decisions when the rules cannot cover every situation.

-----------------------------------------------------------------------------------

The Assessment Problem

The school soon discovered that the AI debate could not be separated from another difficult question:

How do we know what a student can actually do when AI is always available?

Traditional assignments had often assumed that the work submitted by a student represented the student's own thinking and ability.

That assumption becomes much harder to make when a student can ask an AI system to explain a concept, generate ideas, improve writing, solve a problem, write code or produce an entire draft within seconds.

One possible response would be to prohibit AI completely during assessment.

But that creates another problem.

If students are expected to use AI responsibly in higher education and eventually in the workplace, should assessment ignore that reality?

Perhaps the important question is not:

“Did the student use AI?”

but:

“What capability are we actually trying to assess?”

Consider a programming assignment.

If the learning outcome is writing code independently, unrestricted AI assistance may undermine the purpose of the assessment.

But if the learning outcome is designing, evaluating and debugging a software solution, using AI to generate an initial piece of code may be entirely reasonable - provided the student can explain, test and improve what was produced.

The same distinction applies to writing.

If the objective is to assess a student's ability to construct an argument independently, generating the entire essay with AI defeats the purpose.

But if the objective is to develop the ability to critique, verify, refine and improve information, AI-generated material could become part of the learning activity itself.

The school therefore faced a fundamental assessment-design problem:

"The availability of AI changes what an assessment can validly measure."

The answer may not be to make every assessment “AI-proof.”

Instead, some assessments may need to become more authentic, process-oriented and transparent.

Students might be asked to:

• explain how they arrived at an answer;
• demonstrate their reasoning;
• critique an AI-generated response;
• compare different solutions;
• defend their decisions;
• document how they used AI;
• complete part of an assessment independently and another part with appropriate tools.

The purpose would be to make the student's thinking visible, rather than simply judging the final product.

When Policy Meets Assessment

This brought the school's original AI debate back into focus.

The school had initially been asking:

“Should students be allowed to use AI?”

But the assessment problem exposed a more fundamental question:

“What are we trying to know about the student?”

A blanket ban might protect the integrity of some assessments while preventing students from developing the ability to use AI appropriately.

Unlimited access might reflect the real world while making some traditional assessments meaningless.

The school's AI policy therefore could not be designed independently of its learning outcomes and assessment strategy.

If the school wanted students to become capable and responsible users of AI, then its assessments would need to distinguish between:

"Using AI to avoid learning" and "Using AI as part of learning."

That meant the institution could no longer treat AI policy, pedagogy and assessment as three separate conversations.

They were becoming parts of the same educational decision.

What we allow students to do with AI must be consistent with what we expect them to learn - and what our assessments are designed to measure.

 ----------------------------------------------------------------------------------------------------------------------------

From Rules to Principles

The school had started with a simple idea:

Create a policy that tells students what they can and cannot do with AI.

But the more deeply the leadership examined the problem, the less useful a simple list of rules appeared. 

A student using AI to translate an unfamiliar word is not doing the same thing as a student using AI to write an entire essay.

A student asking for another explanation of a difficult concept is not doing the same thing as a student asking AI to solve every problem.

A student using AI to generate possible ideas is not doing the same thing as a student submitting those ideas without checking whether they are accurate.

The tool may be the same.
The learning behaviour is not.

The school therefore began to consider whether its AI policy should be built around principles rather than permissions.
Instead of asking:

“Is AI allowed?”

Teachers could begin asking:

• What is the student supposed to learn?
• What should the student be able to do independently?
• What kind of assistance would support that learning?
• At what point does assistance replace the capability being developed?
• What evidence will show that the student actually understands the work?
• What should the student be able to explain, question or defend?

This changed the conversation.
AI was no longer being treated simply as a technology that was either inside or outside the classroom.

It became a tool whose educational value depended on purpose, context and judgement.

 

A Classroom Example

Consider a Grade 8 science lesson on climate change. The teacher wants students to develop two capabilities:

1. explain the basic science independently; and
2. evaluate different claims using evidence.

For the first part of the lesson, AI is deliberately not used.

Students receive a short set of information and are asked to explain, in their own words, why increasing greenhouse gases can affect global temperatures.

The purpose is not to produce a polished answer.

The purpose is to find out:

Can the student explain the underlying concept?

Later, the teacher introduces an AI tool and gives students a deliberately flawed AI-generated explanation of climate change.

Now AI is part of the learning activity.

Students must identify:

• what the AI explanation gets right;
• what it gets wrong or oversimplifies;
• what evidence is needed to verify its claims; and
• how the explanation could be improved.

In this activity, students are not being assessed on their ability to generate an answer with AI.

They are being asked to think about an AI-generated answer.

The same technology has therefore been treated differently at different points in the lesson.

• Without AI → demonstrate independent understanding
• With AI → evaluate and improve information
• About AI → develop judgement about its limitations

Neither “AI everywhere” nor “AI nowhere” would have captured the purpose of the lesson. The important decision was not whether AI was allowed.

It was "why it was being used at that particular moment."

A Different Kind of AI Policy

The school eventually began considering a policy built around four questions:

1. Purpose

Why is AI being used?

2. Independence

What must the student still be able to do without AI?

3. Judgement

Can the student evaluate, question and verify what AI produces?

4. Evidence

How will the teacher know that meaningful learning has taken place?

This approach also allowed the school to recognise an important difference between younger learners and more experienced learners.

A child who is still developing reading, writing, reasoning or problem-solving skills may need more opportunities to practise those capabilities independently before AI begins doing substantial parts of the work.

An older student working on a complex research, programming or design problem may reasonably use AI as part of the process—but should still be able to explain the decisions made and evaluate the quality of the result.

The question was therefore no longer:

“How much AI should students be allowed to use?”

It became:

“How much independence does this particular learning task require?”

That was a much harder question.

But it was also a much more educationally meaningful one.

The school was beginning to move from controlling technology to designing learning around technology.

And that shift raised one final challenge:

If students are expected to make increasingly intelligent decisions about AI, where in the curriculum will they actually learn how to make those decisions?

 -----------------------------------------------------------------------------------

The Teacher's Judgement Matters

The school soon discovered another difficulty.

Even with a principles-based policy, two teachers could make very different decisions about the same AI tool.

One teacher might say:

“AI is useful here because students are learning how to evaluate information.”

Another might respond:

“I don't want students using AI because they are still developing the underlying skill.”

Both could have valid educational reasons.

This meant that a policy alone would not create consistency.

Teachers themselves needed to understand why AI should be used, when it should be restricted, and what students should still be expected to do independently.

Consider a simple writing activity. A teacher asks students to write a short persuasive paragraph.

For one class, the objective may be to develop independent writing fluency.

In that situation, asking AI to rewrite the student's paragraph into sophisticated language could defeat the purpose of the exercise.

But in another lesson, the objective may be to develop critical editing skills.

The teacher could provide an AI-generated paragraph and ask students to identify weak arguments, unsupported claims, inappropriate language or poor structure.

Here, AI becomes part of the learning activity rather than a shortcut around it.

The difference is not the technology.

It is the learning objective.

From AI Rules to Teacher Decisions

The school therefore began thinking about a simple principle:

Teachers should not have to decide whether AI is “good” or “bad”. They should decide whether a particular use of AI supports the intended learning.

That requires teachers to ask questions such as:

• What is the learning objective?
• What capability must the student practise?
• Would AI support that capability or perform it for the student?
• What evidence of learning do I need to see?
• Could the student explain and defend the work produced?

To make this easier to apply consistently, the school developed a simple Teacher AI Decision Framework.

The Teacher AI Decision Framework

Before allowing AI in a learning activity, the teacher asks five questions:

1. What am I actually assessing?
Knowledge? Independent reasoning? Writing? Problem-solving? Creativity? Evaluation? Application?

2. What must the student be able to do independently?
Identify the capability that should remain visible without AI assistance.

3. What role could AI play?
Could it provide an explanation, example, feedback, alternative perspective or practice opportunity without completing the core learning task?

4. What could go wrong?
Could AI remove the productive struggle, introduce inaccurate information, create dependency, expose personal information or make the assessment meaningless?

5. What evidence will I collect?
What will allow the teacher to see the student's own thinking—an explanation, draft, reflection, oral defence, working process, comparison, correction or final product?

The resulting decision does not have to be simply “AI allowed” or “AI prohibited.”

Instead, the teacher can choose among several possibilities:

1. AI not needed → The capability should be developed independently.
2. AI as scaffold → AI provides limited support while the student performs the core task.
3. AI as learning partner → Students use AI to explore, question, compare or practise.
4. AI as object of analysis → Students examine, challenge or correct AI output.
5. AI as production tool → Students use AI as part of a more advanced task, while remaining responsible for evaluating and improving the result.

This creates an important distinction:

The question is not “Can students use AI?” but “What role, if any, should AI play in this learning activity?”

Consistency Without Uniformity

The school also realised that consistency did not necessarily mean that every classroom had to follow exactly the same AI rules.

A mathematics lesson, a literature discussion, a programming project and a primary-school writing activity may have very different learning requirements.

What the school needed was not one identical rule for every classroom.

It needed a shared set of principles that teachers could apply to different learning situations.

The institution could therefore establish common expectations around:

• student independence;
• transparency of AI use;
• verification of AI-generated information;
• privacy and responsible use;
• assessment integrity; and
• evidence of learning.

Teachers could then determine how those principles applied to their particular subject, learners and learning objectives.

The school was beginning to understand something important:

An effective AI policy does not remove teacher judgement. It makes teacher judgement more informed, more consistent and more purposeful.

But there was still a problem.

If responsible AI use depends partly on teacher judgement, how can a school prepare all of its teachers to make those decisions confidently?

That question would take the discussion beyond student AI literacy.

It would force the institution to consider teacher AI literacy.

-----------------------------------------------------------------------------------

Teacher AI Literacy

The school had initially focused almost entirely on the students.

Should students be allowed to use AI?
How should they use it?
What should they be taught about its risks?

But the leadership eventually recognised an uncomfortable reality:

Students cannot be expected to develop responsible AI judgement if the adults guiding them do not have that judgement themselves.

Teacher AI literacy could not simply mean knowing how to open an AI tool or write a good prompt.

A teacher might generate an excellent lesson plan with AI but fail to recognise an inaccurate claim. Another might create an impressive AI-generated quiz without noticing that it tests recall when the learning objective requires analysis.

The school therefore began to distinguish between technical AI skills and educational AI literacy.

What Should Teachers Actually Know?

Teachers need to understand:

• what AI can and cannot do;
• how to evaluate AI-generated information;
• when AI genuinely supports learning;
• how AI changes assessment;
• how to guide students towards responsible use; and
• issues such as privacy, bias and transparency.

But perhaps the most important skill was professional judgement.

A Concrete Example

A history teacher asks AI to generate a Grade 9 explanation of the causes of the First World War.

The response looks excellent. A technically capable teacher might simply use it.

A teacher with stronger educational AI literacy pauses:

Are the claims accurate? What has been oversimplified? Which perspectives are missing? Does this actually support my learning objective?

The teacher then changes the activity.

Students receive the AI-generated explanation alongside two historical sources. They must identify which claims are supported, which are questionable, and what the AI explanation has left out.

AI has therefore changed from the answer into an object for investigation.

The teacher has not merely learned how to use AI. The teacher has learned how to design learning around AI.

From Training to Professional Judgement

The school therefore decided that teacher development should not be limited to a workshop called “Introduction to Generative AI.”

Teachers needed opportunities to discuss real classroom situations:

• A student uses AI to rewrite an entire assignment.
• A struggling learner uses AI for another explanation.
• Students use AI to generate solutions and then critique them.
• An AI-generated resource contains factual errors.
• An assessment deliberately incorporates AI but requires students to explain their decisions.

The question in each case is not simply:

“Is AI allowed?”

It is:

“What is the learning objective, what role is AI playing, and what evidence will show that learning has taken place?”

The school was beginning to see teacher AI literacy differently.

It was not primarily about learning a technology.

It was about developing the professional judgement to decide when technology belongs in the learning process—and when it does not.

And that raised the next institutional question:

If teachers are making these decisions across different classrooms, how will the school know whether its overall approach to AI is actually helping students learn?

 -----------------------------------------------------------------------------------

Returning to the Original Question

The school began with a seemingly simple question:

Should students be allowed to use AI?

After examining the issue from different perspectives, that question no longer seemed adequate.

The real challenge was not to decide whether AI belonged in the classroom.

It was to decide what kind of learning the school wanted to protect, what new capabilities it wanted students to develop, and where AI could contribute without replacing those capabilities.

The school therefore moved towards a more balanced approach.

Students would still need opportunities to think, read, write, calculate, reason and solve problems independently.

They would also need opportunities to use AI as a learning tool, where it could provide support, feedback, alternative explanations or new possibilities.

And they would need to learn about AI itself - its limitations, risks, biases and the responsibility that comes with using it.

In other words:

• Learn without AI.
• Learn with AI.
• Learn about AI.

 ----------------------------------------------------------------------------------

What This Case Taught Me

The balance would not always be the same.

A young learner developing foundational skills may need more protected opportunities for independent practice.

A more advanced learner working on a complex problem may benefit from using AI as part of the process.

• An assessment designed to measure independent reasoning may deliberately restrict AI.
• Another assessment may require students to use AI - and demonstrate that they can question, verify and improve its output.

The goal, therefore, is not to create students who depend on AI. Nor is it to create students who are afraid to use it. 

The goal is to develop learners who can make informed decisions about when to use it, how to use it, and when to put it aside.

That brings us back to the question with which this case began:

What should a young person still be able to do when the machine is available to do almost everything?

Perhaps the answer is not a fixed list of skills. Perhaps the deeper answer is judgement.

• The judgement to think before asking. 
• The judgement to question an answer.
• The judgement to recognise when help has become dependence.

And the judgement to know when the most important learning begins after the machine has given its answer.

Technology policy is ultimately not a technology problem. It is a learning-design problem.

A school does not become future-ready simply by giving every student access to AI. Nor does it become future-ready by banning every digital tool.

The real challenge is therefore not choosing between technology and no technology. It is learning how to use technology without losing sight of the learner.

A school becomes future-ready when its curriculum, teaching, assessment and technology choices work together to develop capable, independent and responsible learners.

-----------------------------------------------------------------------------------

Research Behind This Approach

The ideas in this case are not based on the assumption that AI is either a threat to education or a solution to every educational problem. They reflect a growing shift in how AI and education are being discussed: from technology adoption towards human-centred, responsible and purposeful use.

UNESCO's AI Competency Framework for Students (2024) identifies four broad areas of competency: a human-centred mindset, ethics of AI, AI techniques and applications, and AI system design. It also emphasises critical judgement, human agency and the ability to use AI responsibly rather than simply learning how to operate AI tools. UNESCO — AI Competency Framework for Students

UNESCO's AI Competency Framework for Teachers (2024) similarly goes beyond technical skills. It identifies competencies in AI pedagogy, ethics, human-centred thinking and professional learning—supporting the idea that teachers need judgement about when and why AI should be used, not merely the ability to use it. UNESCO — AI Competency Framework for Teachers

UNESCO's Guidance for Generative AI in Education and Research (2023) advocates a human-centred approach that considers privacy, ethics, age-appropriate use and meaningful learning while recognising legitimate uses of generative AI in curriculum, teaching and learning. UNESCO — Guidance for Generative AI in Education and Research

Taken together, these approaches support the central idea of this case:

The educational question is not simply whether students should use AI. It is whether we are designing learning environments in which students remain capable of thinking, questioning, creating and making decisions when AI is available to them.

What do you think? 👇

I'd be interested to hear from teachers, educators, lecturers, academic leaders and other education professionals who have experience with these issues. If you have a different perspective, a classroom experience, or an approach that has worked for you, please feel free to share it in the comments.

Thoughtful disagreement is welcome too. The aim is to learn from one another and keep the conversation going. Thank you!