Not Against AI: Are We Raising Thinkers or Typists?

MATHIVATION: Are We Raising Thinkers or Typists?

The AI Paradox in Our Classrooms

Seedhi Baat Now Friday Special 

Not Against AI. For the Human Mind.

“If you give a calculator to a child who doesn't know tables, the child may get the answer - but may never learn multiplication. AI presents us with the same question at a much larger scale.”

 

A Question From a Teacher Who Also Uses AI

Let me begin with a confession.

I am not against AI.

In fact,

I work with AI. I learn from AI. I teach with AI. You could call me an AI-augmented teacher. But I remain firmly FOR the human mind.

But precisely because I use AI, I have one concern:

Are we teaching children to use AI - or teaching AI to do the thinking for our children?

Our students are not at fault.

If we give a young learner a tool that makes homework faster, projects prettier, essays easier and answers instantly available, why wouldn't the learner use it?

Parents are relieved:

“Homework ho gaya.”

Teachers may receive beautifully formatted assignments.

Schools may proudly announce AI-enabled learning.

But there is another question hiding underneath the completed homework:

Did learning happen?

That is where my concern begins.

How Did We Arrive Here?

Perhaps the history of educational technology can be viewed as a series of changing relationships between humans and cognitive effort.

Phase 1 - Calculator Era

Calculators reduced the need for lengthy arithmetic.

That was not necessarily bad.

A calculator can free a learner to focus on concepts - but only if the learner already understands what the calculation means.

Phase 2 - Search Era

Search engines made information retrieval almost effortless.

We no longer needed to remember every fact.

Again, not necessarily bad.

The challenge shifted from:

“Can you find information?”

to

“Can you judge whether the information is correct?”

Phase 3 - Generative AI Era

Now something different has happened.

AI does not merely retrieve information.

It can:

• explain,

• summarise,

• solve,

• write,

• brainstorm,

• code,

• design,

• translate,

• and produce a finished answer.

That changes the educational equation.

The danger is not that AI gives answers.

The danger begins when the learner stops doing the cognitive work required to understand those answers.

The Question Is Not AI vs Human

This is where I want to challenge my own argument.

AI can be extraordinarily useful in education.

The OECD's Digital Education Outlook 2026 makes an important distinction: GenAI can support learning when it is used with clear pedagogical intent, but simply outsourcing tasks to a general-purpose chatbot can improve the quality of the finished product without necessarily producing equivalent learning gains.

That distinction is crucial.

AI as a tutor:

“Give me a hint.”

AI as a thinking partner:

“Challenge my argument.”

AI as an editor:

“Find weaknesses in my paragraph.”

AI as a replacement:

“Write the entire assignment for me.”

Same technology.

Four completely different educational outcomes.

What Happens When AI Does the Thinking?

One of the most discussed studies in this area came from MIT Media Lab researcher Nataliya Kos'myna and colleagues in 2025.

The study, titled:

Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing

compared three groups:

LLM group → Search Engine group → Brain-only group

There were 54 participants in the first three sessions, all adults aged 18–39. EEG was used to examine neural activity during essay writing. A smaller group of 18 participants also completed a fourth session in which some conditions were switched.

The researchers reported differences in neural connectivity:

Brain-only → strongest and most distributed connectivity

Search → intermediate

LLM → weakest

The study also reported lower essay ownership and weaker ability among LLM users to accurately recall or quote their own writing.

The researchers used the term:

Cognitive Debt

The idea is simple:

A tool may reduce the effort you pay today - but the learning you postponed may become tomorrow's debt.

But there is an important scientific caution.

This study is not proof that AI makes children less intelligent.

It was a relatively small study of adults, and it is a preprint rather than the final word on the subject. The findings are important enough to investigate—but not strong enough to justify panic.

And that distinction itself is a lesson in critical thinking.

Evidence ≠ Conclusion

Writing Is Not Just Producing Words

Why does writing matter so much?

Because writing forces the mind to perform several operations simultaneously.

You must:

• decide what you believe,

• select relevant information,

• organise ideas,

• choose words,

• construct an argument,

• monitor coherence,

• revise mistakes,

• and imagine how another person will understand you.

Cognitive psychologist Ronald T. Kellogg has studied writing as a demanding cognitive activity involving planning, translating ideas into text, reviewing, working memory and executive attention.

So perhaps handwriting is not valuable merely because it is old-fashioned.

Perhaps the struggle to construct an answer is itself part of the learning.

The Beautiful Paradox

Here is the paradox.

AI makes writing easier.

But education is not primarily about producing writing.

Education is about becoming capable of thinking.

Therefore:

If AI removes unnecessary friction → wonderful.

If AI removes productive struggle → dangerous.

The challenge is knowing the difference.

The Mathivation Equation

Let us create a conceptual model.

Learning = Struggle × Time × Retention

This is not a scientific measurement equation.

It is a Mathivation model - a way of thinking.

Suppose a learner personally engages with an idea:

S = 1, T = 1, R = 0.9

Then:

L = 1 × 1 × 0.9 = 0.9

Now imagine a learner asks AI to generate almost everything, leaving only a small fraction of the cognitive work.

For illustration - not measurement - suppose:

S = 0.1, R = 0.1

Then:

L_AI = 0.1 × 1 × 0.1 = 0.01

That looks like a dramatic fall.

But remember:

The numbers are illustrative, not empirical.

The real lesson is more important than the numbers:

When cognitive effort approaches zero, learning cannot be assumed to remain at 100%.

 

Cognitive Debt - A Different Mathematical View

We can also imagine:

Cognitive Debt = Work Outsourced − Learning Acquired

Again, this is a conceptual model.

If AI helps you understand something better, cognitive debt may be low.

If AI completes something you were supposed to learn, the debt may increase.

And like financial debt, cognitive debt can remain invisible for a while.

You discover it later - 

during an examination,

an interview,

a classroom discussion,

or a moment when the device is unavailable.

The Search Engine Was Different

The MIT study reported a middle position for the search-engine group compared with the brain-only and LLM groups.

Why might that matter?

Searching still requires the learner to:

• choose keywords,

• compare sources,

• open pages,

• reject irrelevant information,

• connect facts,

• and decide what to use.

In other words:

The search engine still leaves some cognitive work with the human.

Generative AI can remove much more of that intermediate work.

That does not make AI bad.

It makes how we use it enormously important.

The Behavioural Economics of Convenience

Human beings naturally respond to incentives.

If one method takes:

60 minutes

and another takes:

6 minutes

we are naturally attracted to the second.

That is not laziness.

It is economics.

But education contains a strange economic problem.

The cheapest route to the answer may not be the cheapest route to learning.

A student who spends 30 minutes struggling with a mathematical proof may produce nothing that looks impressive.

Another student may ask AI for the proof and receive a beautiful explanation in 30 seconds.

Who appears more productive?

The second student.

Who may have learned more?

We don't know until we test what happens when the AI disappears.

That is the hidden economics of education.

The Output Trap

AI creates a dangerous possibility:

Excellent Output ≠ Excellent Learning

A student can submit:

• a beautiful essay,

• sophisticated vocabulary,

• perfect formatting,

• impressive references,

without possessing equivalent understanding.

Therefore perhaps assessment itself must change.

Instead of asking only:

“Show me your answer.”

We should increasingly ask:

“Show me how you arrived at it.”

And sometimes:

“Now explain it without the machine.”

That second question may reveal more learning than the first.

AI and the Human Relationship

Education is not only information transfer.

A teacher notices:

• hesitation,

• confidence,

• curiosity,

• confusion,

• excitement,

• frustration,

• humour,

• silence.

A parent notices when a child is pretending to understand.

A friend notices when another friend has stopped participating.

A machine can assist with information.

But education also involves relationships, judgement, empathy and responsibility.

UNESCO's guidance on GenAI in education therefore advocates a human-centred approach, with attention to human agency, ethical use, privacy and appropriate educational design.

What Should Parents Do?

Not:

“No AI!”

That battle may already be lost.

Instead:

Teach AI discipline.

Rule 1 - Brain First

Before asking AI, attempt the problem yourself.

Even a wrong attempt is valuable.

Rule 2 - Two Drafts

Draft 1: Human.

Draft 2: AI-assisted improvement.

Do not reverse the order.

Rule 3 - Explain Back

After using AI, close the screen.

Explain the answer in your own words.

If you cannot explain it, you probably borrowed an answer rather than learned an idea.

Rule 4 - Ask for Questions, Not Just Answers

Instead of:

“Solve this.”

Try:

“Give me a hint.”

Instead of:

“Write my essay.”

Try:

“Challenge my argument.”

Instead of:

“Do my project.”

Try:

“Ask me questions that help me build the project.”

That is a completely different relationship with AI.

The Two-Draft Rule

Draft 1 → Human

Think. Write. Struggle. Make mistakes.

Draft 2 → AI-assisted

Ask AI to critique, question, improve, or challenge the work.

The student remains the author.

The AI Viva

After submitting an AI-assisted assignment, ask the learner:

“Now close the screen and defend your work.”

Ask:

• Why did you choose this argument?

• What evidence supports it?

• Which part did AI improve?

• Where did you disagree with AI?

• What did you learn?

This is brilliant because AI-generated work becomes an opportunity for deeper assessment rather than simply a cheating problem.

What Should Teachers Do?

Perhaps the classroom needs a new assessment philosophy.

Ask students to:

• write some work without AI,

• defend their reasoning orally,

• show drafts,

• explain mistakes,

• compare their answer with AI's answer,

• identify AI errors,

• critique AI-generated arguments,

• and solve selected tasks without technology.

In other words:

Don't teach students only to use AI.

Teach them to evaluate AI.

The OECD's current guidance similarly emphasises that students need opportunities to develop independent thinking and critical appraisal both with and without GenAI.

A Small Experiment for Every Classroom

Here is a Mathivation challenge.

Give students the same question.

Round 1

Brain only.

Round 2

Search engine.

Round 3

Generative AI.

Then ask:

• Which answer is best?

• Which one do you understand best?

• Which one can you explain without looking?

• Which process taught you the most?

• Which mistakes did each method make?

• What did you contribute personally?

Now the classroom becomes a laboratory.

AI is no longer the enemy.

AI becomes the experiment.

The 90 - 10 Principle

I like a simple Mathivation rule:

Human Thinking ≫ AI Assistance

For younger learners especially, the balance should favour direct human engagement.

So as a practical classroom heuristic:

90% Human Thinking + 10% AI Assistance

But this is not a universal scientific ratio.

The appropriate balance depends on age, subject, task and learning objective.

The principle matters more than the number:

AI should amplify learning - not replace the learner.

 

Bring Back Productive Struggle

We have become uncomfortable with struggle.

A child struggles with multiplication:

“Ask AI.”

A child struggles with an essay:

“Ask AI.”

A child struggles with a science project:

“Ask AI.”

But what if the struggle is exactly where the brain is building the skill?

A little frustration can be productive.

A failed proof can teach persistence.

A badly written paragraph can teach revision.

A wrong hypothesis can teach scientific thinking.

A difficult question can create curiosity.

Not every difficulty is a problem to be removed.

Some difficulties are part of becoming capable.

The Pencil Still Has a Place

This is not nostalgia.

Paper-and-pencil assessments still have an important role in many educational settings.

For example, Princeton's current course materials include supervised, in-person, paper-based examinations with restrictions on electronic devices in some courses.

The lesson is not:

“Technology is bad.”

The lesson is:

Different tools are appropriate for different learning objectives.

Sometimes the best technology is no technology.

What About the Coming Generation?

The real divide of the future may not be:

Human vs AI

It may be:

Human with thinking skills + AI

versus

Human dependent on AI for thinking

That difference could become enormous.

A student who knows mathematics can use AI to explore mathematics.

A student who does not understand mathematics may simply accept whatever AI produces.

A writer who knows how to think can use AI as an editor.

A person who has never learned to organise ideas may allow AI to become the author.

Foundational knowledge becomes more important - not less - when AI becomes more powerful.

Because you cannot critically evaluate an answer if you do not understand the subject well enough to question it.

My Personal AI Philosophy

I do not want children to fear AI.

I want them to become strong enough to question it.

I do not want students to reject technology.

I want them to understand when technology is helping them and when it is replacing them.

And I certainly do not want children to return blindly to a pre-AI world.

The world has changed.

Education must change too.

But change does not mean surrender.

The Mathivation Equation of the Future

Perhaps we can write:

Future Learning = Human Curiosity + Critical Thinking + Foundational Knowledge + AI Assistance

​Not:

​Future Learning = AI Output

The difference is everything.

A Friday Question for Parents, Teachers & Students

Next time your child submits a perfect AI-assisted assignment, don't ask only:

“How many marks will this get?”

Ask:

“How much of this belongs to your mind?”

And perhaps ask yourself the same question.

Final Reflection

Ask for your share, and relationships sometimes become visible.
Leave your share, and even thorns may become roses.

AI is similar.

If we ask AI to do our share of thinking, our dependency becomes visible only later.

But if we leave some struggle for ourselves - 

some reading,

some writing,

some calculation,

some failure,

some questioning,

some silence - 

difficulty can become a rose of wisdom.

I am here to help you write.

But I should never replace your writing.

Use AI as a mirror, not as a substitute for your mind.

Use it as a mentor-like tool, not as a servant doing your homework.

Use it to ask better questions.

Use it to challenge your assumptions.

Use it to find the holes in your reasoning.

Use it to learn.

But then -

close the screen and think.

Because the greatest achievement of education is not producing a perfect answer.

It is producing a human being who can still think when there is no answer button.

Disclaimer

This article is a Mathivation reflection, not a medical, neurological, psychological, or educational prescription.

Research on Generative AI and learning is still evolving. The MIT study discussed here is an important early contribution, but it should not be treated as conclusive proof that AI reduces intelligence or permanently damages the brain.

The mathematical equations used in this article - such as the Learning Equation, Cognitive Debt model, and 90–10 Principle - are conceptual Mathivation models, created to help readers think about learning and cognitive effort. They are not established scientific formulas.

AI can be an excellent tutor, research assistant, editor, accessibility tool, and creative partner when used thoughtfully.

The central message is therefore not “Don't use AI.”

It is:

“Use AI without outsourcing your ability to think.”

Readers are encouraged to examine the original research and form their own informed conclusions.

Curious Question

Imagine that tomorrow every AI tool disappeared for 30 days.

Could you still:

- write an essay,

- solve a mathematics problem,

- explain a scientific idea,

- defend your opinion,

- research a topic,

- and create something original?

What would remain when the tool disappears?

Perhaps that is the real test of education.

How much of the intelligence belongs to the machine - and how much still belongs to us?

AI should sharpen human thinking - not substitute for it.

The Mathivation Closing Line

The future will not belong to AI instead of humans.

It will belong to humans who know how to remain human while working with AI.

With gratitude, curiosity and responsibility,

Rakesh Kushwaha 

Founder Mathivation Research Lab

Where Mathematics Meets Human Behaviour

Seedhi Baat - Now Friday Special

23 August 2026 | Mumbai

Research & Further Reading

1. Kos'myna, N. et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab / arXiv. The study involved 54 adults in the first three sessions and 18 participants in the fourth; it reported differences in EEG connectivity, essay ownership and recall across LLM, search-engine and brain-only conditions.

2. OECD (2026). Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education. The report distinguishes between purposeful educational uses of GenAI and simply outsourcing cognitive tasks to general-purpose systems.

3. OECD (2025). Evolving AI Capabilities and the School Curriculum. The report discusses how writing can support reflection, meaning-making and thinking, while considering how AI may reshape these processes.

4. UNESCO (2023; updated 2026). Guidance for Generative AI in Education and Research. UNESCO advocates human-centred, ethical and age-appropriate approaches to GenAI in education.

5. Kellogg, R. T. (2008). Training Writing Skills: A Cognitive Developmental Perspective. Journal of Writing Research, 1(1). His work examines the cognitive demands of planning, generating and reviewing written language.

6. Princeton University course/examination materials. Examples of supervised, paper-based, closed-book examinations illustrate that technology-free assessment remains part of contemporary higher education.

A Note on Evidence

The scientific evidence on AI and learning is still developing. Some studies find benefits when AI is used as a tutor, scaffold or feedback partner; other studies raise concerns about over-reliance and cognitive offloading. Therefore, this article deliberately argues for responsible use rather than rejection of AI.

That is the Mathivation position:

Don't fear the tool. Understand the tool.
Don't surrender your thinking to the tool.
Learn to think better with it.