Education 2047 #Blog 65 (22 JUN 2026) 

The End of Easy Answers

Why the AI Age Demands ‘Epistemic Friction’ in Education

 

For most of human history, information was hard to get. Books were expensive. Libraries were few. Teachers were the gatekeepers of knowledge. A school’s main job was to give students access to what they could not otherwise reach.

That world is gone.

Today, any student with a phone can access more information in thirty seconds than a university library held fifty years ago. A generative AI tool will not only retrieve the information — it will explain it, summarise it, and answer follow-up questions with patience no teacher can match.

So what is a school for, now?

That is not a rhetorical question. It is the most urgent design problem in education. And the answer, I will argue, is not ‘more content’ or ‘better content’ or even ‘personalised content.’ The answer is friction.

When information becomes free, the institution built to deliver it has lost its reason to exist — unless it finds a new one.

Not friction in the sense of unnecessary difficulty. Not the cruel obscurity of questions designed to fail students. I mean something specific: the deliberate experience of genuine uncertainty. The moment a learner faces a problem that cannot be solved by looking something up. The moment they must think — really think — because no answer is waiting to be retrieved.

That experience is what education must now protect and cultivate. Because AI can do almost everything else.

 

 

A Map of Learning: The Four Quadrants

To see where education is going wrong — and where it must go — it helps to map the learning landscape. We can do this with two simple questions: Is the question already known? And is the answer already known?

The combinations give us four distinct types of learning, each with a very different character.

 

Figure 1: The Four Quadrants of Learning — KAKQ → UAUQ

Quadrant 1: Known Answer, Known Question — Recall

This is where almost all formal education has lived for two centuries. The question is set by the syllabus. The answer is in the textbook. The student’s job is to remember the answer and reproduce it in the examination hall.

There is nothing wrong with recall as one part of learning. Foundations matter. You cannot build on what you do not know.

The problem is that we built the entire house here and called it education.

AI has now taken ownership of this quadrant. Every question with a known answer — every definition, every formula, every historical date, every multiple-choice option — is answered faster and more accurately by a generative model than by a student who studied for three weeks. The cognitive work being assessed in most examinations is precisely the cognitive work AI does best.

The examination was always a proxy. It measured the willingness to memorise, not the capacity to think. AI has made the proxy visible by doing the job better than the student.

 

Quadrant 2: Known Answer, Unknown Question — Inquire

Here, an answer exists somewhere — but the student must first figure out what question to ask. This is harder than it sounds. It requires noticing a gap in your own understanding, naming it clearly, and directing your inquiry toward it.

Think of a doctor who notices something unusual in a patient’s symptoms. The medical literature holds the answer — but only if the doctor asks the right question. Framing that question is itself the skill.

This quadrant is where curiosity becomes something you can train, not just something you are born with. AI is a useful partner here: it can help a student sharpen a vague question into a precise one. But it cannot notice the gap for them. That noticing is human.

Most institutions reach this quadrant only in final-year projects, if at all. Everything before that is Quadrant 1.

Quadrant 3: Unknown Answer, Known Question — Create

The question is given — by a real problem, a client brief, a design challenge. But the answer does not yet exist in any retrievable form. The student must build it.

This is where engineers design, where managers decide, where clinicians diagnose. Bloom’s higher levels — Analysis, Evaluation, Creation — live here. This is also, not coincidentally, where employers have always wanted graduates to arrive. Most graduates arrive somewhere else.

AI is genuinely useful in Quadrant 3. It can generate candidate solutions, flag similar problems that were solved differently, and help stress-test ideas. But it cannot carry responsibility for the judgement that chooses between options. That judgement is the learning outcome. The human is not decorating the process — the human is the decision-maker.

 

Quadrant 4: Unknown Answer, Unknown Question — Innovate

Neither the question nor the answer is given. The learner must first perceive that a question worth asking exists, then formulate it with enough precision to be productive, then navigate toward an answer that has no prior form.

This is the territory of research, of invention, of discovery. It is also the one quadrant AI cannot authentically inhabit. AI can pattern-match toward questions it has encountered in its training. It cannot genuinely not-know — which is the exact condition from which discovery departs.

The student working in Quadrant 4 is, by definition, ahead of any training corpus. They are doing something new. That is irreplaceable.

AI has not made human creativity redundant. It has made the performance of thinking — the retrieval display of Quadrant 1 — permanently obsolete.

 

The Real Problem: We Teach Q1 While Claiming to Develop Q4

These four quadrants are not four levels of difficulty. They are four fundamentally different kinds of cognitive work. And the crisis of modern education is not that it has stayed in Quadrant 1. It is that it has stayed in Quadrant 1 while using language that belongs in Quadrants 3 and 4.

We put ‘critical thinking’ and ‘problem-solving’ and ‘creativity’ in our graduate attribute statements. We assess recall and call the results evidence of capability. We award degrees for Quadrant 1 performance and are surprised when employers find Quadrant 3 thinking missing.

For a long time, this gap was survivable. A student who had memorised a great deal at least demonstrated diligence and intelligence. Employers could make reasonable inferences. The correlation was loose, but it held.

AI has severed that correlation. The student who retrieves and reproduces is now doing exactly what a language model does, only more slowly and less reliably. The credential that certifies only this no longer certifies anything an employer cannot get for free.

The degree once proved that a person had spent years near knowledge. In the AI age, being near knowledge costs nothing. The degree must now prove something harder — or it proves nothing at all.

 

What ‘Epistemic Friction’ Actually Means in Practice

Epistemic friction is not making things harder for its own sake. It is the deliberate design of learning moments that cannot be resolved by looking something up.

A friction-rich learning environment has a few recognisable features. It starts with problems that have no single correct answer — so that the very act of constructing an answer is where the learning happens. It treats the quality of a student’s questions as a measure of their development, not just the quality of their answers. It makes uncertainty visible and productive: not a deficiency to correct, but the raw material from which understanding is built.

It also uses AI wisely. In Quadrant 2, AI helps sharpen imprecise questions. In Quadrant 3, AI helps generate and stress-test options. In Quadrant 4, AI steps back — because the work there is irreducibly human, and using AI to shortcut it destroys the very learning that quadrant exists to produce.

The educator in this environment is not primarily a content expert. They are a designer of productive difficulty. Their skill is not knowing the answer — it is knowing precisely when not to give it.

The great teacher of the AI age will not be remembered for what they explained. They will be remembered for the questions they refused to answer — because answering would have closed the learning rather than opened it.

 

Why Institutions Have Not Made This Shift

None of what is described above is new as an idea. NEP 2020 calls for critical thinking. Outcome-based education frameworks point toward higher-order learning. Bloom’s Taxonomy has been on syllabi for thirty years. The words are right. The structure has not changed.

The reason is not indifference. It is architecture.

The examination is not simply a bad assessment tool that we could swap for a better one. It is a load-bearing wall. The entire institution — its timetables, its gradebooks, its accreditation frameworks, its ranking metrics — is built around certifying performance at scale through standardised testing. To remove the examination is not to change one component. It is to redesign the building.

This is where the new policy infrastructure — the Academic Bank of Credits, the National Credit Framework, APAAR, the move toward continuous and portfolio-based assessment — becomes genuinely consequential. These are not digitisation projects. They are the foundations of a different kind of certification: one that can carry evidence of Quadrant 3 and 4 performance without collapsing it into a percentage.

The institutions that understand this — that the task is not to put the old system online but to use digital infrastructure to build a structurally different one — will be the ones that survive the coming decade with their credibility intact.

 

The Inversion

The industrial university was a content delivery system that also did assessment. The AI-age university must become an assessment system that enables content access — because the content is everywhere, but rigorous, trustworthy evidence of higher-order capability is almost nowhere.

That is the gap. That is the opportunity. That is the friction that remains after every retrieval task has been automated.

The spark that education risks losing is not knowledge — AI has knowledge in abundance. It is the experience of genuine not-knowing, and the discipline of navigating through it. That experience cannot be transmitted. It can only be cultivated. And cultivating it requires, by design, more friction — not less.

We built institutions to eliminate uncertainty. We now need institutions that teach people to navigate it. That is not a reform. It is a structural inversion.

 

 * * * 

Author

Dr. Neeraj Saxena is a former Scientist at TIFAC/DST and co-author of Educational Roadmap of India's Technology Vision 2035, with subsequent advisory roles at AICTE spanning higher education policy and implementation. He is currently Pro-Chancellor of JIS University, Kolkata, and publishes on AI-induced transformation in education through the Education2047 platform [nrj2000.blogspot.com] & [nrjsaxenajisu.substack.com]

©Dr. Neeraj Saxena, 2026 | CC BY-NC-ND 4.0 | Attribution required | No commercial use | No derivatives without permission.

©Dr. Neeraj Saxena, 2026 | CC BY-NC-ND 4.0 | Attribution required | No commercial use | No derivatives without permission.