India's Curriculum Problem Is a Reward Problem

What earns marks, admission, parental approval and professional recognition becomes the real curriculum. The printed curriculum follows behind. NEP 2020 understands much of this, which makes it worth taking seriously, and worth being harder on.

One question stays with me: what does a school know about how a child thinks?

It knows attendance. It knows marks. It knows whether an assignment arrived on time. Those are useful records. But can the child explain why an answer works, notice when it fails, or change direction when the facts change?

That gap is where I want to begin a serious argument about Indian education. It does not rest on the claim that every school is broken, on nostalgia for a perfect past, or on the promise that an AI tutor will fix everything.

My argument is this: curriculum reform will keep disappointing us if we change the syllabus without changing the rewards around it. What earns marks, admission, parental approval and professional recognition becomes the real curriculum. The printed curriculum follows behind.

NEP 2020 understands much of this. That makes the policy worth taking seriously, and worth being harder on.

We have had the diagnosis for decades

Macaulay is an easy villain. His 1835 Minute on Education explicitly sought a class of intermediaries, "Indian in blood and colour, but English in tastes, in opinions, in morals and in intellect." It was an argument about colonial power, language and who would get access to knowledge. [1]

But "Macaulay created today's rote-learning system" is too neat. A nineteenth-century document cannot explain every decision made by independent India, every state's schools, or every family's choices today. Blaming a dead colonial official can become a way of avoiding responsibility for the incentives we still maintain.

Even before independence, there was another imagination of education. The 1937 Wardha scheme put productive craft and the mother tongue at the centre of basic education. Its self-financing ambition deserves separate scrutiny. But learning through doing is not a discovery we made in 2020. [2]

After independence, the purpose changed. Education was meant to build citizenship, widen opportunity and support national development. The 1968 policy, following the 1964-66 Education Commission, called for equal opportunity, the Common School System, science education and a closer relationship between education and work. It also sought education expenditure of 6% of national income. [3]

The 1986 policy, modified in 1992, made a remarkably direct admission: the general formulations of 1968 had not become a detailed implementation strategy with assigned responsibilities and financial and organisational support. That is not a critic's retrospective complaint. It is the next policy diagnosing the previous one. [3]

Then came the 1993 Yash Pal report, Learning Without Burden. Its deeper concern was not the weight of the schoolbag. It was the burden of non-comprehension. "A lot is taught, but little is learnt or understood," the report said. [4]

The national curriculum frameworks of 1975, 1988 and 2000 were followed by NCF 2005, which explicitly called for moving away from rote methods, connecting knowledge to life outside school, going beyond textbooks, and making examinations more flexible. [5]

The Right to Education Act of 2009, in force from April 2010, made free and compulsory education for children aged six to fourteen a legal right. That expansion of the state's obligation matters. Access is not a minor achievement to dismiss because learning remains uneven. [6]

NEP 2020 and NCF for School Education 2023 are the latest major turn. The latter brings ages three to eighteen into an integrated curriculum framework, following the 5+3+3+4 structure. [7, 8]

Across roughly a century, the aims have changed from colonial mediation to citizenship, access, development and now broader capabilities. Yet the complaint about learning without understanding keeps returning.

So the useful historical question is not "Who first said education should change?" It is "Why does the system repeatedly absorb the language of change without reliably changing the child's experience?"

Read NEP before attacking it

A fair criticism starts by acknowledging what NEP says.

The 5+3+3+4 structure recognises different developmental stages and brings early childhood into the frame. The policy itself says these are curricular and pedagogical stages, not a requirement to rebuild schools into new physical units. Section 4.5 calls for reducing content to core essentials to make room for inquiry, discussion and critical thinking. Sections 4.34-4.35 call for formative assessment and a multidimensional progress card, including portfolios, self-assessment and peer assessment. [7]

It also proposes flexibility across subjects, vocational exposure in Grades 6-8, and a ten-day bagless period with local practitioners. It recommends the home language, mother tongue, local or regional language as the medium of instruction through at least Grade 5, preferably Grade 8 and beyond, "wherever possible." That qualification matters. It is not a simple nationwide ban on English-medium education. [7]

NEP even criticises the harm caused by board and entrance examinations and the resulting coaching culture. It does not abolish board examinations. It proposes redesigning them around core capacities and allowing up to two occasions in a year. [7]

There is real movement beyond the text. CBSE's final notification of 25 June 2025 provides for two Class X board examinations from 2026, with the first mandatory and a second opportunity for eligible improvement or compartment cases. That is a CBSE Class X reform, not proof that every Indian board or Class XII has adopted the same arrangement. [9]

The question, then, is not whether the policy contains progressive language. It does. The question is whether its operating conditions can survive contact with the incentives around schools.

Five places where the reform can break

1. The examination can eat the curriculum

An exam is not inherently the enemy of thinking. A well-designed one can test it. Standardised assessments also provide a check against favouritism and inflated school grades.

But if a classroom rewards explanation while the next selection gate rewards rehearsal, parents have a reason to buy rehearsal. Calling them anxious or backward misses the point. They are responding to risk.

Two attempts can reduce the cost of one bad day. They do not, by themselves, change what the questions reward. They can also create two rounds of preparation if the surrounding incentives stay the same. That is a risk to test, not an outcome I can claim has already occurred.

The market around the school shows the same response to risk. In UDISE+ 2025-26, India had about 14.67 lakh schools and 24.72 crore enrolled students. [18] Private unaided schools were 23.3% of schools [19] but enrolled 40.0% of students, up from 33.4% in 2022-23. Government schools enrolled 48.1%, down from 49.2% a year earlier. Between 2023-24 and 2025-26, government schools enrolled about 86 lakh fewer children and private unaided schools about 88 lakh more, while total enrolment fell by about 8 lakh. [20] In pre-primary, the private unaided share was 71.4%. [19]

Families are paying where they believe the school lowers a risk for their child. I cannot say from this data what they believe they are buying: UDISE+ does not record parental motives, a shift in enrolment is not proof that the same families moved, and a rising private share is not proof that private schools teach better. The picture also differs by state. Private unaided schools enrolled 64.5% of students in Telangana, 50.0% in Karnataka and 6.3% in West Bengal. [20] If the paid route is where families see better odds, a reform that leaves selection signals unchanged will meet the same demand.

A reform should be judged by sample questions, marking rubrics and what teachers must do to prepare students, not by how often the word "competency" appears in a circular.

2. A grade is not a learning level

NEP made universal foundational literacy and numeracy by 2025 its highest priority. The NIPUN Bharat implementation guidelines set a 2026-27 horizon. Those are different dates, and neither should be quietly converted into a claim of success. [7, 10]

ASER 2024 gives a more uncomfortable picture, alongside genuine recovery. In its rural household survey, 23.4% of government-school children in Standard III could read a Standard II-level text, up from 16.3% in 2022 and above 20.9% in 2018. Among Standard V government-school children, the figure was 44.8%. Across school types, 30.7% of Standard V children could do the survey's division task. [11]

These are rural estimates and basic tasks, not a census of every Indian child or a measure of independent reasoning. Nor can a before-and-after survey isolate NEP's causal effect. Still, recovery and an unfinished learning problem can both be true.

A child who cannot comfortably read the lesson cannot be helped merely by renaming that lesson multidisciplinary. If syllabus completion outruns comprehension, the next grade inherits the gap.

India already has an answer worth learning from. Randomised evaluations of Pratham's Teaching at the Right Level approach show that instruction organised around actual learning levels, with the right delivery and support, can improve basic skills. This evidence is about foundational learning in particular settings. It is not proof that one method solves the entire secondary curriculum. [12]

Before importing a foreign school model, we should ask why useful evidence from our own classrooms is not the starting point.

3. We ask teachers for a new practice, then measure old compliance

NEP calls for teacher autonomy, less non-teaching work and at least fifty hours of continuing professional development each year. Those are serious commitments. [7]

But a training-hour count cannot tell us whether a teacher can recognise a misconception, lead a discussion or assess a child's explanation fairly.

Imagine being asked to complete the syllabus, keep records, prepare students for selection, and also give each child thoughtful feedback. The arithmetic of teacher time matters. A portfolio does not assess itself. A new progress card can become another form to fill unless time, staffing and practical support change with it.

I would rather see a teacher discuss three pieces of student work with colleagues and improve the next lesson than collect another certificate that says a course was completed. This is a proposal for how to judge professional learning, not a claim that Indian teachers do no such work already.

4. Choice on paper can become privilege in practice

A policy can remove the formal boundary between arts, science and vocational subjects. A school still needs teachers, timetables, materials and a viable number of students to offer the combination.

A wealthy school and a small, thinly staffed school cannot exercise the same menu of choices simply because both received the same framework. Reform should publish what a child can take in practice, and fund the missing capacity. Otherwise flexibility risks becoming another premium feature.

The same applies to language. Understanding through a familiar language and gaining access to English should not be treated as rival ambitions. The hard work is building good materials, preparing multilingual teachers and giving families a credible route to both.

Vocational exposure also needs an equity test: does every child get to connect making with mathematics, design and science, or do some children get subtly directed towards narrower futures? I want practical work to widen possibility, not become an early sorting mechanism.

5. Ambition without a delivery contract becomes a recurring announcement

The financing promise has a long history. The 1968 text spoke of 6% of national income; NEP 2020 renews a target of public education investment reaching 6% of GDP. Those measures should not be casually treated as identical. I am not using the policy's historical expenditure estimate as a current 2026 figure. [3, 7]

My criticism is simpler: an aspiration is not a funded delivery plan.

Which level of government owns each change? What reaches the classroom? What happens when a deadline slips? Which practice stops to make room for the new one?

Education involves states, boards, local institutions and families. A national policy is neither a single timetable nor proof of uniform implementation. We need fewer claims of transformation based on document release, and more public evidence of what changed in teaching and learning.

Other countries offer mechanisms, not a model to copy

Finland is often reduced to a story about freedom from exams. Its national core curriculum still specifies subjects. Schools must offer at least one multidisciplinary learning module each year, and assessment remains tied to subject objectives. There is also a national matriculation examination at the end of general upper secondary education. [13]

The lesson I take is the combination: a shared foundation, local curriculum work and varied evidence of progress. Removing a test is not the same as building the professional capacity that makes other assessments trustworthy.

Singapore offers a different mechanism. Full Subject-Based Banding, rolled out progressively through 2024, replaces separate Express, Normal (Academic) and Normal (Technical) streams with mixed form classes and subjects taken at different levels. [14]

That challenges the idea that one label should describe the whole child. India can learn from subject-level flexibility without pretending that a compact city-state and a large federal country face identical delivery problems.

China's 2021 "Double Reduction" policy targeted homework and after-school tutoring burdens. It shows that the education system includes the market surrounding school, not only the classroom. The ministry's account is an official implementation report, not independent proof that pressure disappeared. [15]

I would not import coercive market controls as a shortcut. The more useful question is what produces demand for tutoring. Restricting supply cannot, on its own, remove the competition families are trying to manage.

The United States is not a single curriculum to imitate: education is primarily a state and local responsibility. Its variation creates room for experimentation, but should not be confused with guaranteed quality. NAEP 2024 recorded further reading declines in both fourth and eighth grades compared with 2022. [16]

That does not prove decentralisation caused the decline. It does puncture the easy story that autonomy alone solves learning.

These examples are institutional comparisons, not a league table. ASER and NAEP test different populations and tasks. I am not ranking children across countries using incompatible measures.

AI makes the missing evidence harder to ignore

When AI can produce a polished answer, the finished answer becomes weaker evidence of what the child understands.

A ban may protect a particular assessment. Unrestricted use may conceal a learning gap. Neither is a complete curriculum strategy.

The distinction I care about is between getting AI to do a task and directing AI while remaining responsible for the result. Direction requires deciding what to ask, what evidence to trust, what constraints matter and when to reject an output. That depends on knowledge. A child cannot reliably challenge an explanation of fractions without understanding fractions.

This is where my work on Ilanooo connects to the larger question, and why our paper argues that children should learn by creating with the machine. The ambition is to help young people direct AI. It is not to replace foundational learning with prompting, and it is not evidence that we have already solved assessment.

I want a learning activity to leave a trail of reasoning: the first attempt, a question asked, an error noticed, a source checked, a revision explained. Then I want the child to handle a changed version of the problem, sometimes without AI. Not every task needs an elaborate record. But some must let us distinguish polished output from transferable understanding.

Consider a proposed task: design a water-saving plan for a school. A child estimates demand, checks the assumptions behind an AI-generated recommendation, compares alternatives, then explains what changes if the budget is halved. The project connects maths, science and civic judgment. The evidence is in the decisions and the explanation, not the beauty of the presentation.

This must not become permanent surveillance of children. More prompts, more tokens or more screen time do not prove deeper thought. Collect only what serves learning, explain it to families, and make teacher judgment central. UNESCO's guidance calls for privacy safeguards, teacher preparation and protection of human agency. A product founder has to take those constraints seriously too. [17]

What I would change first

I would start with a limited, public test, not another promise to redesign everything at once.

First, protect daily foundational learning and allow teaching to respond to demonstrated learning levels. Track improvement without permanently labelling a child as weak.

Second, choose one unit each term where every child must explain, revise and transfer an idea. Allow a written explanation, conversation, diagram or practical demonstration. Reasoning should not be confused with fluent English or confident public speaking.

Third, give teachers protected time to compare student work and calibrate judgments. Keep a modest common assessment alongside richer classroom evidence. Portfolios can be coached or unfairly marked too. We need checks on both systems.

Fourth, connect that evidence to the signals that matter: report cards, board assessment and admissions. Pilot carefully before raising the stakes. Otherwise the new work becomes an extra burden beside the old competition.

Finally, publish the cost, the implementation difficulties and whether children improved, including those who started furthest behind. If richer assessment consumes too much time or widens unfairness, revise it. My preferred idea should face the same scrutiny I am asking of NEP.

India does not need to choose between strong foundations and independent thought. It needs a system in which one makes the other possible.

The most important curriculum question is not how much content we can fit into a childhood. It is what a child can understand, question and do with that content.

NEP has already said many of the right things. Now we need to stop treating the saying as the achievement.

Sources and scope

Selected sources, checked 11 October 2026; enrolment figures re-checked against the UDISE+ dashboard on 15 October 2026. Policy provisions are distinguished from implementation and proposed reforms. ASER 2024 is used as a dated rural learning snapshot, not a claim to be the latest available assessment. International examples illustrate specific mechanisms; they are not causal comparisons of whole systems.

1. Macaulay, Minute on Education (1835), Columbia-hosted transcription.
2. The Wardha Scheme, historical account and scheme analysis.
3. National Policy on Education 1968; NPE 1986 with 1992 modifications.
4. Yash Pal committee, Learning Without Burden (1993), especially the foreword and introduction.
5. NCF 2005, executive summary and historical context; NCERT framework archive.
6. India Code, Right of Children to Free and Compulsory Education Act 2009.
7. NEP 2020, sections 2.2, 4.3-4.6, 4.11, 4.26, 4.34-4.40, 5.12-5.15 and 26.
8. National Curriculum Framework for School Education 2023, chair's introduction.
9. CBSE, final Class X two-board-examination notification, 25 June 2025.
10. NIPUN Bharat guidelines, 2021, vision and implementation horizon.
11. ASER Centre, ASER 2024 National Findings, pp. 46-48.
12. J-PAL, randomised evaluation of Pratham learning camps in Uttar Pradesh.
13. Finnish National Agency for Education, basic education core curriculum and general upper secondary education.
14. Singapore Ministry of Education, Full Subject-Based Banding.
15. China Ministry of Education, Double Reduction implementation press conference, December 2021.
16. US Department of Education, federal role; NAGB, NAEP 2024 release (January 2025).
17. UNESCO, guidance on generative AI in education, September 2023.
18. Ministry of Education, UDISE+ 2025-26 release: PIB, 7 July 2026 and PIB explainer (24.72 crore students, about 14.67 lakh schools).
19. Arun C Mehta (formerly NIEPA), analysis of UDISE+ 2025-26 tables: share of schools and the pre-primary share. Secondary analysis of the official tables.
20. UDISE+ dashboard, official enrolment by school management, all-India and by state, 2022-23 to 2025-26, read on 15 October 2026. Shares are computed from those counts.