Column | What Should We Learn And Unlearn To Bring The Impact of AI To Our Extremely Capable Children?

In April 2026, the Central Board of Secondary Education (CBSE) launched its Computational Thinking and Artificial Intelligence curriculum for Classes 3 to 8, giving formal curricular shape to something that was well-envisioned in the National Education Policy 2020. It set off a familiar debate: is this too much, too early, for classrooms as varied as ours?

Having helped design this curriculum and interacted with the teachers implementing it, I can see that there are challenges. The task before us is clear: what should we learn and unlearn to bring the impact of AI to our extremely capable children?

Key Challenges

First, I start with the question every parent—and every sceptic—asks: can you seriously teach supervised, unsupervised and reinforcement learning to a Class 6 child?

The curriculum answers this with a scene any child recognises. A class teacher collects notebooks after a test. If every notebook carries a name, she knows exactly whose it is; this is akin to labelled data, and learning from it is supervised learning. If the names are missing, she must group the books by handwriting, which is finding structure without labels: unsupervised learning. A chess programme that gets better after every win and loss is learning from rewards—reinforcement learning. Without bringing in intimidating equations or code, the idea is to just map three names to three familiar experiences. This is not a trick but a principle with a distinguished pedigree—one that shifts the burden from the child’s capacity to our imagination in presenting the idea. Psychologist Jerome Bruner argued in The Process of Education: “Any subject can be taught effectively in some intellectually honest form to any child at any stage of development.” Bruner also gave education the instrument for acting on this conviction: the spiral curriculum—the same design our National Curriculum Framework prescribes. Class 6 meets these ideas at the level of basic awareness. Class 7 returns to them through the domains of AI. By Class 8, students train a small model themselves and grapple better with the problem. The concept is not taught once, badly; instead, it is visited thrice, each pass more intellectually honest than the last.

The second thing to emphasise is what AI education is not: tool training. Students are already experimenting with chatbots, image generators and coding tools; while this is valuable exposure, the tools remain sealed black boxes that are possibly more admired than understood. The purpose of AI education is to open that box, and merely honing prompting skills will never achieve that. I find it useful to think of a ladder instead: AI literacy—knowing what these systems are and are not; AI skills—using and questioning them well; AI knowledge—the mathematics and engineering underneath; and AI development—building such systems oneself.

CT, comprising decomposition, pattern recognition, abstraction and algorithms, is the ideal foundation on which to build AI. It primes the child’s thinking, and without it, children can only marvel at AI, never making sense of it. The top rungs belong largely beyond school—and they take care of themselves. The ceiling has never been the problem; the work of a national curriculum is to raise the floor. The key challenge in our classrooms is not ability or aptitude, but exposure—and exposure is exactly the thing a curriculum can manufacture.

Thirdly, AI cannot live only in the computer-science period. The language class is where machine translation is best examined. Biology classes already teach classification. Geography runs on data and maps. A civics discussion on social media is incomplete without asking how recommendations decide what millions of teenagers see next.

CT, admittedly, sits most naturally within mathematics, and the curriculum deliberately aligns it chapter by chapter with the maths textbook. But AI concepts surface everywhere, and this is what makes teaching AI very important, even to the child who has decided that math is not for them. Weaving AI into every class would be a great way to ensure every last student is included.

The spiral curriculum does have one starting cost: a student entering Class 7 in 2026 never rode its lower turns; the foundations her juniors will build across Classes 3 to 6 were simply not on offer when she passed through. Every school introducing the curriculum this decade faces this bridge problem. The teachers will have to creatively ramp up the children, providing them with essential experiences from lower classes.

The success of this curriculum rests on the shoulders of our teachers. In recent orientation sessions with hundreds of school teachers, what I have seen is appetite, not resistance—sharp questions, honest anxieties, real adaptability. But a webinar is not the same as training. Our teachers are adaptable, but our students are even faster. Therefore, teacher development must be sustained and embedded in subjects teachers already teach, not delivered as a one-time inoculation.

Exposure Matters

Fittingly, CBSE has assigned “Computational Thinking and Understanding AI” as the training theme for the current academic year. The curriculum also helps here by design: its activities are illustrative, and its projects suggestive, empowering teachers with the autonomy to contextualise appropriately for their own classrooms.

That autonomy is what my standing advice to teachers requires: design every activity for the student in the room with the least prior exposure, not for the one the syllabus imagines. That student is not a limit on the class; she is this year’s starting line. The whole point of the design is that the line keeps rising.

Finally, the biggest elephant in the room is assessment. If we examine CT & AI with three-mark “write a short note” questions, we will end up with rote learners, rewarding the coached student over the curious one. A child who can recite “machine learning is the ability of machines to learn from data” has just learned a bunch of “strings” that are ephemeral. A child who has trained a model to sort photographs of her school’s waste, documented what confused it and had a classmate review her work has learned something no examination can extract. Projects, portfolios and peer review are not soft alternatives to rigour; in this case, they are the rigour.

Ultimately, the real meaning of introducing AI in schools comes down to three shifts already underway: from using tools to building concepts; from the computer-science period to every subject; and from the three-mark definition to the project and the portfolio. There is no shortage of capability in our students and teachers. Yet, we must raise the floor, and this curriculum strives to deliver just that.

(Views expressed are personal)

Karthik Raman was part of the committee that designed the CBSE CT & AI curriculum for Classes 3-8

(This story appeared in Outlook magazine’s August 3 issue, 'The AI Divide', which focuses on how India's AI education ambitions are colliding with the reality of inadequate digital infrastructure, undertrained teachers and AI tools that are not built around Indian students' cultural context)