A Class 3 teacher writes a simple challenge on the board: “The school library has 24 books. How could we arrange them so that children can find what they need?” Within minutes, pupils propose groups, invent sorting rules, argue about exceptions, and change their minds. No laptop is switched on. Yet the class is practising the foundations of computational thinking: identifying a problem, sorting information, designing a process, and testing whether it works.
For Classes 3–5, the most useful way to teach the new CBSE Computational Thinking and Artificial Intelligence curriculum is to start with concrete, unplugged tasks embedded in Mathematics and The World Around Us (TWAU). Children should manipulate, explain, compare, correct, and reflect before they are asked to use a digital tool. This article provides eight classroom-ready activities, assessment prompts, differentiation ideas, and a five-day teaching sequence.
The official starting point is the CBSE CT & AI curriculum and resource hub. CBSE introduced the framework for Classes III–VIII for the 2026–27 academic session, as confirmed in its April 2026 launch circular and the Ministry of Education launch announcement. The activity sequences below are teacher-designed examples, not a reproduction or claim of official CBSE-prescribed exercises.
What does CBSE actually expect from Classes 3–5?
The CBSE framework introduces computational thinking and AI understanding across Classes 3–8. It is a mistake, however, to teach the preparatory-stage years as a simplified chatbot course. For Classes 3–5, the official curriculum framework places computational-thinking questions inside the existing Mathematics and TWAU subjects and aligns accompanying resources with those textbooks.
In practical terms, teachers can look for six kinds of observable thinking:
Decomposition: breaking a larger problem into smaller, manageable parts.
Pattern recognition: noticing repetitions, relationships, or differences.
Abstraction: choosing which information matters for the task.
Algorithmic thinking: describing an ordered process someone else can follow.
Testing and debugging: finding where a rule or procedure fails and improving it.
Reasoning about data and fairness: interpreting information and noticing that a classification or decision may be incomplete or biased.
These are teaching lenses, not six separate subjects to squeeze into an already crowded timetable. A child can practise several during a familiar mathematics lesson. The first question for a teacher is therefore not “Which AI app must I download?” but “Which thinking behaviour should my pupils be able to show by the end?”
CBSE also released grade-specific student and teacher resource books, linked from its official CT & AI resource page. Schools should match any locally designed activity to the exact chapter, current grade handbook, accessibility needs, and assessment guidance in those official materials. A Grade 5 extension should not automatically be treated as the expected outcome for every Grade 3 child.
How to prepare an unplugged computational-thinking lesson
Start with a problem that children already understand: arranging classroom objects, planning a route, checking a shopping bill, organising a water-saving routine, or interpreting a weather chart. Give them materials that allow visible thinking: bottle caps, paper slips, number cards, coloured shapes, a grid drawn in chalk, or a familiar NCERT chapter.
Plan the lesson around four stages:
Experience: children try a challenge individually or with a partner before a rule is demonstrated.
Explain: they say or show how they chose a strategy.
Test: another child follows the rule, or a new example checks whether it still works.
Improve: children revise one step and describe what changed.
Keep teacher instructions short. “Can someone else follow your rule?” is usually more revealing than “Did you get the right answer?” That shift matters because a procedure that only works while the teacher gives hints has not yet become an independently usable strategy.
For large Indian classrooms, make a task repeatable in pairs rather than designing a unique set of materials for every pupil. Put one challenge on the blackboard; ask half the class to invent a rule and the other half to test it. A written record can be a small sketch, a two-column table, or three numbered steps, rather than an elaborate worksheet.
Activity 1 — Class 3: The missing pattern detective
Focus: pattern recognition and explanation. Time: 15–20 minutes. Materials: chalkboard and number or shape cards.
Write a sequence such as 2, 5, 8, 11, \_\_, \_\_. Invite pupils to propose the next two terms and explain the change between consecutive numbers. Then show a shape sequence in which a triangle alternates orientation while its colour changes every two positions.
Do not immediately announce a rule. Ask: “What changes? What stays the same? Can a different rule fit the first few terms?” Then extend the sequence so that children can test competing hypotheses. This helps them discover that an apparent pattern needs evidence, not just a guess.
For a Grade 3 class, begin with one changing feature. Support a pupil who is struggling by using concrete counters or by drawing arrows between successive values. For a confident pupil, introduce two changing features, or ask them to invent a pattern with a deliberate error for a partner to detect.
Evidence of learning: a pupil can name the rule, extend it, and explain how they checked it. A correct missing number without an explanation is useful information, but not enough to show that the pattern was understood.
Activity 2 — Class 3: Follow my directions, then debug them
Focus: sequencing, clarity, and debugging. Time: 20 minutes. Materials: a floor grid or a grid on paper.
Draw a simple grid with a start square, a finish square, and two obstacles. In pairs, one pupil gives a sequence of moves—“forward one square, turn right, forward two squares”—and the other follows it using a marker. If the marker enters an obstacle, both examine the instructions.
The important learning moment is not the failed route; it is the revision. Ask: “Which instruction caused the problem? Could you change one step instead of beginning again?” Encourage pupils to read the revised sequence to a different partner. If the second partner reaches the finish without guessing, the instructions have become clearer.
A useful extension is to ask pupils which details can be omitted without changing the route. Does the marker's colour matter? Does the direction it faces matter? This is an accessible introduction to abstraction.
Evidence of learning: the pupil can give an ordered set of instructions, identify an incorrect step, and correct it. Avoid ranking children by speed: careful explanation is more important than finishing first.
Activity 3 — Class 4: The classroom data detectives
Focus: classification, representation, and interpreting evidence. Time: 25–30 minutes. Materials: paper slips and a chart.
Ask pupils which method they used to travel to school today: walking, bicycle, school bus, other bus, or car. Use anonymous tally marks; do not collect names, addresses, vehicle registration information, or individual travel routes. Convert the tallies into a simple table or bar chart.
Now ask three different questions. “Which category has the highest count?” is a reading question. “How many more pupils selected one category than another?” requires a comparison. “Does our class chart tell us how all children in the city travel?” introduces a crucial limit: a small class sample cannot automatically represent a much larger population.
Change one ambiguous category: what should happen if a child walked and then took a bus? Pupils will discover that a counting rule must be decided before collecting the data. The purpose is not to create a perfect statistical study; it is to make the rule and its limits visible.
Evidence of learning: children can sort responses consistently, read a representation, and give at least one reason the data may be incomplete.
Activity 4 — Class 4: Design a fair library sorter
Focus: decomposition, categories, exceptions, and fairness. Time: 25 minutes. Materials: cards with invented book titles, topics, and reading levels.
A school library wants a better way to help younger children find books. Groups receive 12 cards and must design a sorting system. One group sorts by topic, another by language, and another by reading level. Each group then exchanges its rules with a partner group.
Introduce difficult cases. Where should a bilingual science picture book go? What if a book has no clear age label? What if a child is interested in advanced science but prefers shorter text? The point is to surface the limitations of any apparently neutral classification.
Explain that automated recommendations also depend on rules and data, but this paper activity is a model for thinking, not proof of how a particular commercial AI system works. Children need not learn machine-learning mathematics to recognise that a decision system can overlook a person or a useful exception.
Evidence of learning: groups can state their criteria, apply them to new examples, and revise a rule when an exception appears.
Activity 5 — Class 5: Plan the school water-saving routine
Focus: decomposition, constraints, algorithms, and iteration. Time: 30–35 minutes. Materials: scenario cards and an observation checklist.
Present a familiar challenge: taps are sometimes left running after lunch. The class must suggest a system for checking and improving water-use habits without blaming individual children. Divide the task into smaller questions. Where could reminders help? Who can report a leaking tap? Which actions are safe for pupils and which need an adult?
Groups draw a flowchart: notice a tap, decide whether it is running unnecessarily, close it when safe, report a leak to an adult, and check later whether the issue was resolved. Give them a scenario in which a tap does not close properly. Their system should not instruct a child to force broken equipment.
Ask every group to test its flowchart against two situations. Is the procedure understandable? Does it handle an exception? Has responsibility been assigned appropriately? This links logical sequencing to TWAU and shared responsibility.
Evidence of learning: pupils can divide the problem, name a constraint, produce ordered steps, and explain one improvement after testing.
Activity 6 — Class 5: Could this prediction be wrong?
Focus: data literacy, limitations, and responsible AI understanding. Time: 25 minutes. Materials: fictional weather-record cards.
Display six simple weather observations, such as combinations of cloudy or sunny skies and whether it rained later. Ask children to devise a rule that guesses whether rain is likely based only on the recorded cards. Test the rule on two additional examples, including a case where the rule gives a wrong answer.
Invite the class to consider: “Is a pattern always a promise? Which information was missing? How would collecting more observations help?” Explain carefully that some AI tools learn relationships from data, but the classroom rule is only an analogy. Real AI models can work differently, and a small exercise cannot establish a system's real-world accuracy.
Avoid recording real pupils' names, attendance, performance, or personal characteristics as prediction inputs. The teaching target is thoughtful uncertainty: a prediction can be useful while still being wrong.
Evidence of learning: pupils distinguish a prediction from a fact and offer a reason to check data or question a decision.
Activity 7 — Classes 3–5: The human instructions challenge
Focus: precise language, iteration, and collaborative reasoning. Time: 20 minutes. Materials: paper, pencils, and simple geometric designs.
One pupil sees a drawing containing four shapes. Without showing it, they describe the drawing to a partner. The partner reproduces it. Then the two compare the images and identify which instructions were unclear: “Draw a square” may need a size and position; “put a circle near it” may need a direction.
Repeat the challenge using improved instructions. With younger pupils, limit the drawing to two shapes. With older pupils, introduce orientation, rotation, and explicit constraints. Let students use their strongest classroom language during first explanations where appropriate, then practise precise English mathematical language together.
The learning is about communication and executable steps, not about pretending that human classmates are machines. A pupil who asks a clarifying question is often doing precisely the thinking the task is intended to encourage.
Evidence of learning: revised instructions enable a new partner to reproduce the design more reliably.
Activity 8 — Classes 4–5: The fairness test
Focus: ethical questioning and recognising incomplete rules. Time: 20–25 minutes. Materials: fictional decision cards.
A school club has ten seats and more applicants than places. Present a proposed rule: “Only pupils who can stay after school may join.” Ask whether it gives everyone a fair opportunity. What about a pupil who takes the last bus or has family responsibilities? Could the club run a lunch-time option or choose participants through a transparent rotation?
Introduce another rule: “Choose only pupils who scored highest on the last mathematics test.” Is that an appropriate criterion for a creative-design club? Pupils practise identifying whose needs were left out and which criteria actually relate to the decision.
Do not ask children to disclose personal circumstances, disability, family income, or other sensitive details. Keep all cards anonymous and hypothetical. The lesson is not that every decision has one perfect solution; it is that rules deserve scrutiny and affected people deserve consideration.
Evidence of learning: pupils identify an exclusion, explain why the rule might be unfair, and suggest a more relevant or inclusive alternative.
Which activities work best at each grade?
This is a suggested progression, not a replacement for CBSE's grade-wise outcomes or official handbooks.
| Class | Good starting challenge | Thinking to observe | Simple evidence |
|---|---|---|---|
| 3 | Pattern detective; direction grid | One- or two-feature patterns, ordered steps, correction | Explained rule and revised route |
| 4 | Data chart; library sorter | Classification, comparisons, ambiguous cases | Consistent sorting and justified exception |
| 5 | Water-saving flowchart; prediction test | Constraints, decomposition, uncertainty, iteration | Tested procedure and identified limitation |
In mixed-ability groups, make the same problem accessible at different levels instead of labelling children as “good at AI” or “poor at AI.” A learner may sort three concrete cards while another creates a more abstract sorting rule. Both can be challenged to explain why their choice works.
A five-day mini-unit for a busy CBSE classroom
A teacher need not begin with an additional, standalone “AI period.” The sequence below can be woven into Mathematics and TWAU lessons, subject to the school's timetable and official curriculum mapping.
Day 1 — Notice patterns. Start with ten minutes of number and shape rules. Ask for written or spoken explanations, not only completed sequences. End with an exit prompt: “What clue helped you decide?”
Day 2 — Give and test instructions. Use a floor grid or drawing directions. Let a partner follow the sequence exactly as written. The reflection prompt is “Which instruction needed changing?”
Day 3 — Classify and count. Run the library sorter or anonymous tally activity. Ask who might be missing or misclassified and how the rule should handle exceptions.
Day 4 — Solve a community problem. Use the water-saving case or a classroom-materials distribution problem. Groups design a procedure and test it against an unexpected event.
Day 5 — Explain a revision. Return to one previous task. Pupils compare their first and revised solutions, explain the change, and try a related but unfamiliar example to check transfer.
For each day, reserve a short reflection: “What was difficult at first? What did you change? Where else could that strategy help?” This is how computational thinking becomes a reusable learning habit rather than a new list of vocabulary to memorise.
How should teachers assess computational thinking?
The most revealing evidence often appears during an unsuccessful first attempt. A pupil notices that a pattern rule breaks, challenges a classification, or asks to check the procedure. Those moments show active reasoning. A finished worksheet alone does not show which ideas the learner developed independently.
Use a four-part observation record for a small sample of pupils each day, rotating through the class:
Understands the task: identifies the goal and relevant information.
Explains a strategy: uses objects, sketches, words, or steps to show a method.
Checks a result: tests a new case or detects a mismatch.
Revises with growing independence: improves after a question or a peer's feedback.
Record brief evidence rather than a vague judgement such as “very intelligent.” For example: “Used two clues; changed the rule when the third card did not fit.” That observation tells the teacher what to scaffold next.
A practical feedback sequence is notice → question → test → transfer. Notice the student's approach. Ask one question that focuses attention. Invite a test. Finally, present a different problem and reduce support. The teacher learns much more about independence from the final transfer task than from a correct answer copied after a demonstration.
If a learner needs an accessible format—large print, tactile shapes, oral explanation, more processing time, or peer-supported instructions—adapt the means of expression while preserving the core thinking goal. Do not use unfamiliar English vocabulary or rapid handwriting as a hidden test of computational reasoning.
How can teachers explain AI responsibly to young children?
Begin with an age-appropriate distinction: a computer may follow instructions; an AI system may use learned patterns or other computational methods to generate classifications, suggestions, or predictions; neither should be trusted simply because it sounds confident. This is an introductory explanation, not a complete technical account of AI.
Use examples that children can inspect without personal data. A weather guess, a book recommendation, or a pretend sorting machine can support discussion. Avoid requiring children to create personal accounts, submit names or photos, or chat independently with public generative-AI services. Schools must follow their own data-protection, safeguarding, and age-appropriate technology rules.
A classroom agreement can be brief: “Think first. Ask how the answer was made. Check important claims. Protect private information. Tell an adult if a tool produces something upsetting.” For Classes 3–5, teacher-guided discussion usually matters more than hands-on exposure to an open chatbot.
CBSE's teacher-training notification for 2026–27 explicitly includes play-to-abstraction, assessment, interdisciplinary applications, and ethics. That supports a balanced view: the new subject area is about ways of thinking and responsible judgement, not the number of AI products installed.
What does a teacher-as-coach change?
An effective teacher still models ideas, teaches mathematics, and corrects misconceptions. Coaching adds a deliberate step: instead of immediately rescuing the pupil with a completed solution, the teacher helps the learner notice available resources and decide on the next useful move.
Consider a Class 4 student who insists that a sorting activity is impossible because two cards belong in both groups. Rather than providing the category labels, ask: “What rule did you choose? Which card does not fit? What could you change so the rule becomes clearer?” The student remains responsible for thinking, while the teacher provides appropriate guidance.
This is especially relevant when AI can produce an answer instantly. Schools need children who can ask better questions, detect a flaw, and explain a decision in their own words. Those competencies cannot be inferred from a polished final product.
Learning Coach's approach to learning independence, teacher questioning, and supported problem solving offers educators a useful professional-development direction. Teachers interested in that approach can explore the Learning Coach training pathway. The exact availability and conditions of any specialisation should be checked on the current programme page; no course completion or certification outcome is implied by reading this article. For complementary reading, the English Learning Coach book catalogue includes titles about coaching and learning through problem solving.
Mistakes to avoid when implementing the 2026–27 curriculum
Mistake 1: equating AI literacy with chatbot proficiency. At this age, the ability to reason about a rule or data set is more foundational than memorising prompts.
Mistake 2: turning every activity into a screen lesson. Unplugged work can reveal children's thinking and makes participation less dependent on devices, electricity, or connectivity.
Mistake 3: teaching vocabulary before experience. Pupils understand “find what repeats” through a task before they need the term “pattern recognition.”
Mistake 4: rewarding speed instead of checking. Rapid answers can hide weak explanations. Invite students to test a rule in a new situation.
Mistake 5: treating every classroom activity as official. Teacher-developed examples are useful supplements; schools should consult CBSE's current grade handbook and resource materials for formal alignment.
Mistake 6: making unsupported claims about learning outcomes. An engaging lesson does not prove long-term improvement. Collect evidence of independent explanation and later transfer before drawing stronger conclusions.
Frequently asked questions
Is coding compulsory for Classes 3–5 under the new CBSE AI framework?
For these grades, CBSE's curriculum framework describes computational-thinking integration in Mathematics and TWAU, supported by grade-level resources. Teachers should not presume that the preparatory-stage curriculum requires each child to learn a programming language or operate a chatbot. Follow the latest official school and grade guidance.
Can teachers run the activities without computers?
Yes. The eight activities in this guide use paper, cards, conversation, charts, or a floor grid. Their purpose is to practise the thinking that underpins later understanding of computing and AI, not to claim that unplugged games alone constitute a complete AI education.
How can a teacher manage 40 pupils?
Teach one routine to the whole class, then use fixed pairs or small groups. Rotate roles: designer, tester, explainer, and checker. Collect one short piece of evidence per group rather than a large stack of identical worksheets. The teacher can observe different pupils on successive days.
How do we know whether children are becoming more independent?
Give a related problem they have not rehearsed. Ask them to choose a method, explain why it might work, notice an error, and revise with fewer prompts. Record changes over time. Independence is demonstrated through transfer and judgement, not merely through a faster answer.
The essential takeaway
The most promising question in the CBSE AI curriculum for Classes 3–5 is not “How early can pupils use intelligent machines?” It is “How early can pupils learn to make their own reasoning visible, testable, and improvable?” With a few cards, a grid, a chart, and purposeful coaching questions, teachers can turn an ambitious curriculum announcement into authentic classroom learning.
Use the official CBSE CT & AI resources as the authority for grade-level alignment, and use these suggested activities to make the intended thinking observable in everyday lessons. The aim is not to produce children who can repeat the vocabulary of AI, but learners who are increasingly capable of asking, checking, and learning for themselves.
References and official resources
CBSE — Curriculum framework for CT & AI, Classes III–VIII, academic session 2026–27.
CBSE Circular Acad-15/2026, 1 April 2026 — launch of curriculum framework.
CBSE Circular Acad-18/2026, 9 April 2026 — student and teacher resource books.
CBSE Training Notification TRG-02, 9 April 2026 — 2026–27 teacher development themes.
Ministry of Education / Press Information Bureau, 1 April 2026 — official curriculum launch.
