Teaching AI without a computer lab

In UDISE+ 2023–24, 41.4% of schools in Jammu & Kashmir reported a computer facility and 45.3% reported internet. (UDISE+ 2023–24) Many teachers reading this are in the other half — or in a school where the lab exists on paper and the computers stopped working years ago.

If that is you, this is the most useful thing I can tell you: the big ideas of AI do not need a computer. The computer shows them faster. It does not teach them better.

AI is ideas before it is software

Think about what we actually want a fourteen-year-old to understand:

• A machine learns a rule from examples, not from being told.
• Its examples can be unfair, and then so is the machine.
• You test a model on examples it has never seen, or you are fooling yourself.
• A model that writes fluently can still be confidently wrong.

Every one of those can be taught with chalk, paper and objects the children already know. Computer scientists have done this for decades: the CS Unplugged activities from the University of Canterbury teach algorithms, data, compression and AI with cards, string and games instead of screens. (CS Unplugged)

Three lessons, unplugged

Every mission in the AmanahVale books has a No laptop today? version. Here are three, so you can see the pattern.

Teach a machine to sort apples (Class 7, Lesson 1.1). With laptops, pairs study twenty graded apples in a spreadsheet and write the rule. Without them: "The teacher writes 12 apples on the board (size and red %), 8 already graded. Pairs find the rule from the 8, grade the last 4, and compare with the pair next to them." The children have just done what a classifier does — learned a rule from labelled examples and applied it to new ones — and when two pairs disagree about apple number 11, you have a lesson on why models make mistakes. (You can try the laptop version in a browser.)

Decision trees (Class 6). A decision tree is a set of yes-or-no questions. Draw one on the board for sorting apples at the mandi: Is it bigger than 70 mm? Is it more than half red? Then hand the children a real apple and let them walk it down the tree. That is the whole idea, and they will never forget it.

How a machine "sees" (Class 7). The lesson opens with "Draw a picture using only 0s and 1s." Squared paper, a pencil: fill a square for 1, leave it blank for 0. Ten minutes later every child knows that to a computer, a picture is a grid of numbers.

What to do when you do have one device

Often there is one phone — yours. That is enough for a lot:

• Use it to show, not for every child to use. One chatbot answer, read aloud, and the class checks it against the textbook.
• Print the lesson. Every lesson in the library has a print version, made for classrooms without laptops.
• The AmanahVale app saves the library on the phone, so lessons you have opened once can be read again with no internet.

Spend the time on judgement

When you can't spend a period on software, spend it on the part every curriculum agrees is hardest: judgement. UNESCO's framework for students puts a human-centred mindset and the ethics of AI first among its four dimensions (UNESCO), and CBSE's new Classes 3–8 curriculum names ethical use of technology alongside problem-solving (Morung Express).

Ask the class: A tool says our district has 50 government schools. How would we check? Ask: Who is missing from this data? Ask: Should the app keep every mistake a child makes? Who should see it? These are the questions that will matter when your students are adults, and they need no electricity at all.

Start with yourself

The best preparation for teaching AI is using it — for your own lesson plans, worksheets and letters, with your own judgement on top. If you'd like to start there, the AI tools course is twelve short lessons for adults, and there is a page for teachers at work.