Teaching AI literacy to girls who have never owned a laptop

"AI literacy" is one of those phrases that can mean anything, which usually means it means nothing. Here is what it means for four afternoons in Kathmandu.

The girls train their own simple image classifiers. Not in theory: they gather examples, label them, watch the model get it wrong, work out why, and fix it. The lesson lands in a way no explanation does, which is that a model knows only what it was shown, and someone chose what to show it.

They act out how a machine recognises movement, physically, on the same futsal court where they spent the morning. It sounds like a game, and it is, and it is also the clearest way we have found to teach what a feature is without a single line of code or a device per child.

And they learn to spot misinformation before they share it. This is the part with the shortest path from classroom to daily life. Most of these girls have phone access long before laptop access, which means they meet generated images and fabricated claims years before anyone teaches them how those are made.

Two things make this possible on a small budget. The programme is designed for a room with one projector and shared devices rather than one laptop per child, which is line C2 of the budget, $268 for AV, internet and bilingual printing. And it is taught alongside sport rather than instead of it, so the girls arrive through a football programme they already trust.

The multiplier is the coaches. Five to seven of them take the material back into weekly sessions in their own communities. The clinic reaches around seventy people directly. The coaches are how it carries further, and the $282 stipend line is what makes it reasonable to ask them to lose four days of session income to be there.