AI and Python academy for girls

Patterns are everything

Against a player in a rut, the AI wins four rounds in five. Against a fixed cycle it almost never loses. The more predictable the data, the better the model.

Randomness wins

Against a truly random player the AI does no better than chance: 33.6% wins against 33.0% losses. The simplest strategy of all, playing at random, cannot be beaten.

What students try next

Remembering two moves instead of one, forgetting old games, and playing against each other's programs. Each change is measured the same way.

Why this is a real AI lesson

The table of counts is a tiny model trained on data, the opponent's moves. Its predictions are only as good as the patterns in that data, the same reason a large AI system can be confidently wrong.

Students see that directly, in numbers they produced themselves.

Where it goes next

The same students later train classifiers on real datasets in AI classes for teens, and learn why the training data matters as much as the code.

Younger students build the game first in Python and add the learning part once loops and lists feel easy.

Why choice matters

A project she chose gets finished. Finishing things, and showing them to someone, builds confidence faster than any amount of exercises.

The syllabus skills are the same whichever project carries them.

Sharing the work

Finished projects are shown to the group, and older students can publish theirs online with a parent's agreement.

Some go on to student hackathons or programming competitions; others simply keep building.