AIQ
AIQ
Everything a teacher needs to deliver this lesson — pick your grade's script below once you've read the background.
"AI in Your Pocket" is a scavenger hunt through the phone every student already carries. It uses the same device your class walks around with every day to make an abstract idea — "AI is everywhere" — into something they can point at. By the end of the lesson, across the full 5–18 range, a student should be able to:
These objectives build directly on each other rather than standing alone. Naming examples (objective 1) is what makes the sorting rule (objective 2) concrete instead of abstract; understanding how a feed is personalized (objective 3) is what makes the older bands' tradeoff discussion (objective 4) land as something real rather than a warning tacked on at the end. A student who can only do the first — list a few AI apps — has memorized trivia. A student who can sort a brand-new example they've never seen before, using the "does it learn from data" rule, has actually understood the lesson.
You do not need a computer science background to teach this lesson well — you need one distinction, repeated across every example: some phone features run on a fixed set of instructions a programmer wrote, and some run on a model that was trained on examples until it could recognize a pattern on its own. The lesson calls the second kind "AI" and the first kind "not AI," and every item a student sorts comes down to that line. A ruler app measures pixels on your screen and multiplies by a conversion factor — a programmer wrote that formula once, and it never changes. Face ID was never given a formula for "what a face looks like." It was shown enormous numbers of faces during training until it could build its own internal representation of facial structure, which is a fundamentally different kind of process.
Concretely, Face ID works by projecting more than 30,000 invisible infrared dots onto your face and reading the depth pattern they form — how far apart your eyes are, the shape of your nose bridge, the curve of your jaw. A neural network turns that dot pattern into a mathematical fingerprint and compares it to the one stored when you first set up the phone. This is why it still works with glasses, a hat, or a new haircut: those change very little about the underlying geometry the model is actually measuring. It is worth being precise with students about what this is and isn't — it is pattern-matching against a stored template, not the phone "recognizing you" the way a person would, and it is not reading your mind or your identity in any deeper sense.
The second concept this lesson introduces, mostly for older students, is the recommendation system — the AI behind TikTok's For You page, Netflix's row of suggestions, and Spotify's Discover Weekly. These systems work by combining a few signals: what similar users liked (collaborative filtering), what a video or song is actually about (content-based matching), and how you personally behaved — did you finish it, rewatch it, skip it in two seconds. None of this is random and none of it is chronological; it is built continuously from a running record of your own behavior on the app. The specific numbers in the lesson (TikTok adapting within roughly 40 minutes of viewing, Netflix attributing about 80% of watch time to recommendations, Amazon crediting recommendations with a large share of purchases) are industry-reported estimates that companies themselves have published or that researchers have cited — treat them as illustrative of scale, not as figures a student should memorize as exact.
Finally, some of this AI runs directly on the phone's own chip (Face ID and the keyboard, for speed and so your face data never has to leave the device), and some runs on a company's servers, because a recommendation feed needs to constantly retrain on fresh data from millions of users at once. That on-device-versus-cloud distinction is background for you, not something every age band needs to explain — Architect mode's quiz touches it directly (model quantization, on-device inference), while Explorer mode never needs the words "server" or "cloud" at all.
One more thing worth knowing before you teach this: the lesson deliberately includes items that are not AI in every single category — a ruler app, FM radio, a paper shopping list. This isn't padding. Students (and adults) tend to over-attribute "AI" to anything on a screen, so the lesson is teaching the boundary as much as it's teaching the examples. If a student ever asks about an item the lesson doesn't cover — an alarm clock, a calculator, a flashlight app — the same question applies: does it learn a pattern from data about you, or does it run the same fixed instruction every time? An alarm clock, notably, is not AI even though it "knows" when to go off — you told it the time directly; it isn't inferring anything.
A device with a browser and the AIQ app open is normally the entire requirement — there is nothing to print, build, or install beforehand. A few things are worth having on hand or checking in advance: