AIQ
AIQ
Part of the AI Perception Challenge lesson guide. Teaching a different grade? 🌈 Explorer (5–7) · 💻 Hacker (11–14) · ⚡ Architect (15–18)
Open with the line straight from the app, said like you mean it — this lesson only works if students feel like they're being tested on something they've actually earned, not introduced to something new:
"You've learned how AI sees, hears, and reads. Now let's put it all together! Can you figure out which type of AI perception each example uses?"
This lesson doesn't teach a new skill — it's a checkpoint on three skills your class already has. Say that plainly so students don't feel like they're behind if the sorting feels tricky at first: "Nothing today is brand new. You already learned what computer vision, speech recognition, and reading (NLP) mean. Today is just about spotting them in the wild, in real apps you actually use."
Write three headers on the board: Vision, Hearing, and Reading. Ask the class to recall, in one sentence each, what each one means from the earlier lessons: Vision reads an image as numbers (pixels). Hearing (speech recognition) turns sound waves into text. Reading (natural language processing, or NLP) turns text into numbers so a program can find patterns in it. If your class hasn't done those three lessons recently, give the one-sentence version yourself before moving on — the rest of the lesson assumes students have these three labels in hand.
Then raise the stakes: "Today's challenge is that most of the AI you actually use doesn't pick just one of these. I'm going to show you real apps, and your job is to figure out which one — or which two — they're using."
A quick way to gauge where the class is starting from: ask for a show of hands, "Who thinks Instagram filters use AI vision? Who thinks they use AI reading?" You'll usually get a scattering of both answers even after the earlier lessons — that's normal and useful, not a sign the class forgot everything. It means the sorting instinct hasn't fully solidified yet, which is exactly what today's repeated practice is for.
Work through the three categories from the lesson as a "guess before I tell you" round, using real product names so it lands as recognizable rather than abstract.
Keep a running tally on the board of three columns — Vision, Hearing, Reading — and add a tally mark under the right column as each example is revealed. By the end of all twelve, ask the class what they notice about the tally: most categories end up with four examples each, by design, but two or three of the "hearing" and "reading" examples will get an extra mark in a second column once you circle back to them in the discussion below, because they secretly use more than one type. That visual imbalance is a good way to preview the "put it together" theme without stating it outright yet.
Vision examples. Ask for a guess before revealing each fact. Instagram Filters — "Filters detect your face features using computer vision, then overlay effects in real-time!" Checkout-Free Stores — "Amazon Go uses hundreds of cameras with AI to track what you pick up and charge you automatically!" Cancer Detection — "AI can spot early signs of skin cancer in photos with accuracy matching dermatologists!" License Plate Readers — "Parking garages read your license plate with AI cameras — no ticket needed!" Ask: "What's the one thing all four of these have in common?" (They all start with a camera or a photo, and the AI is analyzing an image.)
Hearing examples. Voice Assistants — "Siri, Alexa, and Google use speech recognition + natural language understanding — TWO AIs together!" Pause here — this is the key example for today's "put it together" theme. Music Generation — "AI can compose music by learning patterns in millions of songs — melody, rhythm, and harmony!" Call Centers — "AI transcribes customer calls in real time and suggests answers to the support agent!" Sound Separation — "AI can separate a singer's voice from the music — isolating individual sounds from a mix!"
Reading examples. ChatGPT — "ChatGPT reads your message, predicts word by word what a helpful response should look like!" Review Analysis — "AI reads thousands of product reviews and gives you a summary: mostly positive, negative, or mixed!" News Sorting — "Google News uses text AI to group articles about the same story from different sources!" Content Moderation — "Social media AI reads millions of posts per day checking for harmful content!"
After each round, ask a quick "what's the input, what's the output" question to sharpen the sorting skill further: for Review Analysis, the input is thousands of written reviews (text) and the output is a short summary label like "mostly positive" — that's a reading task from start to finish. For Cancer Detection, the input is a photograph of skin and the output is a flag for "worth a closer look" — that's a vision task from start to finish, even though the final decision still belongs to a doctor. Naming input and output explicitly for two or three examples gives students a repeatable method for classifying any new example they meet later, rather than memorizing this specific list of twelve.
Once all three categories are up on the board, run the sorting practice from the app (or recreate it live): give pairs of items — a face filter versus a plain picture frame, a voice assistant versus a megaphone, a chatbot versus a printed book, content moderation versus a padlock — and ask students to say which one in each pair is actually "AI" doing perception work, and which is just an ordinary object doing nothing smart at all. The point of pairing a real AI example against a look-alike non-AI object is to stop students from assuming anything electronic or app-shaped automatically counts as AI.
Sample exchange for the trickiest pair, content moderation versus a padlock: Teacher: "A padlock keeps something locked. Content moderation keeps something 'locked' too, in a sense — it blocks harmful posts. So why is only one of these AI?" Student: "Because the padlock doesn't have to figure anything out — it just locks or unlocks with the right key." Teacher: "Exactly. Content moderation has to actually read millions of posts and decide, one at a time, whether each one is harmful. A padlock never has to 'decide' anything." Use this pattern — asking what decision or judgment call the AI example has to make that the non-AI object doesn't — as your go-to check for the rest of the pairs.
As a stretch question for students who finish early, ask them to name a real product from home that they think would fit into more than one of the twelve categories, and to defend which perception type(s) it most likely uses without being told the answer.
Notice that three of the four questions have a "sounds right but isn't" trap built in — "text reading" for Instagram, "GPS" for Google News, "cameras" for voice assistants. That's deliberate, and it's worth saying out loud to the class: the quiz isn't just testing whether they remember the right label, it's testing whether they can reject an answer that merely sounds plausible. That's the same skill they'll need later when evaluating any confident-sounding claim, AI-generated or otherwise.
Close with: "You just proved you can spot the difference between AI vision, AI hearing, and AI reading in real apps — and you found the ones that use more than one at a time, which is actually most of the powerful ones. That combining idea has a name, and if you keep going in this app you'll see it called 'multi-modal AI.'"
Extension activity (15–20 minutes): Have students pick one app on their own phone or a family member's phone (or one they know well, if devices aren't available) and write three to four sentences: what does the app do, what type(s) of AI perception does it likely use, and how confident are they in their guess versus how much they're just assuming. Have a few volunteers share, and push back gently on any answer that assumes an app is "AI" just because it feels smart or modern — ask what specific input (image, sound, or text) the app would actually need to process to do what it does.
To make the sharing round more interactive, have the rest of the class guess the perception type before the presenter reveals their own answer — this keeps the whole room actively sorting instead of passively listening. If your class has strong app overlap (many students naming the same two or three apps), group students by app afterward and have each small group compare their reasoning and agree on a final answer together, which often surfaces disagreements worth resolving as a whole class.