AIQ
AIQ
Part of the AI Scavenger Hunt lesson guide. Teaching a different grade? 🌈 Explorer (5–7) · 🔧 Builder (8–10) · ⚡ Architect (15–18)
Teacher: "Your home is basically an AI lab, and you probably don't even realize it. Let's do a systematic audit — every room, every device. How much of your environment is AI-powered?"
This lesson runs about 10–12 minutes inside the app for Hacker age — the audit format and the extra vocabulary take slightly longer than the Builder version of the same activity, so plan your period accordingly and don't feel pressure to rush the room-by-room walkthrough to leave time for everything else.
Frame this as an actual audit, not a game — Hacker-band students respond better to being treated like they're doing real analysis. Before opening the app, ask: "If I told you to define 'AI-powered' precisely enough that two people would always agree on the answer, what would your definition be?" Take a few attempts and write the roughest one on the board. Most first attempts will be too broad ("uses a computer") or too narrow ("acts like a human"). Tell them: "Keep refining that definition in your head as we go through this — by the end, you should be able to defend it."
This works well as a two-minute think-pair-share rather than a cold call: give students thirty seconds to write their own definition privately, thirty seconds to compare with a neighbor, then take two or three out loud. The goal isn't to land on a perfect definition before the audit starts — it's to make the vagueness of "AI-powered" visible up front, so the specific techniques you name during the audit feel like they're answering a real question rather than just adding vocabulary for its own sake.
Run the app's three-room audit (Kitchen, Living Room, Bedroom), but for each AI example, push one level deeper than "it uses AI" — name the actual technique involved. This lesson is where the vocabulary from earlier lessons starts attaching to real products. Have students work in pairs and keep a running two-column list — "device: technique" — as they go through each room, so the vocabulary accumulates visibly rather than getting said once and forgotten.
Kitchen 🍳: Smart speakers combine several AI systems at once — speech recognition (audio → text), natural language processing/NLP (text → meaning), and intent classification (meaning → "which command is this?"). It's worth pointing out that these are three genuinely separate problems: a system could transcribe your words perfectly and still fail to figure out what you meant, or understand the general topic but pick the wrong specific action. A smart fridge that tracks inventory typically relies on computer vision / object recognition pointed at the shelves — cameras plus a model trained to recognize food items, rather than any kind of barcode scanning or manual entry. A toaster and a spoon have zero data pipeline — nothing to recognize, nothing to classify, nothing for a model to be trained on in the first place.
Living Room 🛋️: Smart TV recommendations are an IoT-era application of collaborative filtering ("people similar to you liked this"), content-based filtering ("this is similar to what you already liked"), or a blend of both. Ask students which approach would work better for a brand-new show with no viewing history yet — content-based filtering, since collaborative filtering needs other viewers to have already watched it. A game console's NPCs "have AI" in the sense that their behavior adapts using decision trees and pathfinding algorithms — chase when the player is near, hide when health is low, navigate around obstacles — which is a much lighter-weight kind of "AI" than a trained machine-learning model, but still counts because the behavior changes based on the game state rather than following one fixed script. A couch and a power strip have no sensors and no software at all — there's no data pipeline to even discuss.
Bedroom 🛏️: Fitness trackers and smartwatches run a motion-classification model against a continuous stream of accelerometer (and often gyroscope) data to guess whether you're walking, running, or asleep — this is a straightforward pattern-classification problem, structurally similar to problems covered in later worlds. A phone camera performs face detection and scene recognition, both computer-vision tasks, before you even take the shot — deciding what to focus on, how much light to let in, and whether to apply portrait-style background blur all happen based on what the model detects in the frame. A bed and a desk lamp have no inputs to learn from — there's no sensor feeding either one any data about you at all.
Introduce the term ambient computing once the audit is done: AI that is built into the environment so thoroughly that it becomes invisible — nobody thinks of their smartwatch as "running a model," it's just a watch. Ask: "Is that a good thing, a bad thing, or just a thing?" — there's no single right answer, and that's fine; the goal is to notice the pattern, not resolve it in one class. Some students will land on "it's convenient" and others on "it's a little unsettling" for the exact same example — both reactions are reasonable, and surfacing that disagreement is more valuable than steering the class toward one verdict.
Close with the app's summary: IoT devices with embedded AI are becoming standard in modern homes; voice assistants use NLP and speech recognition, smart TVs use collaborative filtering, fitness trackers use motion classification models — and the overall trend is toward AI becoming increasingly invisible.
The app follows with a short solo practice round — students re-sort eight of the twelve objects individually. Since this band has already built the device-technique list together, use this round as a quiet check rather than new teaching: circulate and look for pairs who can still explain why an object is AI, not just click the right answer from memory of the room they just saw it in.
These questions are meant to push past "identify the AI" into "reason about the AI" — expect longer, messier answers than in the earlier bands, and resist the urge to supply a tidy resolution to every one.
Close by returning to the definition exercise from the hook: "Go back to the definition of 'AI-powered' you wrote at the start of class. Does it still hold up against everything we found today — the fridge, the console, the watch? Revise it if it doesn't." Let a couple of students read their revised definitions aloud, and point out any common thread between them — most refined definitions tend to converge on some version of "learns from data and changes behavior accordingly," which is worth naming explicitly as the class's own version of the test this whole world of lessons has been building toward.
Extension activity (fills the rest of the period): Assign each student (or pair) one real device from their own home that wasn't covered in the lesson — a router, a doorbell camera, a rice cooker with presets, a karaoke machine that scores singing. For homework, have them research (product page, manual, or a quick search) whether it uses AI, and if so, what specific technique. Next class, have each student present their device in one minute: what it does, whether it's AI, and how they decided. This turns the three-room audit in the app into a real audit of their own environment.
Push students to cite something specific when they present, not just an assertion — a line from the product's marketing copy, a feature described in the manual, or a specific behavior they personally observed ("it started suggesting different settings after I used it a few times"). This matters because "uses AI" is also a marketing phrase companies apply loosely, and part of what this age band should be building is the habit of checking a claim rather than repeating it. If a student's research turns up genuine ambiguity — the product page says "AI-powered" but the actual described feature sounds like a fixed rule — that's a great presentation, not a failed one; have them explain exactly what evidence made them suspicious.