Teaching "How Computers See" to Builder mode (ages 8–10)

Hook & Warm-Up

Open with a direct question to get hands up, then use the app's own hook line as your pivot line.

"Quick show of hands — who has unlocked a phone or tablet just by looking at it? [Hands go up.] Okay, keep your hand up if you know exactly HOW that works. [Most hands drop.] That's what we're figuring out today."

Then read the lesson's actual hook, which is designed to land right here:

"Your eyes see colors, shapes, and faces instantly. But how does a computer 'see'? Spoiler: it doesn't see at all — it counts numbers!"

Follow up: "That word 'spoiler' is a big clue. By the end of today, 'counting numbers' is going to make total sense — even for face unlock, even for stuff like Google Photos finding your dog in a hundred pictures."

Optional two-minute warm-up: hold up a phone and zoom into a photo on the camera roll (any photo works) until the image starts to blur into visible squares. Ask, "What are you looking at right now?" Let a few guesses land before saying, "Those squares are pixels — and no matter how good the photo looks normally, that's ALL a computer ever actually has to work with. Not a picture the way you see it. A big grid of colored squares, and underneath each square, just a number."

Main Activity

Walk through the lesson's three scenes as a guided sequence, pausing to ask a question at each stop before students open their own device.

Scene 1 — Pixels

Explain, using the lesson's own facts: a pixel is one tiny colored dot — a phone screen has millions of them. Every pixel is made of Red, Green, and Blue light mixed together (that's how screens can show any color at all). A digital photo is really just a grid of numbers, where each number tells one pixel which color to be. If you zoom into any image far enough, you'll see the squares — that's true of every single image a computer ever shows you.

"So here's the trick question: when your phone shows you a photo, is there really a 'picture' stored inside it? [Let a few answers land.] Not really — there's a giant list of numbers, and the screen turns those numbers into colored squares for YOUR eyes to see as a picture. The computer never sees a picture. It only ever has the numbers."

A good check-for-understanding here: ask students to guess roughly how many pixels might be in a photo taken on a typical phone camera today. Most will guess low. Reveal that many phone cameras today capture 12 megapixels or more — that's over 12 million individual dots in a single photo, each one carrying its own little set of color numbers. Tie it back: "That's 12 million numbers before the AI has even started looking for a pattern."

Scene 2 — Face Detection

Introduce how AI finds faces: it scans a photo for a pattern that looks like a face — two eyes, a nose, a mouth arranged the usual way. Then it measures things like the distance between your eyes and the width of your nose, building something like a "face fingerprint." Your phone stores those measurements and checks them every time you look at it — same idea, applied to Google Photos finding your dog after learning from millions of dog pictures.

"If your phone stores a fingerprint made of measurements, not a photo — why do you think it can still recognize you with different hair, or glasses, or in different lighting? [Guide toward: because the eye/nose/jaw distances barely change even when your hairstyle or lighting does.]"

It's worth pausing here on why this matters for privacy, since this age band is old enough to start caring: a "face fingerprint" made of measurements is very different from a stored photo of your face. If someone stole the raw numbers, they generally can't reconstruct your actual photo from them the way they could from a copied image file — which is one reason face-unlock systems are usually considered safer than keeping a folder of ID photos. That said, it's not risk-free either, which is exactly why AIQ never asks students for a real photo or real name anywhere in the app.

Scene 3 — Object Recognition

Cover the four real examples from the app: self-driving cars recognizing other cars, pedestrians, and traffic signs from camera images; AI spotting patterns in X-rays and MRIs that doctors might miss; Amazon Go stores using AI cameras to see what shoppers pick up, skipping the checkout line; and plant-ID apps like PlantNet identifying a plant from a photo of its leaves.

If your students are more familiar with local examples, you can mention that plant-ID and crop-health apps are increasingly used by Filipino farmers to spot pest damage or disease on rice and vegetable crops early, sometimes through simple SMS- or app-based services — a Philippine-relevant version of the exact same "camera plus pattern-finding" idea as PlantNet.

Ask: "All four of those are really different jobs — but what's the ONE thing they all secretly have in common?" (Answer: all four are just finding patterns in pixel numbers, at very different scales and stakes.)

Then hand out devices for students to complete the lesson (including the practice round sorting AI vision tasks from ordinary display tasks) individually or in pairs.

While they work through the practice round, walk the room and spot-check reasoning rather than just answers. The practice items split into two groups: things that involve AI actually analyzing an image (face detection, self-driving vision, X-ray analysis, plant identification) and things that are just showing or handling an image with no analysis at all (counting pixels as a raw technical step, displaying a photo, screen brightness, printing a page). If a student mixes these up, the fix is usually to ask "is the computer just showing you something, or is it trying to figure out WHAT is in the picture?" — display and printing show pixels; AI vision reasons about them.

One good stretch question for a fast-finishing pair: "Screen brightness isn't AI, but could a company someday add AI to brightness control? What would it need to look at?" (A reasonable answer: a camera-based AI could look at the room's lighting or even your eyes to adjust automatically — some devices already do a simplified version of this with a simple light sensor rather than true computer vision, which is a useful nuance if it comes up.)

Discussion

• "If a photo is really just a list of numbers, what do you think happens if you change a few of those numbers? Would the picture look any different?"
• "Face unlock measures your face instead of storing a photo of it. Can you think of a reason that's actually safer?"
• "Name one everyday app or gadget that probably uses computer vision that we didn't mention in class."
• "The lesson says AI doesn't 'understand' what it sees the way you do. What do you think that means — what's the difference between 'finding a pattern' and 'understanding'?"
• "Self-driving cars and X-ray AI are both 'computer vision.' Which one do you think needs to be MORE accurate before it's trusted, and why?"
• "Amazon Go stores use cameras to see what you pick up instead of a cashier scanning it. Can you think of one good thing and one worrying thing about that?"

These questions work best round-robin rather than hands-up-only — this age group tends to have a few dominant voices, and a quick "let's hear from someone who hasn't answered yet" keeps the whole room engaged rather than just the most confident three students.

Quiz Walkthrough

Students take this quiz on their own device after finishing the three scenes and the practice round. Each question uses the emoji shown below as a visual anchor in the app, so calling out the emoji when you review answers as a class helps students match your recap to what they tapped through.

Wrap-Up & Extension

Close with the lesson's own summary line: "Computers turn pictures into tiny dots called pixels. Each pixel is just a number! AI finds patterns in those numbers to spot faces and objects — but it doesn't really understand what it sees." Ask one student to repeat it back in their own words before dismissing the topic.

Extension activity: Split the class into small groups and give each group one computer-vision example NOT covered in the lesson (e.g., handwriting recognition apps, sports instant-replay tech, a supermarket self-checkout scale-camera, security cameras that count people). Have each group spend five minutes sketching what "pixels → pattern → answer" might look like for their example, then present it back to the class in one sentence: "Our example takes a picture of ___, and the AI is probably looking for the pattern of ___."

A second option, if you'd rather keep it as a full-group activity: play a short "guess the pixelated picture" game. Prepare (or find online) a handful of images that have been intentionally blurred or reduced to large blocky pixels, and reveal them gradually — very blurry first, then a little clearer, then clear. Ask students to guess at each stage and notice how much easier it gets as more detail appears. Tie it back to the lesson: "This is basically the AI's problem too — the less clear or detailed an image is, the harder it is to find a reliable pattern, which is part of why AI vision can make mistakes on blurry or low-quality photos."

For homework or a quick exit ticket, have students write one sentence answering: "Name one machine in your home, school, or a store you visit that you now think probably uses computer vision, and explain what pattern you think it's looking for." Collecting these gives you a fast read on who has really internalized "pixels → numbers → pattern" versus who is still thinking of AI vision as literal seeing.