Teaching "Train Your Own Model" to Explorer mode (ages 5–7)

Hook & Warm-Up

Gather everyone in a circle where they can see a screen or tablet. Hold up both your thumbs and say, in your own words, the app's opening line: "Let's teach A.I. to recognize things! You'll show it pictures, tell it what they are, and watch it LEARN! It's like teaching a pet new tricks!"

Ask: "Has anyone ever taught a puppy, a kitten, or even a little brother or sister something new — like sit, or how to stack blocks?" Take a few excited answers. Say: "You probably didn't explain it with big words. You showed them, again and again, until they got it. That's EXACTLY how we're going to teach the computer today!"

Do a quick warm-up game before opening the app: play "Thumbs Up, Thumbs Down." Call out silly statements — "Ice cream for breakfast!" "Homework on a Saturday!" "A puppy that can fly!" — and have the whole class show thumbs up (yes, good idea!) or thumbs down (no way!) with big, clear gestures. Say: "Hold your thumb up nice and high so a camera could see it clearly — that's exactly what we're about to practice with!"

Play a second quick round where you deliberately mix things up — call out a mushy statement like "Broccoli ice cream!" and let kids show whatever thumb feels right, thumbs sideways included. Point out with a laugh: "See how some of you weren't totally sure? Even people can find some things tricky to sort into just two piles sometimes — keep that in mind for later!"

Explain the plan simply: "Today, YOU are going to be the teacher, and the computer is going to be the student. You'll show it lots of thumbs-up pictures and lots of thumbs-down pictures, tell it which is which, and then test if it learned!"

Before you open the app, do one more short warm-up so the idea of "showing, not telling" really sinks in: hold up a stuffed animal or a picture of an animal the class knows well and ask, "How did YOU learn what a dog looks like? Did someone hand you a list of rules, like 'four legs, a tail, and fur'? Or did you just see LOTS of real dogs and photos of dogs until you knew one when you saw one?" Most kids will land on the second answer without much nudging. Say: "That's exactly the plan for our computer friend today — no rules, just lots and lots of good examples."

Main Activity

Open the AI Playground together (as a whole-class demo on one shared screen works best at this age, with volunteers taking turns; pairs on individual devices works too if every pair has an adult or older buddy nearby). The app's two starting categories are already "Thumbs Up 👍" and "Thumbs Down 👎" — perfect for this age, since no typing or reading is needed to get started.

Before the first tap, remind the class of the one grown-up rule for today: "The camera only looks at us right here, right now — it doesn't save our pictures anywhere or send them anywhere else. As soon as we're done, it's like the pictures were never there." This is true and worth saying plainly, since a "the camera is watching me" activity naturally makes some young children a little unsure, and a simple, honest reassurance settles that quickly.

Scene 1: Show and Tell (Collect Data)

Say: "First, we have to SHOW the computer lots of pictures, just like Show and Tell!" Pick a volunteer to hold their thumb up to the camera and tap the button to capture a picture. Say: "One picture isn't enough — the computer needs to see LOTS of thumbs-up pictures, from lots of different friends, at different angles, so it really understands what a thumbs-up looks like, not just one friend's thumb!" Rotate through 6-8 volunteers doing thumbs-up, then the same number doing thumbs-down.

Stop and ask: "What if we showed the computer 20 thumbs-up pictures but only 2 thumbs-down pictures — what do you think would happen?" Take guesses. Guide toward: "The computer would start guessing thumbs-up for almost everything, because that's mostly what it's seen! We have to show it a FAIR amount of each, like sharing crackers evenly with your friends."

Point out one more thing while collecting: "Notice how some friends held their thumb close to the camera, and some held it far away? Some pictures are bright, some are a little dark? That's actually GOOD — we want lots of different-looking examples, not the exact same picture over and over, so the computer gets really good at spotting a thumbs-up no matter how it's shown." Have the class count out loud together how many thumbs-up and how many thumbs-down pictures were captured, and cheer once the two piles are close to even.

Scene 2: Let It Think (Train the Model)

Say: "Now here's the magic part. We tell the computer, 'Okay, look at ALL those pictures and figure out what makes a thumbs-up different from a thumbs-down!'" Tap the button that trains the model. Say: "It's looking really, really closely at all our pictures right now, just like you look really closely at a puzzle before you know where a piece goes."

Ask: "Do you think it will get it right the very first time, every time?" Let kids guess. Say: "Sometimes yes, sometimes not yet — and that's okay! That's exactly what we're going to check next."

While the class waits, do a tiny "be the computer" game to make the idea feel real: hold up two of the captured thumbs-up pictures and one thumbs-down picture (or point to them on screen) and ask the group, "If I showed you a brand new picture, and it looked kind of like these two but not like this one, which pile would you put it in?" Let a volunteer point. Say: "That's basically what our computer friend is about to do — look at a new picture and decide which pile it looks most like!"

Scene 3: Quiz the Computer (Test & Improve)

Have a new volunteer — one who did NOT take a training picture — hold their thumb up to the camera and watch the app guess. Celebrate loudly if it's right! If the app guesses wrong, say: "Not yet! That's okay — we just need to show it a few MORE pictures like this one, and try again." Capture 2-3 more examples of that tricky case and test again.

Run this a few more times with different volunteers, mixing thumbs-up and thumbs-down, so every child sees at least one correct guess and, ideally, one "not yet" moment followed by a fix — that contrast is the whole point of the lesson.

If a lot of guesses come out right very quickly, add one small twist to keep it interesting: have a volunteer make a "sideways thumb" — not quite up, not quite down — and ask the class to predict out loud whether they think the computer will be sure or unsure about that one. This gently introduces the idea that some examples are just naturally trickier than others, for a computer and for a person, without using any scary or discouraging words about it.

Discussion

Bring the class back to the circle. There are no wrong answers here — the goal is just getting them to say the ideas out loud.

• What did WE have to do before the computer could learn anything? (Guide toward: show it lots of pictures and tell it what each one was.)
• Why did we need MANY thumbs-up pictures instead of just one? (There's no single right answer — steer toward "so it really understands, not just remembers one friend's thumb.")
• What did we do when the computer guessed wrong? Did we give up? (Guide toward: no, we showed it more pictures and tried again.)
• Is this like anything YOU had to practice a lot before you got good at it, like riding a bike or tying your shoes?
• Do you think the computer feels proud when it guesses right? Why or why not? (Most kids will land on "no, it's not really thinking or feeling" — a good early seed for later lessons.)
• If a new friend joined our class tomorrow and did a thumbs-up the computer had never seen before, do you think it would still get it right? Why might it? (Guide toward: yes, probably — because it learned from MANY different friends' thumbs, not just one.)

Quiz Walkthrough

Wrap-Up & Extension

Close with: "Today YOU were the teacher, and the computer was your student! You showed it pictures, told it what they were, and it learned — just like teaching a puppy a new trick, one example at a time." Ask the class to give themselves a round of applause for being such good AI teachers.

Before moving on, revisit the very first question from the hook one more time: "So — does the computer really UNDERSTAND what a thumbs-up means, the way you do?" Let the class answer together. Guide gently toward: "No — it just got really good at spotting the pattern in pictures we showed it. It doesn't know thumbs-up means 'yes' or 'good job' the way you do. But it's still pretty amazing that it learned to spot the pattern just from watching examples, isn't it?"

Extension activity — Silly Faces Sorting Game: In pairs, have students pick two silly, easy-to-repeat faces or poses (like "big smile" vs. "surprised face," or "arms up" vs. "arms crossed"). Using the Playground on a shared or classroom device, have each pair rename the two categories (with a grown-up's help typing) and collect 8-10 example pictures of each pose from several classmates. Then take turns testing it with a brand-new pose from a friend who didn't help train it, and cheer for every correct guess. If there's time, have pairs try to "trick" the computer on purpose with a pose that's halfway between the two, and talk together about why that one might be harder for the computer to guess.

A simple no-device backup, if a pair's camera or connection acts up: play the same game with paper. Give each pair a stack of small drawings or magazine cut-outs of two categories (fruits vs. vegetables, for example). Have one child be "the computer" — they haven't seen the sorted piles yet — while their partner shows them, one at a time, several already-labeled examples of each pile, saying the label out loud each time. Then show "the computer" child a brand-new picture and have them guess which pile it belongs in, out loud, explaining what it reminded them of. This keeps the exact same show-label-guess structure alive even without a working camera.