All activities Student learning
Children train a pretend robot with picture-word cards and discover that AI makes guesses from patterns, and that a guess can be wrong.
Young children already hear talking speakers and see apps that "know" things. In this game, the class becomes the teachers of Robo, a pretend robot that can only learn from the example cards it is shown and can only notice color. Robo makes confident guesses, gets some hilariously wrong, and gets better when the class gives it more kinds of examples. Children leave with a simple, true idea they can say out loud: "AI looks for patterns and makes guesses. Guesses can be wrong, so we check."
Hook
Without talking, clap, stomp, clap, stomp, clap, then freeze. Children will jump in with a stomp. Ask: "How did you know what came next?" Take answers, then name it: "You found a pattern. A pattern is something that repeats, so you can guess what comes next." Try one more: "Pat head, pat head, touch nose, pat head, pat head, …"
Facilitator noteCelebrate the word "guess." You are setting up the big idea that a pattern lets you guess, not know for sure.
Explore
Show Pattern Cards 1–5 one at a time. Read each aloud and have children show the answer with their bodies (arms up high for "big," crouch for "small," and so on) or call it out. After each, ask one child: "How did you know?" Then show Card 6, the Trick Card. Everyone guesses "blue." Flip it: the next color is green. Say: "Your guess was a smart guess. It was still wrong, because the pattern changed. Smart guesses can be wrong."
Facilitator noteKindergartners may need you to point along the pattern and say it together before guessing.
Model
Put on the robot hat and use your robot voice: "I am Robo. I learn from examples. I only know what you show me. And I only notice COLOR." Read each of the 6 Training Cards aloud and let children tell you which sign it goes under. Tape it there. When all 6 are up, Robo "thinks" out loud: "Beep. Fruit is RED. Not-a-fruit is blue, green, or yellow. I have learned!"
Facilitator notePoint out that Robo learned a pattern the class did not mean to teach. That is the heart of AI bias for this age: the examples shape the guesses.
Practice
Hold up each Test Card. First, children predict with thumbs: "Thumbs up if Robo will say FRUIT, thumbs down if Robo will say NOT A FRUIT." Then Robo guesses, using only color. Then the class checks: "Is Robo right?" A banana is yellow, so Robo says "not a fruit" (wrong). A red ball is red, so Robo says "fruit" (wrong). Let two or three children take turns wearing the hat and guessing "as Robo," using only color.
Facilitator noteChildren love catching Robo. Keep the tone kind: "Robo isn't silly. Robo only saw red fruit. What could we do to help Robo?"
Debrief
Ask: "Why did Robo make mistakes?" Guide children to: Robo did not see enough kinds of fruit. Add the two Fix-it Cards (a yellow banana and green grapes) under FRUIT and retest the banana. Robo gets it right. Then say plainly: "Some computers use AI. AI tools learned from LOTS and LOTS of examples, way more than Robo. They make guesses from patterns, too. They can be very smart, and they can still be wrong. So what do we say? 'Let's check!'" Practice the phrase together.
Facilitator noteAvoid saying the AI "thinks" or "knows." Use "guesses" and "learned from examples." Children pick up our words.
Reflect
Children complete Handout C: draw the next thing in a pattern, draw a thing Robo might guess wrong, and draw a person who can help them check. Circulate and ask two or three children, "Why might Robo guess wrong about your picture?" Write their words in the margin as evidence.
Facilitator noteLook for answers like "because it's a different color" or "Robo never saw one." Those show the child understands that examples drive the guess.
The whole lesson runs with the printed cards, the board, and your robot voice. The Pattern Cards can be read aloud instead of shown. For a movement version, tape FRUIT and NOT A FRUIT signs on opposite walls, give each Training Card to a child, and have them walk it to the right wall while "Robo" watches and learns.
After the Debrief, project a district-approved generative AI chatbot from the teacher's device. Type, as the class watches: "Finish this pattern for a kindergarten class: red, blue, red, blue, red, …" Read the answer aloud and have children check it with their own eyes. Then type a riddle the class writes together ("I am yellow, long, and you peel me. What fruit am I?") and ask children to predict before you press enter. Whatever the tool says, ask: "How do we know it's right? We checked." Children never get their own AI accounts; you drive every click.
Does it need a screen? This lesson is stronger unplugged: children need a clear mental picture of "examples in, guesses out," and Robo gives them one without a screen. A short projected demo adds one thing paper cannot: children see that a real AI tool also answers by guessing, and they practice checking it.
What you should be able to see or collect if it worked.
Children train a pretend robot with example cards, watch it misclassify, and fix it with more varied examples, an early model of how AI learns patterns.
At home, children play "Robo at dinner": a family member makes a pattern with spoons, cups, or claps and the child guesses what comes next, then tells the family, "AI guesses from patterns, so we check." Next lesson, connect to Ask Two, where children check a robot's answers with a book and a grown-up.
