How AI Learns — Knowledge Graph Activity

What this activity teaches

Artificial intelligence can feel like magic. This hands-on activity pulls back the curtain by having kids build a knowledge graph out of physical materials — cards, string, and sticky notes — to see how a machine connects ideas, makes predictions from patterns, and occasionally gets things hilariously wrong.

By the end, learners understand three big ideas:

• AI learns from examples, not rules someone typed in.
• Connections carry meaning — the more relationships, the better the predictions.
• Bad or missing data leads to wrong answers — “garbage in, garbage out.”

What you’ll need

• Index cards or sticky notes
• String, yarn, or markers to draw connections
• A flat surface or whiteboard
• About 45–60 minutes

How it works

1. Make the nodes. Each card is a concept — “dog,” “tail,” “barks,” “cat,” “purrs.”
2. Draw the edges. Connect concepts that relate, labeling each link (“has,” “makes,” “is a”).
3. Make a prediction. Cover a card and have the AI (the group) guess it from its connections.
4. Break it on purpose. Remove or mislabel a connection and watch the prediction go wrong — a memorable lesson in how AI fails.

Bring it home

This activity pairs beautifully with conversations about how today’s AI assistants work — including Kuriosity, the Guild’s own kid-safe family AI.

Want the full printable facilitator guide and worksheets? Contact us and we’ll send the complete materials.