AI literacy & education · Educators
How Can I Teach AI Literacy Without Student Accounts?
Turn a few fictional kitchen tickets into an AI literacy lesson. Students use examples to make predictions, then test which clues hold up.
The quick answer
How can I teach AI literacy without making students sign up for an AI account?
Give students a set of labeled paper tickets and ask them to predict the label on tickets they have never seen. Have them compare two clues, test the predictions, and explain why one clue failed. You can run the lesson on a board or with index cards. Students need no logins, and you can use the same material with their usual accessibility tools.
Choose one idea students can explain
By the end, ask students to explain how a pattern can fit every example they have seen and still fail on a new one. A short activity with one observable goal makes it easier to see what each student understands. Ask each student to write or say the explanation in their own words.
The student competency framework from UNESCO includes AI foundations and critical judgment. The exercise below uses fictional tickets to give students a concrete prediction to examine. UNESCO AI Competency Framework for Students
In supervised learning, a system can learn relationships between clues, called features, and the answers attached to examples, called labels. For a classification task, the prediction is a category. In the exercise, students act out a small part of that process by making and testing rules by hand. Google’s introduction to supervised learning
Prepare four examples and two test tickets
Write two clues on each card. For the first four, add the correct label. In this fictional kitchen, orders served on a plate are dine-in, and orders packed in a bag are takeaway. Hold that kitchen rule and the last two labels back until students have made predictions.
For the labeled dataset, use red ticket with plate for soup, red ticket with plate for toast, blue ticket with bag for soup, and blue ticket with bag for toast. Write the dine-in or takeaway label beside each example. Read the color words aloud or print them as text so the clues do not depend on seeing color.
Put two fresh orders aside. A red ticket with a bag is takeaway. A blue ticket with a plate is dine-in. Show those two cards without their labels during the test. Keep the answer key with you.
Have students try two rules
Ask pairs to name two clues that fit all four labeled examples. One rule sorts by ticket color. The other sorts by the plate or bag. Both rules match the examples. Ask each pair to write its predictions for the two hidden-label tickets before you reveal the answers.
On the test tickets, the color rule predicts dine-in for the red bag and takeaway for the blue plate. Both predictions miss the kitchen’s labels. The plate-or-bag rule matches both. Ask which combinations were missing from the first four examples, and which additional examples would have exposed the color shortcut sooner.
Using a learned pattern to make a prediction on a new input is called inference. Here, students make those predictions themselves. In a real model evaluation, keep examples separate from those used to train or revise the model. Once students have used the two test tickets to improve their rule, hand them a different ticket for the next check. Google’s guide to dividing datasets
Listen for the reason behind the answer
Give the class a final ticket with a yellow color word and a bag. Ask for the label and the clue they used. Then give them a ticket that omits plate or bag. Ask what they would need to know before sorting it. The first ticket can be labeled takeaway under the stated kitchen rule. The second is a chance to explain what information is missing.
Look for a response like, “Color matched the first set because the examples left out red takeaway orders. I used the bag on the yellow ticket because the kitchen rule describes how the order is served.” Invite a spoken, written or dictated response. Keep students’ usual accommodations in place, and offer a large-print or accessible text copy if useful.
You can connect this activity to AI literacy by asking students where else a computer might use examples to predict a category. Ask them to name a clue that could be misleading and to describe a fresh example that would test it.
A five-minute learning activity
Run the kitchen ticket challenge
Use the fictional tickets above. Allow about 15 minutes for pairs to suggest rules, compare predictions, and explain one mistake.
- Show the four labeled examples. Ask pairs to write the color rule and the plate-or-bag rule in their own words.
- Show the red bag and blue plate tickets with the labels hidden. Ask pairs to predict both labels with each rule.
- Reveal the two labels and the kitchen rule. Have pairs explain how both rules fit the original cards and why their predictions diverged.
- Show the yellow bag ticket and the ticket with no plate or bag clue. Ask each student to predict when they have enough information and to explain what they would ask for when they do not.
Which missing example would have helped you catch the color shortcut earlier? Explain how you would test a revised rule on a fresh ticket.
Common questions
Do I need to run a chatbot in front of the class?
The tickets, hidden labels, and discussion are enough to run this lesson. You can write the cards on the board, read them aloud, or share accessible text.
Is this what a chatbot does?
The activity models a simple classification task with two possible labels. Chatbots that generate text do a different task. Here, students can see how the examples and the test affect confidence in a prediction.
What should I collect from each student?
Ask for one prediction, the clue behind it, and one fresh test case. Those three pieces show how the student used the examples and where they think the rule could fail.
Sources & further reading
Sources were checked on November 11, 2026. The activity is an AI Word Bank learning exercise, not a published study.