One cricket video leads to another. A drawing tutorial is followed by more drawing tutorials. Your child might ask, “How does it know what I like?”
That question is a useful opening for a conversation about AI. A recommendation is a suggestion based on information available to a system. It can help us discover something interesting, but it is not a complete picture of our interests or everything worth exploring.
A timely conversation for families
Common Sense Education's Digital Citizenship Week runs from 19–23 October 2026. Its middle-school activities include exploring how algorithms shape online experiences and how curated content can be useful while also narrowing what people encounter. The event offers a timely reason to discuss recommendations at home.
The paper-card game below is an original BAIO activity. It is not an official Common Sense lesson or a recreation of any platform's algorithm. No account, device or viewing history is needed.
What information can shape a recommendation?
YouTube's own explanation lists watch and search history, subscriptions, likes, dislikes and feedback among its recommendation signals. It also says different parts of the service use signals differently: the current video is a main signal for Up Next, while homepage recommendations rely primarily on watch history.
These examples describe YouTube, not every app. They help illustrate why suggestions can reflect earlier activity without revealing everything a person enjoys. A recommendation also does not establish that a video is accurate or suitable for a particular child.
Try the paper recommendation game
Set aside about ten minutes and make twelve cards. Write or draw three imaginary video ideas in each of four categories. For example, use cricket, drawing, animals and space. Give each card a category label so the rules are easy to follow.
Possible cards include “How a cricket ball spins”, “Draw a bird”, “Why cats have whiskers” and “Make a paper rocket”. These are invented titles for the game, not links to real videos. One person plays the viewer; another arranges the recommendations.
Round 1: Start with a mixed selection
Place one card from each category in front of the viewer. Ask them to choose one they would like to explore. Record the category on a piece of paper, then return the chosen card to its pile.
Ask what the choice tells you. It shows that the viewer selected that card in that moment. It does not tell you why they chose it or whether they dislike the other three categories.
Round 2: Follow a simple rule
Use this rule for the next selection: show all three cards from the chosen category and one card from a different category. The viewer chooses again. Return the cards and repeat the rule once more, using the latest choice.
If the viewer keeps choosing cricket, most of the cards they see will be about cricket. If they choose the different category, the next selection changes. Keep a record of which categories appeared, so you can discuss the result without relying on memory.
Round 3: Change the rule
Now show two cards from the latest chosen category and one card from each of two other categories. Ask the viewer to choose again. Compare the range of topics offered under the two rules.
Discuss which selection felt more useful and why. There is no single correct preference. Sometimes a person wants depth in one interest; sometimes they want to discover something new. The rule changes the choices available before the viewer makes a decision.
What the game shows—and what it leaves out
Our pretend system follows a rule written by a person. It does not train an AI model, analyse a real account or explain exactly why a particular video appeared. Real recommendation systems can use many signals and much more complex methods.
The game makes one relationship visible: a choice can influence the next set of suggestions, which then shapes the next choice. When some categories receive fewer chances to appear, the viewer has fewer opportunities to select them within that set.
Ask your child, “What might the system miss about you?” They may enjoy space even if they chose cricket today. They may have selected a card for a school project or because its title sounded funny. An observed action can leave the reason behind it unclear.
Three questions to take beyond the game
First, ask: “What did I come here to find?” Choosing a topic before browsing can help a family discuss whether the suggestions match their purpose. A parent or teacher can help identify suitable sources for that topic.
Second, ask: “What else would I like to explore?” Make a small list of interests that did not appear in the paper feed. For a younger child, draw those interests together. This is a way to practise making deliberate choices without opening an app.
Third, ask: “What would make this information trustworthy?” A familiar topic or an attractive suggestion still needs a closer look. Discuss who made the content, what supports its claims and when an adult or another source could help check it.
Adapt it for Classes 3–8
For Classes 3–5, use pictures and compare the categories aloud. Focus on how a rule changes what appears next. Let children swap roles so they can both make choices and follow the recommendation rule.
For Classes 6–8, count how many categories appear in each round and propose another rule. Ask what their rule prioritises: similarity, variety or a balance. Invite them to explain one trade-off and one limitation of their design.
The activity connects everyday media choices with BAIO's focus on AI literacy, reasoning and problem-solving. Continue with our AI mistake challenge or the family AI design activity, linked below. The useful takeaway is a question children can keep asking: “Why might I be seeing this, and what else could I choose?”
Key Learning Self-Check
Test your understanding of the essential insights from this guide:
Did you find this guide insightful?
Click an emoji to share your reaction with the BAIO editorial team
Frequently Asked Questions
Does this game show exactly how YouTube or another app works?
No. It uses transparent rules and fictional cards to illustrate a feedback loop. Real systems use more complex methods and signals that vary across services and features.
Are recommendations always a bad thing?
No. Suggestions can help people find useful or enjoyable content. The activity explores how a selection is made, what it may leave out and how viewers can make deliberate choices.
Does my child need an account or phone to take part?
No. Paper, a pencil and twelve fictional topic cards are enough. Children can explore the idea without sharing personal information or using a real video feed.