At-Home Mini Lab: Build a Snack Recommendation System (Ages 6–10)

Why snack recommendations are the perfect way to explain algorithms

If your child has ever asked, “How does YouTube know what I want to watch?” or “Why does this game keep suggesting the same stuff?”—you’re already halfway to a great conversation about algorithms.

A recommendation system is just a set of steps (an algorithm) that makes a smart-ish guess about what someone might like next. For kids ages 6–10, the easiest way to make this real is to turn your kitchen into a tiny “AI lab” and use snacks as the “data.”

This mini lab is designed to help parents who are searching for:

• how to explain recommendation algorithms to kids without heavy tech talk
• an algorithm activity for elementary students that feels like play
• a family activity to teach AI recommendations using everyday items
• a simple machine learning demo for kids that shows learning from choices

By the end, your child won’t just repeat a definition—they’ll understand the idea: recommendations come from patterns in preferences, and those patterns can be helpful… or a little weird.

What you’ll need (10 minutes to set up)

Keep it simple. The goal is to model the idea, not to build a perfect system.

Materials

• 6–10 snack options (mix sweet/salty/crunchy/chewy)
  • Examples: pretzels, popcorn, raisins, apple slices, crackers, cheese cubes, yogurt bites, mini cookies
• Small bowls or plates
• Paper + marker (or sticky notes)
• A “rating” method:
  • Option A: thumbs up / thumbs sideways / thumbs down
  • Option B: 0–2 points (0 = no, 1 = maybe, 2 = yes)
• A parent “score sheet” (you’ll make it below)

Safety and allergy note: Pick snacks that are safe for your family. If you’re in a classroom or playgroup setting, stick to packaged allergen-safe items.

Kid-friendly explanation (script you can use):

• “We’re going to build a snack recommender—like Netflix, but for your taste buds.”
• “We’ll collect data (your choices), then our algorithm will make a guess about what you’ll like next.”

Mini Lab Part 1: Collect data like a recommender system

Step 1: Create your “snack features”

In real recommendation systems, items have information about them. For kids, call these clues.

Choose 3–4 simple features and label them on paper:

• Sweet
• Salty
• Crunchy
• Chewy

Now label each snack with 1–2 clues. Example:

• Pretzels: Salty + Crunchy
• Raisins: Sweet + Chewy
• Popcorn: Salty + Crunchy
• Apple slices: Sweet + Crunchy (or just Sweet)

Parent tip: Don’t overthink accuracy. The point is to connect “items have traits” to “recommendations use traits.”

Step 2: Taste-test and rate (your training data)

Have your child taste 5–8 snacks (small bites). After each one, they rate it.

Use either:

• Thumbs: Up / Side / Down
• Points: 2 / 1 / 0

Write the ratings down. You can say:

• “These ratings are our data.”
• “The recommender learns from data.”

Step 3: Fill in a simple score table

Here’s a table you can copy onto paper or recreate as you go. You’re tracking what they liked and what clues each snack has.

Snack	Clues (Sweet/Salty/Crunchy/Chewy)	Kid Rating (0–2)	Sweet Score	Salty Score	Crunchy Score	Chewy Score
Pretzels	Salty, Crunchy	2	0	+2	+2	0
Raisins	Sweet, Chewy	0	+0	0	0	+0
Popcorn	Salty, Crunchy	2	0	+2	+2	0
Apple slices	Sweet, Crunchy	1	+1	0	+1	0
Mini cookies	Sweet, Crunchy	2	+2	0	+2	0

How to use the table:

• For each snack, add the rating points to any clue it has.
• At the end, total each clue column.

What you’re building is a tiny “preference profile.” If Salty and Crunchy totals are high, the system should recommend other salty/crunchy snacks.

Say it like this:

• “The algorithm is counting patterns.”
• “It’s not reading your mind—it’s adding up clues.”

Mini Lab Part 2: Make recommendations (and test them)

Now you’ll act like the recommendation engine.

Step 4: Find the child’s “top clues”

Add up the scores for Sweet, Salty, Crunchy, and Chewy.

Circle the top 1–2 clues.

Example outcome:

• Salty = 6
• Crunchy = 7
• Sweet = 3
• Chewy = 0

Your recommender conclusion: This kid tends to like Crunchy + Salty snacks.

Step 5: Recommend a new snack

Pick a snack your child hasn’t rated yet that matches their top clues.

• If they love Crunchy + Salty: try pita chips, roasted chickpeas, salted crackers
• If they love Sweet + Chewy: try dried mango, fruit snacks, marshmallows
• If they love Sweet + Crunchy: try granola, cereal, apple chips

Ask:

• “Our algorithm recommends ______. Do you want to try it?”

Then have them taste and rate it.

Step 6: Compare “prediction” vs “reality”

This is where the learning becomes real.

Ask your child:

• “Did our recommendation work?”
• “What clue did it get right?”
• “What did it miss?”

Explain:

• “Recommendation algorithms make guesses. Sometimes they’re great, sometimes they’re off.”

Step 7: Update the algorithm (simple machine learning demo for kids)

Here’s the “machine learning” moment: learning from new data.

• Add the new snack’s rating into your table.
• Update totals.
• Recommend again.

Kid-friendly line: “The system learns as it gets more examples.”

This is a genuinely useful way to show a simple machine learning demo for kids: the model changes when it receives new information.

Make it feel like real-world AI: 3 quick twists kids love

These are optional, but they help you explain what’s happening in real apps.

Twist 1: “Because you liked…” explanations

After a recommendation, say:

• “We recommended this because you liked popcorn and pretzels (both salty + crunchy).”

This mirrors the transparency features some platforms try to provide.

Twist 2: Popularity vs. personal taste

In real systems, recommendations aren’t only about you—sometimes they’re about what’s popular.

Do this:

• Ask another family member to rate 3 snacks.
• Declare one snack “Trending Today” (even if your child didn’t love it).

Ask:

• “Should the recommender show what you like most, or what everyone likes?”

That question opens a great talk about why recommendation feeds can feel repetitive.

Twist 3: The “filter bubble” game

If your child keeps choosing only crunchy snacks, your algorithm will keep recommending crunchy snacks.

Show the downside:

• “If we only recommend crunchy foods, we might miss something new you’d love.”

Then add a rule:

• “Every 3rd recommendation must be a ‘wild card’ snack.”

This is a gentle way to explain exploration vs. exploitation without using those terms.

Troubleshooting + parent prompts (so it doesn’t turn into chaos)

A snack lab with kids can go sideways fast. Here are practical fixes.

If your child rates everything a 2 (loves everything):

• Switch to a forced choice: “Pick your top 2 snacks.”
• Or use 0–3 points and require one snack to be a 0.

If your child refuses to taste:

• Let them recommend to you first.
• Allow “smell test” or “touch test” ratings.

If your child wants to change their rating:

• Let them—then say: “When the data changes, the recommendations change.”

Questions that deepen learning (without sounding like a lesson):

• “What do you think the algorithm noticed about you?”
• “If your friend used this system, would it recommend the same snack?”
• “What’s one clue we forgot to track?” (spicy, sour, temperature, color)
• “Should the recommender always try to make you happy?”

That last question is sneaky-in-a-good-way: it leads into healthy media habits and mindful choices.

Next Steps: Turn this into a family AI habit in 15 minutes a week

If you want this to stick (and not be a one-time craft), keep it light and repeatable.

Try one of these next:

• Build a “movie night recommender”: track clues like funny/action/animals/short/long.
• Make a “book recommender”: clues like adventure/mystery/pictures/chapters/series.
• Add one new clue each week: teach your child that better data can improve predictions.
• Do a fairness check: ask, “Is it recommending the same thing too often? Who might that leave out?”

If your child enjoyed being the “AI,” they’re ready for more hands-on learning. On Intellect Council, we turn these exact ideas—patterns, data, and smart guesses—into interactive lessons kids can explore safely, step by step.

Pick a day this week, grab 6 snacks, and run your first mini lab. Your child will walk away understanding recommendation algorithms in a way most adults never did: by building one with their own choices.