Case Study: An After-School AI Club for 5th Graders—No Laptops Required

The Challenge: Teaching AI When There Are Zero Devices

This case study comes from a public elementary school after-school club that wanted to add “AI week” to their STEM rotation—except they had a real-world constraint: no laptops, limited Wi‑Fi, and only one shared classroom projector that wasn’t always available.

The club served 18 fifth graders (ages 10–11) across six 45-minute sessions. The facilitator was a veteran paraprofessional (not a computer science teacher) with a co-leader who usually ran robotics.

Their goal wasn’t to make kids “program an AI.” It was to help students understand:

• What AI is (and what it isn’t)
• How AI “learns” from data
• Why bias and mistakes happen
• How to think like an AI designer (asking good questions, testing, improving)

In other words: the club wanted AI literacy—the kind that translates to real life and makes future coding easier.

This is exactly the kind of situation parents ask about: how to teach AI without computers and still keep it fun.

The Plan: A 6-Session Unplugged AI Club That Actually Worked

The club used a simple structure each week:

1. Hook (5 min): a quick scenario kids could relate to (YouTube recommendations, photo filters, game matchmaking)
2. Unplugged AI activity (25 min): hands-on group challenge
3. Debrief (10 min): what the activity “stands for” in real AI systems
4. Exit ticket (5 min): a one-sentence reflection or mini quiz question

Below is the exact sequence they ran—these are practical ai club ideas for elementary school you can copy.

Session	Big Idea	Unplugged AI Activity	Materials	What Kids Learned	Quick Assessment
1	AI vs. “regular” rules	“Robot or Not?” Sorting Game	Scenario cards	Some systems follow fixed rules; AI uses patterns from examples	Students explain one example of AI in daily life
2	Training data	“Teach the Classifier” with picture cards	Animal/food/object cards, bins	AI needs examples; the examples shape the model	Kids predict how changing examples changes results
3	Features (signals)	“Mystery Object: Feature Hunt”	Bag of objects, feature checklist	AI uses measurable clues (features), not “magic”	Teams list 3 good features for an object
4	Errors + iteration	“Confusing Cases” (edge cases challenge)	Card set with tricky items	Models fail on unusual cases; you improve by testing	Students propose one improvement
5	Bias + fairness	“Data Detective: Who’s Missing?”	People cards with attributes, chart paper	If your data is unbalanced, your system can be unfair	Kids identify what’s missing and how to fix it
6	Responsible AI	“Design an AI Helper” poster sprint	Poster paper, markers	AI should solve real problems with limits + safety	Teams present constraints and a testing plan

What made this work wasn’t fancy materials. It was consistent routines, kid-friendly analogies, and a strong focus on discussion.

Inside the Activities: Three Unplugged AI Lessons Kids Loved

1) “Teach the Classifier” (Training Data in 25 Minutes)

The leaders gave each group a stack of cards with images (animals, sports, foods, weather, household objects). The groups had to invent a “sorting machine” by creating two bins (for example: “Healthy” vs. “Not Healthy” or “Pet” vs. “Not Pet”).

Then the twist: the facilitator acted like the “AI” and only followed what the group had shown through examples. If the kids gave confusing examples (e.g., put a tomato in “Vegetable” while also saying “vegetables are always green”), the “AI” would make mistakes.

Key moments that landed with 5th graders:

• Kids realized they were “training” the system by what they included.
• They noticed how a few odd examples could throw everything off.
• They naturally started using the language of testing: “Let’s try new cards and see what happens.”

Parent takeaway: This is one of the best unplugged AI activities for kids because it shows that AI learning is “pattern learning,” not mind-reading.

2) “Mystery Object: Feature Hunt” (Why AI Needs Clues)

A bag of objects went around (mini basketball, spoon, leaf, Lego piece, pinecone, etc.). Teams couldn’t name the object at first. Instead, they had to describe it using features:

• Shape (round, flat, long)
• Texture (smooth, bumpy)
• Material (metal, plastic)
• Weight (light, heavy)
• Color (but with a warning: color can be misleading)

Then teams tried to “classify” the object using only their feature notes.

This led to a simple but powerful discussion: AI doesn’t “understand” like humans do. It uses signals. If your signals are weak, you’ll get weak results.

Pro tip the club used: They banned “vibes-based” features like “looks tasty” and replaced them with measurable ones like “is it usually eaten?” or “does it grow on a plant?”

3) “Data Detective: Who’s Missing?” (Bias Without the Lecture)

This session could have gotten preachy. Instead, the club turned it into a detective game.

Each team got a “dataset” of people cards for a pretend AI that recommends after-school activities. The cards had attributes like:

• Grade level
• Preferred activity (art, sports, reading, coding)
• Availability (days of week)
• Transportation options

But the dataset was intentionally skewed (for example, too many sports-lovers, or mostly students who have rides home).

The teams had to answer:

• Who might this system work well for?
• Who might it fail for?
• What new data should we collect—and how?

This made fairness concrete. It wasn’t abstract. Kids could see how missing data could accidentally exclude someone.

What We Learned: Why This After-School STEM Program AI Unit Succeeded

The club’s biggest win was that students stayed engaged without screens. Here’s what the facilitators reported worked best.

The “secret ingredients” that kept kids hooked

• Roles in every group: “Data Manager,” “Tester,” “Recorder,” “Presenter” (kids rotated weekly)
• Fast cycles: make a guess → test it → update the rule
• Visible thinking: chart paper for rules, examples, and “mistakes we found”
• Short debriefs with real examples: recommendations, face unlock, spam filters

Common pitfalls (and how they fixed them)

• Pitfall: Kids think AI is a robot brain.

  • Fix: Start every session with, “What is the input? What is the output?”

• Pitfall: One confident student dominates.

  • Fix: Roles plus a rule: the “Presenter” can’t be the “Data Manager” the same week.

• Pitfall: Activities feel like random games.

  • Fix: Always end with the same three questions:
    • What did our “AI” do?
    • Why did it make mistakes?
    • What would we change if we had more time/data?

Measurable outcomes (simple, but meaningful)

By week 6, the facilitator shared three signs the learning stuck:

• Students could define AI as “learning patterns from examples” (not “a robot”).
• When the “AI” made a mistake, students stopped blaming the machine and started asking about data and rules.
• Multiple students brought in real-life examples (“My streaming app keeps recommending…”), which showed transfer beyond the club.

This is the core of a strong after school STEM program AI unit: kids leave more curious, more critical, and more confident.

Next Steps: How to Run Your Own Unplugged AI Club (No Laptops Needed)

If you’re a parent, educator, or coordinator building an AI club, here’s a simple way to start.

A starter kit you can assemble in under an hour

• Index cards + markers
• Printed picture cards (animals/foods/objects)
• 10–15 everyday objects for the mystery bag
• Chart paper or a whiteboard
• Sticky notes for “training examples” and “test cases”

A repeatable 45-minute meeting agenda

• 5 min: real-world AI example (one sentence + one question)
• 25 min: hands-on activity with roles
• 10 min: debrief using “input → pattern → output”
• 5 min: exit ticket (one sentence or one multiple-choice question)

If you want to go from unplugged to digital later

Once kids have the concepts, adding devices becomes smoother. Look for beginner tools that:

• Let students experiment with classification safely
• Show how training examples change results
• Encourage testing and iteration

At Intellect Council, we’re big believers in starting with understanding first—then layering in coding when kids are ready.

If you’re planning your own club and want help picking age-appropriate lessons, pacing, or “what to do when the activity flops,” your next step is simple: outline your 6 sessions using the table above, pick one activity to pilot next week, and collect student reflections after every meeting. That feedback loop is the same skill we teach about AI—and it’s what makes a club improve fast.