If you were unable to join us for the Teaching AI Through Play: Exploring Gaimes Lab online community meeting, don't worry! You can catch up on all the content and a recording of the session below.

---

Teaching AI Through Play: A Fresh Approach with Gaimes Lab

---

Key Takeaways

• Gaimes Lab is a browser-based platform using video games to teach pupils how to train AI models, not just label data.

• The platform focuses on reinforcement learning, enabling children to explore how AI can make decisions and improve through trial and error.

• Free and paid versions are available, with eight structured one-hour lessons and teacher resources.

• Pupils can design their own games and race against AI they've trained themselves, deepening their understanding.

• Resources are thoughtfully designed for primary-age learners and accessible for teachers with no prior experience in AI.

---

Main Body

In this CAS Primary Community session, we were joined by Ethan from Gaimes Lab, who introduced a highly engaging tool for teaching artificial intelligence through the medium of games. Designed primarily for KS2 pupils—but also suitable for younger KS3—the platform enables learners to build, test, and refine AI models within familiar game environments like a racing circuit and an alien shooter.

Ethan began by explaining the gap he noticed while teaching coding with Scratch, MakeCode Arcade, and LEGO Robotics: while these tools were excellent for programming fundamentals, there were few age-appropriate ways to help learners train AI models. Most available tools involved labelling datasets for AI to process—but this, he argued, lacked the creative depth and agency that would help students truly grasp how AI functions.

With Gaimes Lab, students learn how to choose inputs (what the AI sees), outputs (what actions it can take), and rewards (what the AI is incentivised to do). These core concepts of reinforcement learning are brought to life as students train AI to complete tasks such as driving a car around a track or shooting down aliens. Importantly, learners get to see the AI improve over generations and are encouraged to observe, reflect, and refine their decisions.

Two sessions—focused on outputs and accessible freely—serve as an excellent entry point. The session highlighted the importance of allowing learners to experience the game environment first-hand before training the AI. For instance, students can manually race a car around a track to better understand the challenges their AI agents will face.

From there, the programme builds on core AI concepts: how AI agents "see" using radar-like inputs, how rewards shape behaviour (sometimes unintentionally!), and how different design choices influence outcomes. It culminates in opportunities for students to create their own levels and race against their own trained AI, adding a competitive and reflective edge.

Teacher resources include slide decks, lesson plans, and helpful analogies (such as training a puppy with treats) to explain abstract concepts in relatable ways. Whether you’re new to AI or an experienced computing teacher, the resources are designed to support you in leading these lessons confidently.

---

Next Steps

Here are some reflective questions to consider for your own teaching:

• How are you currently introducing AI in your curriculum? Is it mostly through data labelling?

• Are students given opportunities to see AI behaviour evolve and change over time?

• What creative ways could you use reinforcement learning concepts in unplugged activities?

You might also try:

• Asking pupils to design their own reward systems and predict how they might influence behaviour.

• Creating a classroom discussion on the "intelligence" of AI, highlighting its limitations and how it differs from human thinking.

• Using Gaimes Lab’s free levels to prompt inquiry-based learning: What does the AI do? Why? How can we help it improve?

---

Further Resources

🔗 Gaimes Lab Website

🎥 [Session Recording – CAS Primary Online Community Meeting, 29 April 2025]

🏫 CAS Primary Community Hub on CAS.org.uk