Breakable Machine: A K-12 Classroom Game for Transformative AI Literacy Through Spoofing and eXplainable AI (XAI)

AI systems are vulnerable to manipulation: The "Breakable Machine" game demonstrates that image classifiers can be fooled through relatively simple changes to appearance or environment, revealing that AI systems are more brittle than they may appear to users—an important reality for understanding AI's limitations in real-world applications.

Understanding how AI makes decisions matters for critical thinking: By using explainable AI (XAI) features that show students which visual cues the model focuses on, the game helps learners see the reasoning behind AI decisions, enabling them to identify potential biases and flaws rather than simply accepting AI outputs as objective truth.

AI literacy should focus on breaking, not just building: Rather than teaching students only how to create AI systems, this approach emphasizes interrogating and testing AI critically—a shift that helps develop "data agency" and prepares young people to be informed citizens who can question AI systems embedded in their daily lives.

Hands-on, playful learning builds ethical awareness: By making AI interrogation collaborative and game-based for ages 10-15, the classroom activity transforms abstract concepts about AI bias and failure into concrete, embodied experiences that help students develop ethical awareness about technology from an early age.

Model failure is a teaching opportunity, not a bug: The game reframes AI misclassifications as pedagogically valuable moments for learning, rather than errors to hide—encouraging students to see AI systems as sociotechnical tools worthy of critical examination rather than infallible technologies.