6-8:06:Hands-on Model Training Simulation and Troubleshooting

Get your hands dirty. Students run a model-training simulation, watch it succeed and fail, and learn to spot and fix the most common problems.

Watching a model fail is more instructive than watching it succeed. Students leave able to name the usual suspects – not enough data, wrong labels, over-fitting – and with the habit of asking “what went wrong?” before giving up.

What students do
• What Happens When You Train a Model? – a live demo with a Teachable-Machine-style tool, with fill-in-the-blanks
• The Struggle Is Real: overfitting, underfitting and inaccurate predictions, and how to spot them
• The Fix-It Crew: debug model problems together using a troubleshooting checklist
• Look Back to Move Forward: “What challenges did I face, and how did I solve them?”, then Recap – Becoming Better Model Builders

By the end, students can
• Overfitting, too little data and bad labels – in plain language
• A debugging mindset
• Improving through iteration

Learning focus (Fink’s Taxonomy of Significant Learning): Application · Learning How to Learn.

Hands-on simulation; no coding needed. Part of Grades 6–8: AI Agents in Action, Module 2 – How Machines Learn – Patterns, Data & Bias, Week 2. Next lesson: “Fix the Bias” Game and Presentations on Ethics and Data Quality.