Case study 01 · Curriculum architecture
A short curriculum that starts with “how does this thing work?” and ends with students judging their own AI decisions.
Kids can use AI confidently long before they understand what it does, why it makes mistakes or what using it responsibly means. A good curriculum should mix real understanding, hands-on investigation and good judgement, not just tool demos.
I’ve drafted a six-session curriculum. It starts with the difference between sensors, automation and AI, then looks at training data and predictions, separates rule-based systems from generative AI, practises prompting and fact-checking, explores bias, privacy and uncertainty, and ends with students making an evidence-based call on when AI should be used.
| Session | Core question | Demonstrable learner output |
|---|---|---|
| 1 · What makes a system smart? | What is sensing, automation and AI? | Annotated system diagram and classification |
| 2 · Patterns are not understanding | How do data and examples influence prediction? | Mini classification test + error notes |
| 3 · Rules or learned behaviour? | When does an if/else rule differ from a model? | Decision flowchart with explanation |
| 4 · Ask, compare, verify | Why do prompts and context change outputs? | Comparison log and verification checklist |
| 5 · Who benefits, who is affected? | What makes an AI decision responsible? | Stakeholder and risk map |
| 6 · Should we use AI here? | What evidence justifies the choice? | One-page decision brief and reflection |
| Criterion | Emerging | Developing | Secure |
|---|---|---|---|
| Conceptual understanding | Labels systems with support | Usually distinguishes types | Explains differences with examples |
| Evidence & verification | Accepts output uncritically | Checks some claims | Checks claims and records sources/limits |
| Reasoning & alternatives | Gives a single unsupported choice | Compares two options | Weighs trade-offs and justifies choice |
| Responsible judgement | Mentions general safety | Notes privacy or fairness | Anticipates risk and proposes safeguards |
The sequence starts with real, touchable systems instead of chatbots. It moves from noticing to explaining, and ends with judgement, not with generating more content. Every session has the same kind of student output, so a teacher can watch reasoning grow over time. The rubric rewards explanation and evidence, not how shiny the final product looks.
