Tomorrow Ready Resources → Evidence of Thinking → Train, Test, Explain in Years 7 to 10
[S] Secondary Evidence of Thinking

Train, Test, Explain in Years 7 to 10

Students stop treating AI as magic when they build and test a classification system themselves. Failure analysis requires situational reasoning, exactly what AI cannot fake.
Classroom Strategy

Run a 30-minute mini-version. Every artefact captures decision-making rather than polished output.

  • Give students 12 example items
  • Students write a label rule card defining their classification categories
  • Test a few new items against the rule card
  • Explain two misclassifications: what went wrong and why

Assess the rule card and the test log. The explanation of misclassifications is the highest-value thinking evidence because it requires situational reasoning that cannot be outsourced.

The Assessed Artefacts

Three artefacts, each capturing a different layer of thinking.

  • Label rule card: the decisions made before testing began
  • Test log with misclassifications and hypotheses: what happened and why
  • Short reflection: what changed when we changed the data?

The reflection question is the integrity move. A student who genuinely ran the classification system can answer it. A student who outsourced the work cannot reconstruct the reasoning.

Adapted for Years 5 to 6

Sort lunch waste into categories. Students define rules for what belongs in each group, test with new items, and explain two sorting mistakes. Focus on the rule card as the thinking artefact.

Secondary — Years 7 to 10

Full workflow: define categories, build a dataset, write a label rule card, test and log failures, analyse three misclassifications, improve labels, re-test, then discuss what should never be automated and why.

Decision Checkpoint

Are your dataset examples and labels culturally safe, privacy-safe, and genuinely representative of your learners' contexts?

Governance Reminder

Ensure images or items are free of identifiable people and personal information. Allow defined roles so all students can contribute regardless of technical confidence. Check that datasets represent variety and do not privilege one context over others.

Print or save this resource as a PDF using your browser.

More resources in this theme

Browse all Evidence of Thinking resources →
← Previous: Make AI Errors the Main Learning Event Next: Privacy-First Classroom Routines for GenAI →