REFLECT · Return to the evidence, not the polish

• Bring the exact FIND story, first ideas, final products, DRAFT sheets and sources.
• Include initial AI images or assumption sketches only where students actually made them. Do not manufacture a retrospective “first” image.
• Print the three-page REFLECT sequence.
• Pair groups for evidence-based exchange and identify two with the same or related history for step 4.
• Students answer either the AI audit or the no-AI tool questions, not both.

1. Return · 5 min. Ask what can be defended now that could not be defended in FIND. Require the WE STILL DON’T KNOW box.
2. Two Portraits · 10 min, conditional. Compare the initial assumption with the final representation. Ask what evidence corrected, supported or could not resolve. Focus on assumptions, not visual quality.
3. Exchange · 12 min. Reviewers identify the main idea, one strongly supported part and one “How do you know?” moment. Creators answer KNOW / THINK / DON’T KNOW with reasons.
4. Two defensible stories · 5 min. Elicit differences in perspective, sources, question, emphasis and interpretation while keeping the shared evidence boundary.
5. Tool audit · 8 min. AI users describe help, errors/oversimplification, checks, rejections and work possible without AI. If a prompt does not apply, they say so. No-AI students explain what their tools could and could not decide.
6. Final question · 5 min. Did the class recover the untold story: yes, partly or no? Reasons matter more than a preferred answer.

If step 2 is skipped, use released time for the exchange and discussion. Do not invent a substitute activity merely to fill the period.

Students did not recover everything someone thought, felt or experienced. They investigated a perspective omitted by the starting story and reconstructed what the surviving evidence permits. Some matters became clearer, some remain uncertain and some may never be known.