Goal: record a true prior prediction about a visible feature, compare with an actual approved view if available and preserve mixed/unknown outcomes. Proposed codes: AC9SFI01, AC9SFI02 (actual only), AC9SFI04 (actual prior/result only), AC9SFU03. Prepare: one new approved broad safe sample or L model only, safe source gate, Card 138.
- Gate · 3 min. Teacher pretests material identity, safety and access for a genuinely new large sample; otherwise REAL OBSERVATION NOT AVAILABLE.
- Predict · 4 min. Child predicts “I may see broad ribs / no visible ribs” before any actual view, or chooses NO PREDICTION.
- Inspect/model · 5 min. Reveal/inspect only an approved real sample, or use an explicitly invented L drawing; no forced contact.
- Record · 7 min. Card 138 routes log prior wording/order, exact observed feature or NOT VIEWED/UNCLEAR, material source and access. A model answer is not a real result.
- Compare · 4 min. Compare prediction with later genuine direct feature only when both exist; otherwise CANNOT COMPARE. Do not treat a texture guess as material proof.
- Exit · 2 min. Save one cautious next question, not a success label.