DF-COMP-002 — Perceptual Evidence Path Review

Use this framework to evaluate a composition as a trajectory through evidence and states rather than as a static frame.

Review Procedure

1. Define the task. State what the observer must detect, identify, compare, understand, decide, or do.
2. Define the initial view. Record what is visible before an eye movement, scroll, interaction, or locomotion.
3. Predict the initial hypothesis. Describe the likely scene category, hierarchy, and action possibilities.
4. Identify ambiguities. Mark crowded, hidden, semantically weak, uncertain, or expectation-dependent information.
5. Predict the next sample. State where the observer is likely to look or move and why.
6. Check evidence continuity. Determine whether the next view preserves spatial, semantic, and object continuity.
7. Check violations. Confirm that consequential anomalies and state changes break expectation.
8. Check verification. Ensure identity, state, consequence, and completion can be confirmed when error cost is high.
9. Measure. Record first-fixation target, time to first fixation, fixation count, scanpath length, recognition and relational accuracy, state comprehension, task time, errors, recovery, and confidence calibration.
10. Falsify. Compare an alternative with the same content but a different evidence path and test the predicted advantage.

Interpretation Rules

• Do not equate fixation with comprehension.
• Report first-glance and full-task performance separately.
• Treat observer goal, expertise, culture, ability, and device as explicit parameters.
• Preserve task-relevant violations while reducing irrelevant uncertainty.
• Pair gaze measures with accuracy, memory, confidence, and action measures.

Traceability

Derived from RP-COMP-005 and operationalizes TH-COMP-005.