Reproduction and use-decision worksheet

Copy this blank template into ignored review-notes/ or your institution's approved submission location. Do not commit names, student feedback, credentials or private datasets. Public templates are not a storage location for completed reviews.

• Module and task:
• Verification level: calculation audit / same-environment rerun / independent environment reproduction / reading only:
• Source commit or supplied release identity:
• Python, package versions, OS, kernel and hardware relevant to the result:
• Data source, license, checksum or teaching-fixture identity:
• Exact command, configuration and output directory:
• Prediction before inspecting feedback:
• Artifact and columns/records supporting the claim:
• Independent calculation and expected/observed result:
• Comparison rule declared beforehand (exact counts; numerical tolerance with rationale):
• Agreement or discrepancy, diagnosis and corrective action:
• Training, validation and test boundaries; anything changed after seeing results:
• Intended user, decision and consequence of a false alert or miss:
• Baseline/nonautomated alternative and comparison:
• Who could be harmed or excluded; what evidence is missing:
• Supported conclusion:
• Conclusion the evidence cannot support:
• Next evidence needed, responsible reviewer and pause/rollback condition:
• Assistance used and what I can explain independently:
• Peer or self-review; clarifications needed; unresolved issues:

Pilot observation (optional; confidential)

Record module, setup route (local/WSL/hosted), prior experience category, setup minutes, active task minutes, download/wait minutes, stall point and recovery/help used. Record which pages were skipped and why; ask rather than infer disengagement. Note interface/accessibility barriers without collecting diagnoses. Record whether the task was completed independently, with help or not completed. Use these observations to revise instructions and workload; do not treat AI execution time as learner time.