Definition
The AI score is the original automated result. The effective score uses the latest saved manager decision for each reviewed criterion and the AI grade for criteria without a decision. A manager decision does not erase the AI grade.
Context
Open a completed run from QA Analysis · QA Grades and select the target being graded. The run shows AI and effective scores, review progress, and a separate critical result when a critical check applied.
Attributes
- Approve AI grade: Confirms the AI grade. The effective grade stays the same, but the criterion becomes reviewed.
- Disagree: Replaces the grade used for the effective result with another grade allowed by the scorecard version frozen for that run. A change to a weighted criterion can change the numeric score; a change to a critical criterion can change the critical-failure result.
- Not applicable: Removes that criterion from scoring and renormalizes the remaining applicable weighted criteria. If nothing applicable remains to score, the result is Not scored, not zero.
- Clear decision: Removes the current manager choice so the effective result follows the original AI grade again. The decision history remains available.
- Publication: Saved decisions affect the manager's effective result and QA reporting. An agent sees a snapshot only when a full Quality manager publishes the agent grade. Later corrections stay internal until a new version is published.
The criterion's Show change history records decisions over time. See How To Review And Adjust Criterion Grades for the editing steps and How Corrections Affect A Published Result for what the agent sees.