What this does: Explains each part of the Evaluation panel so you can make an informed decision about whether to approve, edit, or reject the AI suggestion.
Parts of the Evaluation panel
| Element | What it shows |
|---|---|
| AI Score | The score the AI assigned, displayed prominently at the top of the panel. |
| Confidence bar | A 0–100% indicator of how certain the AI is about its evaluation. Higher means more confident; lower means the AI found the response ambiguous or hard to interpret. |
| Overall rationale | A paragraph explaining why the AI assigned that score. |
| Criteria breakdown | A row for each rubric criterion showing the criterion name, score out of the maximum, and a brief rationale for that criterion. |
| Feedback box | Editable text that will be sent to the student if you approve. You can change this before approving. |
| Score field | Editable final score. You can change this before approving. |
How to interpret the evaluation
- Read the overall rationale first to understand the AI’s general reasoning.
- Check the confidence bar. A low confidence score (for example, below 60%) is a signal to read the student’s submission carefully before deciding.
- Review each criterion in the breakdown and compare it against the student’s actual work.
- Decide whether the AI score matches your professional judgment.
Important reminder
The AI evaluation is a suggestion only. The AI can be wrong, especially on handwritten responses, math work, and answers that fall outside typical patterns. You review everything before any grade is sent.
If something looks off
If the rationale seems unrelated to the student’s work, the submission may not have been read correctly (for example, a photo that was blurry). You can use Re-run to try again, or override the score and feedback manually before approving.