Changes apply to all in-flight assessments. Historical completed assessments keep their original question text via version snapshots.
Every question is tagged at the database level to exactly one construct. This is the chain a consultant or client can show the stakeholder asking “where did this question come from?”
1 · Item
How are model performance regressions discovered today?
كيف يتم اكتشاف تراجع أداء النماذج اليوم؟
2 · Pillar
3 · Score map
| Option | Score (1–5) |
|---|---|
| Customer complaint or external incident | 1 |
| Quarterly manual review | 2 |
| Monthly dashboards reviewed by team | 3 |
| Automated alerts on threshold breach | 4 |
| Automated alerts + auto-rollback + post-incident review | 5 |
No validation evidence captured for this question yet. Click below to ask Claude for an AI-suggested anchor against the curated bibliography in the methodology brief. Suggestions are saved as ai_proposed and require admin review before they appear in any client report.
For multiple choice / yes-no, enter options and score_map as JSON.