evaluation-design

To design an AI safety evaluation using a rubric, you align instrument inputs and scoring against a 7-dimension framework, generating a complete design document with an operational checklist and final scorecard.

A 7-dimension rubric for AI safety risk assessment is a structured scoring framework that evaluates evaluation designs across multiple operational criteria to ensure consistent and defensible risk-grounded decisions.

To review an existing AI safety evaluation design, score the current instrument against the 7-dimension rubric, verify exclusion criteria, and apply an operational checklist to produce a final Assessment Scorecard.

Providing Phase 2 targets is required to generate a Phase 3 evaluation design, ensuring your evaluation question and instrument inputs align directly with established risk assessment targets.

An evaluation design document for AI safety should include the evaluation question, instrument design with inputs and scoring, item generation, tier justification, operational checklist results, exclusion criteria, and a final scorecard.