1. Rubric alignment

Each assessment has explicit rubric dimensions that define what understanding looks like.Rubric criteria distinguish between surface-level recall and demonstrated comprehension.Rubric dimensions map to the specific skills or knowledge the assignment is designed to assess.

2. Artifact submission

Students submit their actual work product (essay, code, slides, report) as the starting point.The submission format captures enough context for meaningful follow-up questions.Submission deadlines allow time for the follow-up step before grading begins.

3. Follow-up design

Follow-up questions are generated from the student's specific submission, not generic prompts.Questions target rubric dimensions and probe for understanding beyond what the artifact shows.Students can respond in the mode that best demonstrates their understanding (text, audio, or video).Accommodations (extended time, alternative formats) are available and documented.

4. Evidence structure

Student responses are linked to specific passages or elements in the original submission.Each rubric dimension has a separate evidence score with supporting excerpts.The evidence record connects artifact, response, rubric finding, and instructor decision.Evidence is timestamped and attributable to a verified student identity.

5. Instructor review

Instructors can view the submission and evidence side by side before assigning grades.AI-generated scores are suggestions that instructors can accept, adjust, or override.Instructors can annotate, flag concerns, and add notes to the evidence record.The final grading decision is always made by the instructor, not the system.

6. Institutional readiness

Academic integrity policies reference evidence standards, not just detection scores.Appeals processes can reference the evidence record rather than relying on memory.Data retention and privacy policies cover follow-up response data.Pilot criteria are defined: which courses, how many sections, what success looks like.

Capability Evidence Checklist for AI-Era Assessment