Provenance Before Proficiency: Scoring Competencies When Students Work With AI

Generative AI has broken the link between the artifact a student submits and the learning it is supposed to demonstrate. A polished answer no longer tells you who reasoned, what they decided, or why. Institutions are responding with AI policies they cannot enforce and with scoring tools that grade the one thing AI writes best.

This session argues for a different starting point: provenance. If we can preserve how a student framed a problem, what they asked, what they decided, and what they took from AI versus contributed themselves, then competency can be scored from that trail — with every score traceable to the student's own words and the instructor holding the final judgment.

Drawing on a live deployment as enforceable course-level AI policy at Babson College and faculty pilots at CSU Channel Islands, we'll cover: what a provenance record needs to contain, how to map it to existing rubrics, what makes an AI-assisted score defensible under accreditation review, and a vendor-neutral checklist for interrogating any tool that claims to score reasoning.