Answers are grounded in this paper and cite their sources.
Generative artificial intelligence (GenAI) can produce fluent academic work without establishing that students performed the reasoning a submission appears to represent. This study examines how assessment can distinguish student reasoning from its written reconstruction and from the intellectual work supplied by AI. Across 14 structured AI-supported assignments, students submitted a Core Reference Extract (CRE), a concise, student-authored metacognitive reconstruction of consequential reasoning, alongside the full interaction transcript. An instructor-aligned audit generated a constrained counterfactual CRE: the strongest faithful reconstruction supported by the student’s demonstrated contributions in the transcript, evaluated against assignment requirements and disciplinary criteria. Comparing these representations disclosed three recurring relationships: the CRE understated reasoning demonstrated in the dialogue; a polished CRE overstated student control where AI supplied much of the intellectual architecture; or the CRE faithfully represented inquiry governed by student judgment despite substantial AI assistance. The contribution is a method for testing the evidentiary warrant of a submission, not a means of detecting AI use or independently observing cognition. It distinguishes subject-matter reasoning, metacognitive reconstruction, and evaluative governance, and shows how their misalignment can direct grading and feedback. AI organizes and reconstructs process evidence; the instructor retains responsibility for the interpretation and consequential assessment judgment.
No takes yet. Share an insight, caveat, or question.
Lars R. Jones (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: