Both Sides of the Desk: Generative Artificial Intelligence in Production and the Judgement of Student Work

Abstract

Higher education has responded to generative artificial intelligence largely as a problem of student conduct. Policy, regulation and scholarship address what students may do with these tools, how such use should be declared, and how qualifications can be assured when detection is unreliable. Far less attention has been directed to the other side of the desk. Generative systems are now used to draft feedback, apply rubrics, screen submissions and support moderation, and this use is governed largely by the assertion that final responsibility rests with the human marker. The result is an asymmetry: student use is regulated by declaration, surveillance and sanction, while institutional use of the same technology in judging that work is regulated by a statement of intent. Developed through a narrative review of the assessment security, automated assessment and algorithmic accountability literatures, this conceptual paper argues that the two developments form a single phenomenon. Generative artificial intelligence now occupies three positions in the assessment relationship simultaneously, as producer of student work, as arbiter of it, and as the subject of what is assessed, and each position generates a distinct problem of validity or accountability. The paper proposes a four-dimensional framework of assessment trustworthiness comprising validity, transparency, security and equity, and finds institutional practice well developed on security, asymmetrical on transparency, under-examined on validity and largely silent on equity. Six propositions and a research agenda follow, with particular attention to non-self-accrediting providers, which are largely absent from the current evidence base.