Beyond Plagiarism: Reconstructing Academic Integrity Frameworks under Conditions of Human–AI Co-Production

Abstract

Existing academic integrity policies remain anchored in plagiarism detection, rendering them conceptually inadequate for contexts of routine human–AI co-production. This study develops and empirically validates a multidimensional integrity framework suited to AI-augmented writing environments. We analyzed 72 institutional integrity policies across Southeast Asia and conducted policy-practice alignment audits in five Indonesian universities following the introduction of generative AI guidelines. Additionally, 312 student writing samples were examined using process-tracing software that captured drafting histories, AI interaction logs, and revision trajectories. Findings reveal a structural mismatch between policy language (emphasizing authorship originality) and actual compositional practices characterized by iterative AI prompting, selective acceptance, and human rhetorical reshaping. Integrity breaches were less associated with AI usage per se than with opacity of contribution disclosure. Experimental testing of a “transparent co-authorship declaration” protocol reduced policy violations by 38% without decreasing AI engagement. We propose a shift from prohibition-based integrity models toward accountability-centered frameworks grounded in disclosure, process documentation, and epistemic responsibility. The study reconceptualizes academic integrity as governance of collaboration rather than surveillance of textual similarity.