Assessment Validity, Authenticity, and Academic Integrity in AI-Pervasive Higher Education: Evidence and Design Responses

Abstract

AI-pervasive higher education is used in this entry as a working descriptor, adopted by the authors rather than drawn from an established terminology, for assessment environments in which students can readily use generative systems to produce or transform text, code, images, audio, and multimodal artefacts. Their availability weakens a familiar inference: that an unsupervised submission records the unaided capability of the named student. Because validity is a property of the inferences drawn from results rather than of the task itself, it raises a question about what assessment evidence can support and therefore concerns validity as well as academic integrity. Under the conditions examined in the available evidence, detector scores and unaided academic judgement have not proved sufficient, on their own, to restore that inference once text has been edited, and detection-led responses raise risks for fairness, trust, and due process. A defensible response separates three purposes that are often conflated: assessment for learning, in which generative tools may legitimately be used; independent certification, which warrants a claim about what a student can do unaided; and AI-augmented professional performance, in which working with such systems is itself the assessed competence. Permitted tools are aligned with each purpose, evidence of process and judgement is gathered alongside the final artefact, and identity-verified conditions are reserved for consequential claims. Authenticity then concerns both participation in professional practice and credible evidence of the student’s own contribution, while integrity requires transparent permissions, attribution, verification, accountability, and equitable access. Drawing on a state-of-the-art review of 48 sources, 41 of them published between 2022 and 2026, analysed through three methodological pathways matched to the research questions of the entry, this entry uses DESIGN (Declare, Evidence, Secure, Include, Guard, and Navigate) as an author-developed mnemonic that summarizes these established principles and is derived from the findings of those pathways.