<![CDATA[Generative artificial intelligence (GenAI) can provide rapid explanations, feedback, ideation, and language support, but the same affordances can displace cognitive work that assessment is intended to reveal. This integrative review synthesizes recent empirical studies, systematic reviews, and policy analyses on GenAI in higher education. Twenty traceable scholarly and institutional sources published primarily between 2022 and 2025 were analyzed through five themes: learning support, cognitive outsourcing, feedback, AI literacy, and academic integrity. The evidence suggests that learning benefits are strongest when students interrogate, verify, revise, and explain AI outputs rather than submit them as finished products. Risks increase when GenAI performs core reasoning, source evaluation, or composition without learner oversight. The review proposes a three-zone assessment architecture: AI-free tasks for measuring independent competence, AI-assisted tasks requiring process disclosure, and AI-required tasks when AI literacy itself is an intended outcome. Effective integration therefore depends less on universal permission or prohibition than on preserving cognitive ownership, transparent attribution, verification routines, and assessment designs that make students’ reasoning visible.]]>
Copyrights © 2026
