From Principles to Practice: A Systematic Review of Generative AI-Responsive Assessment Interventions in Higher Education

Abstract

Generative artificial intelligence (GenAI) challenges higher education assessment because a polished submitted artefact may no longer provide transparent evidence of the learner's knowledge, reasoning, or competence. This systematic review examines empirical studies in which assessment practices were deliberately configured in response to GenAI and implemented with higher education students. It synthesizes evidence concerning assessment design, the role assigned to GenAI, evidence of student learning, educational outcomes, and implementation conditions. Five recurring functional responses are identified: regulating AI use, structuring human-AI collaboration, requiring critical evaluation of AI-generated material, making the human learning process visible, and verifying competence through oral or performance-based evidence. Across these approaches, the central design issue is not simply whether GenAI is permitted, but how cognitive responsibility is distributed between learner and technology and how the learner's contribution is evidenced. The review therefore proposes an AI-Human Assessment Responsibility Model linking AI delegation, human cognitive responsibility, and evidentiary mechanisms. Sustainable GenAI-responsive assessment should be designed from intended learning outcomes and evaluated in relation to learning, validity, integrity, equity, workload, accessibility, and scalability rather than seeking a universally 'AI-proof' format.