Authentic e-assessment in South African higher education: A systematic review of AI-driven assessment practices and challenges

The growing use of AI in higher education has heightened interest in genuine e-assessment to improve assessment efficacy, inclusivity, and academic integrity. Nonetheless, the implementation of AI-enhanced assessment techniques in South African higher education is inconsistent, influenced by structural disparities, ethical dilemmas, and differing degrees of institutional preparedness. This study offers a qualitative systematic literature analysis that investigates the implementation issues and pedagogical ramifications of AI-enhanced genuine e-assessment at South African Higher Education Institutions (SAHEIs). The evaluation, structured using a PICo-informed framework and according to PRISMA principles, synthesised evidence from 49 peer-reviewed publications published from 2020 to 2024. A narrative theme synthesis was utilised to examine persistent patterns concerning implementation obstacles, ethical considerations, assessment change, inclusion, and governance. AI-driven assessment technologies have significant advantages, such as higher grading consistency, scalability, tailored feedback, and increased student involvement. Nonetheless, infrastructural constraints, ongoing digital disparities, inadequate faculty readiness, and the lack of cohesive institutional policies hinder these advantages. Ethical concerns, especially algorithmic bias, data privacy, and academic integrity, exacerbate the implementation of AI in varied and resource-limited settings. This analysis finds hybrid evaluation models that integrate AI automation with human judgement as the most pedagogically sound and ethically responsible method. The study indicates that AI-enhanced authentic e-assessment possesses transformative potential for SAHE, contingent upon context-sensitive policies, ongoing professional development, equitable infrastructure investment, and strong ethical governance frameworks.