Author
Master’s student, Computer Science, College of Computer Science, King Khalid University, Kingdom of Saudi Arabia
[email protected]
Professor, College of Computer Science, King Khalid University, Kingdom of Saudi Arabia
Abstract
Background. Artificial intelligence (AI) is increasingly embedded across K–12 and higher education for assessment, tutoring, feedback, predictive analytics, student support, content generation, and institutional decision-making. The rapid diffusion of generative AI has compressed adoption timelines and intensified pre-existing ethical debates surrounding privacy, fairness, transparency, accountability, learner agency, and academic integrity.
Objective. This PRISMA-informed systematic literature review synthesizes recent peer-reviewed scholarship and policy-oriented evidence concerning the ethical dimensions of AI applications in education, with particular emphasis on the 2023–2025 period.
Methods. The review was conducted in accordance with the PRISMA 2020 reporting framework. Predefined review questions, eligibility criteria, a reproducible search-string template, a screening protocol, and a thematic synthesis approach were applied. Searches drew on Scopus, Web of Science, ERIC, IEEE Xplore, ACM Digital Library, ScienceDirect, and SpringerLink, supplemented by policy literature from UNESCO, the OECD, and the European Union. A thematic synthesis was conducted across recurring ethical domains, and a quality-appraisal plan suited to the heterogeneity of the corpus was specified.
Results. Six interdependent ethical domains were consistently identified across the corpus: (1) privacy and surveillance; (2) fairness, bias, and inclusion; (3) transparency and explainability; (4) accountability and governance; (5) learner and educator agency; and (6) academic integrity. Five cross-cutting tensions characterize the field: personalization versus surveillance, efficiency versus pedagogical autonomy, innovation versus fairness, automation versus accountability, and assistance versus integrity. Persistent gaps include limited longitudinal and comparative empirical work, underrepresentation of K–12 and Global South contexts, insufficient attention to procurement and platform governance, and a tendency to treat heterogeneous AI technologies as interchangeable.
Conclusions. Ethical AI in education should be conceptualized not only as a technical design problem but also as a pedagogical, institutional, and sociopolitical challenge. Responsible adoption requires participatory governance, transparent procurement, meaningful human oversight, redesigned assessment, and context-sensitive policy. Future research should prioritize longitudinal and implementation-focused studies, technology-specific risk–benefit assessment, and stronger representation of underrepresented educational contexts.