AI-Driven Decision Support Systems in Education: A PRISMA 2020 Systematic Review of Applications, Adoption Conditions, and Decision-Making Outcomes
- Phollawat Chantaramongkol
- Thongpan Pariwat
- Phanuwat Ruangkulsap
- Jaturong Chitiyaphol
- Thada Jantakoon
- Wantipa Unarat
Abstract
Artificial intelligence-driven decision support systems (AI-DSS) are increasingly used to inform the decisions of learners, teachers, and administrators, yet evidence on how these systems are applied, adopted, and evaluated remains fragmented. This systematic review, reported in accordance with the PRISMA 2020 statement, synthesized research on AI-DSS in educational settings. Searches of PubMed and Google Scholar (614 records), supplemented by 92 records identified through other methods, yielded 706 records; after deduplication and a relevance-ranked screening cap, 500 records were screened, 56 full texts were assessed, and 54 studies were included in a narrative synthesis that graded each reported outcome by the type of evidence supporting it. Six application domains were identified, led by adaptive and personalized learning and by academic performance prediction and early warning. Machine learning was the predominant approach (89% of studies), whereas explainable AI was incorporated in fewer than one quarter of studies. Adoption depended on institutional support, technical infrastructure, data quality, user-centered design, and stakeholder trust; recurrent barriers included data privacy, model opacity, algorithmic bias, implementation cost, and resistance to change. Reported benefits included more efficient assessment, earlier identification of at-risk learners, and better-informed decisions; however, comparative evidence came mainly from health-professions education, and several headline effects derived from predictive scenarios, technical reports, or single-site evaluations. Subgroup-disaggregated evidence on algorithmic equity was scarce. The review proposes an integrated decision-centered framework linking technological features, institutional conditions, human factors, and calibrated trust to appropriate use, decision quality, and educational outcomes. Advancing the field will require explainable, ethically governed, human-centered systems evaluated through longitudinal, comparative, and equity-sensitive research.
- Full Text:
PDF
- DOI:10.5539/hes.v16n4p838
Index
Contact
- Sherry LinEditorial Assistant
- hes@ccsenet.org