A systematic mapping review at the intersection of artificial intelligence and self-regulated learning

Synthesis: Banihashem, Bond, Bergdahl, Khosravi, and Noroozi systematically map the intersection of AI and self-regulated learning (SRL), reviewing 84 studies via a systematic mapping review across Web of Science, Scopus, IEEE Xplore, ACM Digital, EBSCOHost, Google Scholar, and OpenAlex. Using the "Who–What–How–Why" framework (stakeholders, theory, methods, objectives), they find AI–SRL research predominantly targets higher education students with minimal attention to primary education and educators. AI is implemented primarily as an intervention — through adaptive systems and personalization, prediction and profiling, intelligent tutoring systems, and assessment and evaluation. The direct impact of AI on SRL focuses mainly on the metacognitive and cognitive aspects, while the motivational aspect of SRL remains underexplored, and over one-third of studies did not specify an SRL theory.

Key Findings

• 84 studies mapped across six years. The review covers AI–SRL research growth over roughly 2019–2025, examining study levels, participant demographics, and geographical distribution.
• Higher-ed dominance, primary-ed gap. AI–SRL research predominantly focuses on higher education students; there is minimal attention to primary education and to educators as stakeholders.
• AI as intervention, in four main forms. AI is primarily implemented as an intervention to support students' SRL and learning processes: (1) adaptive systems and personalization, (2) prediction and profiling, (3) intelligent tutoring systems, and (4) assessment and evaluation.
• Metacognitive and cognitive focus, motivational gap. The direct impact of AI on SRL is concentrated on the metacognitive and cognitive aspects, while the motivational aspect of SRL remains underexplored — a key research gap.
• Theory often unspecified. Over one-third of the AI–SRL studies did not specify an SRL theory, pointing to a need for stronger theoretical grounding.
• A holistic research-and-practice agenda. The review identifies both research and practical gaps in the AI–SRL nexus to guide future work.

What this means for practice

• Learning designers. Design AI support for the motivational dimension — Self-Efficacy, engagement — alongside the cognitive and metacognitive AI Regulation in Education the literature already targets, since motivation is the aspect the mapping found most underexplored.
• Instructional designers. Name the SRL theory a tool is built on and state which phases it supports; over one-third of the mapped studies specified no SRL theory, which leaves effects hard to interpret or reuse.
• Learning designers. Plan AI-SRL support across a term or longer rather than a single module and treat short-term gains as provisional: only about one-third of the mapped studies extended beyond two months.
• Instructional designers. Use AI profiling and prediction to identify at-risk students for timely intervention, and adopt the implementation forms the record supports best — Adaptive Learning and personalization, Feedback, and assessment and evaluation.
• Researchers. Recruit educators and primary-level learners as participants; the mapping shows both are nearly absent from AI-SRL research, and the authors ask for individual- and group-level effects to be reported with effect sizes.

Limitations

• The 84 mapped studies were found with a search string limited to permutations of "self-regulation" and requiring the term in the title or abstract, and the authors state the protocol was not pre-registered because of the project's timeline; work using other terms for goal-setting, self-monitoring, or mastery learning may have been excluded.
• The search strategy carries a Global North and linguistic bias that the authors name: while authors came from 30 countries, 48% were from North America, and other academic platforms were not searched.
• No formal quality appraisal of included studies was performed, as systematic mapping reviews do not require one; the authors recommend a full appraisal in any future systematic review.
• The mapped evidence is mostly short-term — only about one-third of studies extended beyond two months — and 71% quantitative (N = 60), so long-term effects and the qualitative experience of learners are thinly evidenced.

Citation

Banihashem, S. K., Bond, M., Bergdahl, N., Khosravi, H., & Noroozi, O. (2025). A systematic mapping review at the intersection of artificial intelligence and self-regulated learning. International Journal of Educational Technology in Higher Education, 22, 50.