Adaptive Learning Systems Powered by Artificial Intelligence for STEM Education Improvement
Linda Ifechukwu Onwubuya *
Department of Higher Education, School of Education, College of Liberal Arts and Human Sciences, Virginia Tech, USA.
Yetunde Omoyiwola Fawehinmi
Department of Educational Administration and Human Resources, Texas A&M University, College Station, Texas, USA.
David Sunday
Emory University, Atlanta, Georgia, USA.
Manfred Obinwanne Igwenagu
Computer Information Systems, Prairie View A&M University, USA.
Adebimpe Oluwaseun Adeniran
Tennessee State University, Nashville, Tennessee, USA.
Evans Asante Agyapong
Department of Communication, University of Colorado, Boulder, USA.
*Author to whom correspondence should be addressed.
Abstract
AI is evolving into a revolution within the education sector especially in the process of designing adaptive learning systems that customize learning using individual learner information. It is a systematic review of empirical and theoretical research evidence on the role of AI-based adaptive learning technology in increasing engagement, personalization, and knowledge retention in science, technology, engineering, and mathematics (STEM) education.
In line with the PRISMA guideline, the identification of the relevant studies was done in databases such as scopus, web of science, ERIC, IEEE xplore, and Google Scholar. Inclusion criterion centered on AI-based adaptive or intelligent tutoring systems that were implemented in K-12 and higher-education STEM settings and non-AI or non-STEM platforms were excluded. The selection criteria included twenty-six peer-reviewed articles published in 2019-2025.
Thematic synthesis identified how AI-based adaptivity in the form of machine learning, deep neural networks, intelligent tutoring architectures, and so on, has steadily enhanced learning performance, engagement, and retention, especially when based on constructivist and self-regulated learning models. Nevertheless, the review also discovered severe challenges such as algorithmic bias, the presence of data privacy threats, infrastructural constraints, and inequality of access in a low-resource environment. Comparative evidence indicates that there is a gap in research that is longitudinal and cross-platform in the measurement of the persistence and generalizability of learning gains.
In general, the findings indicate that although the application of AI-enabled adaptive learning systems has a high potential of providing personalized STEM education, their long-term implementation presupposes transparency, ethical governance, and equitable, fair-minded algorithm design.
Keywords: Adaptive learning, artificial intelligence, personalization, STEM education, student engagement