A Meta-Analysis of Adaptive Learning Technologies and Their Effect Sizes on Mathematics Achievement: Moderating Roles of Implementation Fidelity and Socioeconomic Status

Abstract

Adaptive learning technologies have been increasingly adopted in mathematics education to provide personalized instruction and improve student achievement. However, inconsistent empirical evidence regarding their effectiveness and limited understanding of factors influencing success have created uncertainty for educational practitioners and policymakers. This meta-analysis systematically examined 89 peer-reviewed studies comprising approximately 45,000 students to determine the overall effectiveness of adaptive learning technologies and identify key moderating variables. Using random-effects meta-analytic models with Hedges' g as the effect size measure, results revealed a moderate overall effect size of 0.67 (95% CI: 0.58–0.76), indicating significant positive effects on mathematics achievement. Implementation fidelity emerged as a critical moderator, with high-fidelity implementations producing effect sizes of 0.89 compared to 0.38 for low-fidelity implementations. Socioeconomic status also significantly moderated effectiveness, with larger effects observed in lower-SES contexts (g = 0.82) compared to high-SES contexts (g = 0.54), suggesting adaptive technologies may promote educational equity. Secondary analyses revealed that intelligent tutoring systems produced larger effects than assessment-only systems, and effects increased substantially with longer intervention duration. These findings indicate that well-implemented adaptive learning technologies represent effective interventions for improving mathematics achievement, particularly for disadvantaged students, though success depends critically on implementation quality, instructional design, and attention to contextual factors.