AI-Powered Adaptive Blended Learning and Self-Regulated Learning: A Quasi-Experimental Study

Abstract

ABSTRACT
The proliferation of learning management systems has normalized blended instruction, yet most implementations still deliver uniform pacing regardless of individual learner readiness. This study examines whether an AI-powered adaptive blended learning (AI-ABL) model, one that dynamically adjusts content sequencing and pacing based on real-time performance data, enhances students' self-regulated learning and knowledge retention relative to standard blended learning. A quasi-experimental nonequivalent control group design was employed with 130 undergraduate students (experimental n = 66; control n = 64) across an eight-week intervention. Data were collected through the Motivated Strategies for Learning Questionnaire (MSLQ), platform interaction logs, and delayed knowledge-retention tests administered immediately, at four weeks, and at eight weeks post-instruction. Analysis combined ANCOVA, paired-samples t-tests, and thematic coding of reflective journals. Results showed the experimental group achieved significantly greater SRL gains (17.4-point MSLQ increase versus 5.7 points, p < .001) and retained substantially more content at the eight-week follow-up (74.3% versus 60.1%). Engagement logs revealed that adaptive pacing and AI feedback prompts accounted for over half of student time-on-task. These findings indicate that algorithmic differentiation, when designed as an advisory rather than fully prescriptive mechanism, cultivates metacognitive regulation rather than replacing it. The study contributes an empirically grounded framework for balancing automation with learner agency in adaptive blended design.