Fostering self-regulated learning through adaptive learning technology: A differentiated perspective on the role of feedback

Synthesis: Mejeh and Fromm investigate how differentiated feedback types — directive, informative, and transformative — delivered through an adaptive learning technology (ALT) and taken up by learners relate to self-regulated learning (SRL) across its pre-actional, actional, and post-actional phases. In a study of 194 students in a stochastics course using an ALT over eight weeks, analyzed with self-report + trace data via hierarchical linear modeling, they find that task value, Self-Efficacy, goal orientation, and positive emotions in the pre-actional phase positively influence regulatory behavior in the actional phase, while negative emotions hinder it. However, adaptive feedback's moderating effects were limited and sometimes negative — transformative feedback may contribute to cognitive overload — underscoring that feedback effectiveness is context-dependent and must be tailored to learners' needs.

Key Findings

• Pre-actional factors drive actional AI Regulation in Education. Task value, self-efficacy, goal orientation, and positive emotions in the pre-actional phase positively influenced regulatory behavior in the actional phase; negative emotions hindered it.
• Post-actional reflection feeds forward. Post-actional satisfaction was positively associated with motivation, metacognitive activity, and positive emotions in the subsequent pre-actional phase — evidence of the cyclical nature of SRL.
• Adaptive feedback's effects are limited and context-dependent. Feedback's moderating effects were limited and, in some cases, negatively affected regulatory outcomes; transformative feedback may contribute to cognitive overload.
• Differentiated feedback types matter. Directive, informative, and transformative feedback delivered through ALT are taken up differently by learners, with variable effects on SRL components across phases.
• Design implication. While ALT feedback can promote SRL, its effectiveness requires tailoring to learners' needs; future work should optimize feedback design and address individual differences in SRL development.

What this means for practice

• Designers. Tailor adaptive feedback to learners' regulatory needs and to the phase they are in instead of maximizing feedback density: directive, informative, and transformative feedback were taken up differently, and transformative feedback was associated with cognitive overload rather than better regulation.
• Designers. Instrument an ALT to record uptake rather than delivery: the study logged only whether a learner encountered each feedback affordance at least once in a week, not how often, so frequency and dosage went unmeasured.
• Instructors. Prime the pre-actional phase before learners start work — task value, Self-Efficacy, goal orientation, and positive emotions predicted actional regulation, while negative emotions such as anxiety hindered it — so address motivation and affect before content.
• Instructors. Build end-of-task reflection into the design: post-actional satisfaction predicted motivation, metacognitive activity, and positive emotions in the following week's pre-actional phase, which is the cyclical mechanism the study documents.
• Instructors. Combine self-report with trace data when studying Self-Regulated Learning in a technology-mediated course, since the study paired weekly questionnaire responses with log data on dashboard use and neither source alone would have shown how feedback was engaged.

Limitations

• The sample was 194 students in a single stochastics course over eight weeks; many observed effects were only marginally significant, and the authors state the sample size, while acceptable, may have limited statistical power for detecting subtle interaction effects.
• Feedback was operationalized as weekly dummy-coded indicators of whether a learner encountered each variant at least once in a week, not how often, so feedback frequency and dosage are not measured.
• Adaptivity was primarily rule-based and threshold-triggered, engagement with feedback resources was voluntary, and the dashboard served an informational function; regulatory behaviors were assessed only after task completion rather than in real time.
• Self-report data were used to prompt the directive feedback, which the authors note introduces potential bias from inaccuracies in learners' self-assessments.

Citation

Mejeh, M., & Fromm, Y. M. (2026). Fostering self-regulated learning through adaptive learning technology: A differentiated perspective on the role of feedback. Learning and Instruction, 105, 102394.