Human–AI Co-Regulation in Adaptive Learning: Developing GPT-Supported Self-Regulated Learning Models

Source raw items (1)

• Semantic Scholar2026-07-30 07:16:12
  Human–AI Co-Regulation in Adaptive Learning: Developing GPT-Supported Self-Regulated Learning Models

  Background The rapid integration of generative artificial intelligence into higher education has created new opportunities for supporting adaptive learning and self-regulated learning. However, existing adaptive learning systems primarily emphasize automated personalization and feedback, with limited attention to how learners and artificial intelligence collaboratively regulate learning processes. This study aimed to develop and evaluate a GPT-supported Human–AI Co-Regulation model to enhance self-regulated learning, metacognitive reflection, and adaptive engagement in higher education. Methods A Design-Based Research approach integrated with mixed methods and learning analytics was employed. The quantitative phase involved 214 undergraduate students, while 24 participants were included in the qualitative phase through interviews, reflective journals, classroom observations, and analysis of artificial intelligence interactions. Quantitative data were analyzed using descriptive statistics, paired-sample t-tests, structural relationship analysis, and learning analytics visualization. Qualitative data were analyzed using thematic analysis. Results The findings demonstrated significant improvements across all dimensions of self-regulated learning following implementation of the GPT-supported Human–AI Co-Regulation model. Metacognitive regulation showed the largest improvement (Δ = 0.98, p < 0.001), followed by reflective thinking and self-monitoring. Learning analytics revealed substantial increases in GPT interaction frequency (105.2%), adaptive pathway utilization (51.9%), reflective prompt responses (48.3%), and task completion rates (27.2%) across iterative implementation cycles. Structural relationship analysis indicated that metacognitive reflection (β = 0.45, p < 0.001) and Human–AI interaction quality (β = 0.42, p < 0.001) were strong predictors of self-regulated learning outcomes. Qualitative findings showed that learners increasingly perceived GPT as a collaborative cognitive partner that supported planning, monitoring, reflection, and adaptive learning decisions. Conclusions The study extends self-regulated learning and adaptive learning theories by introducing a Human–AI collaborative regulation perspective. The findings highlight the value of integrating generative artificial intelligence, learning analytics, reflective prompting, and adaptive feedback mechanisms to support sustainable, learner-centered, and adaptive higher education environments.