A psychological dynamics model of AI-supported learning: examining the mediating and moderating effects on preservice early childhood teachers’ instructional design ability

Introduction: Generative AI is reshaping higher education by acting not only as a learning tool but also as a psychological catalyst for self-regulation and creative expression. Drawing on Self-Regulated Learning Theory, Social Cognitive Theory, and Self-Determination Theory, this study examines how AI feedback quality and AI scaffolding influence the instructional design ability of preservice early childhood education students through the mediating roles of self-regulated learning (SRL) and creative self-efficacy (CSE), as well as the moderating role of professional identity in early childhood education (PIEC). Methods: A cross-sectional online survey was conducted with 439 undergraduate students majoring in early childhood education at universities in mainland China. All constructs were measured with validated 7-point Likert scales. The hypothesized model was tested using Partial Least Squares Structural Equation Modeling (PLS-SEM) in SmartPLS 4.0, with 5,000 bootstrap resamples to assess direct, indirect (mediation), and interaction (moderation) effects. Results: Both AI feedback quality and AI scaffolding significantly enhanced SRL and CSE, which in turn improved instructional design ability. SRL and CSE each exhibited significant mediating effects, with SRL demonstrating the stronger indirect influence. Professional identity displayed a dual moderating pattern: it strengthened the effects of AI feedback on SRL and CSE but weakened the effects of AI scaffolding. Discussion: These findings reposition AI as a psychological facilitator that triggers strategic and creative processes rather than merely a technical tool. The study contributes a dual-process model of AI-supported learning and a professional congruence moderation framework, offering implications for teacher preparation, AI tool selection, and instructional design training. Limitations and directions for future research are discussed.