BRIDGING THE AGENCY GAP IN DIGITAL TEACHER EDUCATION: A STRUCTURAL EQUATION MODELLING AND LEARNING ANALYTICS INVESTIGATION OF PRE-SERVICE TEACHERS' SELF-REGULATED LEARNING TRAJECTORIES.

While digital learning environments have become ubiquitous in higher education, their capacity to foster genuine learner agency and self-regulated learning (SRL) among pre-service teachers remains a contentious empirical question. In many developing academic infrastructures, university learning management systems inadvertently reinforce an 'agency gap'—operating as passive file-distribution repositories that perpetuate reproductive, surface-level compliance. This study investigated how an integrated digital architecture combining multi-tiered instructional scaffolding, structured peer assessment, and real-time learning analytics shapes pre-service teachers' self-regulatory trajectories. Employing a comprehensive two-step structural equation modelling (SEM) approach integrated with process-oriented behavioural analytics, data were collected from 384 undergraduate pre-service teachers engaged in a 16-week educational technology course utilising the specialised MustaqilTa'lim.uz platform (https://publibruary.uz). Confirmatory factor analysis verified a multidimensional measurement model across five latent constructs: Perceived Scaffolding Affordance, Metacognitive Monitoring, Peer Feedback Engagement, Digital Resource Orchestration, and Technological Pedagogical Self-Efficacy. Structural model evaluation revealed that Perceived Scaffolding Affordance exercised a profound direct effect on Metacognitive Monitoring (beta = .52, p < .001) and an indirect effect on Technological Pedagogical Self-Efficacy (beta = .31, p < .001). Peer Feedback Engagement directly catalysed iterative revision depth (beta = .48, p < .001), serving as a crucial metacognitive trigger. Furthermore, a two-step cluster analysis of granular platform telemetry classified participants into three distinct behavioural profiles: Proactive Iterators (38.5%), Strategic Compliers (43.2%), and Reactive Minimisers (18.3%), which exhibited statistically significant disparities in self-regulation and academic mastery (F = 74.12, p < .001). These empirical findings demonstrate that digital technology fosters teacher autonomy only when deliberately structured around dialogic peer review, iterative task refinement, and actionable behavioural feedback. Theoretical contributions and practical design heuristics for teacher education programs are discussed.