Enhancing deep learning in blended learning environments by investigating the interplay of self-regulated learning, social presence, cognitive load, and learning outcomes among pre-service teachers

Blended learning has become central to higher education, yet its capacity to promote deep learning depends on how cognitive, self-regulatory, and social factors are designed across online and face-to-face components. This study examined the relationships among cognitive load, self-regulated learning (SRL), social presence, deep learning, and learning outcomes among 450 Indonesian pre-service teachers in a structured blended learning environment. Data from survey responses and course-performance indicators were analyzed using partial least squares structural equation modeling (PLS-SEM) to test direct and mediated relationships. The results showed that cognitive load positively predicted deep learning but did not directly predict learning outcomes, indicating that cognitive demands can support achievement when they stimulate meaningful processing rather than overload learners. SRL also positively predicted deep learning but had no direct effect on outcomes, suggesting that regulation strategies improve performance mainly when translated into higher-order engagement. Deep learning strongly predicted learning outcomes and mediated the effects of cognitive load and SRL on achievement. Social presence did not significantly predict deep learning and negatively predicted learning outcomes, implying that interaction may be insufficient or distracting when not pedagogically structured. These findings identify deep learning as the central mechanism linking cognitive, self-regulatory, and social factors to academic performance. The study recommends blended learning designs that optimize cognitive challenge, scaffold SRL, and structure social interaction around purposeful inquiry, feedback, and knowledge construction.