Using multimodal learning analytics to understand the combined effects of dynamic self-regulated scaffoldings and learning from examples on computer programming: an empirical study

As computational thinking and programming skills become increasingly essential across STEM disciplines, secondary schools worldwide are integrating programming education into their curricula. However, many students encounter significant challenges when learning programming, due to the high cognitive demands of writing code and the need for effective self-regulation. Prior research has demonstrated the potential of worked and erroneous examples, as well as self-regulated learning (SRL) scaffolding, in supporting conceptual understanding and strategic learning. However, few studies have adopted adaptable SRL scaffoldings and further investigated the combined effects of multiple instructional strategies. To address this gap, this study conducted a quasi-experimental investigation to examine how different combinations of learning from examples and SRL scaffolding influence students’ programming performance, learning behaviors, and motivational experiences in a secondary education context. First, our results showed a significant interaction effect between SRL scaffolding and example type on knowledge acquisition. Specifically, without SRL scaffoldings, students learning from erroneous examples outperformed those using worked examples, whereas worked examples combined with SRL scaffoldings led to the higher knowledge acquisition than those without SRL scaffoldings. SRL scaffolding also produced a significant main effect in enhancing students’ SRL strategies. Second, behavioral analysis showed that students in C1 (worked examples with SRL scaffolding) and C4 (erroneous examples without SRL scaffolding) engaged in more iterative learning processes. Third, although no significant main or interaction effects were observed on students’ learning attitudes, thematic analysis indicated that students in C1 and C4 reported more active engagement than those in the other conditions. This study offers implications for instructors designing programming learning environments that incorporate SRL scaffoldings and learning from worked and erroneous examples, and for researchers seeking to enhance adaptable support through AI-driven techniques and multimodal data in programming and broader STEM education.