Progression of students' SRL processes in subsequent programming problem-solving tasks and its association with tasks outcomes

Learning programming is difficult, and many students fail or have poor outcomes. To learn to program means to master steps in the complex problem-solving activity. Previous research uncovered a rich set of domain-specific and generic cognitive and metacognitive strategies students use when they learn to program. The processes that problem-solving experts demonstrate are very similar to those studied by self-regulated learning researchers. This study proposes Self-Regulated Learning (SRL) process types derived from the SRL phases indicators developed from log data captured from students' interaction with the instructional scaffold for programming assignments in LMS. The process types were defined from theoretical and pragmatic perspectives, with the aim to indicate concrete interventions for improving problemsolving skills. We have observed and quantified students' use of the SRL processes of distinct types in the series of five problem-solving assignments. We have also observed the progression of SRL processes used by each student in the assignments. Our modelling showed that students with domain knowledge at the same level achieve higher assignment marks when they demonstrate SRL processes at the higher level; importantly, weak students benefit the most. • We have determined SRL processes on five consecutive programming assignments using a validated classifier of SRL phases • We have defined four SRL process types motivated by theory and interventions used to improve students' SRL process • Students' SRL process types between assignments fluctuate within a narrow range and provide an opportunity for interventions • Our models show the highest positive effect of more advanced SRL process types for students with the lowest programming proficiency