Abstract
<p>The increasing adoption of self-directed programming practice environments creates new opportunities to examine how learners regulate their use of instructional supports and how these decisions influence learning outcomes. In this study, we analyze fine-grained log data from 115 students enrolled in an introductory programming course to investigate learners’ preferences and transitions between scaffolded tasks (e.g., tracing and code completion problems), code writing tasks, and worked examples. Using problem-level regression analyses, we investigate how prior knowledge, recent performance, and use of examples relate to learners’ regulatory decisions and subsequent performance outcomes. We found that higher prior knowledge was associated with higher success rates and productive struggle. Additionally, students with higher pretest scores were also slightly more likely to transition between task types. In general, switching behavior was symmetric: students were equally likely to move from code writing to scaffolded tasks as from scaffolded tasks to coding. However, failure on the prior problem significantly increased the likelihood of switching, and failure-triggered switching was positively associated with success, suggesting that why learners switch matters more than whether they switch task types. Our results highlight the importance of incorporating both learner characteristics and recent performance signals into self-regulated sequencing decisions. Productive practice may depend not only on the completed problems, but also on how learners adapt in response to failure, providing suggestions for programming environments that support strategic sequencing.</p>