Exploring the role of errors in programming learning

Computer science students often perceive programming errors as indicators of failure rather than learning opportunities, resulting in frustration, anxiety, and dropout. This study investigates how early programming errors can contribute to students’ subsequent learning. We hypothesise that the number of errors made by students in the initial learning stage, which is distinct across different learning concepts, is positively correlated with their future learning. Our hypothesis is tested using a real-world dataset consisting of 601 programming questions that were practised by 632 students, with results validated using the ITAP dataset. Through statistical analysis controlling for pre-practice understanding, we identify error-based intervention indicators—specific error thresholds that correlate with maximum improvement rates for different programming concepts. These findings provide evidence-based perspectives on the pedagogical role of errors in computer science education, suggesting concept-specific and proficiency-based instructional approaches for programming education.