Mining student coding behaviors in a programming MOOC: there are no actionable learner stereotypes

Abstract

Education often involves categorizing students into broad groups based on perceived attributes like academic abilities, learning pace, and unique challenges. However, the validity and applicability of these stereotypes require closer examination. This research investigates student grouping factors, exploring both conventional variables like gender and education level, as well as innovative methodologies that utilize students’ problem-solving behaviors. The study critically evaluates the effectiveness of these grouping techniques in capturing and distinguishing students’ diverse learning patterns. Through a comprehensive analysis of ten methodologies used to cluster students in traditional programming courses and programming MOOCs, we aim to reveal how students from different cohorts exhibit varying learning behaviors and outcomes. By examining diverse models of student learning, we assess whether students in distinct groups indeed demonstrate discernible disparities in their educational journeys. Our meticulous data analysis uncovers compelling insights that challenge the notion of predefined student stereotypes and their practical utility within group-based adaptation settings. This research contributes to the discourse on student grouping by highlighting the limitations of traditional categorizations and introducing innovative approaches to understanding student diversity and tailoring educational interventions accordingly. By transcending simplistic generalizations, we strive to foster a nuanced understanding of students’ individual strengths, challenges, and potentials, promoting inclusive and effective educational practices.