Computing Students' Help-Seeking Approaches and Behavior Throughout the Curriculum: 2021--2025

Abstract

Background.Academic help-seeking is an effective self-regulatory and metacognitive learning strategy that benefits postsecondary computing students’ learning. Help resource selection, where students decide what help resource(s) to interact with, is a key stage in their help-seeking processes. The rapid evolution of the help ecosystems in modern computing classrooms has necessitated re-examination of both students’ help-seeking behavior (what they do) and approaches (what they think they do or claim they do). Existing theoretical literature has an overfocus on formal and social help resources (Chapter 2), while empirical works in computing education often study one help resource at a time and overfocuses on introductory, programming-heavy course contexts (Chapter 3). The interaction between different help resources and the applicability of existing findings to upper-level, theoretical, or interdisciplinary course contexts are less studied.Objectives.This dissertation aims to quantitatively characterize computing students’ help-seeking approaches and behavior throughout the curriculum, with an emphasis on students’ sequential transition behavior among multiple help resources. We seek to study students’ help-seeking approaches and behavior not only collectively but also individually, then investigate the relationships between the identified individual differences and various student characteristics including demographics, prior relevant experience, and learning-relevant psychoconstructs. We also document the temporal trends in this ever-evolving subject.Methodology.On the theoretical side, we synthesize a set of eight known important factors on students’ help resource selection into the Help Resource Landscape, a framework that characterizes help resources and explains students’ sequential help resource usage. On the empirical side, we collect and quantitatively analyze students’ approach and behavioral data from eight courses (47 total offerings, 6,115 total enrollments) across two research-oriented institutions over a four-year timespan (Chapter 4). We design and deploy three instruments to measure students’ perceptions of the Help Resource Landscape and quantify their approach in utilizing help resources sequentially. We triangulate this approach dataset with students’ behavioral logs in their use of course-affiliated discussion forums, office hours, and program autograder systems. We employ descriptive statistics, nonparametric statistical inference techniques, frequent rule mining algorithms, and various types of regression models to answer suitable research questions. We make conscious methodological choices to emphasize context effects in our analyses (Chapter 5).Findings.At a high level, this dissertation reports: four tiers of important factors in students’ resource selection decision process (Chapter 6); a progression of help resource clusters where resources in the same clusters are similar in their characterization by the Help Resource Landscape framework and see similar sequential utilization (Chapter 7); an overall high agreement between students’ self-reported approaches and behavior in utilizing course-affiliated help resources sequentially (Chapter 8); substantial individual differences in students’ help-seeking approaches and behavior, with some seemingly context-agnostic and some apparently context-dependent (Chapter 9); many nuanced relationships between the individual differences and student characteristics such as gender, prior relevant experience, and confidence (Chapter 10), some of which again subject to context influences; and finally, a non-universal decreasing trend in students’ utilization of course-affiliated help resources amid the age of generative artificial intelligence (Chapter 11).Implications and Contributions.This dissertation is an attempt to holistically understand computing students’ help-seeking approaches and behavior throughout the curriculum beyond what has been revealed in single-resource or single-context studies. Based on these insights, we design and describe an intervention for boosting first-year computing students’ help-seeking efficacy and help-seeking approach richness (Chapter 13). In addition to informing help ecosystem design, resource allocation, and teaching staff training, this dissertation provides a set of mindsets to rethink students’ help-seeking (Chapter 14). Finally, our methodological choices throughout this dissertation may be instrumental for future context-aware quantitative educational research.