Online Archive of University of Virginia Scholarship
Responsible and Equitable Use of AI Code Generators in Computer Science Education157 views
Author
Gardella, Nicholas, Systems Engineering - School of Engineering and Applied Science, University of Virginia0009-0000-6912-7462
Advisors
Riggs, Sara, EN-SIE, University of Virginia
Abstract
Generative Artificial Intelligence (AI) is highly effective for software development tasks. GitHub Copilot and other such AI-driven Development Environments (AIDEs) are widely available for students and professional programmers alike. Institutions need to react swiftly to this technology by implementing evidence-based updates to policies and curricula in computer science (CS) education. While research in this area is on the rise, there is limited empirical work that systematically compares individual, AI-assisted, and human-human pair programming paradigms with respect to both well-being and performance over time. Furthermore, despite many attempts to model human interactions with AIDEs, there are several underexplored topics. These include reconciling the numerous and disparate qualitative interaction models, quantifying the relevance of trust to interactions with AIDEs, and exploring implications with respect to educational equity and political economics. This dissertation provides foundational understanding to foster inclusive student success by (1) evaluating and explaining novice-AIDE interactions in terms of performance, well-being, subjective experience, and trust, (2) situating novices' experiences with AIDEs against the societal backdrops of racial exclusion and capitalist economics in education, (3) comparing AI assistance to a social pair-programming alternative, and (4) synthesizing the findings of the prior three steps into a list of evidence-informed recommendations for educational use of code-generating AI.
Degree
PHD (Doctor of Philosophy)
Keywords
Computer Science Education; Artificial Intelligence-driven Development Environment (AIDE); Introductory Programming; CS1; CS2; AI Code Generators; GitHub Copilot; Novice Programmers; Generative AI; Artificial Intelligence; Trust in Automation; Compliance and Reliance; Software; Human-Automation Interaction; Qualitative; Interview; Diversity; Equity; Capitalism; Critical Political Economy; Dehumanization; Alienation; Theory of Rationalization; Pair Programming; Emotion; Affective Computing; AI-Assisted Programming
Gardella, Nicholas. Responsible and Equitable Use of AI Code Generators in Computer Science Education. University of Virginia, Systems Engineering - School of Engineering and Applied Science, PHD (Doctor of Philosophy), 2026-04-22, https://doi.org/10.18130/dean-t677.
