A Scoping Review of Vibe Coding and AI-Assisted Programming in University Computing Education
Abstract
Background: Generative AI coding tools are increasingly used in university computing education, including emerging practices commonly described as vibe coding. However, evidence on their educational effects is scattered across studies that differ in interaction patterns, instructional designs, assessment practices, and learning contexts. Methods: We conducted a scoping review using a PRISMA-guided process and synthesized 37 studies published between 2022 and July 2026. Five research questions examined student–AI interaction patterns, learning outcomes, task-level instructional strategies, assessment and academic integrity, and curriculum alignment with emerging professional practice. Results: The reviewed articles reported that students commonly relied on debugging-oriented prompts, common-case testing, and short edit-run cycles, while stronger performance was associated with more exploratory and evaluative interactions. Reported outcomes included conceptual learning, programming productivity, and affective benefits. Emerging instructional strategies included planning with hints, adaptive scaffolding, dual-mode assignments, and process visibility. Conclusions: The educational value of AI-assisted coding appears to depend less on tool access than on how student interaction is scaffolded, assessed, and connected to emerging competencies in specification, verification, testing, and critical evaluation. Therefore, computing education should treat AI-assisted programming not simply as accelerated code generation, but as a practice requiring deliberate reasoning and responsible oversight.
