The impact of vibe coding on students’ programming self-efficacy and attitudes toward artificial intelligence

Vibe Coding is a generative AI–based programming approach that may support novice learners who lack confidence in their programming abilities, particularly given the cognitive demands of traditional coding. This study examined the impact of a Vibe Coding intervention on students’ programming self-efficacy and attitudes toward artificial intelligence (AI). Employing a single-group pretest–posttest quasi-experimental design, 50 undergraduate students participated in a four-week intervention involving natural-language–mediated AI-assisted programming using GitHub Copilot. Data were collected using the Computer Programming Self-Efficacy Scale (CPSES) and the General Attitudes Toward Artificial Intelligence Scale (GAAIS). Results indicated a significant increase in programming self-efficacy following the intervention ( p < .001, d = 0.93). Subscale analyses showed significant improvements in both simple ( p < .001, d = 0.54) and complex programming tasks ( p < .001, d = 0.90). No significant change was observed in students’ attitudes toward AI; both positive ( d = 0.20) and negative ( d = 0.28) attitude subscales showed small effect sizes. These findings suggest that participation in a natural-language–based AI coding environment was associated with short-term improvements in programming self-efficacy within the four-week intervention period, while short-term changes in attitudes toward AI were limited. This study contributes to the growing field of generative AI–based learning by highlighting the potential pedagogical relevance of Vibe Coding for programming self-efficacy. The large effect size ( d = 0.93) indicates a substantial within-sample change in perceived competence.