More AI, Less Learning? The Role of Technological Self-Efficacy in STEM Education

Crisostomo R.A.; Sanchez C.; Nunez A.-M.; Nunez N.A. · ICACIT 2025 - Proceedings: 11th International Symposium on Accreditation of Engineering and Computing Education (2025) · DOI: 10.1109/icacit66767.2025.11369511

Resumen

The increasing adoption of generative AI tools, such as ChatGPT and Copilot, is transforming how STEM students approach their studies. Yet, the key question remains: Does the use of AI enhance academic performance? This research analyzed survey data from over 250 undergraduate STEM students at a private university in Lima, Peru, to investigate whether there is a correlation between AI usage and self-reported academic achievement. Using ordinal logistic regression, we explored three models: (1) AI use alone, (2) AI combined with self-perceived technological Self-Efficacy, and (3) a moderation model considering prior AI experience. The findings suggest that frequent use of AI does not lead to improved academic results; in some cases, it even correlates with lower performance. In contrast, students with higher technological Self-Efficacy generally performed better, regardless of how often they used AI. These results underscore the importance of higher education in going beyond mere access to AI, emphasizing the development of digital skills for effective tool use. The study discusses the implications for STEM education, highlighting the value of competence-based approaches to integrating digital technology.

Instituciones: Pontificia Universidad Católica del Perú