“It Makes Me More Efficient": Student Perceptions and Experiences with LLMs in Computer Programming Courses in Mexican Higher Institutions

Abstract

The rapid integration of Large Language Models (LLMs) into computer science education necessitates understanding student perceptions in diverse contexts like Mexican higher education. This study investigates the experiences of 311 computer science students from public and private Mexican universities. A hybrid methodology was employed, first validating the survey instrument through Confirmatory Factor Analysis and reliability tests. Statistical methods were then combined with machine learning in a comprehensive analytical framework. Results reveal a clear narrative: students overwhelmingly perceive LLMs as transformative, significantly enhancing programming efficiency. This core perception is deeply embedded within an interconnected network of positive attitudes and behaviors. Strong correlations between perceived efficiency, positive attitudes, future use intentions, and increased usage frequency create a virtuous adoption cycle. The study concludes that LLMs are firmly rooted in Mexico's programming education, urging a pedagogical shift from "if" to "how" they are used, thereby informing ethical curriculum design and AI literacy initiatives.