Transformation of the University Classroom: Applications and Perspectives of Generative Artificial Intelligence in Teaching

Abstract

This study examined the integration of generative artificial intelligence into university teaching at a public Andean institution. The purpose was to explore innovative strategies that enhance engagement, autonomy, and the quality of academic outputs in response to the need to refresh conventional educational models. The main objective was to explain the shifts experienced by students and lecturers when language and visual generative tools were introduced as assistants for lesson planning, doubt resolution, feedback provision, and multimodal resource creation. A convergent mixed-methods, non-experimental design gathered platform logs, course grades, learning artefacts, reflective journals, and semi-structured interviews. Quantitative data were treated with parametric tests and mixed-effects models, whereas the qualitative strand followed a reflexive thematic approach to enable triangulation. The findings revealed sustained improvement in academic performance, extended time devoted to formative activities, and lower early withdrawal. Three overarching themes emerged: guided autonomy, expanded curiosity, and reconfiguration of feedback routines. Generative artificial intelligence thus acted as a catalyst for more dynamic pedagogy, fostered critical reflection, and streamlined course management, while raising ethical challenges related to originality and the cultural footprint of generated content. The investigation concluded that responsible prompt design, transparent data practices, and inclusive language options are pivotal to maximising benefits and limiting risks. Future research should adopt longitudinal perspectives, quantify environmental costs, and test multilingual models that incorporate indigenous languages to secure inclusion, sustainability, and cultural relevance. Moreover, institutional policies ought to embed digital critical literacy programs ensuring that both learners and faculty fully retain agency within algorithmic learning environments.