Generative AI in Vietnamese higher education: Pedagogical innovation, learning outcomes, and academic integrity

Abstract

The rapid penetration of Generative AI is reshaping the global higher education ecosystem; however, its integration in developing countries like Vietnam faces significant institutional and cultural constraints. This study utilizes an explanatory sequential mixed-methods design to examine how AI influences pedagogical innovation and student learning outcomes. During the quantitative phase, survey data from 970 participants (students, lecturers, and administrators) were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). The qualitative phase integrated insights from 30 semi-structured interviews. PLS-SEM results indicate that perceived usefulness strongly promotes teaching integration intentions (β = 0.525) and learning outcomes (β = 0.315). Furthermore, pedagogical integration directly enhances academic performance (β = 0.468). However, academic integrity risks pose a substantial barrier to pedagogical innovation (β = -0.215). Qualitative findings reveal a critical paradox: while AI optimizes short-term efficiency, it threatens critical thinking and fosters algorithmic dependency. Crucially, institutional responsibility emerges as a decisive factor in mitigating these integrity risks (β = -0.385). The study concludes that the principal obstacle to AI integration is the absence of comprehensive digital ethics frameworks, not technological limitations. Consequently, universities must shift toward transparent governance policies and systematically restructure assessment practices to ensure responsible AI integration.