This study examines whether artificial intelligence (AI) use practices are associated with perceived learning improvement among higher education students in Côte d’Ivoire. It focuses on AI-use frequency and personal integration strategy while controlling for age, gender, gender study level and AI-related concerns.
The study uses survey data from 850 license, Master’s and doctoral students. Binary logit models estimate the probability of reporting better learning with AI. A complementary quadratic specification examines whether the association between AI-use frequency and perceived learning improvement exhibits diminishing returns.
AI-use frequency and personal AI integration strategy are positively and significantly associated with perceived learning improvement. The nonlinear analysis indicates diminishing additional benefits: predicted probability rises substantially from non-use to several uses per week but reaches a plateau between several uses per week and daily use. Other controls are not statistically significant.
The cross-sectional, self-reported and non-probability sample prevents causal interpretation and does not measure objective academic performance. Future research should use representative or stratified samples, longitudinal designs, objective learning outcomes and more detailed measures of AI-use strategies.
Universities should promote AI literacy, purposeful academic use, critical evaluation of AI-generated content and clear academic-integrity guidelines. Policies should emphasize the quality and purpose of AI use rather than simply encouraging more frequent use.
Responsible AI integration can support perceived learning while limiting risks related to excessive dependence, misinformation, unequal access, privacy and academic misconduct.
The article provides original individual-level evidence from Côte d’Ivoire, an under-researched African higher education context. It distinguishes between frequency and strategic integration of AI and identifies possible diminishing additional benefits at higher usage levels.
