Human–AI Collaboration in Higher Education: Effects of Generative AI Support on Academic Performance, Critical Thinking, and Learning Autonomy

Abstract

Generative artificial intelligence is increasingly reshaping higher education by supporting academic tasks, information processing, and learner decision-making, yet the factors associated with students’ educational use of GenAI remain insufficiently understood. This study was conducted with a cross sectional secondary quantitative approach that relied on a survey of 663 students at participating Saudi universities. The analysis developed and validated a structural model comprising of four factors: Institutional Support, Risk/Integrity concern, Educational Value/Acceptance, and Educational GenAI Use. The analytical method used was parallel analysis, exploratory factor analysis, independent confirmatory factor analysis, reliability and validity test, structural equation modeling, 1000-replication bootstrap inference, and sensitivity analyses. The factor analysis model had very good psychometric properties (4 factors x 22 items) and the final structural model had good fit (CFI = .987, TLI = .985, RMSEA = .031, SRMR = .062). The strongest correlate of Educational GenAI Use was Educational Value/Acceptance (β = .688, p < .001) with 48.4% of the variance accounted for by the model. Risk/Integrity Concern was negatively associated with Educational Value/Acceptance (β = −.145, p = .002) and showed a supported negative indirect association with Educational GenAI Use (β = −.100, 95% CI [−.185, −.018]). Institutional Support showed smaller and less stable relationships. Overall, students’ perceptions of the educational value and legitimacy of GenAI appear central to its academic use, underscoring the importance of responsible and well-guided Human–AI integration in higher education.