Academic Stress, Self-Efficacy, and Generative AI Dependency: A Cross-Sectional Study at a Saudi University
Related Experiment Videos
Ayoob Lone1, Imtiyaz Ahmad Dar2, Arooj Arshad3
1Department of Clinical Neurosciences, College of Medicine, King Faisal University, Al Ahsa, 31982, Saudi Arabia.
Background:
Generative artificial intelligence (GenAI) offers academic benefits but raises concerns about student dependency. Limited research has examined the psychological factors underlying GenAI dependency among university students.
Aim:
This study examined how academic self-efficacy and academic stress relate to GenAI dependency, testing performance expectancy as a mediator and academic self-efficacy as a moderator of the stress-GenAI dependency relationship.
Methods:
This cross-sectional study surveyed 305 undergraduate students recruited through convenience sampling from a public university in the Eastern Province of Saudi Arabia between March and May 2026. Participants completed an Arabic measures of GenAI dependency, academic self-efficacy, academic stress, and performance expectancy. Direct, mediating, and moderating pathways were tested with partial least squares structural equation modeling (PLS-SEM), using 5,000 bootstrap resamples and a supplementary robustness check controlling for demographic covariates.
Results:
Academic self-efficacy was negatively associated with GenAI dependency (β = -0.245, p <0.001) and positively associated with performance expectancy (β = 0.237, p <0.001). Academic stress was positively associated with both performance expectancy (β = 0.315, p <0.001) and GenAI dependency (β = 0.287, p <0.001). Performance expectancy was positively associated with GenAI dependency (β = 0.411, p <0.001) and showed statistically significant indirect associations of academic self-efficacy (β = 0.098, p <0.001) and academic stress (β = 0.130, p <0.001) with GenAI dependency. Academic self-efficacy significantly moderated the relationship between academic stress and GenAI dependency (β = 0.102, p =0.010, 95% CI (0.024, 0.179); contrary to the hypothesized buffering effect, this positive association became stronger at higher levels of self-efficacy. The model explained 36.3% of GenAI dependency and 13.8% of performance expectancy variance.
Conclusion:
GenAI dependency was associated with academic stress and academic self-efficacy, while performance expectancy showing significant indirect associations. Self-efficacy reduced GenAI dependency directly but amplified the stress-dependency relationship. These results support encouraging balanced GenAI use, managing academic stress, and calibrating students' expectations of GenAI.
