Educational anxiety and student mental health in the era of artificial intelligence: a multi-source data fusion analysis in smart education

Introduction: Against the backdrop of the rapid advancement of educational digitalization and the growing prominence of student mental health issues, this study develops an integrated model of "technology perception-educational anxiety-student mental health" to examine the mechanisms through which perceived usefulness, perceived fairness, and perceived controllability are associated with student mental health. Methods: A cross-sectional questionnaire design was adopted, and data were collected from university students through a combination of online and offline surveys from July to September 2025, yielding 380 valid responses. Based on these quantitative data, structural equation modeling and artificial neural network techniques were employed to test the proposed hypotheses. Results: The results indicate that perceived usefulness, perceived fairness, and perceived controllability are all significantly and positively associated with student mental health, whereas educational anxiety is significantly and negatively associated with student mental health. All three dimensions of technology perception significantly reduce educational anxiety and indirectly promote student mental health through this mediating pathway, with perceived fairness and perceived controllability showing relatively stronger associations. Sensitivity analysis using the artificial neural network further reveals that perceived fairness is the most important predictor of educational anxiety, followed by perceived controllability and perceived usefulness. Discussion: This study enriches the theoretical understanding of student mental health in AI-enabled educational contexts and provides empirical implications for educational technology optimization, algorithm governance, and student mental health interventions.