Understanding compulsive generative AI use among college students through a dual system perspective: An integrated SEM-ANN-NCA approach

The powerful capabilities of generative artificial intelligence (GAI) have propelled its adoption in higher education in recent years. GAI offers considerable benefits, but its rapid integration has prompted growing concerns regarding compulsive GAI use, characterised by persistent and uncontrollable engagement. Compulsive GAI use has been linked to adverse psychological and academic outcomes, underscoring the importance of understanding the factors associated with this maladaptive use. However, most existing studies have focused on GAI's functionalities and how its applications support various domains (e.g., education), as well as factors associated with the adoption and continued use of GAI. In contrast, much less attention has been given to identifying the antecedents of compulsive GAI use. Drawing on Dual-System Theory (DST), this study developed a model to examine whether GAI features (intelligence, interactivity, and personalisation), perceived usefulness, impulsive urge, habit, and impulsivity are associated with compulsive GAI use. Data were collected from 593 college students from a Chinese university and analysed using an approach integrating Structural Equation Modelling (SEM), Artificial Neural Network (ANN), and Necessary Condition Analysis (NCA). PLS-SEM results showed that perceived interactivity and personalization were positively associated with perceived usefulness, which was further associated with impulsive urge and habit. Impulsive urge and habit were also associated with compulsive GAI use. ANN results revealed that habit had the highest normalised importance, followed by impulsivity and impulsive urge. NCA identified perceived interactivity, perceived usefulness, habit, and impulsive urge as necessary conditions for compulsive GAI use to emerge. This study provided significant practical insights for developing interventions to reduce compulsive GAI use among college students.