1123 - THE LONGITUDINAL RELATIONSHIP BETWEEN AI DEPENDENCE AND DEPRESSIVE SYMPTOMS IN ADOLESCENTS: A NETWORK PERSPECTIVE

Session: P_D08S004 - Poster Session 4 - Division 8
AUTHORS:
Wang Huanlei (State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University ~ Beijing ~ China) , Huang Shunsen (Institute of Sociology, Chinese Academy of Social Sciences ~ Beijing ~ China) , Lai Xiangxiong (Institute of Future Education, China National Academy of Educational Sciences ~ Beijing ~ China) , Ke Li (State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University ~ Beijing ~ China) , Xiang Kexin (State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University ~ Beijing ~ China) , Zhang Yifan (State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University ~ Beijing ~ China) , Wang Yun (State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University ~ Beijing ~ China)
Abstract text:
Current findings on the relationship between AI dependence and depression are inconsistent. Existing studies some overemphasize the harms of AI dependence, while others highlight the potential value of AI as a tool for mental health interventions, suggesting that AI dependence does not impair mental health. Most researches rely on total depression scores, overlooking that different nodes of depressive symptoms may exhibit differential sensitivity to AI dependence. Using Cross-lagged Panel Network Models (CLPN), we aimed to reveal the relationship between AI dependence (as defined by five symptoms: salience, tolerance, functional impairment, withdrawal and loss of control, adapted from the Smartphone Addiction Scale) and the development of nine depressive symptoms as measured by the PHQ-9. 4,647 adolescents (Mage = 14.6 ± 1.6, 48.2%male) were assessed twice at 12-month intervals in 2023 and 2024. The network structure was estimated using a mixed modelling approach (glmnet package: LASSO regression for variable screening and lavaan package: structural equation modelling), and stability was assessed using Bootstrap sampling. The CLPN model showed moderate accuracy and stable estimates of node centrality. The results suggest that: (1) Salience has high mediation centrality and acts as a bridging symptom between AI dependence and depressive symptoms. (2) There is a vicious cycle between loss of control, anhedonia and sleep disturbances. Furthermore, sleep disturbances, concentration difficulties and suicide ideation are significant precursors to the development of AI dependency. (3) Sad mood and salience negatively predict each other, with salience negatively predicting appetite and withdrawal negatively predicting appetite. Results suggest that relying on AI to alleviate depression may be counterproductive as its effects are potentially harmful. Furthermore, preventing AI dependence should involve identifying and alleviating an individual's depressive emotions rather than simply restricting their use of AI. Future research could explore the underlying psychological mechanisms further, such as motivations for using AI.