Evaluation of Prevention and Treatment Paths of Digital Empowerment of Teenagers' Internet Addiction

Introduction

In recent years, with the continuous increase of internet penetration rate, the online activity time of teenagers has been prolonged, and it has become normal to deeply integrate into digital life (Paat & Markham, 2021; Zilka, 2020). However, this trend is accompanied by the rapid spread of social media, short video platforms and online games, and the problem of Internet addiction is increasingly characterized by universality and younger age (Al-Samarraie et al., 2022; Geisel et al., 2021). Many studies show that excessive use of electronic devices not only seriously interferes with the normal sleep rhythm of teenagers, leading to distraction and decline in academic performance, but also significantly increases the risk of emotional disorders such as anxiety and depression (Al Salman et al., 2020; Li et al., 2022; Neophytou et al., 2021).

Although educational institutions, families, and communities have gradually established an intervention system that focuses on publicity and education, time management guidance, and psychological counseling services, these traditional methods generally have the problems of lag and static, and they lack real-time tracking and dynamic response mechanism for the process of behavior change. In addition, it is difficult to identify addiction behaviors and intervene during their embryonic stage; as a result, most intervention measures still remain at the level of post-event remedy, and the overall prevention and control effect is limited. More forward-looking and accurate solutions are urgently needed (Chadha et al., 2024; Lin, 2020; Wang et al., 2025). In order to break through the bottleneck of the existing intervention mode, the development of information technology provides a new path. By integrating physiological data collected by wearable devices, behavior logs of mobile terminals, and multimodal big data analysis technology, researchers can build high-precision portraits of individual behaviors and realize continuous monitoring and risk early warning of network usage patterns (Cho et al., 2024; Yao et al., 2023). The adaptive intervention system based on an artificial intelligence algorithm can push personalized tips or conduct situational guidance at key nodes, thus effectively blocking the vicious circle before the behavior gets out of control (Ajayi, 2025; Shahzad et al., 2024). However, at present, most studies are still limited to the design and verification of a single functional module, such as only focusing on screen time control or online cognitive training, lacking a systematic framework that integrates risk perception, intelligent evaluation, dynamic intervention, and effect feedback; even less research has carried out long-term and multi-dimensional empirical tests in a real campus environment.

In view of this research gap, this paper adopts a random cluster quasi-experimental design, selects 15 classes of students from two middle schools as the research objects and divides them into intervention group and control group according to grades, constructs a closed-loop intervention path of perception–evaluation–intervention–reevaluation, realizes the dynamic prediction of addiction risk by relying on the two-way long-term and short-term memory (LSTM) network model, combines differential privacy technology to ensure data security, and comprehensively evaluates the actual effect of technology-driven intervention through 10-week longitudinal tracking (Khan et al. 2025; Schurz et al., 2021). This study focuses on three core issues and puts forward three prior hypotheses before data analysis: H1 is that technology-driven closed-loop intervention can significantly reduce the severity of internet addiction (CIUS-11) and the average daily screen use time of teenagers, H2 is that the intervention effect has significant individual and subgroup heterogeneity and shows differences between grades and urban and rural groups, H3 shows that the two-way LSTM-attention model can still maintain good prediction stability and robustness under the conditions of missing data and concept drift in real scenes. The standardized index system and data governance framework finally formed can provide preliminary practical reference for similar schools in urban-rural fringe, but its large-scale popularization needs further verification by multi-site research.