Scholarlink

<p>Internet addiction (IA) has emerged as a growing mental health concern, characterized by impaired self-control and maladaptive decision-making. The aim of this study is to investigate how reinforcement learning abilities and specific personality traits influence the decision-making patterns of individuals with IA. Gaining insights into these factors will deepen our understanding of the mechanisms underlying IA and support the development of personalized diagnosis and treatment strategies. A total of 61 participants with IA and 61 healthy participants completed the Internet Addiction Test (IAT), the Behavioral Inhibition System/Behavioral Activation System (BIS/BAS) scale, and the two-step Markov decision task. Through hybrid modelling and multivariate regression, the relationships among the reinforcement learning parameters, personality traits, and IAT scores were analysed. Results showed that individuals with IA had significantly lower beta model-based (MB) values, indicating reduced goal-directed control, and higher learning rates than healthy controls. Correlation analysis revealed that IAT score was positively associated with both BIS (r = 0.327, <i>p</i> &lt; 0.001) and learning rate (r = 0.239, <i>p</i> = 0.008), while betaMB showed no significant correlation. Multiple linear regression identified learning rate (β = 6.391, <i>p</i> = 0.022), BIS score (β = 1.176, <i>p</i> = 0.009), and craving (β = 0.652, <i>p</i> &lt; 0.001) as significant predictors of IA severity, explaining 46.9% of the variance. Additionally, support vector machine (SVM) analysis using these key features and betaMB achieved 83.3% classification accuracy. These findings highlight the role of reinforcement learning and personality traits in IA, offering valuable insights for targeted interventions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Internet addiction from a computational model: the predictive value of reinforcement learning parameters and behavioral inhibition systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月11日

查看全文
  • Jin Chen,
  • Ze-Yuan Xie,
  • Tian-Zhen Chen,
  • Jian-Xia Lu,
  • Ning-Ning Zeng,
  • Hui Zheng

摘要

Internet addiction (IA) has emerged as a growing mental health concern, characterized by impaired self-control and maladaptive decision-making. The aim of this study is to investigate how reinforcement learning abilities and specific personality traits influence the decision-making patterns of individuals with IA. Gaining insights into these factors will deepen our understanding of the mechanisms underlying IA and support the development of personalized diagnosis and treatment strategies. A total of 61 participants with IA and 61 healthy participants completed the Internet Addiction Test (IAT), the Behavioral Inhibition System/Behavioral Activation System (BIS/BAS) scale, and the two-step Markov decision task. Through hybrid modelling and multivariate regression, the relationships among the reinforcement learning parameters, personality traits, and IAT scores were analysed. Results showed that individuals with IA had significantly lower beta model-based (MB) values, indicating reduced goal-directed control, and higher learning rates than healthy controls. Correlation analysis revealed that IAT score was positively associated with both BIS (r = 0.327, p < 0.001) and learning rate (r = 0.239, p = 0.008), while betaMB showed no significant correlation. Multiple linear regression identified learning rate (β = 6.391, p = 0.022), BIS score (β = 1.176, p = 0.009), and craving (β = 0.652, p < 0.001) as significant predictors of IA severity, explaining 46.9% of the variance. Additionally, support vector machine (SVM) analysis using these key features and betaMB achieved 83.3% classification accuracy. These findings highlight the role of reinforcement learning and personality traits in IA, offering valuable insights for targeted interventions.