ISSN 0439-755X
CN 11-1911/B
主办：中国心理学会
　　　中国科学院心理研究所
出版：科学出版社

心理学报

• •    

大语言模型干预网络欺凌受害者心理韧性：以自尊自悯为靶点（压力、韧性与健康专刊投稿）

岸本鹏子, 郝熙鸣, 艾尼卡尔·艾斯卡尔, 夏雨飞, 白麒钰   

  1. 北京师范大学心理学部, 北京 中国
    南开大学社会学院, 天津 中国
    北京大学新媒体研究院, 北京 中国
  • 收稿日期:2025-06-11 修回日期:2026-02-11 接受日期:2026-02-12
  • 基金资助:
    国家自然科学基金(72304018); 青年人才托举工程(2023QNRC001)

Large language model-facilitated interventions for enhancing psychological resilience in cyberbullying victims: Targeting self-esteem and self-compassion

KISHIMOTO Tomoko, HAO Ximing, ASKAR Ankar, XIA Yufei, BAI Qiyu   

  1. , , China
  • Received:2025-06-11 Revised:2026-02-11 Accepted:2026-02-12

PDF (PC)

168

摘要： 本研究以自尊和自悯为靶点，探索大语言模型(LLM)在干预网络欺凌受害者心理韧性中的效果与作用路径。研究1(干预效果检验)采用随机对照设计，比较自尊-自悯对话、心理教育对话和心理教育阅读三组的干预效果。结果显示两组对话式LLM均显著提升了心理韧性。研究2(作用路径分析)进一步采用网络干预分析揭示干预作用路径，发现自尊-自悯对话干预除了可以直接提高心理韧性外，还可以靶向孤立感、正念和社交维度自尊，通过自尊自悯多维度的相互作用，促进干预效果的进一步提升。本研究首次聚焦自尊与自悯路径，基于LLM构建具备共情交互能力的对话干预平台，为网络欺凌受害者提供个性化心理支持，并为智能化心理干预工具的开发提供理论依据与实践路径。

关键词: 大语言模型, 网络欺凌, 心理韧性, 自尊, 自悯

Abstract: Enhancing the psychological resilience of individuals who have experienced cyberbullying not only contributes to improved mental health but also effectively mitigates the negative impact of victimization. Interventions targeting self-esteem and self-compassion may represent a feasible pathway for strengthening resilience. Online interventions offer the advantages of scalability and timeliness, while the integration of large language models (LLM) can further reduce costs and enhance intervention efficacy. In Study 1, 59 participants were randomly assigned to one of three groups: a self-esteem and self-compassion dialogue intervention group (Experimental Group 1), a psychoeducational dialogue intervention group (Experimental Group 2), and a psychoeducational reading control group. Results indicated that both LLM-facilitated dialogue interventions (Groups 1 and 2) significantly improved psychological resilience. Study 2 replicated the design with a new sample of 105 participants and employed network intervention analysis to explore the distinct mechanisms underlying the two LLM-facilitated interventions. Results revealed that the self-esteem and self-compassion dialogue intervention not only directly improved psychological resilience but also targeted feelings of isolation, mindfulness, and self-esteem in social dimensions. Through the multi-dimensional interaction of self-esteem and self-compassion, it promoted further improvement of the intervention effect. These studies provide preliminary evidence for the application of LLM in resilience-focused interventions for cyberbullying victims. The findings demonstrate that targeting self-esteem and specific components of self-compassion can effectively enhance psychological resilience and mental well-being. Moreover, the research identifies novel intervention targets, offering both theoretical insights and practical implications for future studies.

Key words: Large language models, cyberbullying, psychological resilience, self-esteem, self-compassion