登录 EN

添加临时用户

面向青少年心理韧性的智能交互测量与培育系统研究

Research on the Intelligent Interactive Measurement and Cultivation System for Adolescent Mental Resilience

作者:胡子卉
  • 学号
    2021******
  • 学位
    硕士
  • 电子邮箱
    hu-******.cn
  • 答辩日期
    2024.05.08
  • 导师
    倪士光
  • 学科名
    电子信息
  • 页码
    125
  • 保密级别
    公开
  • 培养单位
    599 国际研究生院
  • 中文关键词
    青少年;心理韧性;大语言模型;对话代理;积极心理干预
  • 英文关键词
    Teenagers; Mental Resilience; Large Language Model; Conversational Agents; Positive Psychological Intervention

摘要

心理韧性(Psychological Resilience)有助于青少年成功应对逆境、维持心理健康,提升心理韧性对于预防心理疾病和促进潜能发展具有重要意义。传统心理韧性培育方法存在成本高、耗时长等局限,目前仍缺乏大规模、数智化的高效培育方法或工具。因此,本研究结合了积极心理学理论与大语言模型(LLM)等人工智能技术,面向青少年群体积极发展,使用人机交互研究方法,探索了智能心理韧性测量与培育系统的设计实现。研究的主要内容和结果如下:在理论层面,通过系统性综述方法梳理了国内外青少年数字心理韧性交互应用,从心理韧性干预效果和技术接受模型两个角度进行分析;通过质性研究归纳了青少年心理韧性成长的需求与障碍,提出了以AI对话代理为核心的心理韧性培育新机制,总结了面向青少年心理成长的对话代理的六条设计原则。在方法层面,实现了传统心理韧性培育与测量的智能化方法,设计了情感支持对话和积极心理干预两个面向培育的对话任务模型;并将传统量表改编为情景式问答测量,构建了基于LLM的智能心理测量新范式。在应用层面,构建了包含420组情感支持多轮对话的数据集,并通过提示工程研究、构建数据集微调等方法进行LLM模型优化;设计了用户友好的系统原型,将LLM对话代理部署在虚拟卡通形象中,实现完整可用的网页系统。最后,招募了48名青少年受试参与定性与定量用户实验,对系统干预有效性、测量有效性、系统整体的可接受性进行了评估。结果显示,单次干预任务将青少年心理韧性测量得分提升了16.13%,具有显著促进作用;智能测量与传统量表的得分结果一致,且主观评分的五个维度上均显著优于传统量表,在吸引力维度提升了93.5%,趣味性和启发性提升超过65%,系统SUS可用性得分超过82.9分。本研究结合了积极心理学前沿的心理韧性理论与干预方法、自然语言处理技术和人机交互方法,探索了青少年数字心理韧性成长需求,提出了基于智能对话的心理韧性测量与培育的新方法,开发了一套网页系统原型,通过用户实验初步验证了系统的效果,为未来进一步研究与推广应用提供了基础。

Mental resilience can assist teenagers in coping with difficulties, maintaining mental health, and promoting future development. Enhancing psychological resilience plays a crucial role in preventing psychological diseases. Traditional resilience cultivation methods are costly and time-consuming, and yet lack efficient cultivation methods or tools that cater to a larger body of teenagers. Therefore, this research merges psychological theory with computer technology like large language models (LLM) and utilizes human-computer interaction research methods to explore the design and implementation of an intelligent system for measuring and cultivating psychological resilience for teenagers. The main content and results of this study are as follows:On a theoretical level, this study organizes teenagers’ digital psychological resilience interactive applications through a systematic method. The material is analyzed from the perspectives of the effect of psychological resilience intervention and the technology acceptance model. It consolidates the demands and obstacles for teenage psychological resilience growth through qualitative research. The study proposes a new mechanism for nurturing psychological resilience centered around AI conversational agents (CAs), and it outlines six design principles for CAs aimed at the psychological growth of teenagers.On a methodological level, this research has implemented an intelligent approach to traditional resilience cultivation and measurement. It designs emotional support dialogue and positive psychological intervention as dialogue task models oriented towards cultivation. Moreover, it adapts traditional psychometric measures into scenario-based questions and answers, establishing a novel intelligent psychological measurement format based on LLM. On an application level, it constructs a dataset containing 420 groups of emotion-support multi-round dialogues and optimizes the LLM model through prompt engineering and fine-tuning. A user-friendly system prototype has been designed. It deploys LLM dialogue agents in virtual cartoon images and implements a fully functional web system. Through user tests, the study assesses the effectiveness of system intervention, the validity of measurements, and overall system acceptability with 48 children volunteers. Results demonstrate that the intervention tasks promote teenagers‘ psychological resilience positively. The intelligent measurement scores highly align with traditional scale scores, significantly surpassing traditional scales across five dimensions of subjective ratings. The system‘s overall usability received high evaluations.This study integrates psychological theories, natural language processing techniques, and human-computer interaction methods to explore the digital psychological resilience growth of teenagers. It proposes a new method for psychological resilience measurement and nurturing based on intelligent dialogues and develops a web system prototype. The preliminary verification of the system‘s effectiveness via user tests provides a reference for further optimization in the future.