图书情报知识 ›› 2026, Vol. 43 ›› Issue (1): 11-24.doi: 10.13366/j.dik.2026.01.011

• 学术聚焦(1)·AI 与用户信息行为 • 上一篇    下一篇

人智交互中用户的提问行为与信息使用行为研究

李雨佳, 祝嘉淇, 刘畅   

  1. 北京大学信息管理系,北京,100871
  • 出版日期:2026-01-10 发布日期:2026-03-24
  • 通讯作者: 刘畅(ORCID: 0000-0002-9183-6385),博士,长聘副教授,研究方向:信息行为分析、交互式信息检索,Email: imliuc@pku.edu.cn。
  • 作者简介:李雨佳(ORCID: 0000-0002-8959-5420),硕士研究生,研究方向:交互式信息检索,Email: liyujia76@163.com;祝嘉淇(ORCID: 0009-0007-5497-3106),硕士研究生,研究方向:交互式信息检索,Email: 2301211755@pku.edu.cn。
  • 基金资助:
    本文系国家社会科学基金重点项目“智能时代提升全民数字素养的理论和实践研究”(24AZD030)的研究成果之一。

Users' Prompting and Information Use Behaviors in Human-AI Interaction

LI Yujia, ZHU Jiaqi, LIU Chang   

  1. Department of Information Management, Peking University, Beijing, 100871
  • Online:2026-01-10 Published:2026-03-24
  • Contact: Correspondence should be addressed to LIU Chang, Email: imliuc@pku.edu.cn, ORCID: 0000-0002-9183-6385
  • Supported by:
    This is an outcome of the Key Project "Research on Theoretical and Practical Approaches to Fostering Digital Literacy in the Intelligent Era"(24AZD030)supported by National Social Science Foundation of China.

摘要: [目的/意义]以ChatGPT为例,聚焦用户的提问行为与信息使用行为,探究用户与生成式人工智能交互行为的具体特征,以促进生成式人工智能产品的优化改进。[研究设计/方法]首先通过问卷收集3个场景和6个认知层次下的ChatGPT真实任务,然后采用用户实验法,结合定性的主题分析法与定量的统计检验对问答数据进行分析。[结论/发现]归纳了用户在提问行为和信息使用行为方面的特征,提问行为方面,用户在完成高认知层次任务时的提问技巧更加丰富;信息使用行为方面,用户以复制为主,创新较少;而新手用户与熟练用户的提问行为和信息使用行为也存在差异。[创新/价值]在理论层面丰富了生成式人工智能使用场景下的用户行为研究,在实践层面为生成式人工智能提供改进建议。

关键词: 生成式人工智能, ChatGPT, 提问行为, 信息使用行为, 人智交互

Abstract: [Purpose/Significance] Taking ChatGPT as an example, this study focuses on users' prompting behaviors and information use behaviors to explore the specific characteristics of users' interaction with generative artificial intelligence, with the aim of promoting the optimization and improvement of generative artificial intelligence products. [Design/Methodology] In this study, the relative data about real tasks of ChatGPT in three scenarios and six cognitive levels were firstly collected through questionnaires. Then, adopting a user experiment method, combined qualitative topic analysis and quantitative statistical tests, we conducted the analysis of these Q&A data. [Findings/Conclusion] This study summarizes the behavioral characteristics of users' prompting and information use behaviors. In terms of prompting behaviors, users' prompting skills are more diverse when engaging in tasks that require higher cognitive processing. Regarding information use behaviors, users mainly rely on copying information with less innovation. Moreover, there are differences in prompting and information use behaviors between novice users and experienced users. [Originality/Value] Theoretically, this paper enriches the research on user behaviors in the actual use scenarios of generative artificial intelligence. Practically, it proposes optimization suggestions for generative AI to provide users with higher-quality services.

Keywords: Generative artificial intelligence, ChatGPT, Prompting behaviors, Information use behaviors, Human-AI interaction