图书情报知识 ›› 2023, Vol. 40 ›› Issue (2): 10-19.doi: 10.13366/j.dik.2023.02.010

• 专题·人智交互体验度量与智能产品设计 • 上一篇    下一篇

人智交互体验度量:受控实验的应用及启示

姜婷婷 1,2 ,田慧溢 2 ,许艳闰 1 ,傅诗婷 1   

  1. 1.武汉大学信息管理学院,武汉,430072;
    2.武汉大学信息资源研究中心,武汉,430072
  • 出版日期:2023-03-10 发布日期:2023-05-09
  • 通讯作者: 姜婷婷(ORCID:0000-0002-5310-2073),博士,教授,研究方向:人智交互、信息行为、用户体验度量、人本数据分析,Email:tij@whu.edu.cn。
  • 作者简介:田慧溢(ORCID:0000-0002-7527-7691),博士研究生,研究方向:人智交互、信息行为,Email:tian_huiyi@foxmail.com;许艳闰(ORCID:0000-0003-0641-0329),博士研究生,研究方向:人智交互、信息行为,Email:yarunx@whu.edu.cn;傅诗婷(ORCID:0000-0002-5636-7931),博士研究生,研究方向:人智交互、信息行为,Email:fushiting@whu.edu.cn。
  • 基金资助:
    本文系国家社会科学基金重大项目“人本人工智能驱动的信息服务体系重构与应用研究”(22&ZD325)和国家自然科学基金面上项目“信息茧房的形成机理与联觉干预研究”(72074173)的研究成果之一。

Measuring Human-AI Interaction Experience: Implications for Conducting Controlled Experiments

JIANG Tingting, TIAN Huiyi, XU Yanrun, FU Shiting   

  • Online:2023-03-10 Published:2023-05-09
  • Contact: Correspondence should be addressed to JIANG Tingting, Email:tij@whu.edu.cn, ORCID: 0000-0002-5310-2073
  • Supported by:
    This is an outcome of the Major Project "Information Service System Restructuring and Application Driven by Human-Centered Artificial Intelligence"(22&ZD325)supported by National Social Science Foundation of China and the project "The Formation Mechanism and Synesthesia-Based Intervention of Information Cocoons"(72074173)supported by National Natural Science Foundation of China.

摘要: [ 目的 / 意义 ] 人智交互是人机交互领域的前沿课题,人智交互体验研究关注人类使用人工智能系统完成特定任务的过程和结果并探讨各要素在其中的作用机制。揭示受控实验这一重要方法在国外已有人智交互体验研究中的应用情况,将为未来人智交互体验实验研究的开展提供有力指引,从而推动人本人工智能的发展与普及。[ 研究设计 / 方法 ] 采用系统性综述方法,收集了发表于2007年至2022年间的114篇国外人智交互体验实验研究文献,构建了包括实验样本、实验任务、实验材料、实验观测维度、数据分析方法这五个一级类目的编码表,基于该编码表对文献的实验设计部分开展内容分析。[ 结论 / 发现 ] 以往人智交互体验实验研究表现出实验样本抽样方式单一、实验材料可复用性低、客观的实验观测手段使用不足、实验任务前瞻性强等特点。[ 创新 / 价值 ] 聚焦于人智交互体验这一新兴研究方向,首次在研究方法层面进行了全面梳理与分析,贡献了一套完整的实验设计基本框架,有助于提升人智交互体验研究的规范性和科学性。

关键词: 人智交互体验, 研究方法, 受控实验

Abstract: [Purpose/Significance] Human-AI interaction(HAII)has become a frontier research theme in the field of human-computer interaction. The research of Human-AI interaction focuses on the processes and outcomes of human using artificial intelligence to accomplish specific tasks and explores the function and mechanisms of each element in the process. This study aims to reveal the application of controlled experiments in existing foreign HAII research and to acquire implication, which would provide guidance for future experimental studies and accelerate the development and popularization of human-centered AI. [Design/Methodology] This study conducts a systematic review on 114 experimental research articles on Human-AI interaction experience, which were published between 2007 and 2022. It proposes a five first-level categories coding table including experimental samples, experimental tasks, experimental materials, experimental observation dimensions and data analysis methods, based on which the content analysis on the experimental design part of the literature has been conducted. [Finding/Conclusion] Previous researches have shown the characteristics such as single sampling method, low reusability of experimental materials, insufficient use of objective experimental observation tools, and prospective experimental tasks. [Originality/Value] The research focus on the emerging research direction of human-AI interaction experience, is the first to comprehensively review and analyze the research methodology and contribute a complete set of basic framework for experimental design, which helps to improve the standardization and scientificity of human-AI interaction experience research.

Keywords: Human-AI interaction experience, Research method, Controlled experiment