理论探讨

从ELIZA到ChatGPT: 人智交互体验中的AI生成内容(AIGC)可信度评价

  • 宋士杰 ,
  • 赵宇翔 ,
  • 朱庆华
展开
  • 1 河海大学商学院 江苏 211000; 2 武汉大学信息管理学院 湖北 430072; 3 南京大学数据智能与交叉创新实验室 江苏 210023; 4 南京大学信息管理学院 江苏 210023)
宋士杰,男,1989年生,河海大学商学院副教授,武汉大学信息管理学院博士后。 赵宇翔,男,1983年生,南京大学信息管理学院教授,博士生导师(通讯作者)。 朱庆华,男,1963 年生,南京大学信息管理学院教授,博士生导师。

网络出版日期: 2023-07-12

基金资助

本文系国家自然科学基金青年项目“社交媒体失真健康信息的特征识别与协同纠偏机制研究”(批准号:72204076)、教育部人文社会科学研究 青年基金“社交媒体环境中失真健康信息的协同纠偏研究”(项目编号:22YJC870013)、中央高校基本科研业务费专项资金资助(项目编号: B220201054)的研究成果。

From ELIZA to ChatGPT : AI-Generated Content ( AIGC ) Credibility Evaluation in Human-Intelligent Interactive Experience

  • Song Shijie ,
  • Zhao Yuxiang ,
  • Zhu Qinghua
Expand
  • 1 Business School, Hohai University, Jiangsu, 211000; 2 School of Information Management, Wuhan University, Hubei,430072; 3 Laboratory for Data Intelligence and Cross-Innovation of Nanjing University, Jiangsu, 210023; 4 School of Information Management, Nanjing University, Jiangsu,210023)

Online published: 2023-07-12

摘要

[目的/意义]人工智能的发展不断形塑着人智交互体验。从早年的聊天机器人ELIZA到近期的AI应用 ChatGPT,由AI驱动的各类应用在自然语言理解与推理上的能力愈发强大。当AI生成内容(AI-Generated Con? tent,AIGC)逐渐逼近甚至超越人类创建的内容时,自然的人智交互体验为用户提供了便捷高效的信息服务,但也 引发了不容忽视的可信度问题。因此,亟须从社会技术视角重新审议AIGC的可信度评价。[方法/过程]文章采用 规范性分析,在整合可信度研究与AI研究的基础上形成规范性命题。首先,解析了AIGC可信度的概念内涵,继 而从AI作为信息源、交互方式、社会行动者、用户算法隐喻四方面建构人智交互体验中AIGC可信度评价的研究 框架。[结果/结论]首先,文章认为AIGC可信度概念应当涵盖AI的可解释性维度。继而,本文提出AIGC可信度 评价应当建立AI内容生成与技术媒介的整合性视角、重视AI自有示能性以及与交互界面内嵌所产生的多级示能 性叠加、融合AI作为社会行动者的拟人化视角、考虑用户算法素养并重点识别、关注低素养人群。

本文引用格式

宋士杰 , 赵宇翔 , 朱庆华 . 从ELIZA到ChatGPT: 人智交互体验中的AI生成内容(AIGC)可信度评价[J]. 情报资料工作, 2023 , 44(4) : 35 -42 . DOI: 10.12154/j.qbzlgz.2023.04.004

Abstract

[Purpose/significance] The development of AI has been constantly shaping the experience of human-AI in? teractions. From the early stage of chatbot ELIZA to the recent AI application ChatGPT, various AI-powered applica? tions demonstrate significant progresses in human language understanding and generating and reasoning. When AIGenerated Content (AIGC) becomes increasingly close to human generated content, the nature experience of human-AI interaction not only provides users with convenient information services but also invites challenges of credibility issues. Therefore, there is a pressing need to revisit the assessment of AI credibility based on socio-technical perspectives. [Method/process] The study employs a normative analysis approach, synthesizing the literature of credibility and AI research to achieve several normative statements about AIGC credibility assessment. First, the article discussed the conceptualization of AIGC credibility. Moreover, the article pinpointed AIGC credibility by examining AI functionatin? gas information sources,interactivity, social actors,and users’algorithmic metaphor. [Result/conclusion] First, the arti? cle proposes that the concept of AIGC credibility should incorporate the dimensions of AI explainability. Moreover, the article proposes that AIGC credibility assessment should compass the holistic view of AI functionating as content gener? ator and technological medium, integrate view of AI affordances and generative affordances from third-party interfaces, underline the anthropomorphism view of regarding AI as a social actor, and encompass the consideration of users’algo? rithmic literacy by identifying and taking care of the low-literacy users.
文章导航

/