生成式智能搜索结果可信度研究  

The Credibility of the Results of Generative Intelligent Search

作　　者：吴丹[1,2] 孙国烨 WU Dan;SUN Guoye

机构地区：[1]武汉大学信息管理学院,湖北武汉430072 [2]武汉大学人机交互与用户行为研究中心,湖北武汉430072

出　　处：《中国图书馆学报》2023年第6期51-67,共17页Journal of Library Science in China

基　　金：国家自然科学基金创新群体项目“信息资源管理”(项目编号:71921002);湖北省自然科学基金创新群体项目“以人为本的人工智能创新应用”(项目编号:ZRQT2023000026)的研究成果。

摘　　要：以ChatGPT为代表的生成式人工智能取得突破性进展,使生成式智能搜索在运行机制和结果呈现形式上与传统搜索引擎的差异更加显著,也使得生成式智能搜索这一新型搜索工具的结果可信度面临新的挑战。本研究以SOR模型为基础,基于可信度研究的成果和想象可供性理论,从用户信息获取的视角出发,运用卡片分类法和叙事研究法,通过生成式智能搜索典型代表New Bing开展实验,旨在厘清生成式智能搜索结果可信度从刺激、机体到反应的个体心理链路,探究其形成机理与影响因素。以生成式智能搜索结果界面的信息线索为外部刺激,以用户的内在感知和信任意愿为机体,将导致的用户情绪及行为作为反应,最终形成了生成式智能搜索可信度的SOR模型。对生成式智能搜索可信度的形成机理进行分析后认为,系统的研发者可以面向用户,根据使用体验催生的搜索需求,从内容整合性、问答交互性、工具拟人性三个维度出发,优化生成式智能搜索系统。Along with the breakthrough of generative AI represented by ChatGPT,generative intelligent search is more significantly different from traditional search engines in operating mechanism and result presentation form,which makes the credibility of the results of this new search tool face new challenges.Since users'trust and willingness to accept generative intelligent search is particularly important,based on SOR model,combined with the results of credibility research and imagined affordances theory,this study focuses on the users,and makes up for the research gap on the credibility of generative intelligent search results from the perspective of individual information acquisition.This study adopts card sorting and narrative research methods,and conducts experiments through New Bing,a typical representative of generative intelligent search,aiming to clarify the individual psychological link of the credibility of generative intelligent search results from stimulus,organism to response,and explore its formation mechanism and influencing factors.First of all,this study summarizes three characteristic dimensions of generative intelligent search,which are content integration,question-and-answer interactivity,and tool imitation humanity.Secondly,this study takes the information clues in the interface of generative intelligent search results as external stimuli,takes the users'internal perception and trust willingness in three characteristic dimensions as the organism,and takes the users'emotions and behaviors caused by credibility changes as responses,to form the SOR analytical framework of generative intelligent search credibility.Subsequently,this study divided the elements of the search result interface of New Bing into seven types of information cue stimulus labels.Participants were asked to classify the cards corresponding to the clues into different trust perceptions,and answered which type of characteristic perception was stimulated by these clues,as well as whether they trusted New Bing or traditional search engin

关 键 词：生成式人工智能 信息检索 可信度 Microsoft Bing 

分 类 号：G254.9[文化科学—图书馆学]