交互失误情境下交互式人工智能拟人化的负面影响——基于ChatGPT和搜索引擎的实验证据  

The Negative Impact of Interactive AI Anthropomorphization in Interaction Failure Contexts:Experimental Evidence Based on ChatGPT and Search Engines

作　　者：孟韬[1] 陈梦圆 张天锴 赵非非 关钰桥 Meng Tao

机构地区：[1]东北财经大学国际商学院,辽宁大连116025 [2]东北财经大学工商管理学院,辽宁大连116025 [3]辽宁大学亚澳商学院,辽宁沈阳110136

出　　处：《情报理论与实践》2024年第1期84-91,共8页Information Studies:Theory & Application

基　　金：国家自然科学基金面上项目“大众生产视角下的分享经济组织形成机制研究”(项目编号:72072026);辽宁省“兴辽英才计划”项目“数字经济时代企业高成长的机制与促进政策研究”(项目编号:XLYC1904017);辽宁省社科规划基金重点项目“技术赋能辽宁数字政府治理机制研究”(项目编号:L21AGL009)的成果。

摘　　要：[目的/意义]对人机交互失误情境下用户产生的情绪和态度展开研究,旨在验证人工智能拟人化理论的负面效应并探索交互式人工智能拟人化产生负面影响的缓解机制。[方法/过程]以人机交互失误为情境,基于期望一致性理论和拟人化理论,通过实验法分别从交互的结果失误和调用失误两个角度,探索了交互式人工智能不同的拟人化程度对用户产生负面态度的影响机制。[结果/结论]研究发现,人机交互失误时,高拟人化的交互式人工智能会加剧用户对其的厌恶感和负面态度。验证了算法可解释性能够缓解厌恶感和负面态度。综合上述分析,从人机交互健康长远发展角度,提出了人工智能应采用有限拟人化的设计框架和具有可解释性的算法架构等建议。[Purpose/significance]In order to study the emotions and attitudes generated by users when the interaction fails,to validate the negative effects of the theory of AI anthropomorphism,and to explore the mitigation mechanism of the negative effects generated by interactive AI anthropomorphism.[Method/process]Taking the human-computer interaction failure as a context,based on the expectation consistency theory and anthropomorphism theory,the mechanism of interactive AI with different degrees of anthropomorphism on the generation of negative attitudes of users is explored through the experimental method from the perspectives of the resultant failures of the interaction and the invocation failures.[Result/conclusion]It was found that highly anthropomorphic interactive AI exacerbates users’disgust and negative attitudes towards it when human-computer interaction fails.It is verified that algorithm explainability can alleviate disgust and negative attitude.Combining the above analyses,from the perspective of healthy long-term development of human-computer interaction,it is proposed that AI should adopt a design framework with limited anthropomorphism and an algorithmic architecture with explainability.

关 键 词：交互式人工智能 交互失误 拟人化 厌恶感 可解释性 

分 类 号：H08[语言文字—语言学] TP18[自动化与计算机技术—控制理论与控制工程]