The rise of conversational agents: can they keep up with humans? Evidence from a meta-analysis
作者
Sepideh Ebrahimi,Samira Farivar,Kamran Eshghi
出处
期刊: [Figshare (United Kingdom)] 日期:2026-04-24
标识
DOI:10.6084/m9.figshare.32088171
摘要
The proliferation of conversational agents (CAs) has sparked a significant body of research on the consequences of users’ interactions with them. While many studies document positive user responses, some also advise against switching from human agents to automated CAs. To address this debate, this meta-analysis employs the Tripartite Response Model and combines findings of 72 studies to compare users’ cognitive, affective, and behavioural responses to interactions with CAs versus human agents. The results, surprisingly, indicate a tie between the two types of agents except for cognitive responses, which tend to favour CAs over humans. CAs outperform humans even further when the cognitive response is of a descriptive type – users’ understanding of objects and events – as opposed to their subjective evaluation. Our findings also indicate that the performance of CAs exceeds that of humans when it comes to influencing behaviours that are temporally distanced from the time of interaction, giving CAs an advantage in influencing users’ long-term behaviours. By theoretically combining 345 effect sizes to study various user responses across different contexts, such as healthcare and commerce, this study offers deeper insights into the circumstances in which switching from human agents to CAs can benefit individuals as well as firms.