|
|
|
| The Impact of Anthropomorphism in Artificial Intelligence Agents on Human-Agent Trust and Potential Moderating Effects: A Meta-Analysis |
| ZHOU Lulu，ZHAO Yinshan，ZHAO Shuming |
| 1.Southeast University,Nanjing, China;2. Nanjing University, Nanjing, China |
|
|
|
|
Abstract Based on the three-factor model of human-agent trust, a meta-analysis of 62 valid studies with a total of 18,878 samples was conducted to thoroughly investigate the relationship between four anthropomorphic features of artificial agents—appearance, movement, personality, and voice—and human-agent trust, along with their boundary conditions. The results demonstrated a moderately positive correlation between anthropomorphic features and human-agent trust. Movement, personality, and voice anthropomorphism also showed moderate positive correlations with human-agent trust. Compared to physical robots, studies using non-physical agents found that overall anthropomorphism as well as appearance, movement, and personality anthropomorphism had more significant effects on human-agent trust. In non-organizational work settings, overall anthropomorphism along with appearance and personality anthropomorphism had more significant effects on human-agent trust, while in organizational settings, voice anthropomorphism had a more significant effect on human-agent trust. When comparing relationship types, the servant-type human-agent relationships showed more significant effects of overall anthropomorphism as well as appearance and personality anthropomorphism on human-agent trust compared to the collaborative relationships. Finally, in research designs using remote interaction compared to on-site interaction, overall anthropomorphism along with appearance, movement, and voice anthropomorphism exhibited more significant effects on human-agent trust.
|
|
Received: 28 October 2024
|
|
|
|
|
|
|
