Detrimental effects of anthropomorphism on the perceived physical safety of artificial agents in dangerous situations

Abstract

Designers of artificial agents often give them humanlike features, reflecting assumptions that humanlike agents evoke more positive evaluations than machinelike agents do. However, through four studies, the current article reveals a detrimental effect of anthropomorphizing embodied artificial agents. This effect occurs because these agents appear physically less safe in dangerous situations, which leads to consumers’ diminished self-safety perceptions and less favorable downstream consequences, both attitudinal (e.g., quality and trust perceptions, consumer evaluations, willingness to pay) and behavioral (e.g., information search, donation behavior). However, this detrimental effect is mitigated in non-dangerous situations or for artificial agents that usually do not operate in dangerous situations. The findings also reveal some theoretically important and practically relevant moderators. Specifically, when consumers receive marketing messages that direct their attention to artificial agents’ humanlike minds (e.g., cognitive and socio-emotional capabilities) rather than their humanlike bodies, the negative effect of anthropomorphizing artificial agents disappears. In addition to advancing emerging research on embodied artificial agents, this study provides practical guidance for marketers who plan to integrate artificial agents with humanlike features into their operations.