Gender Bender Bot - The Effect of (Not) Following Gender Stereotypes in Conversational Agent Design

Abstract

Conversational agents (CAs), are increasingly becoming a common presence in our daily lives (e.g., Alexa or ChatGPT). Research has shown that designing CAs humanlike (e.g., through social cues such as a human name or avatar) results in a higher perception of humanness, which increases and service satisfaction by the user. In this context, CAs are exclusively designed to portray stereotypical genders (e.g., combining a female name and avatar). To challenge this quasi-standard, a 2x2 experiment (male/ female avatar x male/ female name) with 262 participants was conducted to investigate the effect of gender-mixed CAs. Our results indicate that users of CAs with a stereotypical gender report higher service satisfaction and a partially higher perception of social presence for male CAs. However, the results do not reveal any differences in perceived empathy and competence. Thus, it appears that users prefer stereotypically CAs, which is in sync with current practice.