Racial mirroring effects on human-agent interaction in psychotherapeutic conversations

Do You Have Time for a Quick Chat? Designing a Conversational Interface for Sexual Harassment Prevention Training

“…Building on to our research, the interaction between the gender of the protagonist and of the user could be an interesting direction for future research. Other demographic factors such as race should be further tested in the design as Liao and He found that racial mirroring influence human-agent interaction [38].…”

Hey There! What Do You Look Like? User Voice Switching and Interface Mirroring in Voice‐Enabled Digital Assistants (VDAs)

“…Inclusivity in terms of race/ethnicity in conversational agents is another factor users desire in digital assistants. Focusing on chatbots, Liao and He (2020) involved 212 participants who interacted with a chatbot that was designed to appear as White/Caucasian, Black/African, American, Latinx, and Asian (male and female). They found that a user's engagement and perception of conversational agents was highly influenced by the perceived race and ethnicity of the agent.…”

“…However, despite the prevalence of VDA use in daily life, scarce research has investigated user voice switching behavior in general or preferences for voice interfaces that are embodied in images representative of people of diverse backgrounds beyond race/ethnicity. This study begins to address this gap by extending Liao and He's (2020) “racial mirroring” to “interface mirroring,” which in addition to race/ethnicity, explores a gamut of user preferences for voice interfaces embodied in diverse images within the context of voice interface switching behavior. Such embodiment could contribute to creating blueprints for designing and developing future voice interfaces that are diverse and more tangible, capable of meeting the needs of a wide range of users.…”

“…Additionally, we are yet to understand user voice switching behavior (VSB) in VDAs and whether the availability of voices embodied in certain identities and characteristics may provide reasons for users to switch the voice interface. Liao and He (2020) found that in using conversational agents, users prefer that the agents match their own race and ethnicity. They characterized this user preference as “racial mirroring,” “a match between the user and agent's race/ethnicity” (p. 431).…”

“…The findings from this study have implications for humanizing VDAs through inclusive design that supports voice options extending beyond gender, language, and accent, contributing to fairness and equality in designing this AI technology. As Liao and He (2020) note, technology should be designed to appeal to different genders, races, and ethnicities in order to successfully meet the needs of diverse members of society. Other implications are that the voice interface in VDAs should support images representative of a wide range of diverse people as embodied voice interfaces that provide more concrete than abstract voice interactions.…”