A Kansei Engineering Approach to Virtual Personality of Embodied Voice Assistants

As speech recognition and artificial intelligence technologies mature, virtual assistants (VAs) capable of interacting with information services through spoken language alone have become mainstream voice user interfaces (VUIs). Platforms supporting VA functionalities, such as mobile phones, smart speakers, and service robots, exhibit diverse trends, with the degree of VA embodiment emerging as a design consideration for VUIs. Among the myriad applications of VAs, imbuing them with distinct personalities to facilitate social and emotional connections with users constitutes a crucial design strategy for VA development. However, suitable theories or design criteria for virtual entity embodiment still need to be developed. Consequently, this project endeavors to investigate, from a Kansei engineering perspective, how various design elements in physical voice assistant human-computer interactions influence the perceived virtual personalities of VAs. Adopting a Kansei engineering approach, this study employs user characteristics (gender, age), VA gender, and degree of embodiment as design factors in human-computer interactions, utilizing the five dimensions of the brand personality scale as Kansei vocabulary. An online experiment was conducted using snowball sampling. The results of analyzing data from 159 participants indicate that design elements related to users and VAs significantly impact users’ perceptions of VA virtual personalities, though the degree of influence varies across different personality aspects. These findings contribute insights into the design of engaging and resonant virtual personalities for embodied voice user interactions.