Social responses to conversational TV VUI: Apology and voice

The study investigated whether apologetic synthetic gendered voices affect users' perception of an error-prone VUI. In a TV viewing task, participants interacted with the conversational TV, and executed eight menus in a 2 (apologetic error message: yes vs. no) by 2 (voice gender) by 2 (subject gender) gender balanced, between participants experiment. When participants encountered errors, the TV provided verbal error messages, with or without an apology. The results revealed significant two-way interaction effects of apology (yes) and voice gender (male) on perception of the TV, and the voice. Irrespective of gender, participants responded to a male voice more, when it offered apologies for errors. It is interpreted that the context in which genuineness of apology was regarded as important made participants perceive a male voice as being more trustworthy than a female voice. The participants seem to have applied gender stereotypical perceptions to gendered VUI, as they do to other humans.