Gendering Social Robots: Voice
Voice
Male vs. Female voice genders a machine (Nass et al., 1997; Siegel, et al., 2009)
Childish, often seen as gender neutral (Obaid et al., 2016)
Ethnicity: African American, European American, Asian, etc.
Language and Accent (Andrist et al., 2015). English, for example,
US—differs by region, slang, speech patterns, vocabulary
UK—differs by class
Australian
South Asian—Hinglish
Caribbean—not available in speech technology
South African
Voices are full of cultural information. The choice of a robot voice is a major gender cue (Nass et al., 1997). Pitch of a voice indicates whether it is a male, female, or child. Lower voices carry more authority in Western culture. For example, Margaret Thatcher, the first woman prime minister in the U.K., sought training by a vocal coach from the National Theatre to lower her voice and make it more authoritative (Atkinson, 1984).
Numerous voices in assistive technology for the elderly are childish. Although we could find no analysis of this choice, a childish voice is culturally non-threatening (Obaid et al., 2016).
Voices also encode cultural information via accents. For English, virtual voices currently include the following accents in an adult male and female version: US, UK, Australian, and South Asian (Hinglish), but not the great varieties of Caribbean or South African (Knight, 2016). The US accent is typically not Southern or Texan.
