Sexualised synthetic personas encode and amplify gendered power asymmetries through voice

The paper investigates the implications of sexualized AI-generated voices from a commercial platform, revealing that these systems perpetuate gendered power dynamics. Through a listening experiment, it was found that female-coded voices are often described with submissive and sexualized attributes, while male-coded voices are linked to dominance and positive traits, highlighting a narrow and largely binary representation of gender. This research is significant for AI practitioners as it underscores the need for critical evaluation of voice AI systems to address inherent biases and promote a more diverse representation of gender.