The representational voice space: Understanding trustworthiness and competence in natural and artificial voice perception

• Applicant: Dr. Claudia Roswandowitz
• Subject Area: Human Cognitive and Systems Neuroscience
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 559451439

When listening to a voice, various information about the speaker is extracted, such as identity and social characteristics like trustworthiness and competence. This information forms the basis for effective social interaction between humans. Until recently, the voice was considered an exclusively human characteristic, but advances in speech technology have changed this. This progress could change the use and perception of artificial voices in the future. Even though artificial voices are widely accepted by the general population, artificial voices are perceived as less trustworthy. What remains unresolved is the crucial and fundamental question of which natural vocal features convey social trait information and are thus critical for the mental representation of social voices. With this knowledge, fundamental research questions in cognitive neurosciences will be answered, and the design of socially convincing artificial voices can be facilitated. In this project, I will characterise the features of the representational space of vocal trustworthiness and competence (WP1, methods: computer modelling, behavioural tests, acoustic analysis) and plan to use voice synthesis to causally test the validity of the representational space dimensions (WP2, methods: voice synthesis, behavioural testing). By applying neuroimaging, I will test the extent to which artificial voices have the potential to elicit a rewarding neural response comparable to natural voices (WP3, methods: functional magnetic resonance imaging, behavioural testing).

• DFG Programme: Research Grants