Set the tone: Trustworthy and dominant novel voices classification using explicit judgement and machine learning...

Area of Science:

• Psychology
• Acoustics
• Machine Learning

Background:

• Voice evaluations at zero-acquaintance are primarily driven by valence-trustworthiness and power-dominance.
• Understanding the link between vocal acoustics and perceived personality traits is crucial for various decision-making processes.
• The complex relationship between acoustic properties and vocal impressions requires further investigation.

Purpose of the Study:

• To replicate and validate the bi-dimensional structure of voice-first-impressions (VFI) based on acoustic features and sex.
• To explore non-linear relationships between acoustic parameters and VFI using machine learning models.
• To identify specific acoustic correlates of trustworthiness and dominance in voices.

Main Methods:

• Study 1: Assessed personality judgments and correlated acoustic voice features (e.g., center of gravity, shimmer, pitch) with VFI dimensions.
• Study 2: Employed machine learning classifiers (Support Vector Machine, Random Forest) to predict VFI from acoustic parameters.
• Utilized a dataset of novel voices evaluated at zero-acquaintance.

Main Results:

• A bi-dimensional model of VFI (valence-trustworthiness, power-dominance) explained 80% of the variance.
• Trustworthiness was associated with brighter, smoother, and louder voices.
• Dominance was linked to vocal roughness, high-frequency energy, and pitch.
• Machine learning models achieved above-chance classification of vocal profiles, confirming predictive power.

Conclusions:

• Voice-first-impressions exhibit a robust bi-dimensional structure.
• Acoustic features significantly influence perceptions of trustworthiness and dominance.
• Machine learning techniques are valuable tools for analyzing vocal impressions.
• Sex plays a role in the formation of voice-first-impressions.