AI Emotion Measurement from Voice

What if the voice itself could tell us how consumers feel?

We develop and evaluate AI-based methods for estimating consumer emotion from vocal prosody, including rhythm, pitch, intonation, pace, and other nonverbal signals. This work complements self-reports with richer, real-time evidence of how people feel during human-AI interactions.

About the programme

A pair of over-ear headphones, drawn in small squares

Empathy research is only as good as its emotion measures, and self-reports capture just a fraction of the story.

Leveraging recent advances in computational psychology and AI, we develop novel voice-analytics methods that read consumer emotions directly from prosody: rhythm, intonation, pace, and the micro-signals beneath the words. We conceptually argue and empirically demonstrate that AI-based emotion measurement offers marketing research a richer, more dynamic window into consumer feeling than traditional approaches allow.

Working paperSSRN

The Voice of Emotion: Advancing Marketing Research Through Real-Time AI Emotion Measurements

Read the work sample

Related publications

2023Nature Human Behaviour, 7(2), 240–250Deep learning reveals what vocal bursts express in different culturesBrooks, J. A., Tzirakis, P., Baird, A., Kim, L., Opara, M., Fang, X., … & Cowen, A. S.

2023Current Directions in Psychological Science, 32(3), 242–249Semantic space theory: Data-driven insights into basic emotionsKeltner, D., Brooks, J. A., & Cowen, A.

Related explainers

TerminologyWhat separates empathic AI, emotional AI, and affective computing?One is a research field, one reads emotion, and one responds to it. Where they overlap, and where the law tells them apart.

MethodsHow do you measure empathy in a machine?Researchers score what the system says, ask people what they perceived, and track what they do next.