Affective Responses to Digitally Embodied Conversational Agents: Measuring Emotional Activation Through Voice Prosody

In plain English

In a 118-person between-subjects study, people solving collaborative problems with a dynamically embodied conversational agent showed more high-arousal emotion in their own speech than people working with a static-image agent, and reported more positive activation and engagement — though not more positive feeling overall. Perceived humanness and attractiveness mediated several of the effects. The transferable finding is methodological: voice prosody picked up activation differences that self-report questionnaires missed, which makes it a cheap, unobtrusive way to measure whether your avatar's animation is actually doing anything.