Proceedings of the Annual Meeting of the Cognitive Science Society

Humans are now regularly speaking to voice-activatedartificially intelligent (voice-AI) assistants. Yet, ourunderstanding of the cognitive mechanisms at play duringspeech interactions with a voice-AI, relative to a real human,interlocutor is an understudied area of research. The presentstudy tests whether top-down guise of “apparent humanness”affects vocal alignment patterns to human and text-to-speech(TTS) voices. In a between-subjects design, participants heardeither 4 naturally-produced or 4 TTS voices. Apparenthumanness guise varied within-subject. Speaker guise wasmanipulated via a top-down label with images, either of twopictures of voice-AI systems (Amazon Echos) or two humantalkers. Vocal alignment in vowel duration revealed top-downeffects of apparent humanness guise: participants showedgreater alignment to TTS voices when presented with a deviceguise (“authentic guise”), but lower alignment in the twoinauthentic guises. Results suggest a dynamic interplay ofbottom-up and top-down factors in human and voice-AIinteraction.