Morphing Vulnerable Machines: Paralinguistic Cues in Digital VoiceAssistants Shape Perceptions of Physicality, Vulnerability,And Trust
Author: Fotios Efthymiou
Paralinguistic features in the human voice are reliable cues to detect discrete momentary emotions and personality inferences in other humans. By listening to different paralinguistic features in the human voice people infer differences in personality (Mohammadi, Origlia, Filippone, & Vinciarelli, 2012). The current work provides evidence from a series of tightly controlled experiments that altering digital voice assistants along the vibrato dimension (i.e., systematic changes in the pitch of a synthesized voice) causes systematic changes in personality perception and trust, while holding critical dimensions of message content, syntax, and other paralinguistic cues constant. Our results demonstrate that humans attribute greater submissiveness, lower dominance, and reduced perceptions of power to a digital voice assistant with increasing vibrato. We further show that these perceptions are explained by altering the perception of physicality such that greater vibrato results in perceiving the voice assistant as older and smaller. Moreover, we show that these changes in perception cause subsequent attributions of trust and are robust across a broad range of consumer demographics and psychological trait measures.