Let voice assistants sound like a machine: Voice and task type effects on perceived fluency, competence, and consumer attitude

Voice Assistants (VA) are increasingly penetrating consumers' daily lives. This study aimed to investigate the effects of synthetic vs. human voice on users' perception of the voice, social judgments of the VA, and attitudes towards VAs. Drawing from CASA(Computers-as-socialactors) framework and social perception literature, we developed a theoretical model that explains the psychological underlying mechanism of the voice effects. Through two online experiments, we rejected our initial hypotheses that human voice would increase users' perception and evaluation of the VAs. Instead, our findings support that the VAs were favored when they spoke with a synthetic voice only when the users engaged in functional tasks. There was no difference between the voices for social tasks. A further investigation revealed that participants perceived the synthetic voice to be more fluent when VA responds to functional tasks. This enhanced perception of fluency increased competence perception and attitudes. The findings imply that VAs should not be designed to closely resemble humans. Rather, consideration of usage contexts and consumer expectations should be prioritized in developing most likable VAs.