ScholarWorks@성균관대학교

초록

As multimodal voice-based AI proliferates, UX guidelines for optimal modality selection are urgently needed. This study examined how voice AI should deliver refusal messages when handling inappropriate user requests. We conducted a 2×2 between-subjects experiment (N =144) testing the effects of source (AI agent vs. company) and modality (text vs. voice) in refusals by a voice AI. Results showed that shifting from AI to company and from voice to text both increased negative emotions and psychological reactance while decreasing social presence and trust. Qualitative findings revealed that changing the messenger more effectively discouraged persuasion attempts and reduced harmful intent, but increased feelings of disruption and discomfort. Overall, our findings demonstrate that beyond the specific choices of modality and messenger, the very act of shifting away from the established communication channel and source significantly impacts user experience.

키워드