Designing Emotional Expressions of Conversational States for Voice Assistants

The use of voice-activated virtual assistants (VAs) to execute tasks, request information, or entertain oneself on the smartphone is on the rise. However, insufficient feedback on the states of VAs may impair the interaction flow. We propose to use nonverbal emotional expressions to indicate a VA's conversational states and promote user engagement. We introduce three emotional expression designs of VA, iconic facial expressions, a text box with body movements, and emotional voice waveforms. Our user study results show that a VA with an expressive face or text body movements can evoke stronger user engagement than the one with voice waveforms.