Evaluating Emotional Design in ChatGPT Voice Modes: User Experience and Gender Effects in the Language Learning Scenario

This study tests the relevance of emotional design principles in conversational AI voice modes during language learning, with a specific focus on user gender differences. Drawing on Don Norman’s three emotional levels—visceral, behavioral, and reflective—we evaluated three ChatGPT voice modes (Juniper, Cove, Spruce), selected through acoustic analysis for their distinct pitch, tone, and intensity profiles. Sixty expats living in Germany (balanced by gender) engaged in German learning sessions through voice interaction and rated their experience using a 12-item Likert scale. Results show that while overall emotional impact did not significantly differ among the voices, Juniper and Spruce exhibited greater differentiation across emotional levels, while Cove demonstrated consistent user experience. Male participants consistently rated all voices higher than female participants. These findings offer insights into how voice design can affect emotional user experience and highlight the importance of gender-sensitive design in AI-powered language learning tools.