The empathic voice penalty: Vocal delivery reduces perceived empathy in humans and AI

Artificial intelligence (AI) technology is increasingly capable of delivering high-quality empathetic messages to human users. As this technology has continued to advance, new modes of engagement with these artificial entities have become possible. While prior research on empathic AI has focused almost exclusively on text-based empathy, we explored how more immersive vocal delivery of empathetic messages might enhance or ultimately undermine the perceived quality and emotional consequences of AI versus human empathy. Across multiple studies (N = 799) we examined how the mode of expression (voice vs. text) and the empathy source (human vs. AI) jointly shape people’s perceptions of empathic quality and empathy recipients’ experiences of feeling heard. We found evidence of an empathic voice penalty: Voice delivery reduces perceptions of quality and experiences of feeling heard. Critically, this voice penalty appeared for both human and AI empathizers—though was especially strong for AI empathizers. We further explored how this mode of expression impacts feelings of uncanniness, and the mediating role that these feelings of uncanniness play in making people feel empathized with. Taken together, these findings challenge assumptions that more immersive, human-like modes of emotional expression might ultimately boost empathic effectiveness and highlight uncanniness as a possible impediment to effective AI empathy.