Impact of anthropomorphism in AI assistants’ verbal feedback on task performance and emotional experience

Artificial intelligence (AI) assistants are increasingly deployed across various fields to replace human operators in providing feedback to users, making anthropomorphism a pivotal topic in AI-assisted interactions. We investigated the impact of varying degrees of anthropomorphism in verbal feedback on task performance and user experience. A total of 30 participants were recruited, and their task performance and subjective experiences in response to different levels of feedback were measured. Moderate-level anthropomorphic verbal feedback elicited significantly lower self-efficacy and pleasure compared to high- and low- level feedback. Notably, under high-level feedback, participants demonstrated a significant increase in response speed compared to lower levels. These findings suggested that high-level feedback of AI assistants could enhance participants' performance and experience, thereby improving training or learning efficacy. However, moderate-level anthropomorphic feedback appeared to be less effective. The study outcomes could offer insights for future research and design of AI voice systems.