Less Anxiety and Stress, More Artistic Guidance: A study on How Artificial Intelligence Supports Traditional Voice Lessons

Lesson-related anxiety and stress may reduce students’ effectiveness and improvement during traditional university voice instructions. Recently, AI-based intervention exhibits great potential in students’ voice singing trainings. However, how it influences traditional voice instruction still remain elusive. To do so, we conducted a one-semester study to explore the role of AI in traditional voice instructions compared to human mock-jury intervention and regular self-practice. Across four time points (week 4, 8, 12, and 15), AI-exposure group (AI-EG) showed a progressive reduction in state anxiety and gradual improvements in autonomic regulation (higher RMSSD and lower LF/HF) during voice lessons. More importantly, video-based meaning-making feedback analysis showed that the AI-EG exhibited a higher negotiating feedback ratio over time, indicating a greater proportion of feedback episodes that provided space for conversation, exploration, and reflective meaning-making. These findings suggest that AI may be most valuable not as a replacement for voice instructors, but as a supportive evaluative scaffold associated with reduced evaluative strain, more favorable autonomic regulation, expanded reflective feedback space, and higher performance outcomes within traditional voice lessons.