The Potential of Conversational Agents for the Realisation of Attention Bias Modification Training Approach to Address Performance Anxiety in Vocal Education

Performance anxiety remains a persistent barrier in higher vocal education because it can undermine both technical control and expressive delivery during public singing. This study examined whether a chatbot-delivered, ABMT-informed intervention could reduce performance anxiety among student vocalists and improve selected performance outcomes under routine training conditions. The study included 203 fourth-year vocal students assigned to either a psychologically active Shim chatbot condition or a neutral-script chatbot control condition over a four-month intervention period. Pretest–posttest outcomes were evaluated for situational distress, music performance anxiety, and three rated dimensions of vocal performance, while post-intervention measures assessed self-disclosure, perceived interpersonal closeness, and total interaction time. The revised analysis treated pre–post outcomes using mixed repeated-measures models and post-intervention interaction outcomes using between-group comparisons. The active Shim condition was associated with a marked reduction in music performance anxiety (53.38 to 29.46 vs. 52.67 to 51.78) and a stronger increase in performance confidence (3.67 to 4.21 vs. 3.58 to 3.64) than the neutral-script control condition. The active condition was further associated with lower situational distress, stronger vocal accuracy gains, and higher self-disclosure, perceived interpersonal closeness, and total interaction time, whereas musical expression did not show a clear intervention-specific benefit. These findings suggest that a psychologically active chatbot may be a useful adjunct for reducing short-term performance anxiety in higher-education vocal students and for supporting conditions associated with more confident performance. However, the results do not indicate that chatbot use alone improves expressive interpretation, and the relational outcomes should be interpreted cautiously because the two dialogue conditions differed in content as well as psychological activity.