Evaluating the Effectiveness of ChatGPT as an AI-Assisted Tool in Music Performance Pedagogy (Menilai Keberkesanan ChatGPT sebagai Alat Berbantukan Kecerdasan Buatan dalam Pedagogi Persembahan Muzik)

Abstract

Artificial intelligence (AI) has produced measurable benefits in general education, yet quantitative evidence remains
limited in the embodied and interpretive domain of music performance pedagogy. This one-group pretest-posttest
intervention study with correlational analyses examined changes in students’ blind-scored music performance during a
ChatGPT-assisted period, teachers’ perceptions of preparation time, lesson accuracy and assessment reliability, and
associations between technology-use variables and teaching-learning outcomes. A four-phase 12-week study, including
a four-week ChatGPT-assisted phase (Weeks 7-10), involved 247 students and 52 teachers from five Chinese highereducation
institutions selected through stratified purposive sampling. Data were collected using adapted student and
teacher outcome questionnaires, three study-specific technology-use indicators, and a five-dimension performance
rubric, blind-scored by three judges at pretest and posttest, and were analysed using t-tests, ANOVA, correlation,
regression and MANOVA. Students’ performance increased significantly (dz = 0.74), with a significantly greater gain in
musical expression than in the embodied technical dimensions. Teachers rated all instructional-efficiency outcomes
significantly above the neutral midpoint. Among students, usage frequency predicted self-reported outcomes; among
teachers, only perceived usefulness was significant. The linear association between usage frequency and blind-scored
performance gain was nonsignificant. The findings support ChatGPT’s effectiveness as a supervised aid for interpretive
learning, practice planning, lesson preparation and rubric-aligned work, but not as a substitute for embodied technical
feedback from human teachers.