Author(s) :

Shreya Chandrasekar.

Volume/Issue :

Abstract :

This study aimed to identify which features of three AI violin tutor applications—Trala, Violy, and Caidence—high school violinists perceived as most supportive when learning a new piece of music. Traditional violin pedagogy often depends on individualized instruction, which can limit access for students who face financial and logistical barriers to private lessons and high-quality coaching. Prior studies on AI-assisted violin learning have largely emphasized listener-based evaluations or technical system performance rather than students’ direct experiences of using the tools. The present study addressed that gap through a student-centered, mixed-methods, quasi-experimental design with 12 high school violinists from one competitive high school in the San Francisco Bay Area. Each participant completed a single 45–60 minute session in which all three applications were used sequentially, followed by a structured reflection survey. Data sources included Likert-scale survey responses, open-ended survey responses, and observational field notes. Qualitative analysis used deductive coding focused on visual design and interface, feedback clarity and accuracy, and perceived effectiveness. Quantitative analysis was descriptive and exploratory. Trala was the most preferred application, selected by 8 of 12 participants, and was described as strongest for immediate intonation feedback. Violy was perceived as more useful for complete run-throughs of longer excerpts but less clear in locating errors. Caidence was least effective for technical correction but was valued by some participants for stylistic and interpretive guidance. Across applications, the most salient characteristics were immediate feedback, clarity of explanation, and interface simplicity. These findings suggest that the educational value of AI violin tutors depends not only on technical capability but also on how clearly feedback is communicated to learners. Because the study used a small, geographically homogeneous sample and a single-session design, the findings should be interpreted as preliminary and descriptive.