Empowering Intercultural Performative Competence: An Empirical Study of LLM-Assisted Vocal Learning in a Multicultural Context

Abstract

Multicultural vocal education requires students to deeply understand foreign cultural contexts and linguistic nuances, which often poses a significant cognitive challenge. Traditional empirical learning often leads to mechanical imitation rather than authentic emotional connection. This study investigates the impact of Large Language Models (LLMs) as cognitive pedagogical tools on enhancing vocal students' intercultural performative competence, specifically focusing on textual comprehension and aesthetic expression rather than vocal mechanics. A 2×2 within-subjects crossover experiment was conducted with eight Chinese vocal majors (four postgraduates and four undergraduates). Participants were tasked with learning two Western classical art songs (Italian and German) under two conditions: traditional desk work and LLM-assisted learning. Performance was evaluated through double-blind expert scoring based on aesthetic dimensions, triangulated with subjective student questionnaires. The results indicate that the LLM intervention significantly improved students' language diction, textual comprehension, and emotional empathy (p < 0.05). Qualitative feedback revealed that LLMs effectively reduced cognitive load and helped students decode deep cultural metaphors, thereby transforming mere linguistic memorization into authentic aesthetic expression. Based on these empirical findings, this paper proposes a three-stage "Perception-Internalization-Expression" pedagogical strategy, providing a novel and highly adaptable framework for integrating generative AI into multicultural music education.