The impact of AI- and peer-assisted interaction modes on oral ability, interactive experiences, and perceptions among adolescent Chinese Tibetan students

The integration of artificial intelligence (AI) into education has created new opportunities for language learning, yet little is known about the role of embodied AI agents in supporting adolescent ethnic minority students' acquisition of Standard Spoken Chinese (SSC). This study compared AI- and peer-assisted interaction among ninth-grade Tibetan students using a sequential explanatory mixed-methods design. Forty students were randomly assigned to interact with either a customised embodied AI agent or peers in four decision-making tasks, followed by interviews. Quantitative analyses examined speaking skills, interactional skills, and flow experiences, while qualitative analyses explored students' task-related emotions and perceived learning gains. Results showed that AI-assisted interaction significantly enhanced accuracy, fluency, lexical complexity, turn-taking, listener responsiveness, and flow, but was less effective for syntactic complexity and students' agency in leading interactions. Qualitative findings indicated that AI-assisted interaction elicited more positive emotions, whereas peer interaction relatively triggered more anxiety due to social pressure. Overall, students perceived the embodied AI agent as useful for linguistic, pragmatic, and (meta)cognitive skill development. The study highlights the potential of embodied AI agents in supporting SSC learning and suggests that a hybrid 'AI + peer' interaction mode may best integrate the complementary strengths of both to facilitate language learning.