Affective Trajectories and Learning Outcomes in Russian Language Learning among Chinese-English Bilingual Learners Using Personalized AI Learning Software

Generative artificial intelligence (GenAI) is entering the field of language learning. However, longitudinal evidence remains limited for Chinese-English bilingual learners studying a third language such as Russian. This study examines affective trajectories and learning outcomes in foreign language learning supported by personalized AI software. A four-week longitudinal pilot study followed 30 Chinese-English bilingual learners as they used a personalized AI application to learn Russian. Weekly questionnaires measured enjoyment, anxiety, motivation, perceived personalization, and perceived progress, and a Russian test was administered at the end of the study. The results showed that, during AI-supported learning, learners' anxiety declined, enjoyment and perceived progress increased, perceived personalization rose slightly, and motivation remained stable. Multilevel models showed that perceived personalization was associated with a faster decline in anxiety (interaction term = -0.178, p < .001), and this result remained robust after baseline anxiety was controlled. These findings suggest that visible personalization in AI-supported early L3 learning may function less as a direct route to short-term achievement and more as an affective support mechanism, especially for anxiety reduction. The study provides practical evidence for third-language teaching and for the design of AI-based educational tools that attend to learners' emotional experience.