Code-switching—mixing languages within a single conversation—is a familiar feature of multilingual communication and has been studied extensively in natural language processing. But its role in language learning, especially when learners practice with AI chatbots, has received far less attention.
The new arXiv paper addresses that gap by looking specifically at Korean-English code-switching in LLM-based speaking practice. Rather than treating code-switching as a mistake or a failure to stay in the target language, the authors examine what functions it serves for learners and what pedagogical value it might have.
Because the paper is a preprint, its findings are not yet peer-reviewed. Still, it highlights a growing research area: as AI conversation partners become more common in language education, understanding how learners actually use multiple languages with them will be important for designing better practice tools.