etranslation an d comparison in AI-mediated Korean language instruction: A pilot exploration with multilingual learners
This study explores the design and preliminary application of a Korean language class incorporating a retranslation-comparison activity in an AI-assisted translation environment. Grounded in the AI-in-the-loop (AI²L) principle, whereby learners serve as the primary agents of judgment while AI translation functions as a mediating text rather than an authoritative source, a three-session structure was developed and piloted with 30 multilingual international students at a Korean university. In the first session, learners activated background knowledge and critically reflected on AI translation tools. In the second session, they retranslated a Korean text from AI-translated versions in their mother tongues using only a dictionary, without AI tools. In the third session, they compared their retranslations with the original Korean text and presented their translation choices for peer and instructor discussion. The findings indicate that the activity fostered learner agency, metalinguistic awareness, and cross-linguistic contrastive thinking. Variation in retranslation outcomes across language backgrounds—Chinese, Japanese, Vietnamese, Russian, and English—generated rich opportunities for meaning negotiation and interaction. The study also identified several limitations, including uneven participation by proficiency level, learner resistance to AI-free translation, and time constraints within the three-session format. It suggests that retranslation-comparison activities offer a pedagogically viable structure for Korean language instruction in an era of generative AI, and calls for expanded implementation and empirical validation in future research.