Using AI to Support L2 Pragmatic Competence in Interpreting Context‐Specific Meanings

ABSTRACT This study investigates whether large language models (LLMs) can provide a cost‐effective alternative to human‐generated metapragmatic input for the explicit teaching of L2 pragmatic meaning. Building on previous research into Korean first‐person pronouns, the study examines whether GPT‐4‐generated explanations of the honorific pronoun ce can support learners’ understanding of both stereotypical and contextual meanings. Twenty learners of Korean at Australian universities participated in individual and group interview sessions centred on Kim Jong Un's use of ce in a 2018 inter‐Korean summit speech. Prior to instruction, learners primarily interpreted ce through broad notions such as politeness, respect, and formality, differing from L1 speakers’ stronger emphasis on relational meanings such as lowering oneself and elevating others. Learners were then exposed to GPT‐4‐generated metapragmatic explanations derived from repeated prompting and coding procedures. Following the intervention, participants increasingly incorporated the concept of “lowering oneself” into their interpretations of contextualized pronoun use, producing explanations more closely aligned with L1 speakers' relational and hierarchical understandings of ce . The findings suggest that LLM‐generated input can effectively support the development of L2 metapragmatic understanding when embedded within guided pedagogical interaction, while also offering scalable possibilities for pragmatics‐focused instructional materials development.