ChatGPT as a Conversational Partner: Korean as a Foreign Language Students’ Perceptions and Challenges

Abstract

The perceptions of learners of Korean as a foreign language (KFL) regarding the use of ChatGPT-3.5 Voice as a conversational partner and their interactional challenges were explored in this study to address the lack of research on real-time voice interactions with generative artificial intelligence (GenAI) in the context of less commonly taught languages (LCTLs), as most studies have focused on text-based interactions in English language-learning contexts. Guided by the adapted Technology Acceptance Model-3 (TAM3), this exploratory, classroom-based study combined survey data with transcript-based analyses to explore learners’ perceptions and interactional challenges. The findings revealed that students appreciated ChatGPT’s availability and its potential to support fluency, but reported issues with speech recognition, response adaptability, and coherence. Learners showed more agency in structuring conversations compared to prior chatbot studies, but also disengaged more readily during conversation breakdowns. ChatGPT’s persistence often led to circular exchanges that disrupted conversation flow. The findings indicated the need for targeted pedagogical support and technical refinement to optimize artificial intelligence (AI) mediated speaking practice. The study contributes to understanding how real-time AI tools can be integrated meaningfully into LCTL instruction.

Keywords

ChatGPT-3.5 Voice, chatbots, Korean as a foreign language (KFL), Technology Acceptance Model-3 (TAM3), AI-mediated speaking practice

How to Cite