earticle

영어

The purpose of this study was to analyze and compare the characteristics of Korean college students’ conversations with ChatGPT and peers in the classroom, in terms of turn-taking, adjacency pairs, repair sequence, and sequence organization of their conversations. A total of 20 conversations from ChatGPT- and peer-based English speaking practice were analyzed through the Conversation Analysis (CA) procedure qualitatively (Heritage, 1984; Sacks et al., 1974). Also, Python 3.14 (2025) was employed to analyze the number of turn-taking, adjacent pairs, repair sequence, and sequence organization, quantitatively. Results of data analysis indicated that the students exchanged more turns in conversations between peers than in those with ChatGPT. Adjacent pairs of conversations were also found to have used a larger number of adjacent pairs in peer-based conversations. The students repaired 3.8 times in the conversation with ChatGPT while repairing 13.4 times with their peers. In the part of sequence organization, there is no significant difference between conversations with ChatGPT and peers. Based on the results of data analysis, pedagogical implications and suggestions for effectively utilizing AI devices in the Korean EFL classroom are presented.