김경민 Kyungmin Kim , 이장호 Jang Ho Lee

DOI: Vol.27(No.2) 1-22, 2024

In recent years, ChatGPT has integrated itself into applied linguistics and language education research, with studies proposing different ways to use this new technology in second language teaching. As part of these efforts, the present study aimed to investigate ChatGPT’s ability to solve the domain-specific questions of high-stakes exams. To this end, the dataset for this study consisted of multiple-choice questions taken from the teacher certification exams for prospective English teachers, developed by the Korea Institute for Curriculum and Evaluation between 2009 and 2013. The authors then asked ChatGPT to solve each question by entering them as prompts. The results showed that ChatGPT’s average percentage of correct answers was 66%, with the percentage slightly lower for the English Literature and Linguistics subsections than for the English Education Theories and General English subsections. The qualitative analysis also revealed four different patterns in the errors made by ChatGPT when answering the target questions, which we illustrate with specific examples. Based on these findings, we have provided implications for pedagogy and future research.

ChatGPT 4.0, domain-specific knowledge, teacher certification exam, English education

Na-young Kim

DOI: Vol.27(No.2) 23-45, 2024

This paper examines whether and to what extent an AI tutor affects TOEIC learning of English learners. For the study, a total of 110 students were recruited and randomly assigned to an AI tutor group (n = 57) and a human tutor group (n = 53). The study also investigates the difference in English proficiency among students in the two groups. Before the experiment, a TOEIC pre-test was administered. Based on the pre-test scores, participants were grouped into three CEFR proficiency levels: A1, A2, and B1. During the experimental period, the AI tutor group self-studied English with an AI tutor. The human tutor group took a traditional English class offered by a human tutor. After the experiment, a post-test was administered. The study found that the human tutor group increased their TOEIC test scores at all proficiency levels at the 5% significance level. On the other hand, the AI tutor group showed significant increases only at A1 and A2 levels. The group difference was statistically significant for the A1-level participants, indicating the superior effects of the AI tutor over the human tutor. On a basis of the findings, pedagogical implications are made.

AI tutor, human tutor, English learning, TOEIC, proficiency, CEFR

Sangmin Lee

DOI: Vol.27(No.2) 46-59, 2024

As mobile technologies proliferate and offer significant technical and pedagogical advantages, an increasing number of learners are turning to mobile applications for language learning. However, despite the convenience and potential benefits, maintaining motivation for long-term language learning remains a challenge. This study developed a speaking app tailored for middle school students learning English as a foreign language, with the goal of increasing and sustaining their motivation through the integration of playful constructivism. This approach incorporated engaging content, interactive activities, and gamification strategies into the design of the app. A total of 75 middle school students participated in this study, using the speaking app over a 10-month period and completing a post-use survey. The research focused on examining the students’ use of the app and their perceptions of learning to speak English through the app. In addition, the study explored the impact of students’ self-directed learning (SDL) skills on their app use and perceptions, given the integral role of SDL in mobile language learning. The results indicated that students had positive attitudes and perceptions towards learning English through the app. In addition, a positive correlation was found between SDL skills and their use of the app.

English speaking, mobile learning, middle school EFL students, self-directed learning, learner perception

Shinjae Park

DOI: Vol.27(No.2) 60-70, 2024

Recent studies on second language acquisition indicate that various metrics―speed, breakdown, repair, and composite measures―effectively evaluate fluency and correlate with proficiency. However, existing research does not sufficiently identify which metrics most accurately depict fluency at distinct proficiency levels or differentiate among these levels. This study examined the fluency of 187 monologues, assigned across five proficiency levels (A1-C1, according to the Common European Framework of Reference for Languages), using the software Praat to analyze aspects of utterance fluency. The findings reveal that the composite measures most accurately predict proficiency levels. Additionally, an analysis of gender differences in speed showed that males spoke faster than females, with significant statistical differences noted. These results suggest that a thorough analysis of fluency metrics provides a more dependable understanding of the concept and its association with proficiency assessments.

fluency, measuring fluency, assessing fluency, proficiency, rating scale validation