Presentation Designing and Evaluating a Custom GPT for Conversation Practice with University Students more

This study explores how Japanese university students used a custom GPT chatbot to support English speaking practice in a first-year Basic English Communication course. The chatbot was built using conversation scripts and target vocabulary from the course textbook (Conversations in Class by Alma publishing), allowing students to practice class content through structured AI-mediated dialogue. Participants were 14 first-year students. Over an eight-week period, two groups completed different forms of chatbot practice. One group conducted short daily conversations (3–5 minutes) using free ChatGPT accounts outside class. A second group participated in longer weekly sessions (10–15 minutes) in the instructor’s office, which were video-recorded for analysis. The chatbot provided immediate feedback through prompts, compliments, reformulations of student responses, and suggestions for more natural expressions. Quantitative data included chatbot usage logs, pre/post speaking tasks measuring fluency, and Likert-scale surveys on motivation and perceived usefulness. Qualitative data came from reflection prompts and interviews, and analysis of students’ behavior and body language during the recorded speaking sessions. Results showed that students were generally enthusiastic about using the chatbot to improve their speaking ability. However, some participants experienced difficulties interacting with the technology, which may reflect varying levels of technological literacy among students.