What Are They Asking? Analyzing Multilingual Student Questions to ChatGPT in Computer Science Education

Abstract

Large language models (LLMs) like ChatGPT are increasingly used by students to support their learning, yet little is known about what kinds of questions students ask and how they engage with these tools, especially in multilingual contexts. This study analyzes around 600 messages collected from students in an introductory computer science courses taught to international liberal arts students at a Japanese university, where participants were invited to interact with ChatGPT in either English or their native language. Using language detection, sentiment analysis, topic modeling, and intent classification, we explore the linguistic choices, thematic focus, and emotional tone of student inquiries. Contrary to expectations, students rarely used ChatGPT for basic or factual questions—instead, they predominantly posed in-depth, exploratory questions and sought creative ways to apply course content—topics that might not fit within typical course interactions or where students may fear judgment. Our findings reveal a preference for English, a wide range of course-related topics, mostly neutral sentiment, and a notable tendency toward creative and advanced engagement. These insights shed light on how LLMs can foster deeper learning and creativity in multilingual computer science education, informing the design of supportive AI-integrated learning environments.

Ähnliche Arbeiten