Speaker
Description
This study examines how first-year Japanese ESL students use generative AI during academic writing by analyzing the types of prompts they produce and how these relate to writing development and ethical clarity. Using prompt logs and draft revisions collected over six weeks, the study identifies prompt types that support learning-oriented revision and those associated with surface-level correction or ethical uncertainty. The findings offer practical guidance for AI use in ESL writing classrooms.
Abstract
Generative AI tools such as ChatGPT are increasingly used in university EFL writing classes, yet it remains unclear how students’ prompting behavior influences learning outcomes. This six-week classroom-based mixed-methods study examines how first-year Japanese university EFL students use generative AI during academic writing tasks, with the aim of identifying common prompt types and exploring how different prompting patterns relate to writing development and students’ perceived understanding of appropriate AI use. Participants were 40 students enrolled in two intact first-year academic writing classes. Students completed one process-writing assignment involving an initial and revised draft, as well as timed writing tasks completed without AI in Week 1 and Week 6. During AI-permitted stages, students submitted brief prompt logs documenting how they interacted with ChatGPT. Prompts were coded into seven functional categories, and inter-rater reliability was calculated. Results showed that students primarily used AI for language-level support such as grammar correction, vocabulary assistance, and translation, while higher-order prompts related to organization, explanation, and evaluation were less frequent. Prompt types were associated with different outcomes: correction-focused prompting aligned with surface-level accuracy, whereas organization- and evaluation-focused prompting aligned with deeper revisions. Prompts requesting text generation or rewriting were associated with greater uncertainty about appropriate AI use. These findings suggest that the educational value of AI in ESL writing depends on how students prompt the tool.
Short summary
This study examines how first-year Japanese ESL students use generative AI during academic writing by analyzing the types of prompts they produce and how these relate to writing development and ethical clarity. Using prompt logs and draft revisions collected over six weeks, the study identifies prompt types that support learning-oriented revision and those associated with surface-level correction or ethical uncertainty. The findings offer practical guidance for AI use in ESL writing classrooms.
Keywords
Generative AI
ChatGPT
ESL writing
Prompting behavior
| Scheduling preference | Anytime on Saturday |
|---|---|
| Title | A Taxonomy of AI Prompt Types in First-Year ESL Writing in Japan |