Exploring Learner Prompting Behavior and Its Effect on ChatGPT-Assisted English Writing Revision

This study aims to identify EFL learners’ prompting behaviors observed during ChatGPT-assisted English writing revision and examine how these behaviors affect the quality of their revised writings. Specifically, we investigated learners’ objectives for using ChatGPT during revision, their prompt-writing approaches, and the alignment between these objectives and approaches. Data from essays, reflection logs, and ChatGPT logs were gathered using a one-group pretest–posttest experiment, with thematic analysis, network analysis and Wilcoxon signed-rank test used for analysis. Results indicated that learners predominantly focused on surface-level features, displayed a lack of prompt approaches, and often employed general, one-size-fits-all, prompts that did not align with their specific objectives. Consequently, while improvements were noted in surface-level aspects of writing, higher-order elements like content, organization, and cohesion showed minimal enhancement. These findings underscore the necessity of incorporating prompt-writing instruction into English writing pedagogy to enhance learners’ ability to craft specific, goal-aligned prompts, leading to more productive feedback from AI tools like ChatGPT and facilitating meaningful improvements in higher-order writing skills. This study provides valuable insights for instructors and students, highlighting the pedagogical potential of integrating prompt literacy training in the use of chatbots to enhance English writing skills.