EFL secondary students’ use of ChatGPT for writing task completion pathways

This study examines the content and sequencing of prompts that English as a foreign language (EFL) secondary students used with no prior instruction to complete a writing task using ChatGPT. Four cases illustrated students learned to craft prompts to guide the chatbot’s text generation through individual trial-and-error, taking pathways that varied in sophistication. We found six distinct types of prompt content and some sophisticated combinations of that content such as natural language instruction with a question and an input data exemplar. We observed students take from five to 12 turns to complete the writing task. However, we also observed that no student had used all prompt types and each student had used a different combination of prompt types to complete the task. Findings evidence the need for explicit prompt engineering education in EFL writing classrooms rather than leaving students’ prompting to individual exploration. With personalized and collaborative instruction on prompt content, sequence and capabilities, students can become more strategic, sophisticated collaborators with ChatGPT on writing tasks.