Speaker
Description
Simple requests for help from generative AI (genAI) can lead a tool to perform work that teachers intended to be learning experiences for students. The purpose of this study is to investigate the pedagogical guardrails of commercial genAI tools by examining how they respond when a learner makes requests under different instructional boundaries. Using prompt-based discourse elicitation (PBDE) as a methodology, the study presents different genAI tools with the same writing task under contrasting conditions. First, an open-ended request for help establishes the tools’ default behavior. Other trials introduce teacher-provided instructions that restrict genAI use and test whether the tools recognize those boundaries. Finally, PBDE examines explicit learner requests for guidance that ask the tools to provide feedback and ask leading questions rather than rewrite the text. Responses are analyzed according to how far the systems intervene in the learner’s work and whether they respect the stated instructional boundary. Expected findings are that underspecified requests will encourage substantial direct intervention, while explicit constraints will produce more pedagogically supportive behavior. The study approaches these differences as questions of genAI literacy. The implications for TESOL are that responsible educational use of genAI depends not only on learners understanding instructor boundaries, but also on their ability to use prompt engineering to shape the kind of assistance they receive. The findings may inform genAI literacy and ethics by offering specific suggestions for writing pedagogy in language learning classrooms.
Keywords: generative AI; AI literacy; learner agency; writing assistance; prompt-based discourse elicitation
