TEACHING SCIENTIFIC WRITING WITH AI TECHNOLOGIES: TYPOLOGY-DRIVEN ERROR CORRECTION

Abstract

Using AI tools as writing assistants has become a reality. However, there is a problem of students’ over-relying on technology. Considering specific characteristics of scientific writing, students need to be taught to revise LLM-edited texts to avoid errors and maintain the individual author’s voice. We developed a framework for reviewing LLM-edited texts to guide students in evaluating a text’s morphology, lexis, syntax, semantics, and pragmatics. This framework categorizes mistakes to clarify which are effectively edited by LLMs and which require human judgment. In a study with L2 English post-graduate students, we implemented this framework during the revision of scientific abstracts. A post-intervention survey showed that the framework effectively organized the editing process and improved writing outcomes. However, despite these benefits, participants maintained a preference for direct instructor feedback over LLM-generated suggestions.

Ähnliche Arbeiten