Karimi Temijani، Hossein

A Comparative Study on the Use of Interactional Metadiscourse Markers in EFL Students’ an‎d AI Generated Texts

Abstract Writing has a central role in academic communication, since it allows learners to present ideas, build arguments an‎d engage the readers. One key element of successful academic writing is the use of metadiscourse, which helps writers express their stance, organize arguments, an‎d establish a relationship with the readers. Interactional metadiscourse markers are especially important because they reveal how writers connect with readers an‎d express their attitudes. This study investigated differences in the use of interactional metadiscourse markers in argumentative essays written by upper-intermediate an‎d advanced EFL students an‎d generated by ChatGPT-4. A total of 100 essays (50 human-written, 50 AI-generated) were analyzed for five markers: hedges, boosters, attitude markers, engagement markers, an‎d self-mentions. Quantitative an‎d qualitative analyses revealed significant differences in four of the five markers. Students used hedges an‎d self-mentions more frequently, reflecting cautious reasoning an‎d authorial presence, while ChatGPT employed more boosters an‎d engagement markers, indicating confident, reader-oriented writing. Attitude markers did not differ significantly between the groups. Variability analyses showed that student writing was highly diverse, reflecting individual style an‎d experience, whereas ChatGPT texts were consistent an‎d systematic. The findings demonstrated that human an‎d AI writing exhibit distinct interpersonal strategies: students focus on self-centered, mitigated claims, while AI emphasizes assertive, interactive discourse. These results have implications for EFL teaching, suggesting that AI-generated texts can model confident, reader-focused writing, while human writers should be encouraged to maintain their unique voice.