Speaker
Description
This study investigates the extent to which AI (ChatGPT 5.5, Thinking standard) can replicate the fine-grained clausal and phrasal complexity of advanced L2 Vietnamese IELTS learners in argumentative writing. A comparative corpus analysis of 60 human-made IELTS Task 2 essays (20,267 words) and 60 AI-generated essays (20,398 words) was conducted using the TAASSC tool, with statistical analyses performed via SPSS and Excel (p < 0.01). The findings reveal distinct syntactic patterns between the two groups. Human writers consistently exhibit greater overall clausal and phrasal complexity, utilizing denser clauses with varied subordination and heavier noun modification through prepositional phrases and determiners. Conversely, AI texts demonstrate a narrower, targeted complexity, relying heavily on adjectival modifiers and existential constructions. Ultimately, while ChatGPT mimics isolated advanced structures, it fails to replicate the deeply embedded, multi-layered syntactic architecture characteristic of advanced human writing. These divergent syntactic fingerprints offer practical implications for the teaching and learning of argumentative writing by highlighting complex multi-layered structures for explicit instruction. Furthermore, they contribute to syntactic complexity research by underscoring the necessity of fine-grained phrasal analysis and inform research into language and AI by exposing the current syntactic boundaries of large language models.
Biography
Lien Tran-Hong is an English language educator and researcher based in Ho Chi Minh City, Vietnam, specialising in advanced EFL writing, assessment, and teacher development. Drawing on extensive classroom experience with high‑proficiency learners, she focuses on helping students and teachers understand how lexical and grammatical choices shape clarity, nuance, and argumentation in academic and test‑prep contexts. Tran-Hong has designed and led curricula for IELTS preparation, academic writing, and TESOL teacher training, often integrating corpus‑informed materials and data‑driven learning tasks. She is especially interested in supporting teachers to move beyond formulaic band score strategies toward evidence‑based approaches that foreground rhetorical purpose, stance, and audience expectations. In her professional practice, she regularly experiments with using learner corpora and AI tools as pedagogical resources, encouraging critical comparison between human and machine‑generated texts as a way to deepen metalinguistic awareness. Academically, Tran-Hong's interests span syntactic complexity research, language assessment, genre‑based pedagogy, and the evolving relationship between language learning and artificial intelligence. She aims to contribute to a more nuanced understanding of what counts as advanced writing in English today and how teachers and learners in expanding‑circle contexts can engage with global academic norms without losing local voices.
| Affiliate type | Others |
|---|