Presentation Examining Linguistic Shifts in Student Writing Before and After the Launch of ChatGPT more

The prevalence of Large Language Models (LLMs), such as ChatGPT, has sparked debate regarding their influence on student academic writing. While existing literature has predominantly investigated LLM usage through quantitative methods—such as word frequency statistics and probability-based detection—few studies have systematically assessed the potential impact of these tools on the linguistic characteristics of student writing. To bridge this gap, the researchers analyzed 1,000 research papers written by fourth-year students in humanities disciplines at universities in Hong Kong. Using a self-compiled corpus, this study compared papers written before and after the launch of ChatGPT, specifically examining two linguistic features: namely, the frequency of LLM-preferred vocabulary and syntactic complexity. This study uses the Tool for the Automatic Analysis of Syntactic Sophistication and Complexity (TAASSC) to calculate 14 measures of syntactic complexity, which can be categorised into five types: length of production unit, amount of subordination, amount of coordination, phrasal structures, and sentence complexity. The results indicate a significant rise in LLM-preferred lexical patterns, demonstrating the influence of generative AI on student argumentative writing. Preliminary findings also indicate that syntactic complexity declined, pointing toward simpler sentence structures. These findings provide valuable insights for educators seeking to distinguish between human and AI-generated text.