Interactional Metadiscourse Markers in the Era of Artificial Intelligence: An Analysis of Human and GPT-Generated Academic Writing

As the boundaries between human and machine-authored academic texts continue to blur, there is a compelling need to evaluate how artificial intelligence (AI)-generated content navigates the landscape of interpersonal engagement and persuasion in research articles. This investigation examined the functional role of interactional metadiscourse markers in academic texts authored by humans versus those generated by ChatGPT. Relying on Hyland's (۲۰۰۵) metadiscourse framework, this research study compared the frequency, rhetorical function, and contextual use of five key interactional markers in the conclusion and discussion sections of ۴۰ applied linguistics research articles. Results from a mixed-methods corpus-based discourse analysis revealed significant differences in rhetorical style and communicative intent. The human-authored texts employed more hedges, engagement markers, and attitude markers, indicating a more cautious, evaluative, and reader-aware stance. In contrast, the AI-generated texts overused self-mentions and lacked contextual subtlety, often resulting in a more expository and impersonal tone. These findings highlight AI's current limitations in replicating the rhetorical sensitivity required in academic writing and suggest the continued need for critical human oversight in AI-assisted scholarly communication.

کلیدواژه ها:

AI-generated text ، human-authored texts ، ChatGPT ، metadiscourse