Options
Can OpenAI write as rhetorically as humans do? a discourse study of human- and ChatGPT-generated research article abstracts in applied linguistics
Date Issued
2024
Citation
Kwan, B. S. C., Chan, T. H. T., Ko, Belinda W. S., & Kwan, C. Y. H. (18 Oct 2024). Can OpenAI write as rhetorically as humans do? a discourse study of human- and ChatGPT-generated research article abstracts in applied linguistics. LSP & Professional Communication Association Webinar 2024, Online.
Type
Conference Paper
Abstract
Since its release by OpenAI in 2022, ChatGPT has shaken up the entire world. With claims that it can replace humans in performing various writing tasks, a new line of research has emerged to examine the extent to which ChatGPT can generate human-like research articles. Results of the studies have so far been mixed, with some pointing to high degrees of similarity while some showing notable differences. The inconclusive findings are attributable to the various evaluation methods employed, some of which involve human-judgement that relies on expert intuition while some adopt detection methods that untrained automatic or machine-learned algorithms with parameters focusing on different linguistic features. The various methods have their own strengths and pitfalls, but one common issue that stands out is that their evaluation parameters rarely reflect scholarship in the field of English for Research and Publication Purposes (ERPP) that forms the bedrock of current research writing instruction. Hoping to address this theoretical lacuna in the existing studies and exploring what the field of ERPP can offer, we employed a discourse analytic approach to evaluate ChatGPT-generated article abstracts by tapping into theory of rhetoric in research communication and existing empirical findings about the rhetorical move structures of human-generated abstracts. Our data comprise abstracts of articles published in applied linguistics journals and those generated by ChatGPT. The results reveal noticeable differences between the two corpora. In this webinar, we will elaborate some of the observations and draw implications for future evaluation studies. We will also share thoughts for critical instruction in ERPP in contexts where students are allowed to use ChatGPT.
Availability at HKSYU Library
