Li, H., Lee, S., & Botelho, A. F. (2024, July). This Paper Was Written with the Help of ChatGPT: Exploring the Consequences of AI-Driven Academic Writing on Scholarly Practices. Proceedings of the 17th International Conference on Educational Data Mining, 542--550, Atlanta, GA, USA. https://doi.org/10.5281/zenodo.12729880 [HTML] [bib] [pdf] [detail]
Learn about other projects on Computer-Supported Collaborative Learning (CSCL) that we presented at the same venue.
(4:00 pm - Collaborative Learning, Tuesday, July 16 - Room 236)
2024.EDM-short-papers.55.pdfHere's the full text of our study.
Learn more about the code and the dataset via our OSF (Open Science Framework) project.
Five AI content detectors were evaluated:
Key Points
Current AI content detectors are not accurate enough; estimates of these detectors are not proportionally aligned with the amount of AI-generated content in the text.
Educators could lead to overly loose or overly punitive consequences if the only detection scores were to be used to assess students’ writing.
