Centre for Teaching and Learning

Something borrowed, something true? Detecting and addressing ChatGPT-generated texts

Africa/Johannesburg
Hybrid
Description

While ChatGPT is still evolving, the way it writes can still be discerned from human writing (if one looks closely). Here at the Write Site, where we deal with both types of texts, we have noticed tell-tale signs that indicate when a student has over-relied on the system to produce their text. These signs include flawless grammar, longer sentences combining ideas, the absence of a personal voice or writing style, errors related to recent events, questionable citations, and an idiosyncratic structure. Therefore, during this session the importance of detecting AI-generated texts in the context of the University will be highlited. Subsequently, the distinctive characteristics of texts generated by ChatGPT will be explored. Finally, approaches to address AI generated texts submitted as student texts will be discussed.

Organized by

Centre for Teaching and Learning

Academic Staff and Leadership Development