AI Text Detection and How to Better Address the Issue
(Article Updated: April 2024)
Text detection will be an ongoing “cat and mouse” game of trying to identify AI text as evasion techniques continue to develop, yet there are better ways to address the issue. Although many detection techniques are available (as well as planned AI text watermarking), there are also many ways to evade these detectors. It is important that we show the importance of academic integrity by pursuing these types of detection technologies while at the same time focusing more on knowing our specific students, instructional processes (learning activities), and assessment types/procedures.
AI text detection, with an accuracy of between 65% and 90% (Singh, 2023), continues to develop and improve using a variety of detection techniques as well as AI systems themselves including AI watermarks to improve later AI text detection. “Watermarking text in ChatGPT involves cryptography in the form of embedding a pattern of words, letters and punctuation in the form of a secret code” (Monti, 2022, para. 8). However, an MIT Technology Review article expressed that watermarking is not living up to expectations, “Watermarks for AI-generated text are easy to remove and can be stolen and copied, rendering them useless, researchers have found. They say these kinds of attacks discredit watermarks and can fool people into…
