Since OpenAI’s ChatGPT has catapulted generative AI into practical use, 17% of students have used ChatGPT for an assignment.answerOn the other hand, teachers alsoAI detectorConflicts between students and teachers over AI are becoming more and more intense, with students and teachers trying to detect cheating using AI. When the overseas media Bloomberg tested mainstream AI detectors using human-written texts, they reported that false positives occurred at a rate that was not high but could not be ignored.
Do AI Detectors Work? Students Face False Cheating Accusations – Bloomberg
https://www.bloomberg.com/news/features/2024-10-18/do-ai-detectors-work-students-face-false-cheating-accusations
Below, Bloomberg lists 500 essays written before the release of ChatGPT on mainstream AI detectors.GPTZeroThis is a diagram showing the results of applying this to Copyleaks. As a result of the test, 3 out of 500 articles were determined to have been generated by AI, and 9 articles were incorrectly identified as being partially written using AI.
The essay used in the test was submitted to Texas A&M University in the summer of 2022, before ChatGPT was publicly available, so it is highly unlikely that it was generated by AI. Also, since it was obtained through a request for information, it has never been used for AI training.
Among the essays that were incorrectly detected as being output by AI, there were some that were determined to be almost 100% accurate.
Students with developmental disabilities who tend to use the same words and phrases frequently and students learning English as a Second Language (ESL) are particularly likely to be probed. is.
A study conducted by Stanford University found that while AI detectors had near-perfect accuracy on essays written by American-born eighth-graders, more than half of essays written by non-native English speakers were AI-generated. It was identified as text.
OpenAI was previously able to detect text written using ChatGPT with 99.9% accuracy, citing concerns that it could negatively impact certain groups, including ESL students.toolIt has been announced that the release of the movie will be postponed.
Moira Olmstead, who has autism spectrum disorder and tends to write formulaic texts that can easily be mistaken for AI-generated texts, is one person who has experienced this problem. After taking a leave of absence from university to give birth, Ms. Olmsted returned to school to balance childcare and her dream of becoming a teacher, but due to a misidentification by the AI detector, one of her compulsory subjects received a zero point. .
Mr. Olmsted said of his shock when his professor told him about checking the AI detection tool, “It was like being punched in the stomach.”
It is no longer clear how far AI has penetrated educational settings. When Bloomberg ran 305 essays submitted in the summer of 2023, after the release of ChatGPT, using the same AI detector as the above test, 9% were determined to be AI-generated.
Bloomberg’s review also found that AI detectors can sometimes be fooled by so-called “AI humanizers,” automated tools created to disguise AI-generated text as human text.
Specifically, when an essay that was misidentified as “AI with 98.1% accuracy” by GPTZero was corrected using an AI humanizer called Hix Bypass, the judgment result drastically decreased to 5.3%.
According to Bloomberg, this situation risks turning into an arms race between competing AI technologies, creating a deep rift between educators and students with little educational benefit.
Rather than banning the use of AI, some educators are looking for ways to encourage students to properly utilize AI.
Adam Lloyd, a professor of English at the University of Maryland, who grades assignments based on his own senses rather than AI detectors and talks directly with students when in doubt, told Bloomberg that “like it or not,” Regardless, artificial intelligence will be essential in the future, and it would be a mistake to think that AI should be kept out of the classroom or that students should be discouraged from using it.”
An attempt to incorporate “how to use and evaluate” AI into student education rather than banning it – GIGAZINE
Copy the title and URL of this article