No AI-generated posts detected

Up to 94% of submissions generated by artificial intelligence (AI) went undetected by teachers. Even when stricter criteria were applied and the word “AI” was specifically mentioned in the filtering process, 97% of AI submissions went undetected.

More and more homework and tests are being written by AI and students are submitting them to receive grades, credits and degrees - Photo: Pixabay

A student who used the most basic AI code without any modifications had an 83% chance of getting a higher grade than their classmates, compared to just a 6% chance of being caught if the teacher did not use AI detection software.

This is what a team of researchers from the UK has found. Teachers are completely unable or unable to detect academic products created by chatbots. A paper by Peter Scarfe and colleagues at the University of Reading looked at what happens when researchers create fake student profiles and submit AI-generated papers without teachers knowing.

Education is perhaps the most heavily and negatively impacted sector since ChatGPT went public. More and more homework and tests are being written by AI and then submitted by students to receive grades, credits, and degrees. According to Forbes , this is a serious problem because the lack of real-world knowledge of students can have serious consequences.

Preventing academic cheating with AI is not yet a priority for most schools or educational institutions. Some schools are even making it easier for AI to be used by not investing in AI detection technologies.

This isn’t the first time there have been warnings that humans can’t independently detect AI-generated writing. Last year, a study from the University of South Florida concluded that linguists couldn’t differentiate between AI-generated text and human-written text.

Another study found that AI detection tools were much more effective than teachers. The team found that the Turnitin AI detection system correctly identified 91% of papers as having AI-generated content, while teachers reported only 54.5% of papers as “potential instances of academic misconduct.”

The results of the study showed that when humans were not warned and AI detection tools were not used, almost all AI-generated articles went undetected. Worse, the study found that AI-generated articles scored higher than human-generated articles.

If a school or teacher does not have the support of detection technology, using AI to cheat is almost certain to improve student scores with little risk. Online courses are more vulnerable to cheating, especially AI cheating, because teachers do not know their students and have little visibility into their learning process.

Source: https://tuoitre.vn/khong-phat-hien-bai-viet-do-ai-tao-ra-20250111215437798.htm