AI cheating detectors are wrong too often, US schools worry

A student at Adelphi University in the US was wrongly accused of using AI to write a paper by Turnitin's detection tool, which said it was 100% AI-generated. He won a court case in January, proving his innocence. The court said the school didn't follow its own rules or give him a fair chance to appeal.

Studies show 32% of students in the UK admit to using AI for homework without permission. Since ChatGPT came out, many schools worldwide started using tools like GPTZero and Turnitin to catch AI writing. But these tools often make mistakes, especially with students who don't speak English as their first language, according to a 2023 Stanford study.

Several top schools, including Yale and Vanderbilt, have stopped using AI detectors. Experts say these tools should only be used as a hint, not as proof. Turnitin itself says its tool is just a starting point for investigation.

Many educators now think schools should redesign how they test students, instead of relying on detectors. A vice-chancellor at Queen's University Belfast said the real question is what skills schools want to evaluate, not how to stop AI use. Student groups worry that detection systems are like "black boxes" that aren't fair or clear.