Turnitin should be required to share GPT-4-updated, peer-reviewed studies of the accuracy of their AI text detection tool, including the false positive rate, before releasing it. Unacceptable that we are supposed to take their word that it has fewer than 1% false positives.
Can someone please explain to me how Turnitin has a 1% false positive rate when OpenAI itself can only manage 30%? It's clear to me that Turnitin is very wrong, and with a tool like theirs, wrong means harm to students. Since it is a black box, how will instructors know better?