Many years ago, Richard Pipes, a librarian at Wingate University revealed his secret to successfully in tracking down the sources used by students who plagiarize. He went to the most obvious source, he explained, because if a student were willing to put the effort to find an obscure source, they would be willing to do the hard work of writing a paper.
That was a different age, of course. The hard copy books and periodicals on the library shelves were still as equally accessible — if not still more accessible — to students as those found on the internet.
Nevertheless, his logic still holds true. A student who is actively trying to plagiarize their way out of an assignment is different from one who does not understand how and when to cite a source -- whether that confusion arises from poor preparation in their prior education or is, in part, culturally determined.
Right now, I want to set aside those students who want to do it correctly (or are at least willing to do it correctly).
Right now, I want us to consider those who are plagiarizing intentionally with malice aforethought.
Catching these students has always been and will always be a cat and mouse game. It is only when confronted in this light that practical approaches can be considered. For several years now, plagiarism detection tools have made the task of documenting their efforts easier.
But plagiarism detection services have always been hit-or-miss at best and actively problematic at worst and things have not improved with the arrival of large language models.
For those who are hoping Turnitin will save you from the threat of an AI generated paper, please know that they will almost certainly be one generation behind. At the time of writing, this means Turnitin believes it can identify work generated by ChatGPT 3.5 but is less certain it can detect work generated by ChatGPT 4.0.
It is possible for Turnitin to catch cases where ChatGPT 4.0 has been used but it comes at a cost. It increases the odds of generating false positives.
Turnitin makes a point of talking about this risk on their pages devoted to AI and what they have written there is worth reading for those trying to wrap their heads around our new normal.
I would stress one point in closing, though -- something those who are looking to the hills for Turnitin or something similar to arrive and solve your problems.
You are choosing to trust an AI with your work instead of engaging in the hard work of adjusting your pedagogy.
That formulation should give you some pause.