The ubiquitous availability of AI has made plagiarism detection software utterly useless, argues our Hans de Zwart in the Volkskrant.
The old way of checking for plagiarism provided some form of certainty (it could show if some text already existed somewhere else). This certainty has disappeared with the comeuppance of AI. Turnitin, the market leader in originality detection, as the cop shit industry likes to call its product, is still pretending all is fine. This is them talking about their “AI detector”:
It is impossible for software to reliably predict whether a text was written by AI. OpenAI admits this (and so does Turnitin if you listen to them carefully).
Moreover, research at Stanford has shown that people who don’t write in their first language are more likely to be flagged falsely (being accused of having used AI, when they’ve written a text themselves) than native speakers.
Educational institutions can therefore no longer justify the use of this technology. Teachers will have to create new types of assignments and new ways of validly assessing the progress of their students.
See: Opinie: Leuk of niet, door de opkomst van AI zul je als docent je onderwijs moeten aanpassen at the Volkskrant. Can’t access the Volkskrant? Then read the piece on Hans de Zwart’s blog.