I reread two Turnitin help pages this morning. Neither of them is written for students. Both are written for the instructor. As rendered plain text, with the site nav and footer counted in, they are 2,081 and 1,824 characters. Under four thousand for the pair.
We import Turnitin AI writing reports into the tool I work on, so I read that help centre more than is healthy. The habit it left me with is one plenty of developers already have with APIs: when something confusing happens, go read the docs for the other role. Admin docs, not user docs. The part that decides what happens to you is usually in the version nobody sent you.
Almost everything written about a high AI writing percentage is aimed at the student. Two of the nine articles listed under AI writing in Turnitin's own guides are about the conversation instead of the report: "What should I do if the AI writing percentage is high?" and "What questions do I ask my student if AI is detected?"
Three things in there would change what you would bring to that meeting.
The first instruction to the instructor has nothing to do with the score. It is to "refer to your institution's policy on plagiarism and the use of generative AI in the classroom," and only then to work out a plan for the classroom. So the document that governs the outcome is not on Turnitin's site at all. It is your school's, and two schools can treat an identical number very differently.
Then there is the framing sentence, which is worth reading verbatim:
"Rather than viewing the results of our AI writing detection as a punitive measure, the overarching purpose of the tool is to facilitate student conversations and interventions on the use of AI writing tools that may not otherwise occur, and use it as one strategy in your toolkit."
The same page puts the decision somewhere explicit. The final call on whether misconduct occurred "rests with the reviewer/instructor," and Turnitin "does not make a determination of misconduct, rather it provides data."
The second page is the concrete one. It gives instructors four "conversation jump starters" and tells them to personalise each one. Two of the four:
"Explain your process for completing this assignment. Let's look at areas that you are proud of and areas that you think could use some extra improvement."
"Let's examine your Turnitin AI writing detection report together. Let's look at the highlighted areas and discuss your choices."
Count what is missing. Not one of the four asks whether you used AI. Two ask how you worked. One proposes opening the report side by side and walking through the marked sections.
So the useful preparation is an account of how the file got made. Not a rebuttal of a number. Which parts you drafted first, what you threw out, which sources were open while you wrote, and version history with dates on it if you have any. Read the report before you are in the room, too. The second line above assumes you both have it open and are talking about particular highlighted sections rather than the total.
Two limits, because it would be silly to read too much into a couple of help-centre pages.
Nothing on those pages binds your institution. Your instructor may not know they exist, and plenty of integrity processes are formal from the first email onward. Reciting the vendor's own guidance in the room is also a weak move, since it argues about how the meeting should run when the meeting is about your writing.
The bit I keep coming back to is the unit. "Let's look at the highlighted areas and discuss your choices." Not the percentage. In the conversation Turnitin describes, the thing being discussed is a passage, and a passage is something a person can actually account for.
That is the scope rule we ended up building into HumanPen. Upload the file and the report together, it locates the flagged text inside your document, and the rewrite stops there. Anything the report did not mark is left untouched. That part is for afterwards, though. The conversation is yours.
Originally published at HumanPen: https://humanpen.net/en/blog/the-conversation-after-an-ai-flag
Top comments (0)