What are the indications in D2L quizzes that AI / WA has been used?
I have quizzes in which students have clearly used AI to populate answers:
-when I 'copy-paste' from D2L into WORD, there is a black background behind the text: is this an indication of CHAPT-GPT/ AI?
-I see the same answer many times, and for some reason, the font is in RED: is this copy paste from WhatsApp?
What other indications are there?
Thanks
Answers
-
Hi Philip!
We cannot confirm if an answer is copied from a particular app/software as the black background is very common today for accessibility.
I would recommend using Respondus Lockdown Browser for quizzes and TurnItIn for assignments.
I am sorry that there is no easy answer to this.
Best,
-
Hi Philip,
Thank you for asking this question. I am not aware of indicators of learners using AI in their answers. You might follow Pedoro's instructions to avoid CHAPT-GPT/ AI copy and paste. -
Hi Philip,
I was going through the press release and noticed the following. I hope this will help you in the near future.D2L Announces New Partnership with Copyleaks to Better Detect Plagiarism and AI-Generated Content
Thank you,
Furkan
-
I know that Turnitin's AI detection has a minimum word count of 300 words, they feel that below this the accuracy is too low and it's likely to be throwing up false positives.
I suspect that for many use cases of the quiz tool written response question type, the responses are likely to be below this threshold.
I think ultimately it comes down to writing questions that AI tools can't answer rather than getting into this arms race.
-
Thanks all;
I agree that it's time to re-think assessment techniques.
but in case you have not seen it yet, here is what I was referring to:
this is AI copy/paste yet there is no mechanism to prove it…
-
I suppose the argument is that there is no way to detect standard copy/paste plagiarism from published sources in a quiz either, so if that is a concern you might be better using a different approach.
Turnitin's detection wouldn't do anything with that passage because it is too short.
My experience is that the tools that are available won't "prove it" in any case, there are so many false positives with AI detection that you are always going to want some additional evidence before making any kind of allegation, and on such short pieces of writing I struggle for what other evidence you might have.