Plagiarism and ChatGPT

How do we deal with this emerging technology? Short answer questions are far too easy to simulate. For example, “What are the elements of X” or “Describe Y concept.” A four-page fact pattern, followed by specific prompts, may also be hard to jam into ChatGPT. I think we need to think long and hard about take-home exams. It is too easy for students to use ChatGPT, over and over again, to mix and match answers. Also, any in-class exam should eliminate access to devices only paper sources. (That is my usual policy.) Finally, we should give some serious thought to oral examinations, which cannot be hacked.

Moreover, universities should revisit plagiarism policies in light of ChatGPT. There should be explicit language that using these tools is a violation of academic integrity standards. I imagine some policies may be framed in terms of getting help from “another” person or something to that effect. ChatGPT is not a person not yet at least. Students will argue that ChatGPT does not fall in the plain language of a policy designed to prohibit sentient-cheating. And the burden of proof to determine plagiarism may be shifted, since traditional tools are not effective. There is a real/fake detector, that uses the ChatGPT engine, but I haven’t tested how accurate it is.

via reason.com

Josh Blackman.

I would say just go back to old school exams. If you insist on open book essay exams, eliminate access to the internet, a wise precaution anyway. Take home exams have always involved a lot of cheating. I believe there have been studies that show this.

What I would like to see is an AI tool that can grade essay exams, or write decent multiple choice questions. I’ve experimented with some ChatGPT MPCs. It is pretty lame at question drafting so far, but I expect that to get better. Grading essay exams will presumably take longer, but that is coming as well.