As the debate rages over ChatGPT and its potential use by students to answer all kinds of academic questions; Is there an algorithmic way to identify an answer generated by ChatGPT.
Personally I would not mind a student using ChatGPT as a springboard to enhance their own ideas, However it might be a nice way to distinguish the user who solely used ChatGPT to perform their work vis a vis those students who used their own original thought or supplanted their own ideas and content over it.
It’s kind of ironic really. These tools that are being released to check if it’s AI-generated are the tools that students are using to cheat better.
The best way? Know your students. Their writing style would noticeably change if they completely relied on GPT. Could start the semester with some hand-written essays to identify them.
You could also take advantage of its weak points to lure it out. For example, using newer information that it doesn’t know. Or finding a question that when asked, produces a hallucinated answer.
Thanks @RonaldGRuckus. Some very practical ideas there.