4 ms·

I really think we need a standard for citing AI generated/assisted content so people can use it in their work and call out that pieces of the work may be AI gen

I really think we need a standard for citing AI generated/assisted content so people can use it in their work and call out that pieces of the work may be AI generated. It would be something of a gentleman’s agreement but it would make its use more transparent in academia and maybe even slow the current and ongoing onslaught of indeterminately generated content on the web. I also feel like any prof claiming cheating based solely on an AI or AI detection model needs to be hauled in front of an academic review board on the same criteria they are judging their students. There is no excuse for this level of uncritical thinking at the University level.

conductr 3y ago
Given misinformation provided in some GPT responses, what exactly are you citing? I think the method would be to use GPT for a rough draft, then do fact checking and/or find sources for the claims the GPT was using. In essence, it's a research tool. It seems like if people are just copy/pasting in results they aren't trying at all and very likely didn't even try to comprehend the outputs. It should be considered cheating to do that. At this point, cheating is so pervasive (even before ChatGPT), I feel like they need to change the testing approach; no specific ideas or solutions but seems like bringing back in person hand written testing would work for a lot of cases and subjects (not all).
- flatline 3y ago
  It would be plagiarism to copy much of anything from someone else’s original work without citation. Even using GPT as a drafting tool will result in a lot of AI content making it into the final work verbatim. I think citation would legitimize this process, to an extent. Right now I feel like using the tool at all would count as “cheating” in many people’s minds, and I share some of those feelings if there is no disclosure. I also think LLMs are a legitimate tool for many purposes and their widespread use is inevitable.

conductr 3y ago

Given misinformation provided in some GPT responses, what exactly are you citing? I think the method would be to use GPT for a rough draft, then do fact checking and/or find sources for the claims the GPT was using. In essence, it's a research tool. It seems like if people are just copy/pasting in results they aren't trying at all and very likely didn't even try to comprehend the outputs. It should be considered cheating to do that. At this point, cheating is so pervasive (even before ChatGPT), I feel like they need to change the testing approach; no specific ideas or solutions but seems like bringing back in person hand written testing would work for a lot of cases and subjects (not all).

flatline 3y ago
It would be plagiarism to copy much of anything from someone else’s original work without citation. Even using GPT as a drafting tool will result in a lot of AI content making it into the final work verbatim. I think citation would legitimize this process, to an extent. Right now I feel like using the tool at all would count as “cheating” in many people’s minds, and I share some of those feelings if there is no disclosure. I also think LLMs are a legitimate tool for many purposes and their widespread use is inevitable.