4

Today, several answers were flagged for mod attention on Web Applications SE as being generated by GPT / ChatGPT. Due to the date that these posts were posted, it's fair to assume that they were generated by ChatGPT.

Contrary to what happened on other sites, most of the posts flagged today were not completely bad. Some of them failed to properly address the use case explained by the OP, and they don’t mention the use of ChatGPT. Without any attribution to the tool used to generate them, they count as plagiarism under Stack Exchange and OpenAI guidelines.

Should we ban GPT / ChatGPT answers outright on Web Applications SE?

Other sites in the Stack Exchange network have banned its use. The most notorious case is Stack Overflow, as it added a help centre article describing its ban on GPT and ChatGPT-generated answers. 

Most of the other SE sites have discussed this: Is there a list of ChatGPT or other AI-related discussions and policies for our sites?

In the meantime, GPT / ChatGPT answers, as with any other post, should give appropriate attribution. Ref. How to reference material written by others.

2
  • 3
    Wonder if webapps change too much for chatgpt to keep up. The problem I've seen with ChatGPT is it can spout 'answers' that seem convincingly true when actually not, and the lack of SMEs can result in disseminating wrong infromation
    – Sathyajith Bhat
    Commented Jan 27, 2023 at 9:32
  • I think that most of the major webapps doesn't make radical changes with few exceptions (Twitter?). Our current top tag, google-sheets, the last year introduced several features but most of the features are still working the same way. Also I had some conversations with ChatGPT about Google Sheets macros / macro recorder... very deceptive responses but have to say that I'm still a begginer "prompt designer". I don't think that webapps.se have being flooded with GPT generated answers... perhaps the most important ongoing issue might be the lack of SMEs.
    – Rubén - Volunteer Moderator - Mod
    Commented Jan 27, 2023 at 17:18

3 Answers 3

Reset to default
8

More than outright banning chat GPT, I would recommend that the policy should be that if ChatGPT (or any other source for that matter) is used, it should be checked by the author for accuracy before posting it here as fact. Also it should be credited, in the same way we would source a Wikipedia post

2
  • Thanks for your answer. ChatGPT can't be credited inthe sameway as a Wikipedia post as the response might be not the same over the time which might make imposible to verify the veracity of the reference. AFAIK there is no clarity about how generated text by GPT tools like ChatGPT should be credited. Here is a related Q/A from Writing Do I need to cite ChatGPT in published writing?
    – Rubén - Volunteer Moderator - Mod
    Commented Feb 3, 2023 at 22:27
  • 2
    ChatGPT can be credited, regardless of your ability to verify the attribution or not, and I believe our attribution rules already require that ChatGPT be credited. Whether it is exactly the same format as a Wikipedia article or in a different format seems secondary to me.
    – D.W.
    Commented Feb 7 at 17:39
4

The problem with ChatGPT is that it that it doesn't know when it has very little or no knowledge of the subject. It just works out what word is most likely to come next and as long as the sentence is grammatically correct it will output that sentence.

Therefore, without a subject matter expert checking the results there is no way for the average user - either as the person posting the answer or the person reading the answer - to be confident that the answer is correct without potentially doing a great deal of research.

While users should be verifying answers for themselves, they currently have the expectation that the answer has been provided in good faith and that the author believes what they've written to be true. Neither of which is necessarily true for ChatGPT generated answers.

Therefore, on balance, I'd say it's prudent to ban ChatGPT generated answers.

0
2

The content of a user’s post should be the main criterion for moderation, not how the user got their information.

Posts that plagiarize ChatGPT’s responses, that are simply copypasted without giving credit, should be disallowed and deleted. These posts are often useless, gibberish, nonsensical, and non-factual.

Users who consistently post such AI-generated content should face mod action, such as warnings or suspensions.

Quoting OpenAI, the developer of ChatGPT:

Limitations

  • ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging […]

We should allow it if the answerer specifies which parts of their answer came from an AI tool and states that they verified or tested it themselves, as the answer may contain correct and useful information if verified by a human. It would also be helpful if the answerer included the prompt they used so that others could verify the response. A good example is this answer: How can I word wrap code block output in ChatGPT? The answer is properly credited, not gibberish/nonsensical, and actually answers the question.

Other AI search engine or chatbot responses, e.g., those by Bing's AI chatbot, Bard by Google, you.com's YouChat, etc., should be treated similarly. Unlike ChatGPT, these AI-based search engines or chatbots may include sources in their responses and present information more like traditional search engines (i.e., more likely to present more factual content).

I meant this answer to be future-proof. As AI-based tools continue to improve and become more reliable, their usage is likely to become more prevalent, and they may eventually replace traditional search engines. It may not be beneficial to impose blanket bans on their use, especially since they seem to be improving at a fast rate.

1
  • The problem with this that it's not always obvious which elements of a ChatGPT answer have been 'generated' and which have just been copied wholesale from a specific source
    – Valorum
    Commented Feb 27 at 7:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .