Policy on ChatGPT generated posts

Question

A new Internet phenomenon has reached our own corner of the Stack Exchange network. Yesterday, we had for the first time multiple posts of which we very much suspect they have been generated by ChatGPT.

The network has been plagued by this kind of answers since the tool came out last year. They tend to be at the same time convincingly written and of dubious veracity. That is, it is a coin toss as to whether they are right or wrong, but they come across as written by somebody who knows what they're talking about.

The main issues that stem from these answers are:

• reduction of content quality (due to many wrong and/or vacuous answers)
• interference with the vote usefulness (due to many misguided upvotes for impressively written wrong answers)
• rep farming (due to the same upvotes and the ability to post large volumes of answers)
• unmanageable increase in moderation load (only on SO so far)
• copyright and plagiarism (the situation is quite muddy there)

Stackexchange-the-company has decided to leave it to each site to decide how to deal with these answers. Many sites, notably Stackoverflow itself, have decided to ban ChatGPT generated answers, but this is not a default policy network-wide.

What should we do with these posts, as a community?

Answer 1

I suggest to follow the policy of Stack Overflow in this case:
Temporary policy: ChatGPT is banned - Meta Stack Overflow

The reasons are well articulated on the page above, and on the pages linked to it. Additionally, most of us strive for accuracy of the parental advice on this site. I would not trust ChatGPT to give me accurate parental advice, since it so often gives correct-sounding information that is actually very, very wrong. Our kids are just too precious to hand them off to ChatGPT - at least at present.

Answer 2

(TL;DR at the end)

I'm not someone who frequents the site a lot, but I ended up here after seeing this answer, and wondering why it had been deleted. (I still don't know for sure, but it certainly does look a lot like a ChatGPT-generated answer.)

I actually thought it was a decent answer, and was at a loss to determine why it had been deleted. (Hence the investigation, and ending up here.)

I know there can definitely be concerns about the quality of content generated by ChatGPT, but there are also concerns about the quality of content generated by human users of the site. A lot of human-written posts are downvoted or deleted because they are wrong/vacuous, or because of rep farming.

I used to be a moderator on Puzzling.SE, and we had a similar (though not identical) problem where people would post puzzles they found online as a way to attempt to quickly gain recognition. There, we adopted a policy that any puzzle that was not original needed to be attributed, and puzzles from sources that did not allow copying or reposting would be deleted.

When a puzzle was suspected of being non-original, the user would be asked to provide attribution, or to assert that it was in fact original. If they did not, the post would be deleted.

Obviously, those not complying with the requirements, or repeatedly posting disallowed content after being warned, would be suspended.

I think a similar policy would be useful here. Require any ChatGPT-generated posts to be clearly labelled as such ("this answer was generated by ChatGPT" or something along those lines). If an answer is suspected of not being original, ask the user whether it was written by ChatGPT, and if so, to add a label.

There could be a policy that a user can only post x ChatGPT-generated responses per day/week/month/whatever to prevent users from rep-farming.

I think this addresses all the points mentioned above except the increase in moderation load, but I think that is unavoidable no matter what solution is chosen. The moderators will have to deal with ChatGPT-written posts no matter what policy is adopted. I suppose if it becomes overwhelming, additional moderators could be added.

TL;DR: Many AI-generated answers can be helpful (and many human-generated ones are not), so why not let them be "moderated" by voter opinion? But they should be clearly labelled as AI-generated, and users attempting to pass them off as original should be dealt with appropriately.

Answer 3

After posting my other answer, I decided to ask ChatGPT what it thought about this issue (using a slightly-modified version of the question above, to provide it with the appropriate context):

As an AI language model, I strive to provide objective and informative responses to your question. However, the decision to allow or disallow AI-generated answers on an internet question and answer site is ultimately up to the site's management and community.

That being said, it's important to consider the potential issues that stem from allowing large volumes of AI-generated answers, as you have outlined. Here are some potential solutions that could help mitigate these issues:

1. Implement stricter moderation: This could involve increasing the number of moderators on the site or using automated tools to flag potentially low-quality or plagiarized content. This would help to reduce the burden on moderators and improve the overall quality of the site's content.

2. Require verification of sources: To reduce the risk of plagiarism, the site could require AI-generated answers to cite sources or provide evidence to support their claims. This would also help to improve the overall quality of the content.

3. Limit the number of AI-generated answers: The site could limit the number of AI-generated answers allowed per user, to prevent rep farming and reduce the burden on moderation.

4. Use AI-generated answers in combination with human answers: The site could use AI-generated answers as a starting point, but require human moderators to review and edit them before they are posted. This would help to ensure that the answers are accurate and informative, while still taking advantage of the efficiency and speed of AI-generated content.

Overall, it's important to find a balance between allowing AI-generated answers and maintaining the quality and integrity of the site's content. By implementing some of the solutions outlined above, the site can ensure that AI-generated answers are a valuable addition to the site, rather than a source of issues.

Its answer doesn't seem too far off from my own. (?!)