10
$\begingroup$

We've had our first batch of ChatGPT answers and, no surprise, they were blatantly incorrect.

Stack Overflow has recently created a policy banning answers generated by AI. This policy applies specifically to SO, but other sites on the SE network have created their own policies based on the community's response on meta. The company has left it up to us to create our own policies.

The policy I propose is one which is in effect on SO and on multiple other SE sites: Users caught posting AI-generated content will have the answers deleted and a 7-day suspension will be issued.

What should our policy be? Should we categorically disallow AI-generated answers?

$\endgroup$
6
  • $\begingroup$ Feels like an obvious choice; are there any counterpoints that I’m missing? $\endgroup$
    – Daniel S
    Commented Dec 18, 2022 at 9:49
  • $\begingroup$ Posting AI-generated content is just another form of spamming. Usually when we flag something as spam, we consider it as not disclosing affiliation or conflict of interest; but in this case, it's purposed to gain unjustfully. The proposed policy is justified I think. $\endgroup$
    – DannyNiu
    Commented Dec 18, 2022 at 12:11
  • $\begingroup$ @DanielS I've put two community answers down below, so you can now more clearly vote on the two possible ways of handling this (or add a third one, but I think these are pretty clear cut, just like the question). DannyNiu, for you too of course :) $\endgroup$
    – Maarten Bodewes Mod
    Commented Dec 18, 2022 at 19:03
  • $\begingroup$ By the way, here's the Q/A on the network wide Meta. As you might expect, there doesn't seem to be all that much tolerance for ChatGPT (in case you didn't want to click). $\endgroup$
    – Maarten Bodewes Mod
    Commented Dec 18, 2022 at 21:16
  • $\begingroup$ @DannyNiu There are no counterpoints that I am aware of. I'm just following the lead of other sites who are using their own metas to solicit input from the community before setting up and enforcing policies. $\endgroup$
    – forest Mod
    Commented Dec 18, 2022 at 23:41
  • $\begingroup$ I've by now used ChatGPT a lot. And I must say it is one of the most helpful tools that I've seen. The problem is that it can state incorrect answers as if they have been chiseled in stone without any indication of doubt. Actually, that's the one thing that irritates me most of the system. Unadulterated answers should not be retrieved from this and if any information is generated by it it should be annotated as such. $\endgroup$
    – Maarten Bodewes Mod
    Commented Dec 29, 2022 at 11:55

3 Answers 3

Reset to default
15
$\begingroup$

I fully agree that ChatGPT answers cannot be allowed and should therefore be removed immediately when found.

These kind of answers will take away valuable time from both users and mods. As these answers cannot have been posted in good faith either they should be followed up by a 7 day suspension.

If after the 7 days suspension another of such answers is posted then the account should be deleted as the account is only used to post "spam". This will make sure that the account will not take up more time from the users, mods and cause unnecessary harm due for posting subtly wrong answers.

An exception would be the case where there is a clear indication of positive contribution. In that case another suspension may be required at the specific discretion of the moderators.

$\endgroup$
7
$\begingroup$

If that's their only 'contribution' don't waste time with a week - just hit them with a year. A lot of these are throwaway accounts, so no loss and frankly the amount of time these posts wastes means they are vandals and trolls.

$\endgroup$
-4
$\begingroup$

Although ChatGPT answers are disallowed on SO, there is no need to follow the process of SO. Answers, including those largely generated by ChatGPT, should be considered on their own merit. As such, they should be downvoted if they are wrong and upvoted when they are correct. They should only be deleted when they are clearly not answering the question. StackExchange already has a way of blocking posts by a user if too many bad answers are given.

$\endgroup$
0

You must log in to answer this question.

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