5
$\begingroup$

Are we allowed to answer by asking ChatGPT? even if given credit, it can lead to partial answers but without any citations see https://hsm.stackexchange.com/a/15268/6609

Edit: the example above has been improved greatly, but the question still stands to what degree or under what conditions a response by chatGPT should be valid

$\endgroup$

2 Answers 2

8
$\begingroup$

We have no way to tell if the author of a well-written and well-documented answer might have used ChatGPT, but ChatGPT itself is not yet a trustworthy source.

I consider any ChatGPT answer without sources to be a "check my work" answer that I believe should not be allowed, since we'd have to reverse engineer its sources to trust it. The answer you mention may be broadly correct, but it is far, far too long-winded and has at least one obvious error: Calder was not a physicist.

$\endgroup$
1
  • 1
    $\begingroup$ I edited (and shortened) the answer to focus on the sources that I did find and confirmed. There were a couple of strange claims in the ChatGPT-generated text, among them the claim that Isamu Noguchi created a sculpture "Atomo" in 1937, perhaps conflating the Atlas stature at the Rockefeller Center (1937, Lee Lawrie) with Noguchi's work on Hiroshima memorial designs. $\endgroup$ Apr 1, 2023 at 2:56
1
$\begingroup$

This raises the question of what we should do if we find such an answer. Let us suppose that it is honest about the software it used and that it is broadly correct.

If we delete it then the original question may now go unanswered. That sees in conflict with the goal of this network to build a corpus of curated answers to interesting questions. What happens if someone with the privilege to see the deleted answer proceeds to write a human answer using the clues in the AI answer?

I think we need to have an answer to these dilemmas before we decide what to do.

$\endgroup$
1
  • 2
    $\begingroup$ Is it not sensible that the verification lies in the person that answer not on the author of the question? I guess that if we indeed allow for answers that include chatGPT generated text as acknowledged by the answering user, it should be to the answering user to provide additional references that verify what chatGPT has said. $\endgroup$
    – Mauricio
    Apr 2, 2023 at 16:15

You must log in to answer this question.

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