2
$\begingroup$

More Specifically

Should a new question, $Q_2$, be posted when question $Q_1$ has already been answered and there is a distinctly new question that draws substantially on the content of $Q_1$, or should $Q_1$ be extended to keep the content together?

Most Recent of Several Similar Examples

In a recently active AI SE question, moderator BartoszKP commented.

SE format is not a discussion board. You should not keep updating the question, and thus mutating the topic. If you have another question, just post a follow-up as a separate Q&A.

The question author disagreed with that comment, stating that the content was closely related and should be kept together. BartoszKP further commented as follows.

Yes, the purpose of the site is to create a knowledge base for everyone. It should however be a knowledge base consisting of specific questions and specific answers, not chat history between you and the one who answers.

I edited the question referenced above to remove all the conversation with those that answered (which became a small project, much to my dismay), but the previous versions in combination with the discussion in comments displays much dialog and two distinct and significant extensions of the original question requirements.

Related Stack Exchange and Overflow Positions

I notice that the AI Tour sets forth some related principles.

This site is all about getting answers. It's not a discussion forum, ...

Use edits to fix mistakes, improve formatting, or clarify the meaning of a post.

Not Just The Example Above

Several Q&A scenarios in AI SE have gone through sequential changes in this way so that all those who answer had to choose between these two somewhat counterproductive options.

  1. Let their answer go stale in relation to a much extended question
  2. Be dragged along by those in the SO meta term a Help Vampire

Editing a question to ask additional questions renders all the answers stale and effectively drags all those who worked to answer your original question to keep their answers aligned.

Back to the This Main Question

Should a new question, $Q_2$, be posted when question $Q_1$ has already been answered and there is a distinctly new question that draws substantially on the content of $Q_1$, or should $Q_1$ be extended to keep the content together?

Some SO Meta Wisdom

It seems the SO wisdom1, 2 is that a new question $Q_2$ should be posted. That would also solve the issue of choosing between stale answers and help vampirism, but it may not be popular with a Q author and A author that are essentially collaborating on a small project together and are fully engaged in an extended helping process.

References

[1] https://meta.stackoverflow.com/questions/252113/how-to-deal-with-constant-changing-questions

[2] https://meta.stackoverflow.com/questions/300966/dealing-with-solutions-that-end-with-a-new-error

$\endgroup$
2
$\begingroup$

I don't see why AI should be different to SO in this respect. Updates to questions should be limited to clarifying, improving layout, spelling and grammar. They should not add new insights from or progress of the questioner, once answers to the original have been written.

In general, this process needs to be more friendly to question answerers that askers. It is more expedient to get the OP of the question to take the extra effort to frame their problem as multiple separate questions, instead of having volunteers answering questions track changes and try to follow a conversation (and in the meantime often dilute the purpose of the original question).

If someone tries to alter their question or ask lots of extensions in comments, then in my experience, a gentle/friendly push back and suggestion to ask a separate question is often all that is required. It is more helpful to show what the question OP should do as opposed to telling them that they are doing something wrong.

If the OP of the question ignores such a suggestion, then the best next step is to walk away. There is no point arguing with them if they think they know better how the site should work. Just let their extensions to the question go unanswered. If that is disappointing to you (because you found the question a really good fit to the site, and were excited to answer it), then perhaps help the OP further by opening the new question yourself and pointing them at it - although I personally would not go that far, there are always other good questions.

Regarding this scenario:

it may not be popular with a Q author and A author that are essentially collaborating on a small project together and are fully engaged in an extended helping process

It's not really what the site is for. I would either:

  • Downvote or close the question, if it was clearly too long/confusing and broad to meet site guidelines.

  • Ignore the question if it kept changing, as it would be a waste of time to get involved, and it is only one question. This is not common behaviour.

Given how little rep both the asker and answerer would get for their efforts, as the content becomes too dense for anyone else to work with, it is in some ways self-limiting. I note that as it stands today, the OP questioner got 10 rep in your linked question difficulty in understanding identifiability in Dueling Network paper and the four answers got a total of -2 rep between them. The resulting content is all but incomprehensible to me.

In this case I notice you are one of the affected answerers. I don't think there is much you can do at this stage but chalk it up to experience. There is no way to force collaborating site users to behave according to above - the only tools SE has for moderation at that level are too heavy-handed to apply when the discussion is still technical, on-topic and polite. If you spot the behaviour early enough you can comment that you prefer it another way (and IMO the "correct" Stack Exchange way would be separate questions as you suggest), but if the whole thing has momentum with updates to both questions and answers, just leave the others involved in their discussion.


Some clarifications might still make significant changes, if to answer a question accurately (as opposed to answers with general advice that might apply), the OP needs to add details about their specific situation, including code, data etc. Sometimes this unfortunately can invalidate answers that attack the question in a general sense. This is a tricky area to judge correctly. I think the line is most clearly drawn when the OP unilaterally adds new data, or is obviously changing their question in response to an answer which has already helped them.

$\endgroup$
1
  • $\begingroup$ I strongly agree with this. I remember one instance early on in my modding career where an OP asked an unclear question, and it received valid two answers that interpreted the question differently. I tried to address that by editing the original question which only opened up a can of worms and created more problems. What I concluded was that if it's not the OP providing the clarifications, it's problematic. $\endgroup$ – DukeZhou Oct 21 '18 at 21:56

You must log in to answer this question.

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