I asked a question about distance metrics for DAGs: Distance measure for directed acyclic graphs

I expected an answer which would outline what possible measures there are, so that I could choose which metric is most suitable for my research (which is unfortunately confidential, so I couldn't share all details which I would like).

Instead, I got patronising replies of the kind "it seems to be an XY type of question" (implying that I don't know what I want - untrue) and my question was put "on hold" (which feels like being singled out for some punishment / criticism). I don't think it's a nice way to treat a new member of the community. I added as many details as I could, but it was a bit strange observing how 80% of the responder's effort went into complaining about lack of details instead of trying offer at least a modicum of help.

I was answering questions on other SE sites before, and I never treated askers like that. If I didn't know the details, I responded as best as I could and asked for clarification, respecting that I may not get as many details as I would like. My questions is for research purposes, hence I am unclear on what I will do with the distance measure - I want to know what I could do. It's a bit of a shame that people here prefer to be hostile towards a question is even slightly open-ended.

  • 2
    $\begingroup$ For reference: if I read this right, we routinely close between 3% and 10% of all questions. And that's probably underestimating it, because some closed questions get deleted (immediately by the asker, or later by cleanup tasks). So no, you've not been singled out. Policy was enacted, but unfortunately not communicated; you should have received this or this comment. $\endgroup$ – Raphael Nov 8 '17 at 18:47
  • $\begingroup$ Nobody here is hostile towards open-ended question; it's just that experience has shown they don't work well on this platform. Brainstorming with peers is great, but should be done elsewhere. $\endgroup$ – Raphael Nov 8 '17 at 18:48
  • $\begingroup$ I wasn't brainstorming. I asked about a question which - probably - would require the respondent to at least list a few methods. Such questions are routinely asked on other SE sites and people are not getting worked up about listing a number of methods, see e.g. this question and excellent answers provided to it: math.stackexchange.com/questions/604654/estimating-the-entropy $\endgroup$ – quant_dev Nov 8 '17 at 22:37
  • $\begingroup$ @quant_dev, I don't see the question you link to as comparable. Listing all the reasons why would take more space than is available in a comment, but the short version is that answers to that question can be objectively evaluated, whereas it's less clear to me how to do that for your question. In any case, I'd rather stick to your question, and to this site, and our policies/expectations on this site. $\endgroup$ – D.W. Nov 9 '17 at 0:09
  • $\begingroup$ What makes CS Stack Exchange so special? $\endgroup$ – quant_dev Nov 9 '17 at 9:04
  • $\begingroup$ @quant_dev it's not really "what makes CS Stack Exchange so special", but while all SE sites have a general rule/policy (e.g. "Be Nice"), every site also has different site cultures that may modify the policy/behavior on their site. $\endgroup$ – Andrew T. Nov 22 '17 at 5:06

Thanks for coming here to raise the issue, share your perspective, and ask how to improve your question. I'll share my personal perspective (with the understanding that others might not agree with it).

I'm inclined to agree with closure of that question in its current form. I'd prefer to see this site for focused questions that admit an objective answer -- or at least where it is possible to objectively evaluate proposed answers -- as I think that's where the site format works best. This means that the question needs to define the requirements or criteria that should be used for evaluating answers. Answerers need that information so they can tell what would count as a satisfactory answer for you, and voters need that so they can tell which answers to upvote.

Right now the question feels like it has the form "please give me a list of possible distance metrics; I'm not telling you how I will evaluate them, but please give me a list so that I can choose one, using criteria that I haven't listed in the question". I don't think that's a good fit here. We want you to think through what your requirements or criteria are before asking and articulate them in the question. Sometimes a helpful way to do that is to come up with some possible metrics on your own, see if you're happy with any of them, and if not, think about why you've rejected them and what requirements/criteria that implies you have.

There tends to be some skepticism about "list of X" questions, because they are so open-ended. In particular, there is an infinite number of possible distance metrics, with no clear way provided in the question to select among them. No answers could reasonably enumerate all possible distance metrics, and that's a hint that the question isn't focused enough. It's also unfair to answerers: it asks them to spend a lot of time typing out many candidate metrics, most of which you are going to reject. Since you're asking for help and time from others, I think it makes more sense for you to first figure out how you plan to choose among the candidates, and make that clear in the question.

I do see that you provide some criteria in the question: "easy to compute on graphs of size 10", "invariant to permutations of nodes". That's helpful. Unfortunately, as the comments indicate, these aren't enough to narrow things down sufficiently yet: there are still an unbounded number of distance metrics. As David Richerby explains, essentially everything is easy to compute when the graphs has size 10, so that doesn't really narrow things down. And "invariant with respect to relabelling of nodes" is in some sense part of the definition of a graph metric. It's good that you include it, but it also doesn't narrow things down to a small list of possible answers.

So, this doesn't mean that your question can't work here. It just means we need more information before the question is answerable. It means that if you'd like the question to be open here, you should edit the question to clarify. What do you want the distance metric to measure? How will you evaluate a proposal?

Please don't view the responses as hostility. I don't see it as being hostile or aggressive towards you, personally. Rather, it's addressing the substance of your question: it looks to me like people are trying to keep question quality high and make sure questions are suitable for our site's format, and gather the information they'd need to give a good answer.

If you do see comments that violate our be nice policy, please flag them for moderator attention. Don't engage -- just flag them and move on.

  • $\begingroup$ "Right now the question feels like it has the form "please give me a list of possible distance metrics; I'm not telling you how I will evaluate them, but please give me a list so that I can choose one, using criteria that I haven't listed in the question"." I've added as much criteria as I could (see the updated question). It is false to claim that I didn't want to provide the details. $\endgroup$ – quant_dev Nov 8 '17 at 22:38
  • 1
    $\begingroup$ @quant_dev, I never claimed you don't want to provide details. What I wrote is that, in my opinion, the question does not currently have enough details to be answerable. I did read the updated question before I wrote my answer (my answer was written after you updated your question), and I stand by everything I wrote here. $\endgroup$ – D.W. Nov 9 '17 at 0:02
  • $\begingroup$ Even if you were right, did this justify the patronising tone of "your question seems to be of the XY type" -- i.e. "you don't know what you want, let the adults help"? $\endgroup$ – quant_dev Nov 9 '17 at 9:04
  • 1
    $\begingroup$ @quant_dev, I think there's another interpretation of that comment that isn't patronising. I read it as: based on what you've told us, it feels like there might be a better way to achieve what you're ultimately trying to do without defining a graph metric, so more background might help us give you other answers that would meet your needs without providing a graph metric. I don't know whether I agree (I didn't have that reaction myself), but that's another way you could read that comment. $\endgroup$ – D.W. Nov 9 '17 at 17:25

You must log in to answer this question.

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