I'm an admin on SE:AI. We tend to get a lot of questions that are more properly SE:Data Science or SE:Cross Validated subjects.

Because it's difficult to stem the tide (Stack has a lot of hackers who don't read the instructions, and many seem to feel SE:AI is the place to ask these questions,) we're experimenting with accepting basic questions in these fields.

However, we want to (A) point OPs to the proper stacks for more advanced and follow-up questions, and (B) tag the SE:DS and SE:CV questions to facilitate contributors on these stacks willing to field basic questions on SE:AI.

Apologies if this question has already been asked, but:

  • How should I distinguish Cross Validated from Data Science?

I'm asking here because I want the SE:CV community's distinction. This is not my field, so I need guidance in how to properly tag on AI.

  • 9
    $\begingroup$ Each stack exchange site should be clearly defining its scope. I think ours is pretty clearly defined in our help center. I also think it's incumbent on later sites to clarify where they draw the line vs pre-existing sites with which they overlap, but that delineation doesn't make them off topic at the earlier site. $\endgroup$
    – Glen_b
    Jan 29 '18 at 0:07
  • $\begingroup$ As a mod, I hear complaints about posts that are incorrectly migrated, so we tend to be wary of the process where there is any uncertainty. There's no unilateral solution on AI's side so far as I can see. AI, Data Science and Cross Validated are distinct but inextricably related, so I think there needs to be communication between the communities. A Cross Validated tag is probably inappropriate, since we already have "machine learning", "statistical-ai" and "probabilistic" tags. But it would be good to have some guidance from the CV community on what post should be migrated, if any. $\endgroup$
    – DukeZhou
    Jan 29 '18 at 4:49
  • 15
    $\begingroup$ Frankly, there has been a lot of discussions about DataScience.SE on our Meta (see datascience-stackexchange tag) and there seems to be a strong consensus that the whole DataScience.SE site should not have been created in the first place; that it is overlapping with our site too strongly, and that this is only creating confusion: stats.meta.stackexchange.com/…. Many people hoped it would not graduate from beta, and when it eventually did, were baffled. So I'm not sure you will get useful advice. $\endgroup$
    – amoeba
    Jan 29 '18 at 8:03
  • 9
    $\begingroup$ I only partially agree with @amoeba. My current go to delineation is that DataScience stack exchange is a good place to ask about ML and statistical tools. So things like hadoop, spark, various implementations of neural networks (tensorflow, keras, torch, etc), are more on topic at DS stack exchange than here. While more conceptual, theory driven, or mathematical questions are more on topic here. I would suspect that opinion is not uniformly held though. $\endgroup$ Jan 29 '18 at 21:47
  • $\begingroup$ @amoeba that helps clarify the issue for sure! $\endgroup$
    – DukeZhou
    Jan 30 '18 at 23:39
  • 9
    $\begingroup$ I suppose we should not be surprised that the AI people want a decision rule on this issue. $\endgroup$
    – Placidia
    Jan 31 '18 at 21:10
  • 1
    $\begingroup$ @Placidia :) It's probably mainly just me, but we do have several users who are fond of commenting "This is a DS question" or "This is a CV question". At this point, I'm mainly trying to tag them so folks with the relevant knowledge who may want to answer can find them easily. Although we're taking basic implementation questions, because the tech has become so accessible, we'd certainly prefer the more advanced question to be re-routed to the more optimal stack for the subject. $\endgroup$
    – DukeZhou
    Jan 31 '18 at 21:20
  • 2
    $\begingroup$ @MatthewDrury: "more conceptual, theory driven, or mathematical questions are more on topic here": yes, but I shouldn't like 'more' to be read as 'only'. A practical, applied, or non-mathematical question isn't necessarily about computing tools (as I've commented before on the DS Meta). $\endgroup$ Feb 9 '18 at 11:41

The official statement of what is on-topic on CV is that we address questions about:

statistical analysis, applied or theoretical
designing experiments
collecting data
data mining
machine learning
visualizing data
probability theory
mathematical statistics
statistical and data-driven computing

It is OK if a question about these topics originated as a homework assignment (or is generally comparable to a homework assignment), but we do have a special protocol for such questions. At any rate, you can feel free to migrate such questions.

The biggest trouble spot is questions about programming, or using statistical software. Basically, I think about the issue by asking myself what does the OP need explained? If the explanation would be about statistics, designing experiments, etc., then it is on topic here, but if it is about how the software works, or how to use it, etc., then it is off topic on CV. Such questions are sometimes intrinsically ambiguous; more often such questions aren't really ambiguous, but people assume that since software is required for most such work nowadays, we 'ought' to answer these questions anyway. The issue is that the SE system already has a site for answering coding questions (i.e., Stack Overflow), so coding questions should be sent there, not here. I think it would be fine if you migrated those questions here that were truly ambiguous; we can always close them or migrate them onward.

With respect to Data Science.SE's scope, you would do better to ask them than us (just as you are better off ask us about our scope rather than them). Their scope is here (although my impression is that they are much looser with their requirements than we are). With respect to how to distinguish between the sites, note that they also have a meta.DS thread dedicated to that issue. The truth is that we rarely migrate questions to DS, and we rarely don't migrate but recommend someone check out DS instead of CV. My (admittedly somewhat biased) opinion is that most questions are either (1) better asked here, (2) better asked on SO, or (3) poor questions that don't really belong anywhere on the SE system. Many of the third type do seem to get a favorable reception on DS, because they seem to be looser with their standards—whether that is good or bad in the long run, I don't know. The best I could say is that in the answer to their thread distinguishing DS, there are three questions listed as appropriate for DS but not other sites:

  • Which Big Data technology stack is most suitable for processing tweets, extracting/expanding URLs and pushing (only) new links into 3rd party system?
  • Horizontally scaling a distributed database, what should I use for a simple key-value store? Cassandra, HBase or Riak? what are the pros and cons?
  • How important is data locality to Hadoop and Map/Reduce?

I agree that these questions would be considered off topic on CV and SO.

  • 10
    $\begingroup$ Yup. Your three questions also fit my mental model for what is on topic at DS but not here. $\endgroup$ Jan 29 '18 at 21:48

You must log in to answer this question.

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