What should we have in Help Center > Asking section regarding What topics can I ask about here?

For example Stats SE has this:

CrossValidated is for statisticians, data miners, and anyone else doing data analysis or interested in it as a discipline. If you have a question 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

And here is /help/on-topic at Data Science:

Examples of questions that are likely to be on-topic for Data Science Stack Exchange:

  • Given process monitoring data arriving every 10ms, what statistical tool should I use to best characterize a change in the process - mean? a distribution?
  • When is it suitable to apply L1 regularization for feature selection?
  • I would like to produce a infographic on the 'Brexit' referendum. Given public opinion data across the UK, what are some meaningful techniques to visaualize it in a dashboard?
  • When executing an ARIMA model in Spark, what are the pros and cons of using Python instead of R?
  • Given Facebook Likes, is there an ML technique to predict age and gender?

If we would like to differentiate from the above sites, we should have our unique section about the topics which people can ask about here.

What description of /help/on-topic page for AI site would you suggest?

  • $\begingroup$ I think its still unclear to me what goes on stats overflow vs ai overflow...how is that decided? $\endgroup$
    – Pinocchio
    Commented Jul 29, 2019 at 19:07
  • $\begingroup$ what does: "questions about the mathematics of machine learning should be asked at Cross Validated."? That ambiguous. If one asks for clarification of notation is that mathematics of ML? Or is only a proof of statistical learning theory the "mathematics of ML"? $\endgroup$
    – Pinocchio
    Commented Jul 29, 2019 at 19:17

2 Answers 2


Drawing on these existing discussions:

Also taking some inspiration from the Super User "on topic" page, here's my first stab at it:

If you have a question about...

  • social issues in a world where artificial intelligence is common,
  • conceptual aspects of AI, or
  • human factors in AI development

...and it is not about...

  • the implementation of machine learning, or
  • asking for a development tool or career path recommendation

...then you're in the right place to ask your question!

This is only a draft, but it seems like a good starting point. Please suggest improvements if you see anything that needs adjustment! Specifically, I'm not sure how specific we need to be about what constitutes "implementation" in this blurb. If there are other commonly asked kinds of off-topic questions, those could be worth mentioning too.

  • $\begingroup$ Should we add point about scientific questions about AI? $\endgroup$
    – kenorb
    Commented Sep 22, 2016 at 15:30
  • 1
    $\begingroup$ @kenorb Yes, I was trying to include those in the last two "acceptable" bullet points. I'm a little wary of the term "scientific" because some might interpret it as allowing much more than we have so far. (A related question.) I would welcome better wordings! $\endgroup$
    – Ben N
    Commented Sep 22, 2016 at 20:26
  • 3
    $\begingroup$ @kenorb Since there didn't seem to be objections, I've added the current wording in this post to the on-topic page. Suggestions for improvements are always welcome. $\endgroup$
    – Ben N
    Commented Sep 27, 2016 at 15:59
  • 1
    $\begingroup$ Added a section pointing people to Data Science or Cross Validated if their question is on ML implementation or mathematics. $\endgroup$ Commented Sep 28, 2016 at 21:29
  • $\begingroup$ Should we expand the second bullet of the "not" section to include questions like this one? I think it's in the same general cluster but it also seems like answers that link to good tutorials are a good thing to have around. (That particular question is probably too broad anyway, but you could imagine a "what's the best way to learn about [specific issue X]?" question going well.) $\endgroup$ Commented Oct 4, 2016 at 17:38
  • 1
    $\begingroup$ @MatthewGraves I'm not super enthusiastic about questions/answers dedicated to off-site resources. Adding a big list for beginners is usually something that should be done with community consensus, since it would require continuing upkeep. I would be on board with adding "learning materials" to the list of things we discourage recommendations of. $\endgroup$
    – Ben N
    Commented Oct 4, 2016 at 18:02
  • $\begingroup$ I think it should be clarified what implementation of ML means. Does that mean just coding or does that include the descriptive language of it? e.g. maths (even if its NOT about proofs, but the mathematical language to express ML/DL modern AI). $\endgroup$
    – Pinocchio
    Commented Jul 29, 2019 at 19:04
  • $\begingroup$ I think it would be good to clarify what goes on stats vs AI overflow. Thats unclear. $\endgroup$
    – Pinocchio
    Commented Jul 29, 2019 at 19:12
  • $\begingroup$ what does: "questions about the mathematics of machine learning should be asked at Cross Validated."? That ambiguous. If one asks for clarification of notation is that mathematics of ML? Or is only a proof of statistical learning theory the "mathematics of ML"? $\endgroup$
    – Pinocchio
    Commented Jul 29, 2019 at 19:17

We should drop any reference to implementation specifically being on or off topic. That's really orthogonal to the issue and it makes it too easy for people to justify arbitrarily closing good questions. And as this eliminates so many of the more concrete questions, it makes the site appear as though it's only for science-fiction'ish questions.

  • 7
    $\begingroup$ I think that questions about how to pick the right architecture to solve a particular problem, or about why various architectures are good or bad fits for particular problems, would be interesting and useful to have around. But I also think that debugging tensorflow code or discussing data preprocessing is something that we should try to push off to another SE. Do you have a suggestion as to where to draw that line, and how to communicate it in the help center? $\endgroup$ Commented Oct 4, 2016 at 19:02
  • 1
    $\begingroup$ I agree that there should be a said line and i think that it should place questions about network architectures and methods within the AI site but should leave any deeper discussions outside. A good reference for this line is AI liturature: Most of it doesn't discuss code and rather theorizes, examines, and discusses how aspects of AI can be arranged, defined, estimated, while addressing the features that current AI systems lack. Using this as a guide, perhaps we can start having more answers that reference AI papers than comments which tell people to post in CrossVidated. $\endgroup$ Commented Sep 22, 2017 at 13:30
  • $\begingroup$ @MatthewGraves I think its simple. Anything that involves code is off topic. If its described in the mathematical language of what the ML algorithm is doing that is fine (especially if its brain inspired AI like deep learning). Just explicitly saying code is not allowed sounds good to me or that the questions should be how to code something or not code focused but AI concept focused. $\endgroup$
    – Pinocchio
    Commented Jul 29, 2019 at 19:06

You must log in to answer this question.

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