It's very good to see the community is growing, and we're already through the private beta. Since this session ended, I've seen many questions asking for opinion and recommendation, but I was not sure whether or not they were off-topic.

The point is, datascience is not stackoverflow, where a bug is a bug regardless of who's looking at it, or what is the purpose of the application. And since everyone has their own source of information, some questions asking for source-sharing may be on topic too.

The following questions left me asking myself whether I should vote to close or not. Most of them I didn't, both for others in the community had upvoted the question/answers, and for being not sure whether they were actually off-topic.

So, my question is, what should be the borderline between on/off-topic recommendation/opinion-based questions?


6 Answers 6


Regrading the language comparison questions, I strongly believe that question asking for comparison of programming languages (mainly Python vs R in this site) are opinion-based and not based on real factor-based comparison and are not comprehensive. In fact, there may be a lot of specific use cases in which a language can be better than another, but the answers do not cover these aspects, so they will be misleading instead of providing good guidance.

Additionally, such questions occur over and over, which results in a bunch of questions essentially discussing the same thing, for example, compare these two questions which seems different at first, but have a same set of answers (comparing RAM usage of R and Python over and over!):

So I think to favor quality over quantity, all programming language comparison questions must be closed as being opinion based, or at least just have on such question (like: What is the best programming language for data science) and limit all discussion to that single question and close all others, existing or new.

Note: As we all know, Python is by far the best language for data science, so why bother asking programming language comparison questions knowing that? :)

  • 1
    $\begingroup$ Well, I wouldn't say Python is the best language. I myself rarely use python, and rather c++, R (for graphs), and bash scripting. I do acknowledge how useful python is for data science, though. Anyways, all I can say is that, either the questions are allowed, or all of them should be closed. It'd give space for valid arguments against closing votes. $\endgroup$
    – Rubens
    Commented Jun 16, 2014 at 18:01
  • 2
    $\begingroup$ @Rubens The note about Python was not serious, it was just to show the opinion-based nature of answers in an extreme. I agree with your point about closing all questions if considered off-topic, or none. $\endgroup$ Commented Jun 16, 2014 at 18:10

Opinion-based questions should only be considered "off topic" at such point that hosting them here makes it more difficult for us to have other, more important types of discussions.

As Steve pointed out, there are a great many "off topic" questions on SO that are nonetheless incredibly popular and useful, even if they have since been closed. Today, SO is a huge community, and with that comes a huge workload for moderation, curation, organization, etc. So today, hosting those discussions would detract from SO's ability to be the resource it means to be.

I felt the same way that you did initially, but we need to resist the temptation to treat this community like just another SO. We do have a specific topic for discussion, but today, it's in the best interest of this community to be much more inclusive than SO. (Which is not to say that everything should be considered on-topic.)

Perhaps in the future things will be otherwise, but then again, we are not guaranteed to even make it out of beta, much less have the good fortune to have the "problem" of SO's volume and popularity.


I've been here... less than a week. I see more discussion about what shouldn't pop than I see questions on the site.

One of the most popular questions on StackOverflow is against the rules - "The Definitive C++ Book Guide and List"

If those questions are closed, that wipes out 50% of the activity on the site since I've been here.

My opinion is that if you want to close a question, be prepared to post 2 other questions that will garner participation. Stick to the rulebook when there's actually activity to moderate. Bugging a handful of active users and telling them what they can and can't say is just going to chase them away. The point of being here isn't to get moderation experience. It's to share and pick up knowledge in the arena of data science.

At this point an argument about whether R or Python is better (if anyone actually cared) would at least draw some lurkers into participation and maybe inspire some good questions.

  • 3
    $\begingroup$ An argument on R vs Python is more of an outsider's discussion; a question on linear modelling in R is a question from someone who is actually working in data science. If we want to create a site that practitioners of data science will participate in, then the questions need to be that they face in their day to day work. If we want to create a site for novices or beginners to data science, then questions like best book, best library, R vs Python makes sense. And I find it impossible that there aren't actual data analysis, visualization problems that users who work in data science run into. $\endgroup$
    – asheeshr
    Commented Jun 13, 2014 at 5:13
  • $\begingroup$ So ask them. What I'm saying is that you all need to participate as community members more and as a moderators less. If you want a certain style of question to be the standard, lead the way and provide examples. You can't just shoot down whatever anybody else is trying to do. All that accomplishes is discouraging participation. $\endgroup$ Commented Jun 13, 2014 at 5:21
  • $\begingroup$ I was/am simply providing feedback. You can't just shoot down whatever anybody else is trying to do. Feedback can be interpreted constructively as well. My intention is not to criticize. $\endgroup$
    – asheeshr
    Commented Jun 13, 2014 at 5:27
  • $\begingroup$ @AsheeshR The intended tone doesn't always come across on the screen clearly. I know we've had a little back and forth (here and in that other thread). My intended tone is far from hostile. Have a smilie :). $\endgroup$ Commented Jun 13, 2014 at 5:36
  • $\begingroup$ After having moderated for a year, there is absolutely nothing that can take your smile away :P $\endgroup$
    – asheeshr
    Commented Jun 13, 2014 at 5:52

Pedantry and exclusiveness are not valuable mindsets here. I believe that being overly nitpicky about "no opinion based questions" will simply hurt the site in the long-run, especially as a new'ish site. And I've been through a few of these before, including a few that failed to make it out of Area 51.

In the early days, I believe a more laissez-faire approach is called for, instead of trying to preempt problems that might occur. Better to wait and, IF, "opinion based" questions somehow prove detrimental to the site, THEN change the policy.


as a data science neophyte, I find these questions very useful since they can guide me to good content. In fact, the real problem here is not the question, but the quality of answers. People asking those questions are obviously new to data science and they need experimented people to help them.

I don't think they should be off-topic.


I agree with Amir Ali Akbari about limiting discussion questions where possible. However, data science is not an exact science like programming. If you have an OOM error, then you have a very specific issue.

In the Field Guide to Data Science, it is specifically talked about the art of data science because it is just that - an art and a science. Can you really tell someone you are doing art wrong? Or is it better to suggest other solutions to their perceived issue? It is hard to think of many questions that are not subjective without them belonging on SO.


You must log in to answer this question.

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