17
$\begingroup$

Our usage statistics at http://area51.stackexchange.com/proposals/55053/data-science are all but great.

Unless we get a much larger and more active community going, this site will die in beta, just like enter link description here did.

I must admit that I don't see a future for this site. It is too redundant, and I am not a big fan of having another SE site for every topic. I see too much overlap with stats.stackexchange.com; and I've seen too many questions being cross-posted to three or more sites. That does not make sense to me.

If you're really interested in this site, fill it with life!

$\endgroup$
  • 4
    $\begingroup$ The fact that you care enough to raise this concern is a great sign, but folks may not be clear on what the problem is, or how to reverse it. Can you edit to propose a specific action for the community to take? "Let's vote more!" or "We should narrow our on-topic criteria." gives others more to work with. $\endgroup$ – Ana Sep 9 '14 at 18:38
  • 3
    $\begingroup$ Maybe promotion in relevant communities would help. BTW: I got the "Tumbleweed" badge for this question. $\endgroup$ – Anony-Mousse Sep 12 '14 at 10:39
  • 3
    $\begingroup$ I agree: too much redundancy with other SE. To make it worse, once a SE is cancelled, Q&A are only accessible as a dump and cannot be viewed directly online, so I have always been reluctant to participate in newly-born SE. $\endgroup$ – Franck Dernoncourt Sep 14 '14 at 17:20
  • 2
    $\begingroup$ Many beta sites go through a period of low activity during the first few months. See Earth Science for instance, which is just a month older than this site, the number of visits per day were crawling in the 100s until two weeks ago when the number reached back the 300/400s it had at launch. You can try this query on the data explorer with several beta sites to check how they did. Funny that on a data science site, you didn't thought of comparing with the data at hand :) $\endgroup$ – plannapus Sep 19 '14 at 7:27
  • 1
    $\begingroup$ Chances are that your data won't help the site... Just because you use data doesn't meant the conclusion is right. And IMHO the data doesn't even support your hypothesis. $\endgroup$ – Anony-Mousse Sep 20 '14 at 18:35
13
$\begingroup$

I don't think redundancy has been the problem at all. We all knew about the overlap from the start and I agree with iliasfl's point that users should be able to expect very different types of answers when asking a question that's on-topic on multiple sites, depending on where they choose to ask. The root of the problem in my view is that too many questions like these are being treated as appropriate/acceptable content:

I'm not saying every one of these questions should clearly be closed or that any of them necessarily deserve downvotes, but questions like these at some point came to represent a major portion of the site's content. I used to come read through the new questions here every day (when I had more free time) and I definitely started to feel like the noise was overwhelming the signal. That's really bad for a site that has yet to establish itself; it's a strong disincentive for attracting and retaining expert users at a time when building a strong user base is most critical.

It's not the naiveté of these questions that makes them bad content – beginners are perfectly capable of asking thought-provoking, well-researched, non-trivial questions. For that matter, even a trivial question can make for good content, provided it's well-defined and difficult to answer through secondary research (or the answer is simply not available online).

The main problem with these questions is that they don't fit the format of the network, which isn't something to be taken lightly. The Q&A format of Stack Exchange is a purposeful and central part of its design that has evolved over the years to better support that "repository of useful and accurate answers" mission. We discourage questions like the above – primarily by closing them – because they hurt that mission. To quote from Jeff's blog entry, Optimizing For Pearls, Not Sand:

It’s true that you can’t have Q&A without questions, but having the wrong sorts of questions is far more dangerous. The fastest way to kill any Q&A site is to flood it with low-quality questions....

Now, in our case it's not that we're being flooded with terrible, awful, no-good, very bad questions. These aren't questions nobody should ever ask, anywhere; they're questions that are of practically no interest to the community of experts that we need to attract in order to generate other high-quality content that in turn attracts better questions and builds and retains momentum for a young site.

This is the major flaw with the position that some have held throughout this beta that since we lack content, we should be less choosy about what content is acceptable. Actually, I think the opposite is true in a beta – the less content we have, the less signal we have, and the more important it is to weed out the noise. It's only by being choosy and scoping the site with the ultimate goals in mind that we send the message to the expert community: Stick around because this place is growing into something worthwhile.

This is closely related to the conclusion of the aforementioned blog post:

Answers are the real unit of work in any Q&A system. Therefore, the only logical thing to do is to maximize the happiness and enjoyment of answerers. If this means aggressively downvoting or closing unworthy and uninteresting questions, so be it. Without a community of people willing to answer questions, it really doesn’t matter if there are questions at all, does it?


As to whether the site is "dying"... I think it would be a mistake to think that any beta is easy, if only we have the right policies or cast the right votes. But it may be that DataScience.SE is a particularly hard site to build, at least at the present time. The field itself is very interdisciplinary and very young; how many people really think of themselves as data science experts? Or think of their problems specifically as data science problems? I very much believe this site can and should exist, but the evidence seems to indicate that it's not practical in this incarnation at this time.

$\endgroup$
9
$\begingroup$

I have no idea how this can be solved, for sure this site has lost its momentum.

However, I can tell you why I personally decided to leave this site. Personal opinion but it's time to express it. I was very active and excited when it was first lunched but I didn't want to invest time and effort to a site that from early on it was obvious it will end up to the described situation.

The first reason is simple: people without enough experience don't get what is different about the (overloaded) term data science compared to programming or statistics. Or between DS.se compared to SO or CV. The difference is not on the topics covered by the questions asked. It is about how the engineering and stats should work together when you need to build a solution to a problem. For example, when I ask the question "How can I measure the mean value?" I expect an answer with a simple formula in CV, or the description of a simple algorithm in SO. In DS I would expect to see an answer on how this can be solved efficiently in a Hadoop cluster, or how it can be estimated in sub-linear time by sampling and so on... I expect answers typically not found or discussed in either CV or SO. Nevertheless, people in here were ready to flag questions off-topic based on the keywords used.

This is a simplified example, but I hope you get the point. If you don't get the point then this lead us to the second reason... The quality of the questions AND the answers was very low (I stopped checking at all the last few months, I assume it didn't get much better). This is because questions were asked from software eng. that want to become data scientists. So yes, their questions would get good answers in CV. Also, I have seen few good questions that are answered again by people who don't get the difference between DS/SO/CV leading to naif/low quality answers.

The above are my personal opinion and if you disagree, I am sorry but I don't have the energy or time to change your mind (so don't expect me to answer back). However, I would be really happy to hear what you think is the answer to this question.

$\endgroup$
  • 3
    $\begingroup$ As mentioned above, I do see too much overlap with the existing sites, and I don't like this fragmentation and cross-posting. So I do mostly agree with you. In my opinion, this site is bound to fail. $\endgroup$ – Anony-Mousse Sep 12 '14 at 10:41
9
$\begingroup$

I've been checking this site daily for about two weeks now. There are not enough questions, and most questions are misplaced - things like career questions, or opinion based. Nobody seems to be flagging them or taking them down. I actually answered some before realizing this was the wrong place for those questions. Nobody is even posting responses that those questions are off-topic. It would be great to have a FAQ we could post in response to those, after they are closed.

I'm also having a hard time getting reputation enough to do simple things like leave comments. I'd be happy to help moderate but I can't seem to get the reputation to do so.

I'd like to be active here, but I'm finding it very hard to break into this community.

$\endgroup$
  • 2
    $\begingroup$ I am closing off-topic questions, but am waiting a week or more before doing so to see if the community will flag or close them rather than unilaterally closing them immediately. $\endgroup$ – Sean Owen Oct 14 '14 at 8:04
  • 1
    $\begingroup$ Awesome, I will start flagging them. $\endgroup$ – sheldonkreger Oct 14 '14 at 17:37
  • 2
    $\begingroup$ @SeanOwen Closed questions can always be re-opened. That's part of the reasoning behind displaying them as "on hold" initially - to prompt the asker to revise and improve. Waiting to apply this status to a question that clearly ignores the "don't ask" page is counterproductive because the question gets answers in the meantime and then closing it looks vindictive rather than corrective. $\endgroup$ – Air Oct 29 '14 at 17:15
  • 1
    $\begingroup$ Good point. I suppose the question though is, should they be don't-ask and do they need fixing? I am closing these questions in the end. $\endgroup$ – Sean Owen Oct 29 '14 at 17:35
4
$\begingroup$

One thing that I cannot easily answer: "How is this site different from Cross-Validated?"

It seems that the sites' aims are mostly overlapping. Maybe enough people petitioned from this site without realizing that a reasonable choice was already there?

I have wondered sometimes that Area51 needs to have a better mechanism for suggesting existing communities whilst people are petitioning.

$\endgroup$
3
$\begingroup$

I would suggest a data science flashmob - as was organised for r in stack overflow.

http://www.decisionsciencenews.com/2009/09/03/r-flashmob-2-tuesday-8-september-2009/

perhaps the moderators could come up with a list of "seed questions" and/or answers. ( this should also clarify what this site is about). My own interests for this site is in algorithms/tools for big data. Sadly most of the questions are just basic ML (which I would move to CV).

$\endgroup$

You must log in to answer this question.

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