20
$\begingroup$

I am fairly new to CrossValidated, but during the last month I could not help noticing curiously high popularity of at least two questions that I am missing a right term to describe -- maybe "small talk" questions? I mean the Facebook question and the Amazon question, and you can check https://stats.stackexchange.com/?tab=month to see that they are by far the winners by upvotes, views, and even favourites. Neither of them raises what I would consider to be particularly deep statistical questions, and looking at what sort of questions top users here enjoy answering, I suspect I am not alone. Given plenty of truly excellent questions and enlightening answers appearing here almost daily, I am a bit puzzled. The Facebook question is actually the sixth overall, which I personally find rather ridiculous.

Why does it happen? Is it something like a slashdot effect, or is it internal CV phenomenon? Was it always like that? Do people here perceive it as amusing or annoying?

$\endgroup$
9
  • 19
    $\begingroup$ In large part it's the hot questions effect. This is seen on many other "technical" SE sites as well: A question that is reasonably easily understood by a large population of readers, usually in concert with a provocative/catchy title, and that ends up on the hot questions list will end up garnering an order of magnitude or two more views than a typical question. $\endgroup$
    – cardinal
    Feb 11 '14 at 3:55
  • 9
    $\begingroup$ To answer the second question in your title, I don't think it's a problem per se. There are numerous questions on the Math.SE which some can deem "soft" or "shallow" (and some have provocative titles), gained much popularity very quickly, but bred good discussions and sometimes really great, thoughtful answers (Exhibit A, B). If the result in interesting discussion, I worry less about its genesis. $\endgroup$
    – user5594
    Feb 11 '14 at 13:15
  • 9
    $\begingroup$ cont. Concerning the Amazon question, from the answers it's apparent that the question is deeper than it seems at first glance. If I were to have a problem with it, it'd be that multiple answers arrive at the same conclusion using the same means, i.e. they're not adding anything that isn't already there. I agree with whuber's comment that "The answers that point out the ambiguities and discuss the assumptions are the ones that have merit." $\endgroup$
    – user5594
    Feb 11 '14 at 13:17
  • $\begingroup$ @cardinal: fair enough, but what about vote count? I would expect that people clicking on a catchy title in the list of SE-wide hot questions would not be registered on CV and so would not be able to vote. So I would not be surprised by sky-high view count, it is rather the huge amount of upvotes that puzzles me. (I asked the same question in the comments below, but it directly refers to what you said as well.) $\endgroup$
    – amoeba
    Feb 11 '14 at 20:35
  • 4
    $\begingroup$ @amoeba: But, that's just it. If you have at least 200 reputation on any SE site, you get an automatic association bonus when you click through to CV. This immediately confers voting privileges. Since most of those people are probably coming in through the "hot questions" links, they're very likely to have that reputation on some site already. :-) $\endgroup$
    – cardinal
    Feb 12 '14 at 14:51
  • 3
    $\begingroup$ @cardinal: Interesting, I did not know that. In fact I tried to up/down vote some question on Math.SE (my reputation on CV is above 200), but could not: the system asked me to log in (even though I am logged in on CV). So I thought it is impossible. But I guess if I do log in with my CV account, I would be able to vote. Makes me wonder whether it would make sense to weigh votes by CV reputation of the voter, so that, e.g. your vote counts more than mine or especially more than a vote of a random newcomer from another SE website. But this I guess goes against the whole SE ideology. $\endgroup$
    – amoeba
    Feb 12 '14 at 15:35
  • 3
    $\begingroup$ I think what we want is the ability to sort questions on the basis of votes made by high-karma users. And special badges awarded on that basis if they don't already exist. $\endgroup$ Feb 17 '14 at 21:13
  • 1
    $\begingroup$ Big +1 @MichaelBishop's first suggestion! Don't quite follow the badge idea though; care to clarify? Or, preferably, post that first idea on Meta SO and see if you can at least get a status-declined retag with some upvotes and maybe a user-defined script to support the idea? $\endgroup$ Feb 22 '14 at 0:03
  • 2
    $\begingroup$ @NickStauner, Thanks for the suggestion: meta.stats.stackexchange.com/questions/1963/… ...regarding badges, I don't think its as important, but we could create badges that are earned based on receiving question up-votes/favorites from high-karma users. $\endgroup$ Feb 22 '14 at 22:30
16
$\begingroup$

I'll address the issues that I think are at the heart of your concern separately.

Provocative Titles and Visibility

This section stems from @cardinal's comments above: Part of the popularity of the questions you cite is due to the choice of title. I think we all are quite susceptible to the power of first impressions, and thus can be swayed to read or skip an article/post/answer/book/etc based on its title. Any medium that allows for a piece of work to conspicuously display a title opens the door to the use of sexy, attention-grabbing titles to garner views.

On top of that, the particular questions acquired numerous views quickly, and thus was included in the Network side-bar of 'Hot Questions', increasing the questions visibility and thus, increasing the likelihood that people would click the link to the question, giving it more views (and possibly more votes). And to somewhat belabour the point, but what information is displayed in the 'Hot Questions' sidebar? The title of the question.

Accessibility to non-experts

Also mentioned by @cardinal is that the questions were quite accessible to people who wouldn't consider themselves as statistics- or data analysis-experts. They include real-word concepts that people, regardless of background, are familiar with (Facebook, Amazon, interview questions), and the questions were posed in lay-person terms. Concerning the Amazon question, I'd say the way in which the question is worded is not a problem, but rather a very good thing. The "deep" questions you may be searching for do not necessarily need to be asked in an esoteric fashion. There are numerous questions from the Math.SE which concern deep, philosophical concepts, but are framed in such a way that not only does it not repel non-mathletes, but also reels them in. A really great, recent example is this.

And when you combine accessibility to non-experts with a catchy title and place it in a prominent location, it'll be hard to stop the rise in view count.

Depth, or lack-thereof

I'll mainly focus on the Amazon question for the next couple sections. The Amazon question is deeper than it first appears. To quote @whuber's comment from the original question:

...the point of this question is not to obtain a mathematical answer but rather to see whether the interviewee thinks carefully about what assumptions need to be made in order to obtain a reasonable, defensible answer

Being cognizant of and understanding the assumptions that are explicitly and, more importantly, inherently made when approaching a problem is an important and deep issue. Quoting @whuber again,

Thus, we ought to consider any single, definite answer to this question to be incorrect--or at least not worthy of getting a job offer from Amazon. The answers that point out the ambiguities and discuss the assumptions are the ones that have merit.

I think that if someone provided a clear, detailed answer addressing the issues @whuber brings up, worthy of many positive votes, then the question becomes one we can use as an example of a fantastic question on this site.

Is it a problem?

First, are "soft" or "shallow" questions a problem? I'd say not necessarily. There are numerous questions on the Math.SE which some can deem "soft" or "shallow" (and some have provocative titles), gained much popularity very quickly, but bred good discussions and sometimes really great, thoughtful answers (Exhibit A, B). If the result is interesting discussion, I worry less about its genesis.

I think it's important to note here that SE has the Reversal badge, which is awarded to a user who provides an answer of +20 score to a question of -5 score. We reward those who can provide an excellent answer to a terrible question, so why would we not also condone a great answer to a "shallow" question?

And specifically, are the questions you cite examples of a problem present in this particular SE, or the SE Network in general?

I'd say no. Concerning the Amazon question, from the answers and the discussion above it's apparent that the question is deeper than it seems at first glance. If I were to have a problem with it, it'd be that multiple answers arrive at the same conclusion using the same means, i.e. they're not adding anything that isn't already there. So actually, your issue (and maybe my issue) with it may lie in the answers, not the question itself.

$\endgroup$
6
  • 1
    $\begingroup$ Thank you, that is a nice summary! "And when you combine accessibility to non-experts with a catchy title and place it in a prominent location, it'll be hard to stop the rise in view count" -- fair enough, but what about vote count? I would expect that people clicking on a catchy title in the list of SE-wide hot questions would not be registered on CV and so would not be able to vote. So I would not be surprised by sky-high view count, it is rather the huge amount of upvotes that puzzles me. $\endgroup$
    – amoeba
    Feb 11 '14 at 16:46
  • 3
    $\begingroup$ I'd like to disagree somewhat about the "depth" of the Amazon problem. While whuber's comment has merit, I don't think the ambiguities and assumptions about the problem are really that interesting. How novel is it really to say that you need a representative sample of the population to do statistical inference? This isn't to say that the other comments (mine included) were more interesting or valid -- just that the whole thread wasn't really that interesting (to me, at least). I felt a little weird that I got so much reputation out of an answer that I put little thought into. $\endgroup$
    – ahwillia
    Feb 11 '14 at 21:31
  • 1
    $\begingroup$ @amoeba That's a good question. It'd be interesting to see how many of the (currently) 25k and 15k views of those two questions are from the ~29k registered users of CV. $\endgroup$
    – user5594
    Feb 12 '14 at 0:43
  • 7
    $\begingroup$ @AlexWilliams That's a valid criticism. Concerning your comment, "How novel is it really to say that you need a representative sample of the population to do statistical inference?", I agree with the sentiment. However I was (perhaps unsuccessfully) trying to stress the importance of understanding assumptions and their implications, however seemingly trivial or customary they may be. $\endgroup$
    – user5594
    Feb 12 '14 at 0:56
  • 4
    $\begingroup$ @Ale I believe a good answer ought to at least mention the important but little-noticed lurking issues: the distinction between percentages and counts; regression to the mean; and whether frequentist and/or Bayesian approaches would be useful. It also needs a thoughtful summary of the all-important but all-too-readily dismissed and misunderstood issue concerning the representativeness of the data (and what that even means). Unfortunately, posting such a response now would be useless because it just wouldn't be noticed or upvoted--perhaps leaving newcomers the impression it's unworthy. $\endgroup$
    – whuber Mod
    Feb 14 '14 at 20:19
  • 1
    $\begingroup$ I might be biased, but I see this question vs. this question as another, somewhat less expected instance. In the former, I received 9 upvotes in 6 hours. In the latter, I spent an hour or so referencing new literature for an important basic question that's often handled inefficiently. I don't see a big difference in titles or accessibility, though I admit there are other factors. Both can be improved, I suppose. Good general suggestions whuber! $\endgroup$ Feb 21 '14 at 21:44

You must log in to answer this question.

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