34
$\begingroup$

Lately - just in the last few days for me, but I haven't been paying close attention - I've noticed that the scores on questions seem to be substantially lower. Digging in more deeply, questions like this one on a fairly classic but not immediately obvious complex-numbers question or even this one have started out in the hole - the latter question was at -1 when I initially wrote my (admittedly far-from-perfect) answer to it, though the question fortunately seems to have accumulated several upvotes (and lost its downvote) along the way. I'm definitely not suggesting that bad questions should be rewarded, but it seems like there's been a concerted effort from somewhere to try and raise the bar and that some not-great-but-reasonable questions are getting swept up pretty harshly (IMHO) in it.

Note that this is a distinct question from whether the quality level of the questions has declined - I feel like there's been an overall tapering of quality over the last few years but things haven't seemed particularly worse in the last few months than they were before that, whereas voting definitely feels like it's taken a downturn. Is anyone amenable to running the numbers to see if there's an actual phenomenon here, or just a perceptual issue?

$\endgroup$
5
  • 5
    $\begingroup$ The original version of the first question was broken and had not context at all it said "If n is a natural number then find the value of [z^2012n + z^1006n + 1/z^2012n +1/z^1006n is equal to." and the second looked even more broken "If n is a natural number then find the value of $ z^2012n + z^1006n + 1/z^2012n +1/z^1006n $ is equal to. " It then got fixed. But the first few minutes can be important for votes. Plus a crucial part was and is in the title, which also was broken, and which is generally a practice designed to create confusion. $\endgroup$
    – quid
    Commented Nov 20, 2015 at 19:12
  • 1
    $\begingroup$ It is not very recent but there is some detailed anaysis in a blog post math.blogoverflow.com/2015/05/10/growth-statistics-on-math-se Personally I doubt there is some very recent change. What you should be aware of is just that the opinion that questions typically must contain context/motivation/effort has some following and some users vote based on this, and this quickly, and sometimes including answers to questions failing this. $\endgroup$
    – quid
    Commented Nov 20, 2015 at 19:21
  • $\begingroup$ @quid I've noticed that - and, FWIW, largely agree with that sentiment; I'm no great fan of PSQs myself. I was mostly just curious whether there's been a particular turning of the tide (as it were) in the last few weeks, but it's very plausible that it's strictly perceptual on my part. $\endgroup$
    – Steven Stadnicki
    Commented Nov 20, 2015 at 19:23
  • 2
    $\begingroup$ I thought voting has always been harsh on this site, especially on questions coming from new users. There seems to be a knee-jerk assumption that such questions are of the "do-my-homework-for-me" variety. $\endgroup$
    – Robert Soupe
    Commented Nov 27, 2015 at 18:04
  • 1
    $\begingroup$ Anecdotally, it occasionally happens that someone who doesn't have any mathematical maturity and therefore doesn't know how to communicate mathematically has a question about some mathematics. They largely cannot use the search function, and further would not know how to generalize any answer that they found. It is not fun to answer questions for these people, you have to try to unravel their wrongness, which is frustrating, and if you have empathy, even harder to do without sounding condescending or hostile. Maybe it's better for the site that it be fun for the answerers, but I don't know. $\endgroup$
    – ZKe
    Commented Nov 30, 2015 at 4:33

2 Answers 2

Reset to default
39
$\begingroup$

Interesting. Moderators have access to certain analytical data not present in the 25K site analytics (25K only). Part of this is up-/down-vote data on questions (and answers). Below is a graph showing the sliding seven-day average of up-/down-votes on questions.

sliding 7-day average on up-/down-votes on questions, 2015/05/20-2015/11/19

So, yes, there is an uptick for the last several days. This is coupled with a plateau (or slight decline) in the number of upvotes on questions. Even so, there are five times as many upvotes on questions than downvotes.

(Update: Checking back on these charts for the same period even just a few days later, there are noticeable differences. This can really only be explained by these charts only counting votes on undeleted posts. So, of course more downvotes would be cast on recent questions, because those questions haven't had a chance to be manually deleted or sucked up by the Roomba yet.)

We're also in (another) period of historic highs in the number of incoming questions. The sliding seven-day average has been about 750 questions per day for most of this month, and we're frequently getting more than 800 questions per day (with big drops on Fridays and Saturdays).

sliding 7-day average of number of new questions, 2015/05/20-2015/11/19

I imagine that these two trends are related, as previous years have also seen a fair increase in both the total number of questions asked, and the number of downvotes on questions at about this time of the year.

In terms of question score, a very simplistic SEDE query gives the following picture of average question score by posting date for this (Gregorian) calendar year.

Average question score by date, 2015

This shows a decided downward trend, but it should be noted that this trend has been basically constant throughout Math.SE's existence and could simply be a consequence of too many questions for too few eyes. This would also be affected by recent negatively scored questions still existing, whereas older negatively scored questions are more likely to have been removed.

$\endgroup$
12
  • 3
    $\begingroup$ That uptick in downvotes is pretty striking. In absolute terms it doubled, and relative to the preceding average behavior it must be multiple standard deviations out. $\endgroup$
    – rschwieb
    Commented Nov 21, 2015 at 4:04
  • $\begingroup$ @rschwieb As I mentioned in my answer, in previous years we've seen a similar uptick in downvotes on questions at about this point in the year. For example, in Sept/Oct 2014 there was an average of about 30-35 question downvotes per day, which picked up to the 45-50 range in Nov. While not quite the near doubling we're seeing right now, it is is still pronounced. $\endgroup$
    – user642796 Mod
    Commented Nov 21, 2015 at 8:25
  • 13
    $\begingroup$ Can the data be skewed by the fact that some downvote questions can have been deleted? For instance, some of the posts which were downvoted a month ago might have been autodeleted since then. Would such downvotes still be visible in these graphs? $\endgroup$
    – Martin Sleziak
    Commented Nov 21, 2015 at 10:09
  • $\begingroup$ @MartinSleziak Good question. I'm not exactly certain, but through some research I am beginning to think that votes on deleted posts are removed from the moderator analytic tools (which goes against what I would have originally thought). I'll have to try to find someone more knowledgeable about this. $\endgroup$
    – user642796 Mod
    Commented Nov 21, 2015 at 10:53
  • $\begingroup$ If you do not find somebody to tell you about the deletions, you might check again in a week and compare. $\endgroup$
    – quid
    Commented Nov 21, 2015 at 16:54
  • 1
    $\begingroup$ @rschwieb you are having the Eternal September². $\endgroup$
    – Braiam
    Commented Nov 21, 2015 at 20:44
  • $\begingroup$ I propose that the rolling average number of questions posted is explained by the fact that the academic year tends to begin in September for many educational institutions; thus, there should be a seasonal trend in this statistic. An upswing in the number of downvotes could also be due to more low-quality questions being asked, rather than any change in voting behavior. If more people ask bad questions, you'd most easily see it in the downvote counts even though it's still tiny compared to question volume. $\endgroup$
    – heropup
    Commented Nov 24, 2015 at 5:19
  • $\begingroup$ So it now appears that the uptick in this particular query is an artifact that rolls with us as time goes on, and that there's now some question as to whether or not this query would reveal a real uptick in downvotes? That's the impression I'm getting from the comments and the second snapshot. Could you do a third snapshot? $\endgroup$
    – rschwieb
    Commented Nov 27, 2015 at 5:26
  • $\begingroup$ users can do their own munging at data.stackexchange.com/math/queries $\endgroup$
    – cactus314
    Commented Nov 30, 2015 at 20:29
  • $\begingroup$ Could we see another snapshot? (I'm under the impression that it's available only through your moderator tools. If the query is available for mortals, I would take that instead $\endgroup$
    – rschwieb
    Commented Dec 5, 2015 at 3:38
  • $\begingroup$ I'm very curious if there's any correlation between tags and votes. Are users downvoting particular tags more than they were? $\endgroup$
    – Adrian Keister
    Commented Apr 12, 2018 at 19:56
  • $\begingroup$ Could you recrop the ones that don’t show zero on the y-axis so that it’s included. It makes it much easier to get a sense of the scale of change (without the zero you can’t tell the scale at a glance)? $\endgroup$
    – bob
    Commented Apr 23, 2022 at 20:22
12
$\begingroup$

Exhibit A:

img

If stuff like that wasn't downvoted back in the days, that that was a problem, and I'm glad to see that it's being fixed.

Exhibit B question is a question you've answered. The "broad hint" you wrote made it clear you didn't think the question deserved a full, elaborate answer. Well, that means it was a bad question, so downvotes are in order -- for the question and its answer.

$\endgroup$
18
  • 19
    $\begingroup$ Why would you downvote a good hint to a bad question? It only creates confusion. $\endgroup$
    – wythagoras
    Commented Nov 22, 2015 at 11:59
  • 7
    $\begingroup$ @wythagoras Presumably to discourage answering bad questions, perhaps also to discourage not giving full answers (I know this would be why I would downvote). $\endgroup$
    – Najib Idrissi
    Commented Nov 22, 2015 at 14:54
  • 21
    $\begingroup$ Codswallop. There are many perfectly good questions – perfectly good even by the standards of those who demand a lot of context – that are most appropriately answered by a hint. The notion that every answer should be full and elaborated is pernicious rubbish. $\endgroup$
    – Brian M. Scott
    Commented Nov 23, 2015 at 2:39
  • 5
    $\begingroup$ @BrianM.Scott Please give even one example. $\endgroup$
    – Najib Idrissi
    Commented Nov 23, 2015 at 8:28
  • 15
    $\begingroup$ @BrianM.Scott It is neither polite nor mature to label other points of view as "rubbish" simply because you disagree with them. Other people may disagree with your point of view but don't arrogantly go around talking about how your perspective is "pernicious rubbish." By design, users are allowed--even encouraged--to vote the way they do for whatever reason they personally see fit, short of fraud. That you disagree with this philosophy by implying that only a certain kind of reason justifies a downvote means you disagree with the way this site is intentionally designed. $\endgroup$
    – heropup
    Commented Nov 24, 2015 at 5:23
  • 15
    $\begingroup$ If people want to blanket downvote hints then one cannot do anything about it, but it shows a complete lack of understanding of what the goal of such answers usually is: helping/forcing OP to think for him/herself and in that process actually learn mathematics. Spoon-feeding OP with full answers is often counter-productive to this goal. $\endgroup$
    – Winther
    Commented Nov 24, 2015 at 18:36
  • 5
    $\begingroup$ @Winther If you want to "force the OP to think for himself", don't answer at all and ask for their thoughts in the comments. It seems that people who answer with hints "show a complete lack of understanding" of how this platform works too. It's happened several times now that I can't prove some result, look it up and find the same question on this site, and then I'm greeted with some half-assed hint that I already thought of but couldn't make anything of. Now what? I can't ask a new question, it would be a duplicate. I can't offer a bounty every time that happens. [...] $\endgroup$
    – Najib Idrissi
    Commented Nov 25, 2015 at 9:47
  • 5
    $\begingroup$ [...] If the question is truly bad, close it and either force the OP to improve it or eventually delete it. No one is calling for spoon-feeding. If the OP improves the question, great, post a full answer and contribute to the knowledge repository that math.SE is. If the OP doesn't, please don't contribute to the pollution of this website by putting all the people that will have the same question in the future in a catch-22 situation. [...] $\endgroup$
    – Najib Idrissi
    Commented Nov 25, 2015 at 9:48
  • 5
    $\begingroup$ [...] And I'm not even talking about all the people who take advantage of the "hint" policy to post some one-line hint that they didn't think through and doesn't even work when you try to work out the details -- just two days ago someone posted a solution for some question in that worked in the one-dimensional case and then "HINT: use induction on dimension" or something like that. As you can guess, higher dimension was orders of magnitude more difficult than dimension 1 and the answerer eventually acknowledged that. Sorry for the rant but I'm really unhappy about this issue. $\endgroup$
    – Najib Idrissi
    Commented Nov 25, 2015 at 9:52
  • 2
    $\begingroup$ @NajibIdrissi You started this off with a blanket statement about all hints being bad answers and now you’re attacking a strawman. Nowhere above is there any support for the claims you seem to be arguing against (useless one-line hints, answering truly bad questions, ...). The dramatic catch-22 argument you try to make is just not how it works in practice. Most of the time reason prevails and if OP has a good reason for a closed-duplicate to be reopened then it can and often is reopened (especially if the answer is bad). $\endgroup$
    – Winther
    Commented Nov 30, 2015 at 21:13
  • 2
    $\begingroup$ @Winther It is especially ironic that nowhere in your comment there are arguments to support your position, or that you couldn't explain which "strawman" I'm attacking and which was the real position taken. "Strawman" isn't a magic word that lets you win arguments. I have seen the "catch-22" happen with my own eyes, so simply asserting that it doesn't happen isn't likely to convince me. Anyway, I'm done with this discussion -- I said what I wanted to say. (For the record, yes, I do consider all hint-only answers to be bad answers.) $\endgroup$
    – Najib Idrissi
    Commented Dec 1, 2015 at 9:00
  • 2
    $\begingroup$ @NajibIdrissi I give hints that are not full answers when filling out all the details is neither necessary nor helpful. For example if someone asked "How do I integrate $\int_0^5 x^2 e^x dx$? I tried $u$ substitution but that didn't work." I might reply "Hint, do integration by parts twice. Let $u=x^2$ and $dv=e^x$. Then after let $u=2x$ and $dv=e^x$.". I am not going to go through all the details of plugging in the limits and actually writing everything out because it simply isn't necessary. Anyone who reads my answer would know exactly what the full, complete answer is. $\endgroup$
    – user223391
    Commented Dec 18, 2015 at 1:49
  • 1
    $\begingroup$ @avid19 This isn't what I call a hint. A hint would be "HINT: use integration by parts" with no further information. I too skip tedious computations sometimes. You're talking about a different thing. $\endgroup$
    – Najib Idrissi
    Commented Dec 18, 2015 at 8:51
  • 4
    $\begingroup$ @NajibIdrissi I know I'm a bit late to the party, but I thought I could add something to the other side of the argument. Here is an answer I gave, which I am particularly proud of, which is in the form of a hint, and which I strongly feel was particularly effective because it was shortened to the form of a hint. I think that often, the primary difficulty of a problem is starting in the right direction, not the actual implementation. It is in these contexts that I feel hints are particularly useful. $\endgroup$
    – Ben Grossmann
    Commented Dec 30, 2015 at 2:41
  • 2
    $\begingroup$ @Omnomnomnom As usual, I'm not convinced of two things: that just giving the starting point is better (what metric are we using anyway?) than giving the full answer, and that anyone who finds that question will be able to piece out the full answer from the hint. There's also always this perverse effect of reading a hint, thinking "oh yeah I see how it works", but if you actually tried to write the details you would get stuck because you misunderstood the hint, for example. Anyway, as is evident from the recent meta thread, hints aren't going anywhere, don't worry... $\endgroup$
    – Najib Idrissi
    Commented Dec 30, 2015 at 7:32

You must log in to answer this question.

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