26
$\begingroup$

I would be curious to know whether, now that links to MO appear in the footer of every SE 2.0 site, and in the "hot questions" list of the Stack Exchange button, the rate of inappropriate questions has increased. By inappropriate, I mean things like calculus homework, "what's wrong with this proof that 0=1", etc.

This would presumably need to be measured by an SE team member, although the MO moderators can also offer their qualitative opinion (as far as I remember from my time as a math.SE mod, there are no migration / closing statistics available to site moderators). While any current information about this would be welcome, the switch to SE 2.0 is still quite recent, so perhaps the data should be reviewed in a few weeks to get a better sense of the effect of the switch.

The questions to be counted would be those that were migrated to math.SE (though of course this has only been an option since the switch), and those that were simply closed due to inappropriateness. Now, the range of close options used on inappropriate questions was, and is, wide, and I think several of those close options from MathOverflow 1.0 do not exist anymore as a result of the switch, which increases the difficulty of automating the tabulation of inappropriate questions that were closed. One thing that may help would be to do a site-wide search for comments such as "Crossposted to math.SE [link]" because this would correlate well to questions that were ultimately closed for being inappropriate.

I suspect the answer will be that this rate has increased. I seem to remember that part of the agreement behind the switch was that MathOverflow would be invisible to the rest of the SE 2.0 network (though not the other way around), precisely to avoid this effect. Was this condition of the agreement discarded? Do people feel that it would be good to implement (regardless of its previous status)?


I am using the term "switch", instead of "migration", to avoid confusion with the migration of questions between different sites.

$\endgroup$
  • 3
    $\begingroup$ My recollection is that 'invisibility to the rest of the network' was proposed, and quickly ruled out (by Anton?). $\endgroup$ – Scott Morrison Jun 30 '13 at 10:26
  • $\begingroup$ Since there are probably many questions out there which should be closed but haven't been, it may be beneficial to wait a week before checking this. $\endgroup$ – Manishearth Jun 30 '13 at 11:05
  • 2
    $\begingroup$ Regarding data, just: number of total questions, number of closed (including mig) + deleted; for a couple of weeks is enough, IMO. Finer distinctions seem not necessary. My vague opinion, but I didn't even pay much attention is, perhaps slight increase but not that much. Problems from being in the network IMO are other things than questions (this suggested editing can be somewhat of a nuissance, but also has some good to it) and maybe voting (but this is still harder to tell). If there is problem it's the assoc. bonus, but for this it is too late anyway. And, overall, it's fine as is. $\endgroup$ – user9072 Jun 30 '13 at 11:33
  • 1
    $\begingroup$ @Quid Yeah, I realized that, I'm fetching the data now. I doubt the assoc bonus causes too many problems: anyone can post (bad) Q/A, the assoc bonus doesn't help here. $\endgroup$ – Manishearth Jun 30 '13 at 11:56
  • 6
    $\begingroup$ @Manishearth The problem with assoc bonus is that it allows "anyone" to upvote bad questions... $\endgroup$ – Grigory M Jun 30 '13 at 14:40
  • $\begingroup$ @GrigoryM Oh, that. Forgot about that. $\endgroup$ – Manishearth Jun 30 '13 at 14:49
  • 1
    $\begingroup$ Like quid, I feel like there has been a slight increase, but am prepared to believe that is an illusion if the data do not confirm this $\endgroup$ – Yemon Choi Jun 30 '13 at 17:04
  • $\begingroup$ @YemonChoi See graphs below, the data confirms it. However, for a complete analysis we need to know the number of questions per day too. $\endgroup$ – Manishearth Jul 1 '13 at 3:55
  • $\begingroup$ This has not yet been effected. When it is, compare data from before and after that date. $\endgroup$ – msh210 Jul 1 '13 at 15:21
25
$\begingroup$

Update: Included Questions/day in graph:

(from this data, data from the creation of the universe MO here)

"Doomsday" is the date of migration, June 25th. ("Tuesday" exactly one week before "Doomsday", this is partially to highlight the weekly fluctuation, and partially because it rhymes)

enter image description here

You can clearly see the weekly fluctuation here (we have the same trend on Physics and Chemistry; stats oscillate on a weekly basis).

However, it looks like the percentage of closed questions went up just around migration time (before, in fact, though that may just be an outlier):

enter image description here

Note that the data as shown has a high probability of being skewed:

There is a chance that the autodeletion roomba may have gotten many closed questions older than 9 days, skewing the results.

Also, (h/t @quid), closed questions usually get deleted within 2 days, so that may lend another bit of skewing to this data.

Update (9/16/13)

Closed questions by type over time (interactive graph here, on the sheet "Chart1"):

enter image description here

Questions over time

enter image description here

Percentage of closed questions over time

enter image description here

Questions over time, stacked (cumulative):

enter image description here

(I don't know why the fill isn't working out properly here, but it can be safely ignored)

$\endgroup$
  • 1
    $\begingroup$ Perhaps it would help to mark in your graph the day we switched to MO 2.0? $\endgroup$ – Joel David Hamkins Jul 1 '13 at 10:54
  • $\begingroup$ @JoelDavidHamkins Good idea, done :) $\endgroup$ – Manishearth Jul 1 '13 at 11:08
  • 2
    $\begingroup$ The plots are nice, thanks for making them. But looking at them in detail I think there is a very important piece missing, the deleted questions (I understand you might not be able to get this data); this is just to stress that mainly others should not look at the graph and think there is a huge inc. The point is that on MO the majority of closed questions got deleted often essentially, ASAP, that is 2days + a couple hours after closure (but with some fluctuation). You acknowledge this might be skewed but the skew seems very large to me; also it might explain the going up just before mig. $\endgroup$ – user9072 Jul 1 '13 at 14:02
  • $\begingroup$ @quid I put in a note about the autodeletion roomba, however I'll add that in as well. I'll see if the community team is willing to provide plots and/or data including deleted posts. $\endgroup$ – Manishearth Jul 1 '13 at 14:23
  • $\begingroup$ For some reason, the "Percent" graph seems to show that the majority of the increase was from Tuesday-Doomsday. Any explanation for that? (Could be the roomba) $\endgroup$ – Manishearth Jul 1 '13 at 14:25
  • 2
    $\begingroup$ Thanks for the update! I do not know. There are several conflicting effects one could imagine: on the one hand, I four example did not vote to delete recently, since I was always busy with other things (getting to know the new site, meta and so on), might also affect others so that there is less deletion; also since there is autodeletion know some might have decided not to deleted. In the other direction the new review interface might lead to more closures (even if there were no change in the questions). Hard to tell what is happening; I am just guessing around a bit. $\endgroup$ – user9072 Jul 1 '13 at 16:15
  • $\begingroup$ @quid You're right there, the review interface would help tons. (On Physics.SE it did indeed make things smoother). That's the problem with data, you've got to factor in all of this stuff when analyzing it. Anyway, I'm trying to get data on deleted posts too, so that may at least remove some of the skewing. $\endgroup$ – Manishearth Jul 2 '13 at 4:08
  • $\begingroup$ @Manishearth, could you please update your graph again? I think we've had a long enough 'burn-in' that the graphs will really mean something, and we can decide if we need to do anything. $\endgroup$ – Scott Morrison Sep 16 '13 at 4:57
  • 1
    $\begingroup$ @ScottMorrison Alright, I'll try. I no longer have the scripts used to generate these, so it may take some time. $\endgroup$ – Manishearth Sep 16 '13 at 5:53
  • $\begingroup$ @Manishearth, ah, okay. If it's too much trouble, a hint as to how you did it would let someone else pick up the ball. (But I liked your pretty graphs! :-) $\endgroup$ – Scott Morrison Sep 16 '13 at 6:13
  • 1
    $\begingroup$ @ScottMorrison Not too much trouble, just that I'm studying right now for an exam (after 6 hours or so I ought to be relatively free). I used jQuery to pull the data, converted it to CSV using this fiddle, and then popped it into Google Docs. I'll just need to rewrite the jQuery here :) $\endgroup$ – Manishearth Sep 16 '13 at 6:16
  • $\begingroup$ @ScottMorrison Here is what I have till now, but I feel there's something wrong with the date parsing so I'll have to have another look $\endgroup$ – Manishearth Sep 16 '13 at 15:17
  • $\begingroup$ @ScottMorrison Done (I think). Enjoy! :D Note that the percentage graph shows a spike due to the closed question autodeletion (the same spike is there at the boundary of the earlier graph) $\endgroup$ – Manishearth Sep 16 '13 at 18:13
  • 2
    $\begingroup$ Interesting! Thanks for doing this! Would it be easier to pick up trends if the graphs were using weeks rather than days? Maybe a rolling average to smooth out the data? $\endgroup$ – François G. Dorais Sep 16 '13 at 18:58
  • 1
    $\begingroup$ @Manishearth, I just noticed that something is messed up in your spreadsheeet (the first link update the 9/16 update). If you scroll down to the end, lots of the dates are messed up. I'm wary of using this data further without knowing what's going on there. $\endgroup$ – Scott Morrison Sep 18 '13 at 1:34
10
$\begingroup$

A near duplicate of Manishearth's answer, showing the same graphs but with 7-day moving windows per François' suggestion. I think it's much clearer:

enter image description here enter image description here

My data is copied from Manishearth's with a few columns added to do the calculations. You can grab a copy and do further manipulations yourself.

$\endgroup$
  • $\begingroup$ Thanks Scott! Moderator analytics (closed & deleted posts) doesn't seem to be compatible. There, I see that the number of closed questions is roughly stable in the 10-15 range (which is higher than the roughly stable 5-10 range prior to moving to SE). Any thoughts what the difference between the two data sets? $\endgroup$ – François G. Dorais Sep 17 '13 at 10:45
  • $\begingroup$ Also, the recent rise seen in the graphs here doesn't seem to fit what I would expect from autodeletion according to the description. $\endgroup$ – François G. Dorais Sep 17 '13 at 10:54
  • 3
    $\begingroup$ This is interesting data. But for reasons explained earlier, I would be more interested to know the "closed+deleted" percentage and also over a year or even slightly more. I believe (but have to confess I did not check) this is not easy to get for me as deleted content is not in general dumps/data. $\endgroup$ – user9072 Sep 17 '13 at 12:23
  • $\begingroup$ @FrançoisG.Dorais re deletion: I for one delete a lot less now after the move, there are several reasons: 1. I hope for the autodeletion (but sometimes it should not kick-in as the score will be 'too good') 2. the limitation of delete votes per day is a serious pain for me; deletionn was a thing, essentially the only one, I often did in batches when I had time. (also 3. general lack of time) What could be an issue with the data are the migrated questions that do get deleted but only after a month, which could have serious influence on the 'end' of the graph. $\endgroup$ – user9072 Sep 17 '13 at 12:27
  • $\begingroup$ And also auto-deletion should hardly delete much more than was anyway deleted at least somewhat before the move, say during 2013, where I think really only quite few things fell through the cracks of manual deletion and a very significant proportion of closed questions got deleted (often 48h+epsilon after closure). $\endgroup$ – user9072 Sep 17 '13 at 12:34
  • 1
    $\begingroup$ Okay, I'll ask the SE guys for statistics on deletions. I agree these charts would be much more useful with this. $\endgroup$ – Scott Morrison Sep 17 '13 at 23:25
  • $\begingroup$ Can you redo these graphs now that we have more data? $\endgroup$ – Mariano Suárez-Álvarez Oct 30 '13 at 19:19
7
$\begingroup$

I think the rate is up. I am trying something new, which is voting to migrate to MSE. I'm not sure to what degree the resulting move is visible to the OP, but if it is, the arrival at MSE should probably be more of a wake-up call than simple closing of the question here.

Oh, there are some oddities about closing votes, the action taken may not be the majority type of vote.

For those not watching, the people, and moderators, on MSE who wanted to restrict homework questions lost out, the newly elected moderators are probably moderate, but the effect is that there are no restrictions. There are also no effective requirements that motivation be provided, or effort shown.

For anyone interested, MSE Meta question just before Qiaochu resigned as MSE moderator: https://math.meta.stackexchange.com/questions/9180/proposal-ban-verbatim-homework-questions-which-have-no-accompanying-text

For an early discussion pointing out that MSE also has standards, see http://mathoverflow.tqft.net/discussion/1125/sending-people-to-math-stack-exchange

$\endgroup$
  • 3
    $\begingroup$ Mainly to test it, I just cast my first vote to close since month, a vote to migrate to math.SE However, I would like to say that this was not at all intended as a slap in the face. By contrast the user asking it appeared to be a SO user that just happened to have a math question and searched for the appropriate SE site and picked the wrong one; and this rather is a helpful showing the way to the right place instead of just saying 'not here'. That this happens is perhaps also 'our' fault; as the official description that is easily visible is not yet very telling. $\endgroup$ – user9072 Jun 30 '13 at 19:03
  • $\begingroup$ @quid, changed it to "wake-up call" $\endgroup$ – Will Jagy Jun 30 '13 at 19:14
  • 2
    $\begingroup$ Will are you sure about m.SE? There's several front page questions currently on hold for being direct HW questions. Eg math.stackexchange.com/questions/433104/… $\endgroup$ – Noah Snyder Jun 30 '13 at 19:28
  • $\begingroup$ Noah, one example "It's not our fault that the SE team put "Homework questions must seek to understand the concepts being taught, not just demand a solution" under "off-topic". – Zev Chonoles 38 mins ago $\endgroup$ – Will Jagy Jun 30 '13 at 19:34
  • 3
    $\begingroup$ @NoahSnyder, anyway, you have a point, the needle may not have swung all the way in the direction i dislike. However, for the purposes of deciding what to do with an unsuitable question on MO, my conclusion is that migration to MSE will generally be a suitable vote for most things that are questions rather than rants, should MO people wish to try that out. $\endgroup$ – Will Jagy Jun 30 '13 at 19:38
  • 4
    $\begingroup$ @WillJagy A lot of questions that do not belong on MO do also not belong on MSE. Please keep the popular minimum quality standards at MSE in mind before transferring. As a moderator at MSE, I want also to add that moderators there support closing such questions but prefer ordinary users to do the closing. $\endgroup$ – Michael Greinecker Jun 30 '13 at 21:27
  • 4
    $\begingroup$ @MichaelGreinecker, I believe I understand. I'm active on both sites, and will continue to watch the consensus on homework and related issues. $\endgroup$ – Will Jagy Jun 30 '13 at 21:37

You must log in to answer this question.

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