There's recently been some discussion of the potential impact on voting of users whose only MathOverflow reputation comes from the network association bonus.

I asked the SE team about this, and Joshua Heyer kindly provided the following data.

I'll leave it to the thread below to draw conclusions.

  • 11
    $\begingroup$ The issue I see is that these votes (roughly 4% of all votes) by outsiders are concentrated on a few popular questions. If they followed a distribution similar to the other votes then I wouldn't consider this a problem. But when a large portion of these votes are concentrated on a few questions then it does create a problem IMO. (I am guessing they are concentrated, I don't have access to voting data to confirm it.) $\endgroup$ – Kaveh Jan 1 '14 at 23:43
  • 1
    $\begingroup$ It would be nice to see how these votes are distributed over the posts (i.e. break down of these votes according to posts in place of users, something like: post X, total upvotes 50, total downvotes 0, outsider upvotes 30, outsider downvotes 0). $\endgroup$ – Kaveh Jan 1 '14 at 23:48
  • 3
    $\begingroup$ Dear @cardinal, I wouldn't notice it if the outsider votes were distributed similar to MO user votes (considering that the total number of outsider votes are relatively small) so I think the concentration is more problematic. I feel the concern was a few questions attracting a relatively large amount of outsider votes, which is a systematic issue, not one about an outsider voting with ill-faith on MO. Personally the main issue is I see is that they skew the list of the highest voted posts (which I think of as the representatives of the sites' scope). $\endgroup$ – Kaveh Jan 2 '14 at 0:14
  • 1
    $\begingroup$ @Kaveh, I see. Admittedly, I have not followed this particularly closely here; though, personally I would find association-bonus-only voters exercising this privilege willy-nilly all across the site to be more pernicious than having it confined to a smaller subset of posts. My experience on other SE sites is that the highest voted posts are rarely the most representative and I would maybe go as far to say they are oftentimes among the least representative (though, I suppose it could be argued this is due, in part, to similar such behavior). :) $\endgroup$ – cardinal Jan 2 '14 at 0:19
  • 2
    $\begingroup$ It doesn't make a huge difference but June 24, 2013, is the first day that users could use association bonuses to vote on MathOverflow. $\endgroup$ – François G. Dorais Jan 2 '14 at 1:02
  • 3
    $\begingroup$ Thanks for these data on the cost of the association bonus. Of course, costs alone can't tell us whether the association bonus is a good thing. The costs need to be weighed against the benefits. Personally, I see essentially no benefits, so that even if the costs were very much smaller than your data indicate, I'd still favor eliminating the bonus. $\endgroup$ – Steven Landsburg Jan 2 '14 at 1:48
  • 14
    $\begingroup$ In recent days I've seen several "hot network questions" from MO advertised on other stackexchange sites (Mathematical research published in the form of poems, Teaching homology via everyday examples). Advertising questions like these to non-mathematicians is asking for trouble. Is there any reason why can't opt out of advertising specific MO questions on other sites? Nobody would even notice, while removing the association bonus would surely raise at least a few complaints. (I think it would be a good idea anyway, but this would an even easier first step.) $\endgroup$ – Henry Cohn Jan 2 '14 at 1:49
  • 2
    $\begingroup$ @ScottMorrison: Hmm, it already seems to be proposed at meta.mathoverflow.net/a/1307/4720, with a lot of support. I'd be happy to start a new thread - does that still seem like a good idea? $\endgroup$ – Henry Cohn Jan 2 '14 at 3:20
  • 2
    $\begingroup$ What is this concentration that people are talking about? Unless I have misunderstood, it seems that just a few percent of the 3528 votes in question are concentrated on the posts in the list, and the vast majority must then be not particularly concentrated. $\endgroup$ – Joel David Hamkins Jan 2 '14 at 16:37
  • 7
    $\begingroup$ My perception is that this data shows that this topic is basically a non-issue, since the problematic votes amount to about a hundred or so in all, and only a very small handful of questions affected. There seems to be no serious issue here. $\endgroup$ – Joel David Hamkins Jan 2 '14 at 20:10
  • 4
    $\begingroup$ @JoelDavidHamkins I agree it is not that big a problem in the end. But regarding only a handful of questions being affected: this is not that clear, this new hot list is very recent and we saw significant effects on 'Euclid' on 'not improved.' I would assume we will continue to see such effects on a significant proportion of questions of a certain form (lists and soft). Granted it is perhaps not that big an issue whether Euclid has score 50 or 25 but still it is a noticeable difference. Also I agree it would be interesting to expand from users with no points to users with very few points (con) $\endgroup$ – user9072 Jan 2 '14 at 21:03
  • 2
    $\begingroup$ (cont) Finally we could have some data to support the suspicion of some (or at least me) that a lot of the votes (though not all!) on the so 'popular' questions are due to people having no actual stake in the site. $\endgroup$ – user9072 Jan 2 '14 at 21:08
  • 4
    $\begingroup$ @Sebastian perhaps having a stake was not the best way to express what I mean. But, first, if somebody votes in the same way (as you claim users do) on math.SE and on MO this is already quite likely problematic. Second, this discussion originated from the hot list. One can only vote on questions one can see. So if some q is shown to much more users this will bias the voting significantly (and stronger if assoc voting is in place). I do not like this. And actually primarily I do not like this hot-list things and alike, if things like this would not exists I would not mind assoc voting so much. $\endgroup$ – user9072 Jan 3 '14 at 22:06
  • 2
    $\begingroup$ The question with 20 votes on math videos online was posted on reddit a few months ago; thag's why it had so many votes. $\endgroup$ – Brian Rushton Jan 7 '14 at 15:31
  • 2
    $\begingroup$ Same thing with the question on helping a young person achieve their potential. $\endgroup$ – Brian Rushton Jan 7 '14 at 15:34

I messed around a little with the spreadsheet.


My new columns are:

  1. ratio of votes/external votes. (Note that i used the quotient function which rounds the quotient to the nearest integer)


In the comments i was concerned about questions with ratio at most 4 (the spreadsheet is sorted by the ratio) and with total number of votes at least 10. I marked in the comments when such a question is cw, and added something about the name of the question when it appeared at least twice.

In the process i convinced myself that the problem of external voters is not that big, and opting out of the 'hot list' (the most popular answer here: Measures to separate math overflow from the rest of the stack exchange network) might do the job.


Here is a spreadsheet with more extensive data on the 385 posts with at least two "external" votes.


The columns are:

  1. Post ID.
  2. Total post upvotes minus downvotes (regardless of origin).
  3. Total post upvotes from users whose only points are from the association bonus.
  4. Post link.

In addition to these, there are 2451 posts with exactly one "external" vote as well as two currently deleted posts that had earned two "external" votes (150980 and 137245). Also note that users with only the association bonus cannot downvote since that privilege requires 125 points.

  • $\begingroup$ Is it possible to have the list updated with this question mathoverflow.net/questions/154913/… ? I must say I am curious how that has even stayed open, much less attracted so many upvotes (and I am thus curious as to whether those votes are from outsiders). $\endgroup$ – Tobias Kildetoft Feb 18 '14 at 7:53

You must log in to answer this question.

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