Not signed in (Sign In)

Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.


    I recently noticed that StackOverflow attaches a special notice to old questions with a lot of upvotes that they don't feel reflect the site:

    This question exists because it has historical significance, but it is not considered a good, on-topic question for this site, so please do not use it as evidence that you can ask similar questions here.

    More info:

    What do people think about doing something similar here?


    How do they decide when to put one of those notices?

    If it comes up, I think such a notice is a good idea, but I don't think we should go out of our way to dig up questions to add the notice to. I feel like most of the early questions that would no longer be considered appropriate have been closed for a while, and I haven't seen the "there are greater evils" argument in a while. Maybe I just haven't been paying enough attention.

    • CommentAuthorgrp
    • CommentTimeAug 23rd 2011

    Can we do it (in the special way) without migrating to 2.0?

    While I think it is a good idea in general, I think the community is far from a strong consensus on many such question types that, in my view, might warrant such a notice. If you have a sample question for a test, Qiaochu, that might help focus this thread.

    Gerhard "Ask Me About System Design" Paseman, 2011.08.23


    As far as I can tell, such edits (and locking) can be done manually by moderators without difficulty.

    Note, however, the stats on the question pointed by Qiaochu: 2329 votes, 3912 favorites, 268 answers, etc. Nothing on MO comes remotely close to that!


    @François: On SO, these notices were put in manually by moderators as well, so Qiaochu's proposal is one of social convention. I like the idea of labeling questions which should not be taken to be representative examples, but for the most part, they are already labeled with "[closed]", which is even better than an ad hoc notice.

    Actually, now that I've thought about it some more, I don't understand these notices. Why was the SO question not simply closed (and locked, since SO is more prone to close/reopen wars than MO), rather than having this notice posted and then locked? Given that there are 268 answers, new answers are almost certainly undesirable.


    Maybe to stop people from complaining about it being closed? Who knows, maybe the "special" status with a historical reminder is enough to cool people down.

    • CommentAuthorquid
    • CommentTimeAug 24th 2011

    What I thought was the main goal of such a message, which I too noticed some time ago when starting to browse SO and more generally some SE site in the course of the migrating debate is this:

    It is there for new vistitors of the site. Somebody comes to the site, wants to get a first impression of it, looks at the highest voted questions, and thinks these are the 'best', which they are most often not, and perhaps asks a similar question that then gets closed as off-topic.

    To avoid this a clear messsage seems best, the closed or locked could easily be overlooked, or not understood, by somebody unfamiliar with the site.

    Since also many of the topvoted questions on MO are not representative of the site, I think in principle it could make sense to have this too. Whether there are 10000 or 1000 or 100 votes, is IMO not relevant; what is relevant is whether the topvoted questions are good example question for a new user; if not, clearly mark them as such to avoid misunderstandings.

    Whether it is worth the effort in practise, I do not now.

    Why not just treat the top 5 on the topvoted-but-poor-representatives-of-what-we-want list? Possible objection: "what if someone is filtering searches using tags?". Response: "if they're a newbie looking for a flavour of what is going on, they are probably not doing this".

    [I personally would long to go back to the days when most of the questions were well-formed technical precise mathematical questions at graduate-student-or-above level. I don't see how to steer the ship back this way though.]
    • CommentAuthorgrp
    • CommentTimeAug 24th 2011

    I had the privilege of talking with Scott Morrison on several issues, one of which was a two-tier system of questions. The sense I got from him, and which is likely shared by many of the moderators, is that the risks of introducing such a system to MathOverflow and adversely affecting the existting ecology are too great compared to the possible benefits such a system might obtain.

    Scott may be right, but I still think the idea deserves some discussion, so in response to Kevin Buzzard's longing, I suggest (drumroll please): the 'research' tier and 'other' tier. Most of the present mechanisms that exist would carry over to the new system, but the new features would be that everyone would have three new views to choose from, which would be to look at either or both tiers of questions, that all or many new questions would initially be posted to the "other" tier, and that a certain amount of work (votes, number of sponsors or detractors, reputation) was required to move a question from one tier to another.

    There are several ways to use such an arrangement; my intent in presenting it was to give those like Kevin an opportunity to view a community-filtered primarily research oriented list of questions and the opportunity to ignore a list of other questions about which the community is still deciding whether to tolerate and allow. I see the benefits in such a system and few downsides. I'd like to hear others take on this idea.

    Gerhard "That's Enough Drumroll Now, Thanks" Paseman, 2011.08.24

    • CommentAuthorMariano
    • CommentTimeAug 24th 2011

    I thought MO was the research tier...

    • CommentAuthorquid
    • CommentTimeAug 24th 2011

    Mariano, but not everybody might know or agree what it means.


    I don't like the two-tier idea very much, primarily because I feel like it would cause unnecessary fragmentation. If a question doesn't belong on MO, I don't want to have to answer the defense, "this is meant for the other tier, not the research tier."

    Still, I do feel that there is easily crowd-sourcable information that we're not taking advantage of. I've fantasized about how nice it would be to have a Netflix-style recommendation engine which would show you questions based in part on how your voting history correlates with the voting histories of other people. This would have two awesome benefits:

    1. It would tailer the site to the individual users in a non-fragmenting manner. That is, I think there are enough people in the community with broad interests to ensure that people still end up seeing a variety of questions, not just questions from their field. This would be even more true if there were a good UI mechanism for "recommending" answers as well as questions.
    2. More importantly, it would make votes meaningful. Suddenly people would have concrete a personal incentive to vote for/against the things they do/don't want to see more of. This would make raw votecounts a better measure of quality in general.

    I spoke to Jeff about this idea briefly a couple of months ago. I got the impression that he thought this would be desirable, but is simply technically infeasible. But the more I think about it, the more feasible it feels. Even if the recommendation system is very crude, benefit 2 kicks in right away. On MO, we have about half a million votes which have been cast by about 5000 users. This data set is about two orders of magnitude smaller than the training data for the Netflix prize. On this scale, it is reasonable to maintain a table of correlation coefficients between every pair of users.

    Even if it turns out that the algorithm is costly, it would be reasonable to restrict this feature to users who have cast at least X votes. With X=100, this narrows the list down to about 750 users.

    Edit: actually, the numbers are a bit smaller, since I was counting every kind of vote (including votes to close/delete, as well as accept/flag/favorite, which are also types of votes internally). If you only count upvotes (since this is easy to do in the public dump), there are 3179 users who have cast an upvote, and 646 who have cast at least 100 upvotes.

    • CommentAuthorgrp
    • CommentTimeAug 24th 2011 edited

    Anton, I think that it would be convenient to the reader to have a nicely configurable view that selects the questions most likely to be preferred by that reader. However, I see this as a potential fault from the poster's point of view. In the present system, a post that is made is guaranteed to be seen by a large (and statistically significant fraction of) body of experts who may be capable of answering the question quickly and professionally. This combined with the fact that the entry fee is nonexistent makes MathOverflow a braintrust for everyone, and we've seen the upsides and downsides to this.

    The two-tier system ups the entry fee by requiring new questions to be sponsored before guaranteeing that they will be seen. The self-filtered style you suggest may be appropriate if you can maintain the guarantee, or at least the appearance, that any new question will be seen by enough experts. At this stage, I don't see how you could maintain even the appearance of the guarantee with this addition. I also think the guarantee is too valuable to give up, and I admit concerns about messing with it with a two tier system.

    Gerhard "Ask Me About System Design" Paseman, 2011.08.24


    Returning to the original subject, I think it might be worth considering such special notices. As a thought experiment, which of the following questions are deserving of a special notice?

    These were selected just because they have 80+ votes and aren't obviously on topic.

    Please explain why you think the question should have a special notice instead of being closed as no longer relevant.


    @# François

    How about the memorable titles question?


    It didn't fit my (completely arbitrary) criteria. Why that one in particular?


    @François: "common false beliefs" and "proofs without words" would probably be redirected to math.SE nowadays, so perhaps they should just be closed? I am a little conflicted on "colorful language." The other questions you listed are closed and I think it should stay that way.

    • CommentAuthorquid
    • CommentTimeAug 24th 2011 edited

    Clarified version, and taking Qiaochu's anwer and François's request I initially overlooked into acount:

    From François original list, I'd say: all except 'false beliefs' and (perhaps) 'single paper' should have a special notice. The reason is that to me most of the other ones have a 'for fun' feel. I'd too put 'memorable titles', as while not conceived like this, it turned into a for-fun question.

    Now, I have nothing against the occassional MO-fun question, but the emaphasize is on the occassional; and it is hard to get this right.

    Why I do not think that 'dep closure' is a suitable MO question is document at great (perhaps excessive) length in the depth of this board.

    I agree on 'Thinking and explaining'; nothing against the question, but I'd say it is fair to say that a similar question, or at least a question somebody might consider as similar, would likely not have an easy time when asked (the comments on the question document this), and vaguely similar things do get occassionally asked with much less success (say, this somewhat recent 'flow state' question; yes, of course this is not the same, but to some 'outsider' it could have a similar feel).

    By contrast, 'false beliefs' is to some extent actually useful (if you disagree you claim that you are sure none of the statements could have fooled you, so think twice ;) , and if you pass that test there is still teaching) and perhaps also the 'single paper' as a collection of nice papers, I looked at some that were listed and would otherwise not have done so.

    • CommentAuthorNilima
    • CommentTimeAug 24th 2011
    Francois, and Gerhard, an interesting discussion you have sparked!

    I've drifted out of MO, in part because I find the soft questions irritating. Why? To my eye, many soft questions have answers which reflect opinion or taste. Why don't I simply look at the subset of questions with tags I like? Because I look on MO to learn what is going on in other subdisciplines. I just read. Filtering by my favorite tags means I don't get to see what other people are asking in other areas. However, *not* filtering means there's soft questions to sift through.

    My criteria for assigning a 'special badge': a question which engaged lots of mathematicians, but was about mathematical taste or culture (favorite mathematical jokes/puzzles/papers/titles). Entertaining, no doubt. Reflective of what one hopes the average question will be? No.

    NOT 'special badge' questions: and
    are both excellent community wiki questions. Both speak to how research mathematicians think. The questions are clearly spelled out, and responses do not involve matters of mathematical taste.

    Deserving of special notice because the questions seek answers which are matters of mathematical taste:

    @Nilima Note that you can also filter out undesirable questions, e.g. by the [soft-question]-tag. This way, you would still see mathematical questions from other disciplines.

    • CommentAuthorNilima
    • CommentTimeAug 24th 2011
    @Alex- thanks! Indeed, and I have this enabled.

    Unfortunately, some questions I'd categorize as soft (such as the ones above) don't have this tag explicitly. Filtering out a lot of other tags risks missing interesting questions. For example: 'big-list' includes several soft questions, but a few which are valuable for the intended purpose of collating lists:

    Good point. Maybe it's worth adding that tag to those questions (which would of course have the unfortunate side effect of bumping them to the front page)?


    Nilima's arguments make a lot of sense to me. I'm not too keen on any of the questions in question, but when singling out for a "special notice" then I find Nilima's choices good.

    One comment on the filtering-by-tags: I find this practically useless because very few people ever tag their questions correctly at first so whilst I could exclude "soft-questions" then I'd still see a fair number of them because they don't tend to be tagged "soft-question" at the start.

    • CommentAuthorquid
    • CommentTimeAug 25th 2011

    Nilima, mainly out of general interest, what you refer to by 'taste' would it include questions like this one or is it meant more narrowly:

    • CommentAuthorNilima
    • CommentTimeAug 25th 2011
    Quid, you raise an excellent question. I break up my response into the specific and the general case. The latter is argumentative.

    - The specific question 1420 indeed asks about a matter of mathematical taste. However, the OP provides some qualification about what is meant by 'best'. Further, the range of the question is narrow enough that though it is popular (40 votes), it's nowhere near as readily opined upon (compare with 'Common misconception' with 230+ votes and 53K views). The question, therefore, elicits a polling of taste from a smaller subset of the users: those who know many proofs, and are able to cogently defend a choice between them. Or so I hope.

    - The general question 'what is the best proof of X' can be made clearer. Best how? Does one seek the shortest proof? The proof which relies on the least machinery? The proof which successfully draws together many disparate fields? One that opens up new fields? All of the above? Providing a narrower definition of what the OP mean will have, as always, the effect of getting more relevant answers.

    Otherwise, the question devolves into a matter of mathematical aesthetics. This is interesting in and of itself, but then a clearer math-philosophy question could be asked.