26
$\begingroup$

Reading the site, one finds that there are a number of questions which start out innocently enough: The question-asker is possibly only looking for a straight forward answer to his narrowly tailored, possibly computational, question and is pleasantly surprised -- indeed, we all are -- with a magisterial and deeply insightful answer which is recognized as such and promptly highly voted up by the community. Or, rather more properly, by the community that reads the sorts of questions he answers. These answers, to me, are the stuff of lasting intuition; some of them make me want to get out a pad and paper to immediately begin applying their insights; I'll probably eagerly quote some of those ideas contained therein in the future to my friends.

Now, one can typically find many of these kinds of answers by looking at the profiles of the highest rep users; our two highest-rep users are in particular notorious for them, and there are plenty of gems to be found in the profiles of, say, the top 25 highest rep users. One might also click through to the "nice answer" badges, and one might find a couple of recent ones there as well. (Just like in Google's search results, I'd bet that no one clicks past the first page).

There are, however, a couple of problems with these approaches. For example, clicking through to badge and user pages only gives you the titles of answers, which most of the time, are typically oblique because the people posting their questions are in a hurry for responses. Recently, I've been trying to remedy that situation a bit by renaming the titles of particular questions that have very nice answers to guide future searchers (I think commenter Bill Dubuque, who has is a pedagogical bent, is doing the same same). Moreover, in the badge pane, the questions are rank-ordered by date, not insightfulness, and worse one has no a priori way of telling how useful answers to individual questions might be to one personally. So my very first feature request is that nice answers and famous questions and such should be tagged.

There is also the problem that most people here are not generalists: They only answer topics they think they know well, and, since they are not amongst the highest rep answers, at least some percentage of their answers will be pedestrian. And, because voting tends to be populist, and led by the lowest common denominator of the site's user base, vote rank is not entirely synonymous with quality, as has been noted before, at least between tag specialities. People who post really remarkable answers to questions to set theory, like one user I discovered from one of my recent questions for example, aren't going to be as acclaimed by the community as those who specialize in number theory -- that's just the way it is. And finally and mostly, for some people the perfect question and the inspiration to give a truly classic answers only comes together like that of a lightning strike -- rarely.

This could all be rectified if we had a wiki-like "best-of" page where resident experts, possibly as determined by intra-tag rep ranking or rep thresholds, could highlight particularly nice answers for different communities, as organized by topic. There would be very little or no commentary or discussion there, so as not to create a detrimental "fork" against the main purpose of the site. They would be able to pick out hidden gems, and collate them for the benefit of future readers, which might help attract experts for undeserved communities like, say, algorithms and theoretical statistics. I've noticed that this is already being done organically, through comments to the main answer, and rarely, in community-wiki type posts, but it would be really nice to get something official from StackExchange management.

$\endgroup$
  • $\begingroup$ I received a downvote. I don't care much who or why, but you should make your thoughts somewhere in answer, that's why this is tagged "discussion." $\endgroup$ – Uticensis Mar 8 '11 at 9:04
  • $\begingroup$ amusing typo: "undeserved communities" s/b "underserved communities" :) $\endgroup$ – Mike Jones Dec 21 '11 at 12:49
  • $\begingroup$ This post is (to some extent) related: meta.math.stackexchange.com/questions/10351/star-ranking-system $\endgroup$ – Martin Sleziak Nov 26 '14 at 18:37
10
$\begingroup$

this can be done on the tag wiki pages for the tag, e.g.

https://math.stackexchange.com/tags/number-theory/info

Note that the faq tab here is automatically populated by intra-post links, that is, the more linked a post is, the higher it will appear:

https://math.stackexchange.com/tags/number-theory/faq

$\endgroup$
  • $\begingroup$ Then perhaps we're talking about a design flaw then? Isn't it a good question of why no one is using the tag-wiki, at least on every site that isn't StackOverflow, I think. It's hard to navigate to it from the main page, I had to struggle to find it from the Question pane just now looking for it. Perhaps it could use a bit more structure, like a dialogue menu, to make it easy and pain-free to add an answer? $\endgroup$ – Uticensis Mar 6 '11 at 0:22
  • 1
    $\begingroup$ @bill it's visible by mousing over any tag and clicking "info", so it is one click away from any tag; see screenshot at blog.stackoverflow.com/2011/01/… $\endgroup$ – Jeff Atwood Mar 6 '11 at 5:01
  • 1
    $\begingroup$ I may have been too flowery above, so let's make things concrete: I'm a student, trying to understand the kernel of a linear transformation. I post a question, and someone rewards me with a computation. I've almost got it, but I'd like to know more. Now, it'd be nice to simply click a page titled Linear Algebra, go to a section labeled "Undergraduate", and read various answers arranged topically: Rank, kernel, Jordan decomposition, eigenvectors, the spectral theorem. I wouldn't have to go searching through "I'm having trouble solving this.." questions hoping to stumble on a nice digression. $\endgroup$ – Uticensis Mar 8 '11 at 7:11
  • $\begingroup$ @Jeff Atwood Oh, and thank you for the redirection. I still think that page is too hard to find and naturally edit, though. I think it might be plausible to serve as an OK guide for someone trying to quickly tag his question, but that's not the kind of purpose I'm really appealing to in my OP. $\endgroup$ – Uticensis Mar 8 '11 at 7:19
  • 1
    $\begingroup$ @bill is there any reason math.stackexchange.com/tags/linear-algebra would not work? notice the "related tags" on the sidebar, and the faq | votes tabs at top, and the wiki summary below that with relevant links? $\endgroup$ – Jeff Atwood Mar 8 '11 at 7:21
  • $\begingroup$ @Jeff Atwood "Relevance" as determined by your proprietary algorithms isn't leading me to the most insightful answers. Most of the time it leads to a routine computation, most because that's what the majority of questions ask for. If I click on "votes" in the "linear algebra" tab, I'm similarly not getting what I'm looking for -- that's way too broad! I don't want to learn about every single linear algebra topic; I'm looking for something on only "kernel." And, finally, a frequently asked question queries a different property than insightful answer, they're just not the same thing. $\endgroup$ – Uticensis Mar 8 '11 at 7:30
  • 1
    $\begingroup$ @Jeff Atwood Also, please in the future kindly refer to me as "Billare." I chose that name because I liked it, it has personal meaning to me, and in general I'm not fond of pet names or abbreviations. $\endgroup$ – Uticensis Mar 8 '11 at 7:32
  • $\begingroup$ @Jeff Atwood I think the gist of post can be analogized to the advantange Mahalo and other startups believe they have over Google: that is, leveraging human-powered search. If VCs will fund that idea, it can't be totally off-the-wall. Now, I still think Mahalo is mostly a terrible idea, because it doesn't scale, but you don't have that problem here: Your primary searchers are also going to be your primary compilers, you don't have to pay them -- they're eager to work for zip organizing the information contained in answers to use it as a knowledge base. Wouldn't it be nice to leverage that? $\endgroup$ – Uticensis Mar 8 '11 at 7:41
  • $\begingroup$ @billare sure, but I'm still not clear why the existing tag wiki wouldn't meet that goal? $\endgroup$ – Jeff Atwood Mar 8 '11 at 8:03
  • 1
    $\begingroup$ @Jeff Atwood The fact that the tag wiki feature is not being used on any other site other StackOverflow* -- and its not exactly being used much there if Python, a major language, has only a short blurb -- indicates IMO that it is not "meet[ing] that goal." That's why I suggested a design flaw: If you think you've coded the feature in, and no one is using it that begs for explanation; usability designers are exactly the people whose job it is to figure those problems. I can't be more specific with you than that, because I am not a designer. $\endgroup$ – Uticensis Mar 8 '11 at 8:29
  • $\begingroup$ @Jeff Atwood And really, as you had a hand in programming this site, you should understand how important the design aspects are here. I am still struck at the cleverness of making comments read in small font and making long comment threads collapse to minimize flame wars; those design choices on management's part are what enable your network of sites to be a much better place to get answers than Wikipedia's Answer Desk, though in the essentials one might say you have the same "features." $\endgroup$ – Uticensis Mar 8 '11 at 8:32
  • 1
    $\begingroup$ @billare perhaps you should check out the java tag wiki? stackoverflow.com/tags/java/info the python one is, well, pythonic. It is what you make of it. $\endgroup$ – Jeff Atwood Mar 8 '11 at 8:49
  • 1
    $\begingroup$ @Jeff Atwood Huh, I see. In any case, I've said my piece; if you're unmoved by previous stuff, there isn't much else fruitful I have to offer. Good night. $\endgroup$ – Uticensis Mar 8 '11 at 9:02
9
$\begingroup$

I wrote a couple of Perl scripts a while back to do just this with the database dumps for MO. I'm pretty sure they'd work for MSE, since the dumps are in the same format (though the copyright notice should be changed).

$\endgroup$
9
$\begingroup$

This would help a lot. I have had two occasions in the past weeks to refer to one of Arturo Magidin's posts in another answer and had some trouble finding it as I didn't remember which question it was attached to. Both times I found it, but it took a bunch of searching, even though I had a good idea what I was looking for. But don't be too attached to the highest rep users-I am in that class, but I don't I have a lot of answers that would make the "best in class". I just have a lot that some people (most times) think are decent. One terrific answer doesn't vault you that many points.

$\endgroup$
  • 2
    $\begingroup$ The user: search feature is really useful for this. For example, putting user:742 trigonometry into the search box returns two answers that Arturo Magidin contributed that are about trigonometry. $\endgroup$ – Isaac Mar 8 '11 at 6:58
  • 1
    $\begingroup$ @Isaac: how would I know that Arturo was user:742? I don't even know my own user number. $\endgroup$ – Ross Millikan Mar 8 '11 at 7:03
  • 1
    $\begingroup$ @Ross: Click on his username/icon anywhere on the site (or the meta)—the user list page works for this, too—and look for the number after /users/. For example, clicking on my username here goes to http://meta.math.stackexchange.com/users/72/isaac, so searching for me on meta.math.SE or on math.SE would use user:72. (User numbers are not consistent across different SE sites.) $\endgroup$ – Isaac Mar 8 '11 at 7:06
  • 1
    $\begingroup$ @Ross Millikan I beg to differ on the quality of your answers :) However, my point wasn't that only high rep users give nice answers, not at all, I know there are lots of people who post intermittently on stuff they think they know very, very well. However, I do think high rep users are more likely than average to recognize nice answers and able collate them, and really, I only suggested a rep calculation to avoid possible confusion or chaos on such a page. $\endgroup$ – Uticensis Mar 8 '11 at 7:56
  • $\begingroup$ @Ross: Just out of (morbid) curiosity, which answers? I confess I haven't found the search feature very useful in the past in trying to locate questions. As for user numbers, I always remember mine by looking it up, e.g., clicking on one of the tag counters on my user page. $\endgroup$ – Arturo Magidin Mar 9 '11 at 0:16
  • $\begingroup$ @Arturo: I think there were two, the one I really remember was induction as shown in math.stackexchange.com/questions/19485/… $\endgroup$ – Ross Millikan Mar 9 '11 at 0:34
  • $\begingroup$ @Ross: Like I said, just morbid curiosity... Thanks. $\endgroup$ – Arturo Magidin Mar 9 '11 at 0:42

You must log in to answer this question.

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