In a recent blog post, Jeff Atwood explains their new policy about meta-tags:

From this point on, meta-tagging is explicitly discouraged.

But, what's a meta-tag? How do you differentiate a meta-tag from a normal, innocent tag? Jeff gave two conditions to identify a meta-tag:

If the tag can’t work as the only tag on a question, it’s probably a meta-tag.

Every tag you use should be able to work, more or less, as the only tag on a question. Meta-tags, like [beginner], [subjective], and [best-practices], are useless as single tags on a question. By themselves, they tell you nothing at all about the content of the question.

If the tag commonly means different things to different people, it’s probably a meta-tag.

In a cruel, ironic twist, the meaning of the tag [subjective] itself … is actually subjective. Ditto for [best-practices] and [beginner]. Best practices to whom? Beginner by what criteria? These tags are impossible to define by anything remotely resembling an objective metric. In comparison, the the meaning of tags like [java], [c#], and [javascript] are crystal clear to all but the nuttiest of nutbags.

Our parent site has already its own set of meta-tags: [subjective], [beginners], [best-practices],... What should be our stance about meta-tagging?


It took me a few minutes to really get the point behind that blog post, but I think I get it. It's not that "best practices" shouldn't be asked, or that "subjective" shouldn't be asked.

Just that adding those tags to the questions adds nothing.

The [subjective] tag just means that I can click on the tag and get a list of subjective questions... What am I going to do with that? How is it helpful?

Clicking on the [LuaTeX] tag is useful, because it gives me a list of questions about a specific subject that I might care about or know a lot about (I don't, but that's not the point ;))

The same is true for [beginners]: who would want to click on the [beginners] tag and get a list of beginners questions? Why? Do we have any resident experts on beginners questions who might want to track those questions specifically?

If a question is subjective, or is asked by a beginner, those things are typically obvious from reading the question itself. When someone asks "how do I get started with LaTeX", we know they're beginners, we don't need a tag to tell us that.

So he might be right, maybe we should just discourage all metatags. My litmus test is "would anyone have a reason to click on the tag or follow its RSS feed?"

No one does that for "beginners" or "subjective" or tags like that. So why do we have them?

  • (Re: No one does that) but one can add "beginners" or "subjective" to ignored tags, btw – Grigory M Aug 8 '10 at 8:59
  • I for one completely disagree with this analysis – see my own answer. But specifically, your litmus test presupposes (without arguments) that tags have one purpose only, and that’s trivially wrong, in my view. For example, I’ve never used the RSS feeds, and I rarely click on tags, but I find tags tremendously helpful. – Konrad Rudolph Aug 9 '10 at 8:41
  • @Konrad: helpful with what though? They convey information about the question, sure, and [beginners] may have its uses there. But 99% of the time, it's apparent from the question anyway, or it is irrelevant for the answer. I agree with you that there are metatags that provide useful information about the question, but [subjective] isn't among them, and [beginners] probably isn't either, most of the time. – jalf Dec 2 '10 at 15:07
  • I use tags to quickly classify a question. Tags have an enormously high information content. And of course this also holds for meta tags. [subjective] is telling a lot about a question. For one thing (when perusing things that might help me) that there won’t be one true answer, but rather a kind of cheat-sheet. [beginners] tells me what level of explanation to expect in the answers (again, this is eminently useful when filtering search results). And so on … – Konrad Rudolph Dec 2 '10 at 15:17
  • And I’m sure, if you would use machine learning to cluster the questions on SO (without regard to tags), questions with the same meta tags would form very tight clusters. What do I mean by that? I’m claiming that meta tags are are descriptions of properties that are inherent in the question/answers data, not merely arbitrary labels. (In fact, this would be a very interesting experiment to make. As it stands, I realize that the claim is purely hypothetical and not much use in an argument. Still, I’d be much surprised if this were wrong.) – Konrad Rudolph Dec 2 '10 at 15:22

I completely disagree with Jeff’s decision. I think it was short-sighted and wrong. I think that meta tags do add valuable information.

Take the [beginner] tag – of course most people (but perhaps some?) won’t search for questions using that tag. On the other hand, the tag gives a valuable hint about the skill set of the user, and can be used to formulate the answer accordingly. Not using the tag would mean either of two things:

  1. The users get a bunch of highly technical, unintelligible, effectively useless answers, or
  2. The users have to specify in the question that they’re beginners, using a full sentence (e.g. “please use plain English, I’m just a beginner.”).

Now, first of all alternative 2 is actively discouraged on SO. Many questions are purged of such phrases. Furthermore, just specifying the tag is so much easier, both for the question asker and for potential answerers.

That said, I don’t think we should deviate from other sites’ policies. I.e. meta tags should be discouraged, and they should be removed from questions, provided this leaves the question intact.

  • So, I still don't get it. For example, when is it OK to tag a question as beginners? – Juan A. Navarro Aug 9 '10 at 9:30
  • @Juan: whenever a question is asked that can be misconstrued as asking after details, instead of wanting a very basic answer. E.g. “what does `\newcommand do” elicits the answer “It defines an own macro” instead of a complex technical answer that explains its implementation in terms of TeX. – Konrad Rudolph Aug 9 '10 at 10:18
  • 3
    Ideally it gets answer that says "It defines a macro" in the first paragraph and then explains technicalities in a separate paragraph. Then the answer is useful for everybody. – Caramdir Aug 9 '10 at 16:42
  • 2
    Also from experience, few people identify themselves correctly as beginners (I think there are even studies confirming that). However usually the wording of the question makes it clear what type of answer is expected. – Caramdir Aug 9 '10 at 16:44
  • [beginners] might be useful for the select few that can see that they are and are looking for helpful info? Then again, maybe not ... I haven't really looked at all of those questions. – SamB Dec 2 '10 at 3:28

My sympathies are more with Konrad on this one. On this issue, I think that we're closer to the MO model than the SO model.

That said, if Jeff had said "let's get rid of useless tags and ambiguous tags" then I wouldn't have had any argument with him; of course, there'd've been a discussion about which tags were useless or ambiguous, but that's at a lower level than simply decreeing all "meta-tags" (what they?) as "to be discouraged".

I think that Konrad makes a very good point in his comment on Jeff's blog: that tags are not "one thing" but serve several purposes. Here's where I think that we'll be closer to MO than SO. On SO (I imagine), almost every question should get a "language" tag. Indeed, the few times that I've been to SO, I've almost immediately gone to the latex tag. I do that because only there do I have a hope of contributing, so why would I look anywhere else? (Of course, if I have a question then it might be about, say, PHP and then I'll search within the PHP tag to see if that question has already been asked.) On MO, we have certain "top level" tags which correspond (almost exactly) to the classifications on the arXiv. However, I don't use that to filter the questions. I will scan through all the questions on the main page, and often click through to questions that aren't tagged 'at.algebraic-topology' because there are often questions in other areas where I may be able to contribute.

So what's the difference between MO and SO, and why are we more like MO? In short, the boundaries are not so well-defined and that's MO behaviour, not SO. It's hard to think of any tags that we could have that would match the "language" tags on SO. They would have to fulfil the following criteria:

  1. A questioner should know from their question which of the major tags to use.
  2. An answerer should know from their expertise which of the major tags they can contribute in.
  3. A browser should know from their interests which tags to browse through for interesting questions.

I can't think of any here that match those.

One oft-cited use of tags is as a way of filtering out the stuff you don't want to see. That's certainly true of my use of tags, but because of the lack of definition in the boundaries, tags are simply suggestions and can be freely ignored. Thus I can't trust the RSS feed for a given set of tags simply because I'll miss so much: both opportunities to help and chances to learn. If I ignored all questions tagged "luatex" (since I don't, at present, use it) then I'd miss out on the opportunity to learn about it and learn why I should use it.

So thinking about Jeff's criteria for "what is a meta-tag":

  1. If the tag can’t work as the only tag on a question, it’s probably a meta-tag.

    Frankly, that's ambiguous. What does that mean? I could tag a question latex and that would "work" because the question is about latex, but it doesn't help anyone so it's a useless tag. On the other hand, in conjunction with other tags, latex could help a lot: it says "much as I would like to learn how to do this in ConTeXt, I really only want a LaTeX solution". Why should tags stand alone? Why can't they build on each other and modify each other?

  2. If the tag commonly means different things to different people, it’s probably a meta-tag.

    No, that's just a bad choice and so is already discouraged.

I'd rather say: each tag should slice out an area of the "designated field" of this website. An ambiguous tag is one that slices out an area with a fuzzier-than-usual boundary, and a useless tag is one that doesn't do any actual slicing. But note that both depend a bit on other tags. Clearly, both pdflatex and latex on a question is redundant, but both could stand alone and be useful in the right context. To be even shorter:

A good tag is one that helps a potential answerer come to the right conclusion as to whether or not they should look at the question

In conclusion, I would rather the discussion be about:

  1. Identifying useless tags and dealing with them. If it's a tag that is sometimes used incorrectly, then we (the higher rep users) simply need to be vigilant about them. If it's a tag that is used incorrectly much more often than not, then there's a case for actively discouraging it.

  2. Identifying ways to make the tagging system clear (particularly to newcomers): it's purposes (note the plural) and tips for how best to use it. In particular, if a tag is getting abused, but is still felt as potentially useful, we can consider how to discourage people from using it (perhaps as simple as renaming it).

  3. Figuring out exactly what those purposes are. As I've tried to argue (perhaps a bit rambling), here's one area where importing stuff as-is from SO ain't gonna fit, and we should think a bit more carefully about what would make this site the best for the target audience.

  • I'm tempted to use "TL;DR" for the first time in my life, but I won't. :-) [because I did] – Matthew Leingang Apr 7 '11 at 12:55

For me, at least now, it's not cristal clear that all of these so-called "meta" tags are useless. I agree that some like [subjective] or [soft-question] we could get rid of, but probably some other "meta-tags" do have a clearer meaning in the (La)TeX community.

For example we all know about LaTeX's steep learning curve, and I kind of think that a tag such as [beginners] could be made useful in that context. The "could" is because us, the community, have to make sure that this is the case. But I imagine a new user stumbling with the site being able to click on the [beginners] tag and find a set of questions/answers about how to install the system, how to learn the language, some common pitfalls to avoid.

I could also make a case for [best-practices], which in the LaTeX world also has a somewhat "clearer" definition, e.g. what l2tabu tries to achieve.


A second more radical thought. Yes, lets get rid of all those nasty meta-tags.

  • [subjective] questions shouldn't be asked in the first place.

  • [beginners] should instead be tagged as [installation] or [learning].

  • [best-practices] (or at least part of its meaning) could be saved to mean something like "l2tabu"-ish advice. I was thinking something like [pitfalls] but this could have a negative connotation if a question is retroactively tagged like that. Any ideas for this?

  • [soft-question] and [big-list] shouldn't be used, they are tags about the question. A tag about the content or topic of the question should be used.

  • I disagree with the "subjective" thing. Of course subjective questions should be asked: "what is a good professional font to use in my TeX document" is subjective. It is also answerable, informative and a question people want to know the answer to. Some subjective questions are perfectly fine. But it's possible that the tag should be removed. Apart from that, isn't [learning] a metatag too? Why is [pitfalls] better than [best-practices]? – jalf Aug 7 '10 at 13:07
  • [big-list] is very much meta, and fails the can't work alone test, but it really is crystal clear. That seems to be the crux of the issue. And it's useful. Our 3rd-most used tag, the rigorously objective [macros], seems to cause more confusion in its application (not to mention our most used...) – Charles Stewart Aug 9 '10 at 15:52
  • @jalf: it's much easier to identify pitfalls then figure out which practices are "best"? Who knows... – SamB Dec 2 '10 at 3:34

Looking at the lists of questions tagged with various meta-tags it seems that they are already used inconsistently:

  • [beginners]: One of the questions is something I didn't know (though I guess I should have), one is a discussion, two are actually beginners questions and one is a proposed faq entry. While in theory a [beginners] tag could be useful, I do not see any chance of it being consistently used. People in general do not want to identify themselves as beginners and retroactively tagging something as [beginners] might be seen as an offense by the original author. However, I'd suggest a [basics] tag for the purposes Juan sees in the [beginners] tag. I think less people are offended by that.
  • [subjective]: Two out of four questions are closed, one is (in my opinion) off-topic and one is actually an on-topic subjective but useful question. I think this tag is superfluous.
  • [soft-question]: I don't know what is "soft" about the package question. The other two are fine. The tag might be useful to for marking questions that are related to latex, but do not need any technical knowledge about LaTeX to answer.
  • [best-practices]: The "professions" question seems a bit odd in [best-practices], the others are okay. However, all answers should be best-practices so I don't know when to use this tag. There are lots of questions not tagged with the tag, that should arguably be tagged with it if wee keep the tag.
  • I also started to think along the lines of your "[basics] instead of [beginners]" but that doesn't solve the "problem" (if there is any problem at all) because [basics] is still a meta-tag, as defined in that blog post. – Juan A. Navarro Aug 7 '10 at 10:59

You must log in to answer this question.

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