7
\$\begingroup\$

In a few questions I've raised in the past, it's been pointed out to me that it is not enough to just include the tags you categorize the Q into, but also call it out in the body of the Q itself. I've had this pointed out for a statistics Q, a designer-reasons Q, and most recently, a rules-as-written Q.

This is fair enough: it helps with clarity. But we don't seem to treat a family of tags the same way. Game system tags (3.5e vs 4e vs 5e D&D, for example) are added in without need of mentioning them in the body. And in fact, even if the body makes it clear which system it is using, if the Q is not yet tagged with the specific system (not the same as using the system-agnostic tag), we wait until the querent tags it with a specific game system first before answering.

Using game systems tags is one of the first things a new querent will learn. If they have a 5e question, it is easiest to find the 5e tag. Therefore, game systems have the most used, most visible tags, and they are the most demonstrative of how tags should work. If there is an exception being made for them that the other tags do not share, it is not entirely clear. If adding the 5e tag to a question is enough to mark it as a 5e question, then a new user will understandably rationalize that adding a rules-as-written tag will make it a RAW question. And adding a statistics tag will make it a statistics question. And so it goes.

Has this been discussed before? If so, where? If not, why do we consider game system tags to be different from all the rest, in the regard that you have to be explicit with the latter in the body, but the former does not?

\$\endgroup\$
  • \$\begingroup\$ This is the closest I know of to a discussion about your particular question: Are our implicit-information tagging practices becoming a problem? It's old enough that I think we could stand to revisit the subject. \$\endgroup\$ – BESW May 25 '17 at 3:16
  • \$\begingroup\$ @BESW Thanks! Good points raised there. This Q is bringing up the additional point that it trains new users how to tag, but due to the implied information in some tags (ie, game systems tags), they are being trained improperly. \$\endgroup\$ – user27327 May 25 '17 at 3:19
17
\$\begingroup\$

1. Because in practice most people don't read the tags

Based on many years of experience observing how questions and answers interact here, it is very common (not a majority, but definitely still a common event) for someone to read a title, find it interesting and click to open the question, read the body, and then start writing an answer without ever reading the tags.

(This is a source of a steady drip of answers that are off topic for non-D&D questions, since they skip reading the system tag, assume it's D&D, and write an off-topic answer based on D&D rules.)

Many people don't read tags.

2. Because it's not the answerer's responsibility to check the tags for extra info, it's the asker's responsibility to write a clear question first

Tags are for searching and sorting. Their job is to categorise a question after the fact, for the convenience of future searchers, and to help connect the question with the relevant experts. That's all tags are designed to be for.

When someone clicks through to a question, the behaviour observed and described above — answering based on the title and body, not the tags — is a pretty reasonable response most of the time. We here at Stack Exchange are about leveraging normal human behaviours to best create a repository of expertise on a site's topic. The more we have to fight human nature, the worse we're accomplishing our mission.

So, as a result, we cater to that normal human cognitive process — read question, write answer — and we have a small horde of editors who are ready to revise questions so that they are easily answerable, and we have tools to hold questions that can't be knocked into answerable shape by a 3rd party editor.

To this end, questions are expected to completely describe their problem, before any tags are applied, and be understandable without reading the tags.

(We make an exception for system tags because it works in practice, with the few times it doesn't work being low-frequency and causing little difficulty to fix by 3rd parties.)

3. Because we need to be able to agree how to tag a question without reading any minds

Tags belong on questions only if they categorise the question. It's not enough for the writer to “know” whether a tag applies, because we have to be able to curate it independently of their hidden brain-knowledge.

We need to be able to each individually look at a question, and be able to basically agree on how to tag it according to objective criteria. In the long term, that objective criteria can only exist in the title and body of the question itself, nowhere else, or it gets lost. We don't have permanent guaranteed access to the question-writer every time we need to reconsider how a particular question is curated; we don't have (public) access to old comments; we can't go haring off searching chat or meta on the off chance that the question was once discussed and there exists clarifying material elsewhere.

No, the question itself needs to fully contain its justification for its tags, and tags aren't self-justifying just by their existence (because tags categorise question content, not themselves).

\$\endgroup\$
  • 4
    \$\begingroup\$ It's also very common for answerers to answer the title without reading the body, but we do police that, at least sometimes. \$\endgroup\$ – Please stop being evil May 26 '17 at 21:37
  • 4
    \$\begingroup\$ Yes, the body is the answerer's job to read. \$\endgroup\$ – SevenSidedDie May 27 '17 at 7:34
9
\$\begingroup\$

Game systems are not demonstrative of how tags should work in general: their usage is exceptional and breaks the rules of how tags get used, but we do it that way because it works really, really well without causing any trouble. Every other tag is demonstrative of how tags should work: we tag based on information already in the question.

Tagging basics

Describing the content of the question is how tags work by default. From our help center on tags:

A tag is a word or phrase that describes the topic of the question.

When the text of a question seems to carry no indication of why a tag is there, the tag does not belong and gets removed and never added. Tags exclusively describe the content already there, and do not add new information themselves. A question about casting a fireball doesn't get (or keep) the armor tag unless it's clearly a question that is also about armor, for example.

That's the default, at least.

Game system tags are a useful exception

Some stacks have tags that are an exception to that rule. On RPG.SE it's the system tag: we don't need you to say "I'm playing D&D 3.5e" somewhere in every question if the question is tagged . This is like how on Stack Overflow nobody asks you to tell them the programming language you're using as long as your question is given a programming language tag such as or .

This is an exception to the norm. It's an exception we've developed because it causes no trouble and is pretty intuitive. So you're right, we don't treat them the same way. But this doesn't mean we can start using other tags the same way without causing trouble.

We did treat RAW that way for a while, and we specifically stopped because it was causing trouble. I'll get to that in moment.

And in fact, even if the body makes it clear which system it is using, if the Q is not yet tagged with the specific system (not the same as using the system-agnostic tag), we wait until the querent tags it with a specific game system first before answering.

That's not quite how it is. When it's clear what system they're using, we just add the tag and move on. However there's only one circumstance we count as being clear: they say explicitly what system they're using in the question text. In all other cases, we don't guess, and we do what you're describing there.

Rules as Written is not (any longer) an exception

Until 2016, we handled Rules as Written like system tags: if you add the tag to something then suddenly, regardless of what's in the question text, it's now a rules-as-written question and only rules-as-written answers count. This was a problem because a lot of people didn't understand what using rules-as-written meant, and thought it was equivalent to .

The community collectively endorsed a change to that plan in February 2016: A low-intervention approach [rules-as-written]: back to tagging basics. As the name suggests, we've gone back to the basic tagging practices I described at the top for how the tag gets used: the question gets the tag if it's clear the tag belongs, otherwise, it does not get the tag. If it has the tag despite it not belonging we ask for clarification on why they added it, and then modify the question to clarify why the RAW tag belongs or to remove the RAW tag because it doesn't.

There aren't other exceptions

We judge the content of your question and what tags it should have by what's written in the title and body. It gets tags that correspond to that content, and tags never add new information that wasn't already there. The sole exception is the system tag which is allowed to clarify the game being used all on its own.

Your question is only a statistics question as long as it would be read that way going solely by the question's title and body. Then we add or remove the tag based on whether it is truly a statistics question. Adding the statistics tag does not, itself, make an otherwise not-statistics question a statistics question.

\$\endgroup\$
  • 2
    \$\begingroup\$ Here is something to ponder. If the most used types of tags do not demonstrate how the rest of the tags work, and the exceptional nature of the most used type of tag is unclear, you will have users who believe it is justified that it is enough to simply add the tag in and the question will carry the new implied information. Those people will find their questions put on hold, and if they are new, they may get frustrated. Is this a desired process? Note that some people never discover meta, and those who do will not read back on every thread raised before they found it. \$\endgroup\$ – user27327 May 25 '17 at 14:30
  • 1
    \$\begingroup\$ @markovchain That's not a desirable side effect, but I believe the general feeling is that the alternative (requiring questions to explicitly say somewhere in the body which game system they're using) would be worse, in the sense of causing more distress and friction for using the site “acceptably”. So it's a trade-off. \$\endgroup\$ – SevenSidedDie May 25 '17 at 14:57
  • \$\begingroup\$ @SevenSidedDie Well, that may not be the only alternative. A third option is to accept that some Qs are fully described by their non-game system tags alone without clarification in the body. A fourth option is to include the exception to the rule status of the game system tags in the Help Center, or perhaps even in the Tour (if it's possible). \$\endgroup\$ – user27327 May 25 '17 at 16:03
  • 1
    \$\begingroup\$ @markovchain We have experience with the former, and what happens is that many answers answer only the body without paying attention to the tags and have to be removed. That annoys our base of answering users, which is worse than annoying askers. That experience is what informs the current best practices of requiring the body be a full description of the question. The latter doesn't work either: most people never read tag wikis or the Help or Tour. \$\endgroup\$ – SevenSidedDie May 25 '17 at 16:22
  • \$\begingroup\$ It's not meaningful or accurate to say they're the most used type of tag. Every other tag isn't sorted into "types", they're just "not system tags" -- and every single other tag demonstrates how tags should be used, and non-system tags get used a lot more than system ones since we apply 1-4 per question. The system tags also do not need to demonstrate how tags work -- they're exceptions, we can demonstrate how they work through practice of how we use every other tag in the system. I don't see people having much trouble learning how to tag, so I don't see a problem with this situation. \$\endgroup\$ – doppelgreener May 25 '17 at 16:23
  • 1
    \$\begingroup\$ Right, let me change the verbage on that. System tags are the most visible, or most prominent ones. They contain the most questions of the tags, and in a Q, they are the first tags you will see ordered left to right. If you are asking about a D&D question, it's intuitive to guess we have a tag for it. If you are asking about underwater combat, you may miss that a tag exists that fits that (some Qs are marked only with the game system tag). Re: people having trouble picking up how tagging works, I raised this meta because of 3 Qs I've asked where I ran into this. So you have one user at least. \$\endgroup\$ – user27327 May 25 '17 at 17:00
  • \$\begingroup\$ @SevenSidedDie It seems counterintuitive to me that the best solution is to do nothing about an undesirable result. At the very least, that we encourage people to take the Tour should mean we value what is there and not disregard it. But I am not pushing for a solution, I only ask "why?" If the status quo is the most desired result, then at least I know why. \$\endgroup\$ – user27327 May 25 '17 at 17:05
  • 1
    \$\begingroup\$ @markovchain Sorry, you lost me a bit. Which undesirable result do you mean? (I suspect it's a matter of inherent trade-offs that the SE design has deliberately accepted, but I can't be sure, nor explain it, because I've lost the thread.) \$\endgroup\$ – SevenSidedDie May 25 '17 at 17:12
  • \$\begingroup\$ @SevenSidedDie The one I said in my first comment. I was echoing your sentiments when you said "That's not a desirable side effect [...]" \$\endgroup\$ – user27327 May 25 '17 at 17:13
  • 3
    \$\begingroup\$ I'll point out that we don't tend to try to resolve problems until we can demonstrate they exist, and I don't know of any problem with things being the way they are now. \$\endgroup\$ – doppelgreener May 25 '17 at 17:14
  • 2
    \$\begingroup\$ @markovchain Ah, I see the thread now. Okay, so yes, that's not objectively desirable, but there are other undesirables in play, and it's a deliberate trade-off to sacrifice asker comfort. Question-askers are considered expendable by the design of SE, and the designers have explicitly said so, for good reasons. That essay explains why it's much, much better to annoy an asker than an answerer, within the SE design, and why we prioritise making experts comfortable, even if it comes at the expense of the question-writers. \$\endgroup\$ – SevenSidedDie May 25 '17 at 17:18
  • \$\begingroup\$ @doppelgreener Er, I'm sure you didn't mean to say I don't exist by that statement. Because you are talking to me. I just happen to be someone who visits Meta, I'm surely not a unique snowflake to have this problem on my own. If you did mean to disregard me, well, that's not ideal. I guess I don't mind. \$\endgroup\$ – user27327 May 25 '17 at 17:31
  • \$\begingroup\$ @SevenSidedDie I can accept that. Question-asking is quite jarring sometimes. I've no doubt high-rep users will also encounter these issues if they ask things as well. I suppose it's a bit surprising how the Stack's design disregards question-askers but still manages to pull in questions, but perhaps it just goes to show how truly expendable querents are, which in turn validates the Stack's design. \$\endgroup\$ – user27327 May 25 '17 at 17:38
  • 3
    \$\begingroup\$ Honestly, I really don't like the exception. Or, rather, I think it's fine as an accidental practice (I figure like 1/4 of D&D questions just treat "D&D" and "roleplaying" as synonyms, anyway - and there's no need to slap people's hands and waste their time if we perfectly understand their meaning), but busted once your formalize it and people start writing every question in that style on purpose. Having to look at the tiny little grey boxes on the bottom before you can understand the baseline context of a question is pretty unpleasant UI. \$\endgroup\$ – Alex P May 25 '17 at 21:53
  • 1
    \$\begingroup\$ @AlexP I'm not a superfan of it either. It causes headaches when people start thinking/talking about how tagging policy works (qv. this question, among others). But on balance, I think it works marginally better this way because it's the path of least resistance for new users, old users know more so that shouldn't be a big problem, and in practice most questions are full of “tells” about what game system they're about even when they don't say the name, so it's not quite as starkly information-poor as the imagined binary of tag-only vs. in-post identification. But yeah, not a super fan… \$\endgroup\$ – SevenSidedDie May 26 '17 at 15:27
5
\$\begingroup\$

Eh, I'm just going to repost my answer from the linked Q since we seem to be doing this again here, mildly edited.

This is not a problem specific to tags.

Someone can always make an unclear question. They can do it with words in the title, words in the question, or words in the tags. The site response to lack of clarity is, I trust, well trod enough not to require me to go over it. Get them to clarify, vote to close if needed.

Tags do convey implicit information, which is why they exist, have tag wikis, etc. It's good, and a useful shorthand. If someone wants a system-agnostic technique they don't need to spend a paragraph explaining "I want a technique I can use across the various games I run, man". I can only imagine most attempts at doing this would lead to even more pedantic arguing in comments.

This is a long way to go for a pretty simple answer.

For example, game system tags are pretty self explanatory. If it says [dnd-5e] then it can be assumed to be a normal D&D 5e question. Some uses of tags are clear and don't need further elaboration. (It's possible to have a question tagged [dnd-5e which isn't just trivially about the scope of D&D 5e, in which case you do need more information, see below.)

However, just adding a tag sometimes doesn't really give enough information, or at least is going to create bad answers. Let's take as an example "Give me a way to alternate melee and ranged attacks in one full attack" simply tagged with [feat] isn't going to sufficiently signal to an answerer that you only want feats as an answer to the question. Maybe it's just a "suggestion," maybe it's because you already mentioned a feat in your question, maybe it's because you're a noob and think "well of course the only way to get this would probably be a feat." This question is unclear despite the application of a tag. Your [statistics] question is an example of this - is the tag describing your question, the answers you want, is it just a semi related add on (people do this too much out of tag lust or something)...?

Rules-as-written is a special case, it's a meta tag that we are tolerating that IMO causes more problems than it solves, as it's not a term unambigiously clear to everyone, so you really need to use your words more (scope of sources allowed? what do you personally mean when you say RAW?) when using it as a tag.

If your question is not clear - make it clear, using tags, words in the body, words in the title, etc.

If you do not understand a question - including its tag use - ask for clarification. It's the one actual intended, constructive use of the comment system.

\$\endgroup\$
  • \$\begingroup\$ How did the statistics question create bad answers, or did not give enough information? Is there a case in which a tag describes only the question, but not the answers the querent wants? Is there a case where it describes only the answers, but not the question? How can a tag be a semi-related add-on, and is that accepted practice? Can I just add in a semi-related tag to any given question? \$\endgroup\$ – user27327 May 25 '17 at 14:37
  • \$\begingroup\$ Because people weren't answering with statistics. Tags should always only describe the question. You shouldn't add semi-related tags but people commonly do. No. \$\endgroup\$ – mxyzplk - Justice for Monica May 25 '17 at 14:53
  • \$\begingroup\$ @markovchain I often edit tags, and one of the reasons is to remove semi-related tags. Often people will ask a question and then add tags not categorising their problem, but categorising their reason for asking (e.g., asking a rules question and adding [gm-preparation], when the problem to solve isn't about prep, but they have the rules question about something they're prepping). \$\endgroup\$ – SevenSidedDie May 25 '17 at 14:59
  • \$\begingroup\$ The answers are all statistics-related: 1, 2, 3. Even the deleted answer is statistics-related. The other deleted answer misunderstood the Q, but was quickly corrected by three other users. \$\endgroup\$ – user27327 May 25 '17 at 15:01
  • \$\begingroup\$ @SevenSidedDie I get why some tags are removed from various Qs. But I am asking about this particular statistics Q of mine, in which it apparently was possible to mistake the addition of the statistics tag as a semi-related add-on. Was it, truly? Did its inclusion in the Q as a tag really not give enough contextual information about what the Q was looking for? \$\endgroup\$ – user27327 May 25 '17 at 15:05
  • \$\begingroup\$ @markovchain The [statistics] tag describes the core problem, but that fact wasn't initially clear. Before adding the tag the Q was unclear to a number of people, because people weren't catching that “expected outcome of the roll” meant statistical outcome, rather than “what does this rule do”. After adding the tag it was still unclear, because that didn't make enough people revise how they understood “expected outcome”. After doppelgreener's edit, the body and tags were in harmony and it was fine. It was never only a semi-related tag, but that wasn't obvious until the body was revised. \$\endgroup\$ – SevenSidedDie May 25 '17 at 15:18
  • \$\begingroup\$ @SevenSidedDie The tag was there from the beginning. Doppelgreener edited the body to add the word "statistical" -- which I don't have an issue with, but that word was already part of the tag. The answers it garnered makes me believe the ones who answered, understood the Q from the beginning. Finally, another question has a similar structure before Doppelgreener's edit, and it has a score of 121. \$\endgroup\$ – user27327 May 25 '17 at 15:25
  • \$\begingroup\$ @markovchain Ah, right, I got the timeline a bit wrong. Anyway, the other question's body was more clear because of the word “average”. That's a stats term that's clearly asking for something that is definitely statistical analysis and couldn't be mistaken for “what does this rule do”. There was one answer that was based on that misreading, so that indicates that a bit of clarity was missing, despite the tag. I agree it's a borderline example, but a very small amount of unclarity did have a demonstrable negative effect, and it's avoidable with a bit of an edit. \$\endgroup\$ – SevenSidedDie May 25 '17 at 15:31
  • \$\begingroup\$ @SevenSidedDie Yes, the other Q had the word "average" whereas mine had the word "expected," which are interchangeable in regular English, but also it is a more popular statistics term, as in Expected Value. I am not trying to bog down this line of reasoning to just my own Q specifically, but I am trying to show that sometimes, the tags sufficiently describe the Q itself. As the Q about Adv/Disadv shows, and my own Q shows. Again, yes there was one answer based on a misreading, but three people other than me went in and clarified the Q on my behalf. \$\endgroup\$ – user27327 May 25 '17 at 15:47
  • \$\begingroup\$ @markovchain Both questions had “expected”, the other had “average” in addition. “Expected” only means stats if you're already thinking of stats, which one would as the writer, but not necessarily as the reader. However, at this point, I have to ask: what is your objective in these comments? They seem to be splitting hairs finer and finer, rather than accepting that “make sure the question is clear sans tags” is the expected way to write questions and use tags (and editors will come along and fix it when it doesn't happen, so perfection isn't needed). What's the problem with accepting that? \$\endgroup\$ – SevenSidedDie May 25 '17 at 16:27
  • \$\begingroup\$ @SevenSidedDie There is no problem with the accepted best practice. My original Q was "why do we make this exception?" Tangentially, this A has called out my statistics question of being unclear. I tried to show it was clear pre-edit, with evidence. Pre-edit, the statistics Q did not contain the word "statistics" in its body though it had the tag. To think the statistics Q was unclear pre-edit, to me, seems almost an intentional attempt to misunderstand it. But I have no issue with the Q in its current form, nor its old form. It seems to me it's this A which has an issue with it. \$\endgroup\$ – user27327 May 25 '17 at 16:53
  • \$\begingroup\$ @markovchain One person who wrote an answer misunderstood it not-intentionally. Since we have to think of our readers too, we can take that as a (very) rough idea of the proportion of readers who misunderstood the question (approximately 1 in 5). Since we have orders of magnitude more readers than answer-writers, that's a lot of people potentially misunderstanding the question. Our goal is to have as few as possible people misunderstand our questions, since clarity is achievable and misunderstanding is unnecessary, and understandable Q&A is the ultimate goal of the site. \$\endgroup\$ – SevenSidedDie May 25 '17 at 17:09
  • \$\begingroup\$ @SevenSidedDie Using the new information I gained from you, the Stack prefers to optimize for answerers. In the end, the answers that stuck are the correct ones, and the one answer which misunderstood the Q was self-deleted. As far as the pearls go, there was no problem. The 6 readers who downvoted, and the 3 readers who corrected the self-deleted answer in comments, seem to have understood the Q correctly. Would you have read that Q and thought it was not asking for statistics? But I digress, the "health" of querents is unimportant to the Stack. That invalidates my position here as one. \$\endgroup\$ – user27327 May 25 '17 at 17:50
  • 2
    \$\begingroup\$ Yeah, this seems like a mountain out of a molehill about that one question. "A word was edited to make it clearer?!?" \$\endgroup\$ – mxyzplk - Justice for Monica May 25 '17 at 23:42
  • 1
    \$\begingroup\$ @mxyzplk Your answer was not as clear as you wanted it to be or thought it was, and this mountain of a molehill formed. \$\endgroup\$ – user27327 May 29 '17 at 3:18

You must log in to answer this question.