Discussion about the tag system at MSE

Question

I'm surprised to read what a co-founder of Stack Exchange thinks about the tagging system. From meta.discourse.org :

we saw on Stack Exchange that

1. Users don't understand how to navigate via tags
2. Average users tend to tag incorrectly, when they do tag
3. Even advanced users have strong disagreements about tagging
4. Get enough tags and you start needing a tag "hierarchy" and it's hellishly complicated to understand
5. Tag synonyms must exist, or things get ugly fast
6. The more tags you allow per (item) the worse all the above problems get

Tagging is not a system I am particularly fond of, based on four years eating, sleeping, living, and breathing a tagging system in Stack Exchange as the co-founder. YMMV.

and also:

My main observation is that if I were to do Stack Exchange over again, I would allow a maximum of three tags, not five. Because most people can agree:

• This thing is definitely CATEGORY1.

And there is general consensus around

• This thing is also very likely CATEGORY2

But once you get to that third thing..

• This thing is maybe perhaps also CATEGORY3???

.. It's really sketchy.

Well, I just wanted to react on this, that I almost never see any problem with the tagging system at MSE. But I'm not sure if this is true. At least when I browse the meta here, I don't see much people complaining. I'm interested about what the "community" at MSE thinks about the tag system. The 2 main reasons tags are useful is I think:

1. People that find a tag interesting, can easily search for question with that tag.
2. It gives a short description of what the question is about, if you are browsing questions.

Out of that I can't conclude that it would be better to have maximum 3 tags, instead of 5 tags. The only problems I have with the tag system is that is not intuitive how to hide specific tags, and that the search box doesn't autocomplete if you search for tags. But that is a different topic.

I'm interested if users here also disagree, or agree with with what Jeff Atwood thinks about the tagging system at Stack Exchange.

Answer 1

MSE tagging is fine. Surely, it requires considerable effort from the established user base to keep it that way, but this can mostly be done by plain editing.

The tag system is useful because it allows for two-way sifting through questions: highlighting ones you're interested in, and decreasing time spent on things you're not interested in/don't know anything about. As such, it improves the site experience and the "working pace" for answerers.

---

There is a handful of users that concern themselves more often with tagging. Names coming to mind are Martin Sleziak, Asaf Karagila, and I guess I myself also belong to the group.

With the current volume of questions, this seems to be enough to avoid major derailment of the tagging system, to the point where it would become significantly unreliable.

Some ongoing effort is required in a limited selection of tags, notably (that I am aware of) logic, elementary-set-theory/set-theory, discrete-mathematics, probability and its familiars, and questions in the realm of calculus.

---

So to respond to Mr. Atwood's six observations:

1. I cannot assess this -- but it seems to me that mainly the established user base needs to "navigate using tags", and they do it mainly through the search engine. The search engine takes some practice to unleash all of its power, but this is not so much related to tags as it is to how people use search engines.
2. As pointed out above, on MSE this seems to restrict itself to a small number of tags. More on why that is the case below.
3. Perhaps MSE still has too low a volume for this to happen.
4. idem, and on MSE, the hierarchy is pretty clear, as it naturally resembles the research field hierarchy.

Not much to say on 5. or 6..

But, I disagree that three tags should be sufficient. For, in absence of an automatic "hierarchy", it's useful to have a question about a niche-subfield tagged with the encompassing tag as well, to increase exposure.

Also, since most questions don't use all five tags, it isn't a big problem to keep that number. For, when it is used, it's likely necessary, or done by a new user (and it will be corrected by some editor).

---

To conclude, I think that tags work well on MSE. I think this emerges from the sequential nature of mathematics itself (of which MSE can be expected to be some kind of mirror image): To learn more advanced maths, one needs to understand a lot of simpler maths. This is a double-edged blade:

1. Many users (askers) that are here for elementary maths can be expected to reside in a few tags, viz. algebra-precalculus, calculus, and logic. Many users (answerers) know about this stuff, because everyone starts with it -- hence the maintenance burden arising from the new users (askers) can be spread out over many persons, ensuring a low per capita burden.
2. Many users (askers) that are here for advanced maths already know how to write good mathematics, are likely to know about LaTeX/MathJax, and are in other ways well prepared for interaction on MSE. So the fact that there are fewer knowledgeable users (answerers) to deal with this fraction of maintenance is balanced by the fact that we can expect a lower maintenance burden.

I consider this a strong argument to support the assertion that tagging on MSE will continue to be useful and maintainable, even when the traffic keeps increasing.