Shouldn't you be able to tag with multiple categories?

(sparr) #47

Has it occurred to you that your experience as a co-founder is somewhat unique? If you were quoting statistics about your USERS (which you do, frequently, and I applaud that) then this would be useful data, but your own experience and opinions about how Stack Exchange works are worthless (less useful, even than the experience of even a single user, because you are not just an individual but a UNIQUE individual).

(sparr) #48

You have not here accounted for the different ways that people would look at these questions. Not everyone always browses Newest, and more people browse the top pages of any sort than the rest, so things that tend to stay ‘frequent’ or ‘active’ would be over-represented in a typical experience. By any of these metrics, windows tags on python questions are 40-300% more frequent than you’ve stated.

Frequent 'python' Questions - Stack Overflow has 25,944 python questions, 484 of which are windows. 1.8% is 40% more frequent than the Newest sort. The first page has 2 out of 50, which is 4%, even.

The active sort also has 2 out of the top 50 tagged with windows. I don’t know how to get stats beyond the top 50 without including everything.

PS: 1.3% is not “one in every one hundred”, it’s more like one in every 70, which is closer to one per page (1/50) than to 1/100.

PPS: You also didn’t account for changes over time. Those 2978 windows tags could all be on the most recent 100k python questions, or on the oldest ones. I can’t get those stats, but I bet you can.

(Jeff Atwood) #49

Those comments are based on observing users. Sorry if that was not obvious.
We design for users and observed user behavior – and users consistently
did not understand tagging, or anything about it.

If anything, the desire for “tagging” is an advanced user function itself.

(Erlend Sogge Heggen) #50

So will sub-categories for Discourse work something like that? A maximum of 3 categories per post? (personally I’d count the main category as one taxonomy already, which leaves room for 2 more sub-categories/tags)

Edit: Moved the rest here.

(Bill Ayakatubby) #51

I didn’t infer that from @codinghorror’s post at all. My understanding is that–in the end and assuming no overriding plugin functionality–a post will have, at most, one category OR one subcategory (which implies its own parent category).

(Erlend Sogge Heggen) #52

Oh I see. Yeah that sounds about right.

In that case, I really hope someone puts together a solid tags plugin, because categories are too rigid for my purposes.

For instance, say there’s a software forum with conversations about different kinds of plugins. Any given plugin will have some troubleshooting threads and some feature threads. How do I keep track of or poll/display threads for this plugin?


@codinghorror makes excellent points about the pitfalls and challenges of tagging. Taxonomy is tricky business, especially when involving the masses. However, you do have a perspective which may not be shared by all future Discourse administrators or users. When it comes to tagging, I propose that the baby shouldn’t be thrown out with the bathwater.

Of course, that doesn’t make it your baby. I’m not asking for a free lunch. The Discourse team has built a freakin’ amazing platform. Maybe the Discourse community can raise a bounty or solicit grants to fund a flexible tag plugin? But I get the sense that you have become jaded towards tags due to your experience with StackOverflow. Tags aren’t perfect there, but the Stack Exchange is still phenomenal. Ultimately, I think it would be very helpful for you to recognize that in some contexts, and with some customization to the system itself, tags would be a critical tool for organizing and promoting discussion. I’ll provide an example in my next post which highlights this further.

I also believe that efforts to pursue tagging capabilities are a worthy time investment not just for functionality, but also to help accelerate the diffusion of Discourse. It sounds like many of the people clamoring for tags are the power users that will serve as the early adopters of this innovation, helping promote it through their networks and communities (relevant).

All said, would you concede that tags, in some circumstances, are indeed awesome? And are there any small steps that could be taken by the Discourse team to either 1) further drive this conversation or 2) pursue low-hanging fruit that might accomplish some tagging functionality, e.g. perhaps tweaks to the sub-category system?

While you mull that over, I want to say thanks for building this incredible platform. I’m sure I’ll talk about my own plans more elsewhere, but I see a very real possibility for Discourse, in concert with a trend towards open geographic information systems and other accessible tools, to have a profound impact on the global health and development sector.


I’ll give an example from my own field where tagging could, with care, be extraordinarily useful.
I work in global health as an epidemiologist / statistical analyst. At the risk of oversimplification, people’s interests can typically be broken into three key domains: geographic area (e.g., country or region), disease (e.g., malaria), and methods (e.g., statisticians, immunologists). I personally spend a lot of time dealing with methods due to my work (shout out to Cross Validated), but also try to keep tabs on certain diseases and a few developing countries. Some people are experts on a single disease, others work in specific geographic areas and may be principally focused on a single country, etc.

To be most useful and accessible to my community, discussions would need the ability to be classified in each of those three areas. Further, global health is an arena which requires the intersection of ideas that the current categories does not allow. @codinghorror mentioned advanced search as a potential alternative here, but unless search had a very robust synonym detector (like Embase but more powerful), users could very easily miss relevant content.

Let me provide a specific example. Let’s say I’m designing a randomized trial to test a new malaria vaccine. I’m planning to conduct the study primarily in southern Nigeria. I want to start a general discussion with the community about the study in general, because 1) I’m writing the methods protocol for a grant proposal and 2) I want to potentially collaborate with others.

With only categories, where would I file this discussion? “Study Design Input?” That’s too vague of a classification. Even with sub-categories for different design types, only study design & methods people (like me) would presume to be able to offer advice. But there are many other voices needed. Disease experts could update the researcher about recent developments in prevalence of the disease in the region, which could inform considerations about sample size. Nigeria experts could refer investigators to partners on the ground or inform everyone about important cultural considerations. The possibilities for cross-pollination and collaboration are vast, defying the silos of categories. Ultimately, a discussion like this one (and many others) would benefit greatly from a flexible taxonomy that includes something relating to multiple concepts (in this case: malaria, vaccine, Nigeria, and randomized controlled trials).

Now, I don’t think that purely open-ended tags are needed. A pre-defined controlled vocabulary might also be workable for many people’s needs. But traditional hierarchical organization schemes will have a much more difficult time creating the synergy that can occur when diverse groups of people interact and discuss topics. There are certainly challenges when it comes to tags, but in the long term, I believe that the juice is worth the squeeze.

Does anyone have any thoughts about the key system requirements of a tag feature or plugin? If we can consolidate around some common core options, we stand a much better chance of finding the funding or person time needed to make it happen.

(Cade Roux) #55

As a requirement example of current forums that I frequent with extensive use of structures (not that deep, but extensive) where I don’t believe discourse would be effective in meeting their needs: xda developers forum and forum. Again, I hate forums and I find these forums very difficult to use as they are, but there is a preference there for that structure. I don’t see how Discourse could handle their volume and breadth realistically without effectively making multiple discourses.

(Jeff Atwood) #56

You are aware we support subcategories now, yes? With that in mind, I don’t see why we couldn’t support larger forum structures now.

(Erlend Sogge Heggen) #57

Larger forum structures like that of 30’000 WordPress plugins?

(Matt Bidinger) #58

While I see the position for tagging and understand it as its been laid out here, and I even initially thought it sounded pretty good, this post is the perfect demonstration of how and why tagging breaks down in actual use. In my experience across multiple large properties with tagging enabled, it did as much harm as good and was more neglected than embraced due to these problems.

If one were to approach implementing a complex tagging system that addresses those problems (perhaps as a plugin for those who find need), I think one would also be well served to analyze and understand how reddit handles sub-reddits. While the concept of a subreddit is different than tagging the purpose is half-similar, but more importantly, the challenges faced by the proliferation, management, and presentation of sub-reddits/tags are similar.

That said, I still like the concept of more refined filters to get to what you want. Reddit is a testament to how enabling users to dynamically create their own walled gardens can be really successful in attracting crowds of like-minded individuals within specialty subtopics. Ultimately, I see the only hurdle in identifying an implementation that works well, without suffering the issues @codinghorror identified, or the challenges reddit itself has with the sheer overwhelming number of subtopics. Finding and navigating topics which would be of interest, but aren’t easily discovered, is a real obstacle.

I believe there is a large upside that could be capitalized on by enabling a community to self-organize to some degree, but I’m not sure exactly how rigid Discourse is with its set of defined categories, or if it can be configured to allow the users to develop additional categories organically.

Mainly, if you delegate power to users to do things on their own, the gogetters out there will charge ahead and use those new-found powers to do some cool things you may not have ever even thought of… However the more you make people go through other people in order to do cool things, the less people will be inclined to do cool things. A tagging free for all is too far on one end of the spectrum; on the other end a rigid categorical hierarchy is doom for many forums whose admins fail to organize well or, over time, fail to stay in touch with their audience’s changing trends/interests.

A large challenge historically with successful forums is that the people that run them have to be smart enough to stay ahead of trends and evolve along with the community. Reddit solves that problem by being built in a way where the community evolves the site on its own - as user interest changes focus, they also refocus the site. No forum does that presently, they depend upon omniscient caretakers within forum administration, which often falls short of what it could be. I’m unsure if this is something Discourse could address, or should try to, but I think a better solution exists and just hasn’t been identified yet.


I absolutely agree. Moderation during scale-up is tough, and tagging adds a layer of complexity on top that could (and in many cases would) become too burdensome. I also propose that advanced saved search might be able to incorporate the ease of default hierarchy with the flexibility of tagging.

I also want to point out that you mentioned “large properties.” @codinghorror also seems to be framing his view on this issue from the enormity of the SO perspective. After all, Discourse wants to revolutionize discussion, so thinking about the big picture is important!

However, that’s not the only possible environment for Discourse. If there is to be a revolution in online discussion, I’m hopeful that the community and the Discourse team can consider a wide swath of user perspectives. I, for example, envision a smaller deployment of Discourse with fewer moderation issues, but more user engagement / activity concerns.

Consider this possibility: let’s say one main tag is the only requirement of newly submitted discussions. Most users can probably find the most appropriate category for their topic, as codinghorror mentioned. But I assert that it would be manageable for interested users to also specify a couple of additional tags or details from a closed taxonomy. In my earlier post, I mentioned the possibility of tags referring to countries / geographic areas, specific methods / tools, or specific diseases. I feel that these attributes, if placed at the bottom of a new discussion entry form, could serve as a gentle but optional primer for the organization scheme.

Granted, advanced search might be able to handle some things like “country” reasonably well, but it would be far from perfect. Countries typically have easily identified first administrative divisions (like states, provinces), but the second and third divisions get messy fast. There are related challenges with the other domains I mentioned. Now, perhaps I could include a reminder message in the discussion submission form to mention a certain country or other property in the post itself (perhaps at the end, like hashtags)? That would allow search to index it better. But this option invites user error, sacrificing the standardization and order offered by a controlled vocabulary.

I’m a practical guy. I want to pursue the path of least resistance that will work reasonably well. So maybe I can hack sub-categories to make it work? But I can’t find any details about sub-categories to suggest that they could accomplish these goals. Is there a post that I missed somewhere that provides details that I should investigate?

Thanks to everyone for a super interesting discussion. I wouldn’t have realized ten years ago that so many people could be so passionate by taxonomy!

(German Viscuso) #60

How? I installed bleeding-edge and I can’t find a way to subcategorize…

(Jacob) #61

Just select a parent category when making a new category.

(German Viscuso) #62

Appreciate it! Thx for the screenshot :smile:
Do you know how to move a post categorized in the parent category to a subcategory?

My use case:
Someone created a post and categorized it as A.
I created child category B from parent A.
Now I want to change the post category to B (or A-B)
When I edit the post and look for categories on the drop down I can only see the parent category A (B is unknown)

Thanks for your help.


(Jacob) #63

Edit the title.

(German Viscuso) #64

That’s what I was doing and the child category did not show up in the drop down menu. But now it does. I suppose I had to wait a few minutest after creating the child category. Thx!

(Zachary Lewis) #65

In the case of a community set around a list of high-level topics with similar sub-categories, most users wouldn’t browse at the top level: They would live in one of the high-level topics.

A modder for Snowplow Games medieval survival horror game Night Swords wouldn’t browse the Snowplow Games forum at the highest level looking for modding. He would stay within the Night Swords topic and would see categories like modding, bugs and screenshots.

Is there a reason categories are hierarchical and not a flat tag-based system? Consider the following:

The senior producer of Snowplow Games needs to quickly determine how to allocate staff for the next fiscal year. He looks at the bugs tag and is given a list of all the topics about bugs across all the company’s games.

A reporter for HEADSHOTS magizine is looking to compile a list of the funniest glitches in Snowplow Games latest release, Gun Soldier 7. She searches the forum for topics with the tags Gun Soldier 7, bugs and screenshots.

A player is looking to purchase one of Snowplow Games games with the desire to show off his new gaming rig to his clan buddies. He searches the forum for topics with the tags screenshots and modding.

The only issue to this system is that users could misuse the tagging system, but users miscategorize topics with the current system as well.

(Erlend Sogge Heggen) #66

I think you make some great points, but they’re not right for this thread. This discussion is strictly about visuals for sub-categories.

You should add your arguments here, and feel free to speak your mind about my “special tags” suggestion, which attempts to limit the misuse factor.