@Mooeypoo asked for some examples of the problem with Category searches. Here are a few that illustrate the general randomness.

• Category:Art (this is an entirely empty category, a stand-in for Visual arts)
• Category:Science (there are 19 articles in Science, which is actually worse than none, because a user might think she’s monitoring the topic. Compare to the subtopic Science education, which has 145 articles)
• Category:History (there are 13 articles in History compared to 133 articles in sub-topic Pseudo-history.)
• Category: Technology (there are 90 articles in Technology, compared to 230 articles in subtopic mobile technology and 303 articles in sub-subtopic mobile technology stubs)

Fixing some of it may be pretty complex; we could start with maybe suggesting a couple of the sub-categories, but we can't descend to sub-sub-sub-etc because of technical insanities.

To start, we can add a help popup for "Categories" similar to what we currently have with ORES filters? Something that explains the caveats of categories. To that, we can then later consider adding more features like sub categories, or warnings for empty categories, etc.

In any case, some of these will have to be a sort of "phase 2" of the GSoC project. Let's try and come up with what we would like to see work best, and then we can prioritize and see which would fit the current project scope, and which would have to be prioritized to the 2nd phase of the MVP.

I agree we want to communicate these circumstances better, but I also think it is ok to do it iteratively and I'd not consider it a blocker for the initial category support (Related Changes has already similar limitations).

Some thoughts:

• We may want to explain limitations only when they are relevant. We can inform of the limitations right after the user added a category which is composed of sub-categories that are not added automatically. Conversely, if the user selects a simple category with no sub-categories we may not need to add further distraction.
• Surfacing some of the sub-categories (not descending further to sub-sub-categories), can be a positive way to convey the message that these are not included automatically.
• For redirects, we may want to treat them as if they were the category they are redirecting to.

I can add some mockups to illustrate these ideas better.

I illustrated the idea of showing a follow-up action as a way to communicate that subcategories are not included automatically, and facilitate adding them.

Initial	After selecting a category which has subcategories

• Categories are displayed as a simple list initially.
• If the user selects a category which contains subcategories, an information area is shown below the selected item.
• This area asks users if they want to include sub-categories, surfacing up to two subcategories and a "more" option that goes to the category page (opening in a new tab/window) where the user can explore the existing sub-categories.

The goal is to add categories, or replace the chosen one?

In T195503#4238644, @Trizek-WMF wrote:

The goal is to add categories, or replace the chosen one?

The idea is to add additional categories.

One of the goals to narrow the results and avoid null results would be to suggest a replacement, so that results displayed would be more precise.

Re. the design below:

This is promising, but i was thinking of something clumsier but perhaps simpler. There are 35 subcategories in Science. There are 34 in History, and 39 in Technology. These are too many, clearly, to put in the menu/ But theses lists are not too long to put into a box at the top of the results area. This use of space might be justified by the additional clarity it would provide. And if we're worried about the space, we can make it collapsible.