Last week, RonChrisley said

Should I just put these ideas in with the entry I'm revising, or should I navigate away, and try to find a place to slot these ideas in? What's a good way to do that? Of course, Phil probably has a good idea of the range of topics available, but should all wiki users be forced to study the global structure of the wiki in order to know where to post? And since the wiki is constantly changing, one will have to do this frequently. Perhaps the wiki should be made more modular/hierarchical?

MicroContent/WhyMicrocontent

And today HilanBensusan also pointed out that it's hard to find stuff on ThoughtStorms.

So do we, in fact, have lousy InformationArchitecture? And if so, have I demonstrably failed in my claim on ProgrammersAsInformationArchitects that programmers are likely to do good IA?

Clearly something needs to be done. So I'm thinking how I'm going to evolve the IA here.

Several initial thoughts ...

1) What do users need?

Users of this wiki need to find things in three situations. The first is unproblematic.

  • You want to find a page and you know it's name.

The other two, are tricky

  • You are actively looking for information about X, but you don't know what page(s) have information
  • You aren't actually looking for information about X, you don't even know it's here, but you are interested. Somehow you need to discover that this information is here, and then if you're curiosity is piqued, find it.

Both of these second situations need to be addressed.

2) I don't believe in hierarchy

I really don't believe that there could be one true hierarchical classification system which could organize ThoughtStorms. And I'm not going to waste my time trying to find one. If there were sensible hierarchies, I think we'd have seen them start to evolve intuitively. In other words, it would start to make sense to hook themes such as OnHierarchy, under another theme such as OnOrganizationalStructures. The fact that I've never found an example of this that was obvious, might just be my fault, but suggests that hierarchy will be more problematic than useful.

Local hierarchies are another thing entirely. No reason theme indexes can't show two or three level nested lists. Sometimes it makes sense to use the subpage notation with /. And the front-page tablular overview is already two levels deep. Maybe it could be three.

3) I do believe in promiscuity

Rather than hierarchical organization I believe in dense-linking and SmallWorlds. I believe that most pages here are (on average) within a small number of links of each other. (Maybe I should measure this claim.)

The counter-argument would go, I suppose, that restrictions which prevent links, can actually help direct the user's attention. I'm very resistant to this idea. And I'll need to see a lot of alternatives shot down empirically before accept it.

There are some classic IA things I need to do better

  • There is a real problem of labelling, here. Some important links between pages are not prominent. And page-names are obscure and hard to guess. There's a good reason some page-names are strange. I explicitly want to talk about something which is only hinted at as facets of other topics and I want to make them into AddressableThings in their own right. But sometimes pages have weird names for historic / cute reasons. I want to keep them, because ThoughtStorms should be a surprising, intriguing and beautiful place to just wander around. But I should also use more alternative, sensible alternatives, and #REDIRECT. Consider BangTheRocksTogether and OnSurfaces ... how could these ideas be more findable?)

: See also ThoughtStorms/TooObscure

  • There's a lack of context-setting backlinks. In other words, in many cases page-x answers page-y, and page-y links to it. But not everywhere is it clear on page-x that you should read page-y first to set context. Maybe a standard Up or Context labelled link at the beginning of pages would do wonders.

I need to do some housekeeping

There's out of date stuff here.

And dead connections where I've misspelled a link. Or scattered two different names for the same page eg. NuclearPower and NuclearEnergy around without creating either page. Should fix these broken links.

I do believe in testing

So I have to find ways to test how effective this is. I wonder how to do good testing in this environment where all users read voluntarily and can't be set tasks and where their goals aren't explicit for me?

The only intuition I have is something like FreeSoftwareInterfaceStories.

I need to use some wiki features and conventions more =

  • I can, in fact, make links directly to searches rather than page names (delayed, sort-of LateBound links.) Maybe I can use them creatively

I am open to suggestions

...