Hi @Jkatz (WMF).
Just to note, I disapprove of VE because it seems to me to be about hiding wiki markup, whereas I believe the ubiquity of wiki markup is a technical requisite to the success of wikis (so that hiding it is, in the long range, suicidal); my feelings about Flow are... stronger. I consider it fundamentally anti-wiki. (I tried to write here a succinct list of things it does that I see as catastrophic, but failed because the list kept getting too long; fwiw, I'm finding it almost impossible to write this, in self defense I've resorted to writing plain text.) That's definitely a digression here, and it sounds like we, er, wouldn't easily see eye-to-eye on these matters; just thought I'd mention them, though, so you know where I'm coming from.
Very interesting question, how interactive wiki markup can be brought to bear for information consumers (aka "readers"). I've a few concrete thoughts, which I'll sandwich between a couple of more general ones.
Broadly, I see devising primitive elements for wiki markup as much the same as devising primitive elements for a programming language (my graduate research area): you're trying to devise primitives that liberate users' imaginations, and "success" means they use your primitives in exciting ways you'd never have thought of. It's kind of diametrically opposite from the sort of situation where you start with use cases and design software to handle them. (I'm aware this dissonance can be deeply distressing to folks whose strength lies in the use-case approach).
Specifically, as you say, there are quizzes, and feedback. I imagine anything interactive, such as discussions on talk pages (real ones, not Flow), can likely be made ergonomic, eventually, using wikidialog. A couple of other possibilities occur to me. Interactive dialogs are possible. (I accidentally hit on this because, in developing wikilisp I found I wanted a form where I could enter expressions, click a button, and have the results displayed --- and once thought of, it was dead easy to make happen. I wouldn't dare try to create a link here; you can find it near the top of the module page for wikilisp.) I've also envisioned a page where you can click on different parts of it and wikidialog adjusts the display in various dynamic ways based on what you clicked; no demo to point to on that, but I've long had it in mind for use in a training wizard and, now that you mention readers, I suspect something of the sort might prove useful for general readers as well as up-and-coming wiki contributors. I hope, on Wikibooks, it might be possible to have book-specific assistants to help contributors with that book's idiosyncrasies of organization, style, or what have you (part of why I see Wikibooks as an excellent venue for exploring wikidialog: each book is like a microproject in itself, so you don't just get to try out wikidialog on one new project, you get to try it out on almost three thousand of them).
Broadly again, I've been aware all along that once the low-level tools start to reach critical mass, learning how to use them would open a vast new territory to explore. I expected to have to /learn/ how to write interactive assistants using the tools, and indeed I'm finding I haven't really got the hang of it yet. But eventually I want to write an interactive assistant for aiding in the building and maintenance of interactive assistants, so that just as you read a wiki page, see something you want to change, and just-do-it, when interacting with a dialog you can see something you want to change, and call up a meta-assistant to help you just-do-it. I've put some bits into the low-level tools specifically to support that sort of thing. The trick to all this is that if a tool hides things from the user, that's dehumanizing and therefore bad, but if it overwelms the user with trivia that's bad too; so you need to aid without limiting... not easy, probably there's no general recipe for it. But a useful technique might be for an assistant to have a list of patterns it can offer advice-and-assistance on, and perhaps those ones I've mentioned could be such patterns.
