If you were to peruse the literature on social software, you would think that MUDs and MOOs were a hugely important category of software, given the volume written about them. Meanwhile, back where the users are, social software is primarily the CC: line, multi-player games, and weblogs, and almost no one uses MUDs or MOOs.
What was it about MUDs and MOOs that made them so much more attractive to academics and researchers than their actual use seems to warrant?
Maybe there's new literature I haven't seen, but what I have seen is all old: Howard Rheingold's The Virtual Community and papers you can find at sunsite née metalab née ibiblio. Old as in back when weblogs didn't exist as a named phenomenon and the multi-player games were MUDs. For Shirky's trilogy that leaves cc:ed mail, the absence of which is kind of curious, but in my experience cc:ed mail is task-oriented, so not prone to community-building. Mind the pedestrian reason that it's non-obvious, too, I suppose, and I'm sure one'd find more discussion of it if one looked under mailing list. (This glosses over that The Virtual Community discussed BBSes, and it and several interesting papers were about Lucasfilm Habitat, a visual MUD-like service closer to
multi-player game than a real MUD.)
Self-selection is another argument. The dime/dozen term papers you'll find on ibiblio and elsewhere are about MUDs. That's where the action was back then: the university students who wrote them were probably already MUDders, and self-selected the online community topic because they saw strange things going on there.
As far as new literature goes, The Cluetrain Manifesto and Design for Community come to mind first. Powazek certainly doesn't go off about MUDs (that I'm aware of), and Cluetrain comes from a world of Usenet, the Web, and mailing lists (and several of its authors are now, of course, webloggers). I shouldn't forget that EverQuest demographics study, either.
All the old discussions of community are about MUDs because a large swath of the old community were about MUDs--the users didn't used to be
where the users are, with the exception of task-oriented cc:ed email--and this is borne out by the more diverse tools in contemporary surveys of networked society.