I think the problem of virtual worlds being very hard for the uninitiated may be unsolvable. After all.. anything simple enough that its trivial to use is likely not complex enough to deserve being called a “world”.

It may even be impossible to make virtual worlds really fluid even for the most experienced human operators without evolving what they are or increasing their sophistication until they convincingly emulate “first life”, which we are already well adapted to handle. (Or, alternatively, advancing the ‘humans’)

On this subject I’d recommend reading True Names by Vernor Vinge. It includes a virtual world which is mapped onto “first life”. I’ve always rejected the notion of mapping virtual worlds onto anything too much like “first life” because it’s unlikely that any reasonable virtual world would be sufficiently isomorphic, for example distance in computer networks is very different than distance on earth. When I finished True Names, I still felt like using an emulation “first life” as the user interface for virtual worlds was not a good idea however the novel contains an absolutely brilliant, and rather long, afterward by Marvin Minsky which I found to be quite convincing. I’d highly recommend reading the book (and the afterward!), as it’s not terribly long, I think it took me less than an hour to read it..