Having published one peer reviewed paper on the accuracy of Wikipedia articles, I feel qualified to comment. :)

Hate to say this, but Vandalism studies of wikipedia articles are not new, they have being done since the very beginning of academic research on Wikipedia and generally the results are they are not as bad as you would expect. The study mentioend shows up 2.5%? But I'm not going to squabble about the methodology.

Perhaps people are expecting more out of Wikipedia now, as the reputation of Wikipedia has risen.

Also I find such vandalism studies pointless. Vandalism is crude and obvious to that extent that they were using bots to handle it. Besides it is a simple matter to ignore such segments. That is what most readers do.

The irony is, presence of vandalism might leads users to mistrust an article, which would otherwise be accurate. Another article which is otherwise free of vandalism, might be in fact full of errors added by editors in good faith. .

Anyway, In this users of Wikipedia are no worse off, then if they just blindly trusted a website they got off the net.

In fact, I would argue that because of various features built into wikipedia (history pages, discussion pages, an experienced user of Wikipedia (or should I say editor who knows how things work on the hood because of personal experience) with knowledge of Wikipedia policies and norms would be far more capable of judging the validity of Wikipedia articles than something off a random site.

Given Wikipedia's rise, perhaps this is something Librarians should teach, and not just the same old generic criteria to evaluate websites.

I haven't kept up with Wikipedia studies lately, but there were also proposals to mark certain revisions as stable (based on either explicit voting by users) and also add features that color coded segments based on their reputation to help guide users (based on automatic implicit features) etc...

Some of these features I believe are already in non-english language versions (German?)...