Here's another example of the lack of visibility of scientific literature in online discussions (probably caused by people not knowing if the papers are open access or not, so wanting to link to a readable summary):
Slashdot reports on a PNAS paper about bird flocking behaviour, but only links to a news blog post on Environmental Graffiti, which itself only links to a report in The Telegraph, which doesn't link to anything. You have to actually go to PNAS.org, search for 'starlings' and sort the results by 'newest first' before you find the actual paper (which is, happily, open access after all).
There's a danger in this chain that information can become misinterpreted, or at least that by the end of the chain there's not really much useful information left.
The other problem is that it's hard to pick up conversations around the paper, because few of the posts identified which paper they were talking about (they could still be clustered by text similarity, perhaps).
It may be that only Google [Scholar] is able to condense and canonicalise all this information, as it really needs an index of the whole web.