There's been some discussion on various blogs about the "OMG! NSF is Completely Broken and the World is Going to End!" forum headed up by the disgruntled "Aureliano Buendia." See Prof-like Substance and DrugMonkey for some interesting comments.
Something mentioned in that forum and in a comment at DrugMonkey's piqued my interest. QoQ over at DM's asserted that the NSF is indeed broken and noted in support:
First, the identity of the reviewers is not public and changes from submission to submission -- so you can't target a grant.
You can't target a grant.* I'm not entirely sure what QoQ means by this, but I suspect they want to write their proposals for specific reviewers on the panel. Perhaps so they can try to "butter up" the reviewers by citing their work favorably and often, or to avoid having to write a proposal in more general language that reviewers not experts in the sub-sub-field can understand. Or maybe even both. Or neither.
It doesn't matter really, because targeting reviewers on the panel is WRONG, WRONG, WRONG!!!!!!!
'Cos there ain't no guarantee that the reviewers you are targeting will get your proposal.
In fact the odds are not in the least bit favorable. Review panels at the NSF (and study sections at the NIH) cannot have experts from every sub-sub-field on them. Unless you want either panels with many dozens of members, or many, many more panels than currently exist (and there are already a lot). So the odds are there may be one reviewer at most who is an expert in your particular sub-sub-field (and that person is likely a competitor...).
Targeting panel members would be a particularly stupid thing to attempt (if you could) at the NSF where you could have anywhere from three to ten people reviewing your proposal. Usually three or so on the panel, and the rest as outside, "mail in" reviewers. Even if there is an expert on the panel you could target, one good review isn't even close to being enough to land funding. And let's be realistic - a panel member might be somewhat flattered by some "buttering up" in a proposal, but they're generally smart enough to recognize it for what it is.
Write your proposals for people only somewhat conversant with your corner of the field (and cite all the relevant literature, and none of the irrelevant). If you can't do that, chances are, you won't stand a chance of being funded.
* Actually, the NSF does let you do a form of targeting that a proposer would be foolish not to take advantage of. When you submit your proposal you are given the chance to suggest reviewers. In my experience NSF PO's do actually use some of these suggestions as outside reviewers. Obviously these need to be reasonable suggestions...