On Ted Sider's suggestion, I read Rob Williams' Phil Review paper "Eligibility and Inscrutability" with hopes that it would lead to my being less creeped out by reference magnets. Unfortunately, that didn't happen. I got more worried that we'd have to appeal to the darned things in building theories of reference, but I dislike them just as much as before. Depressing.
I take it that this is how the reference magnet strategy is supposed to work: Say we're trying to reduce intentional relations to something nonintentional. We do the whole Ramseyfication thing, and... uh-oh, we either have too many equally good candidates for what our terms refer to and many of them are counterintuitive, or as Williams' paper suggests, something counterintuitive ends up being the winner. So on the reference magnet strategy we say that the intuitive things have a certain kind of primitive naturalness to them, and that's why they get selected. Now we've reduced the intentional successfully to something nonintentional!
But it's hard for me to see any real theoretical gain here. We've avoided using primitive intentionality by introducing another primitive -- naturalness -- that's just as ontologically extravagant. If naturalness did some other interesting kind of theoretical work so we needed to appeal to it, there would be something to be said for it. But it's hard for me to see what kind of work it's going to do. A notion of natural kinds, which we might maybe be able to get out of the sciences, isn't going to do enough work for us, because we're going to need lots of primitive naturalnesses that go beyond what scientists use in explanation and prediction. We don't just need primitives like 'electron' and 'orangutan', we need 'corset' and 'film noir' and 'Optimus Prime.' We don't have primitive intentionality, but we have a bunch of primitives that are shadows of the intentional relations we were wanting to reduce. So in the end our overall theorizing ends up just as complicated, and the reduction of intentionality is a hollow victory.
I'm quite attracted to the idea of somehow psychologizing the naturalness out of the picture, so that instead of having primitive naturalness in the world, we have some kind of psychological state that does the work of selecting what ends up being the most intuitive referent for a term. There's good explanatory reason to posit something like this -- it explains people's behavior, namely, their yes- and no-saying behavior when you ask them whether this or that thing is the referent of their term (or in complicated cases that support semantic externalism, whether this thing or that thing would be the referent of their term if the world turned out to be a certain way). What's explaining their behavior when you ask them these questions has to be something in their heads, and it's there for us to appeal to.