Monday, October 23, 2006

The Swamping Argument

Since my commentary was rather syncopated I thought I'd paste part of my recent post on the University Sterling's Value Project Blog here.

First, a little background:

The Swamping Argument is specifically meant to falsify the following naïve attempt to designate a method for explaining the value of knowledge over true belief.

(*) Find some property P which distinguishes knowledge from true belief and has value. Then surely EV(TB+P) > EV(TB).

But this is doomed to failure for P such that the *only* value P has is as a means to TB (thus my “pure-means”). The idea is that the property having been reliably formed is such a property.
My bottom line diagnosis is that Kristoffer was confusing the value of a *process* with the value of a *belief* which is the target of the Swamping Argument. In it's barest bones, the Swamping Argument can be put forward as an explanatory argument:

Datum: Necessarily, for any two true beliefs B1 and B2, if B1 is known and B2 is not, then EV(B1) > EV(B2.) [Technically it could be run as an existential claim, but then it's not a very interesting thesis.]
Reliabilism can't explain this datum for all that R-ism has to add to B1 is that it is R-formed. But that's only valuable as a means to truth, so given that we've got truth already ex hypothesi, so it doesn't actually add new value.

Now the comments where I take fairly seriously one suggestion.


This weekend at the 4th Biennial Rochester Graduate Epistemology Conference, I commented on a paper "An Argument Against Swamping" by Kristoffer Ahlstrom from the University of Gothenburg, Sweden.

He said there are very few epistemologists in Sweden, so I'm glad that a significant percentage of them (him!) are working on value-driven epistemology.

Having said that, there seems to be a misconception that advocates of the swamping argument against reliabilist theories of knowledge can't hold that reliable belief forming processes are valuable. I've run into this a surprising number of times, even among people able to formulate a reasonable version of the swamping argument.

I want my belief to have the property of having been formed by reliable belief-forming processes. So I value such processes. What's more, I value them precisely in virtue of my desire to have true beliefs, for it might reasonably be thought that forming beliefs according to such methods is likely to result in my having some more true beliefs (it might even be true by definition).

Still, for any *given* belief, if I know that it's true, what do I care how it was formed? Well, I might care, but not *merely* from the standpoint of my desire to have true beliefs. This seems to me to be the point of Jon's Two Lists Argument (VK, 45-48).

Having said *that* there is an interesting idea suggested by Ahlstrom which few will like, but to which I'm actually somewhat receptive to in other contexts. Ahlstom seemed to want to argue that there was some independent value in the property having been formed by a reliable belief forming process which accrues to a belief B having that property in virtue of that property's having the following property: being such that most instantiations of it are true. Let's call this higher-order property P*.

On the face of it, it doesn't seem like the value of that property can "seep down into" the object-level instantiation. The target belief just is true and an instantiation of P* which has has nothing clear to do with other instantiations of P* which may or may not be true. But consider this example of a common phenomenon: I take especial pleasure in the fact that I'm going to the same grad school as Marshal Swain, Peter van Inwagen, et al (and where Richard Taylor and Keith Lehrer taught). This will no doubt strike most as just pure irrationality, but it's not *abundantly* clear to me that there isn't something to such thinking.

One might think that schools which have good M&E faculty/students will continue to do so (it's certainly been true in this case!). But w.r.t token faculty/students this can at best underwrite the expectation that they are likely to be good at M&E (no comments!). And one might think that processes or agents that have been reliable in the past are likely to continue to be reliable in the future. Still, this can at best underwrite the expectation that the next belief token will be true. So it seems to me at this point that the phenomenon sociologists call "basking in the reflective glory" are, at best, grounds for expectation of some target property which the Swamping Argument shows don't help solve the Meno Problem.

0 Comment(s):