Last week, I laid out some problems for the naive safety account of knowledge that Pritchard presents early in his book Epistemic Luck.
I wanted to get them out so that we could make sure that, whatever revisions Pritchard made to the safety account, we had a safety view that also avoided those worries.
Pritchard’s revised version does avoid those worries (one of my own students actually wrote about this for his weekly – which was awesome).
Here is the final revised version of the safety principle.
Safety Revision
If S knows P, then in nearly all (if not all) nearby possible worlds in which S forms the belief that P in the same way as she forms her belief in the actual world, S only believes P when P is true.
The same belief method clause above avoids both the Bear Beliefs Case and The Shooting Star case from the previous post. Recall that in both of those cases, I was including nearby worlds where the person does not form the belief using the same method (because they don’t form the belief at all).
However, I have a couple of other criticisms to raise against this revised account. First, I think we’ll have generality problem worries that standard reliabilism faces. Second, Pritchard seems to think that the safety account adequately handles Gettier cases – I don’t think it does. Third, Pritchard thinks that safety accounts avoid Kripke-style Fake Barn Country objections that have been raised against Nozick’s Sensitivity Principle. I don’t think his response to these Kripke-style objections are adequate.
I’m not going to get into the generality problem issue. In the rest of this post, I’ll focus on the Gettier Case issue. In the next post, I’ll talk about Fake Barn Country.
Safety and Gettier
Pritchard seems to think that Gettier cases can all be explained because of a bad kind of epistemic luck that he thinks his safety principle rules out. I don’t think this is true.
Sheep in the Field
Imagine Bob wakes up in a field filled with lots and lots of sheep. He opens his eyes and just so happens to look in the direction of a fake sheep. Had he looked in any other direction he would would have looked at real sheep. Further fill in the details so that the only possible worlds in which there are not sheep in this field are VERY FAR OFF.
If Bob forms the belief that there is a sheep in the field on the basis of his visual perception, then it seems that he’s going to count as knowing that there is a sheep in the field. His belief is not veritically lucky as Pritchard defines it. His belief is safe as Pritchard defines it, so it seems that Pritchard’s theory of knowledge runs is not Gettier-proof the way he seems to think it is.
I think Pritchard’s view runs into Barn Country problems as well. He argues that they do not. I’ll post about that in the next post.
This doesn’t go directly to anything in your post, but here is a comment for what it’s worth. It’s been a while since I looked at Pritchard’s book, but if I recall correctly safety seems to do all the heavy lifting in his account of knowledge. That is, it looked at times like he saying that a safe true belief counts as knowledge, but that seems wildly counterintuitive. Is safety supposed to act as some fourth condition to JTB, or can the justification fall away? The problem for safety without the justification is that we can gerrymander all kinds of weird possibilities in which the person forms the safe true beliefs without us thinking that it’s a case of knowledge. (I think this holds for JTB+Safety as well, but Pritchard didn’t even seem to have that.)
Hi Matthew,
He actually isn’t going to go with a pure safety account. He thinks an external safety condition is necessary for knowledge, but he feels compelled to add some kind of internalist component as well.
So to answer your question, I think Pritchard is offering up his version of the safety principle as the elusive fourth condition to tack onto traditional JTB analysis.
As of Chapter 7, however, we haven’t been given his favored account of internalist justification (although he has said some things that make me think he’s not entirely unsympathetic to a Zagzebskian virtue epistemology).
But so far it looks like the account is…
S knows P iff S has a safe true belief that P and S has internalist justification that P. (where the internalist account of justification has yet to be specified)