I hope Ralph won’t mind if I piggyback on his post, but I’m just getting started on a paper that’s partly about normative necessity, and I thought I’d get the old juices flowing with some PEA Soup discussion. (Plus it’s February and my name starts with a D!)
The project is at the intersection of two dialectics. The first concerns normative necessity: Suppose you agree with Fine that metaphysical necessity is a matter of essence. And suppose you think (again, like Fine) that there are no essential connections between normative and non-normative properties. Then there are no metaphysically necessary connections between them.
Obviously, Ralph doesn’t buy the second claim. I don’t either. But I take it the majority of non-naturalists do. (From here I’ll speak as though they all do just for simplicity’s sake.) This means (for one) that non-naturalists have to reject supervenience as a metaphysical claim. But no worries, they can accept it as a claim about normative necessity. (Gideon Rosen has a nice paper where he tries to give a more thorough account of just what the heck this non-metaphysical normative necessity amounts to.)
That’s dialectic the first. The other concerns the reliability of our normative beliefs. Lots and lots has been written about this (obviously), but I’m going to frame things in terms of a recent paper by Justin Clarke-Doane. Keeping it brief: There has been a lot of debate over whether or to what extent debunking arguments like Street’s pose a threat to the reliability of our normative beliefs. Clarke-Doane points out that we need to distinguish the challenge to justify those beliefs from the challenge to establish their reliability. He thinks debunking arguments have no force whatsoever against the latter; indeed they might even help establish reliability. This is because Clarke-Doane thinks we should accept:
Modal Security: Information, E, cannot undermine our D-beliefs without giving us some reason to believe that our D-beliefs are not both safe and sensitive.
Our normative beliefs are safe just in case they could not easily have been false. If our normative beliefs are explained by evolutionary forces, then they are likely to be fairly stable across nearby possible worlds. So given that they’re true (Clarke-Doane goes into why we are allowed to assume this, but I’ll skip over it here), they’re safe.
Our normative beliefs are sensitive just in case had the normative truths been different, our beliefs would have been different, too. No worries here, for there are no metaphysically possible worlds in which the basic normative truths are different.
The fruit is hanging pretty low here: The modal security of our normative beliefs is guaranteed in part by the metaphysical necessity of the basic normative supervenience relations. But on the Finean picture, those relations are not metaphysically necessary. Uh oh!
I think there’s an obvious place to go here: Argue that (assuming Fine is right about metaphysical necessity) non-naturalists have to reject the metaphysical necessity of the basic normative truths. They can avoid giving up supervenience altogether by invoking normative necessity. But this comes with unexpected epistemological costs: They have lost one guarantor of reliability.
I think that’s a fine argument (no pun intended), but it’s not the one I want to pursue. Because I don’t like Modal Security. I think that attempts to cash out our intuitions about reliability in counterfactual terms—e.g., via safety and sensitivity—fail in the face of metaphysical necessity. The fact that my true beliefs would be true in any world is not enough to satisfy my sense of what it is for my belief-forming mechanisms to track that truth when what explains why they are consistently true is not some link between my beliefs and truth but the fact that no matter how I got at the answer, it’s true in all possible worlds. This is far from a new or original complaint, but I think there’s more to say about it, and I think that there’s something helpful about framing it in terms of these two dialectics. (I’m sure not everyone will agree, but it seems at least prima facie odd to me that changes in our understanding of what’s going on solely at other possible worlds would affect our assessment of belief-forming mechanisms which operate entirely within the actual world.)
Ok, I’m tempted to say all sorts of extra things about how much I’m not an epistemologist and how there’s plenty more to say about why I like framing it in terms of these dialectics, about whether this is really a worry about Modal Security or about the epistemological focus on reliability, and so on. But They keep encouraging us to write more exploratory posts, so I’m just throwing the line out and seeing what I can reel in…
Interesting. So what’s the argument you do want to pursue, David? Or are you just looking for comments on the question of whether reliability is appropriately cashed out in counterfactual terms?
Interesting!
First off, you might have noticed this post vanished and then reappeared. I hadn’t realized that people were having trouble with comments. At the request of the editors, I postponed briefly to make room for discussion on Ralph’s original post (sorry Ralph!). If people are still having trouble commenting, I’m happy to postpone again. If not, maybe this will encourage discussion on both posts!
To answer your question, Hille, I think I want to argue something like this:
1. If X is necessary-if-true, then if we believe X and X is true, we are guaranteed to be sensitive to X.
2. If our X-belief-forming mechanisms are stable across nearby possible worlds, then if we believe X and X is true, our belief in X is safe.
3. The modal security of our normative beliefs is derived from the assumption that our normative beliefs are true in combination with:
(a) 1 above;
(b) the metaphysical necessity of the basic normative truths;
(c) 2 above;
(d) the availability of stable genealogies for our normative beliefs (e.g., evolutionary stories).
4. None of (nor any combination of) (a)-(d) is relevant to the question of whether or how our normative beliefs track the normative truth at the actual world.
5. Only considerations relevant to the question of whether or how our normative beliefs track the normative truth at the actual world can establish the reliability of those truths.
6. Therefore, modal security established as above cannot be sufficient to establish the reliability of our normative beliefs.
Before I stop and let people school me, let me just contrast this with the perception case Clarke-Doane appeals to by analogy:
I think Clarke-Doane is right that there can’t be a prohibition on assuming the truth of our beliefs. But I think there’s an important difference between the perceptual and normative cases. For in the perceptual case, the analogy of (3) above is something like:
(3*) The modal security of our perceptual beliefs is derived from the assumption that our perceptual beliefs are true in combination with an evolutionary and neurophysical explanation of how our perceptual appartus came to and function so as to track the subjects of our perceptions at the actual world.
So, obviously, I think the heavy lifting is going to be done with something like (4), since (3*) avoids my objection by establishing modal security in a way that
seems relevant to how our perceptual beliefs track the truth at the actual world.
But that’s all very tentative.
David, thanks a lot for this interesting post. (The link to my paper was broken, but I fixed it.)
You write,
“The fact that my true beliefs would be true in any world is not enough to satisfy my sense of what it is for my belief-forming mechanisms to track that truth when what explains why they are consistently true is not some link between my beliefs and truth but the fact that no matter how I got at the answer, it’s true in all possible worlds.”
I was not trying to cash out our intuitions about reliability. As I note on 95 and 96, there are myriad senses of “explain the reliability”, because there are myriad senses of “explain” and myriad senses of “reliability”. I think that little turns on whether we can “explain the reliability” of our beliefs from an area, D, in some pretheoretical sense of this phrase. What is interesting is whether there is any sense of “explain the reliability” in which *both* of the following are true.
(a) It appears impossible to explain the reliability of our moral beliefs.
(b) The apparent impossibility of explaining the reliability of our moral beliefs undermines them.
You might think that in order to “explain the reliability” of our moral beliefs in a sense which satisfies (a) and (b) it is necessary to show that their contents (or truth) are implied by the best explanation of our having them. Only then could we establish “some link between [our] beliefs and truth”. But if Modal Security is true, then this sense of “explain the reliability” fails to satisfy (b). Learning that we cannot explain the reliability of our D-beliefs in this sense does not, I argue, give us any reason to believe that they are not both safe and sensitive.
I also argue that in order to explain the reliability of our moral beliefs in a sense which satisfies (a) and (b), it does not suffice to show that their contents (or truth) are implied by the best explanation of our having them. If it did, then we could *trivially* explain the reliability of our logical beliefs, because, for any logical belief that p, p is implied by the best explanation of our believing that p — since p is implied by every explanation. However, even if we can relevantly explain the reliability of our logical beliefs, this is not a trivial fact.
What about the challenge to explain the “merely actual correlation” between our moral beliefs and the truths? I do not believe that this challenge is coherent. But even if it were, the apparent impossibility of answering it would not seem to give us any reason to believe that our moral beliefs are not both safe and sensitive (see Section 6 of my paper, “What is the Benacerraf Problem?”). Hence, if Modal Security is true, this sense of “explain the reliability” fails to satisfy (b).