Being Disagreeable

I got into a bit of an argument with Stew Cohen about views on disagreement at lunchtime Friday. (Sorry dear colleagues, and co-residents of St Andrews, for disturbing your lunch unduly – I should have been less, er, disagreeable when arguing.) One of the upshots was I have a slightly clearer idea about what I dislike about “Equal Weight” approaches to disagreement.

Start with the abstract case. A person, call her Z, believes p on the basis of some parts of her total evidence E. Another person, call him RL, believes ¬p on the basis of some parts of his total evidence, which is also E. Z had antecedently, i.e., before finding out this odd view of RL’s, that RL was an epistemic peer on this question. What should Z do?

I say that it depends on what E is, and what p is, on their relation to each other and her relation to them.[1] And so, I think, says everyone.

Let’s consider a particular case of this. E contains the fact that thousands of other people, each of them just as qualified as RL, believe p on the basis of (other parts of) E. Then it would be perverse to let RL’s judgment make more than a trivial difference to Z’s position. If you said that Z had to “split the difference” between her current view and RL’s, that wouldn’t be an equal weight view, it would be a view that the last person you hear from trumps everyone else. And note that’s true even if Z has no independent reason to believe that RL is mistaken, incoherent or otherwise impaired.

So now we’ve established an existential – for some such E and p, it is reasonable to treat RL’s opinion as more or less irrelevant to the correct judgment. I think that kind of situation is not just possible (as we’ve now proven) but reasonably common. In particular, I think there are such situations even when E does not contain any evidence about third party experts. I’m not going to offer here an argument for that position.

But I am interested in what could be an argument that this kind of possibility is rare, though it is possible. Note that any argument, such as the early arguments for the equal weight view, which purport to show that such a situation is impossible are clearly mistaken. We know the situation is possible. The point of the equal weight view must, I think, be that such situations are (a) rare, and (b) only arise in special circumstances. What could those circumstances be?

It might be possible to argue that judgments of others are always trumps, so no E that lacks testimonial evidence could continue to support p in the basis of peer testimony that ¬p, when the peer in question also has evidence E. But we would need a reason to believe that. It seems to me that stated this barely, the position attributes a kind of magical power to evidence about the judgment of others, and we shouldn’t believe in any such magic.

Perhaps the position could be that I misdescribed the case, by including the testimony of others are evidence. Perhaps, as Moran, Hinchliff and others have argued, testimony provides non-evidential warrant. So we should say that Z’s warrant includes evidence E and testimony T. When she hears RL’s position, that changes T, but doesn’t change E. I think this ends up with the result that the equal weight theorists want, namely that when RL is the first person she hears, his opinions should make a big difference to her credences, but when he’s the 1000th, they should only make a small difference. But if it turns out the equal weight view rests on the non-evidential view of testimony, that would be an interesting surprising discovery. Since I prefer something like the Maitra-Nolan view of testimony, which is an evidential view, that would at least explain why I don’t like the equal weight view of disagreement.

Perhaps the argument could be that “second-order evidence” always trumps “first-order evidence”. In this context, second-order evidence is evidence about the force of first-order evidence. Now assume that and when RL is the only peer she speaks to, his opinion is very strong second-order evidence about the force of E, but when she speaks to many peers, his opinions are very weak second-order evidence. Then we get the result that she should pay him a lot of attention iff she hasn’t spoken to many other peers.

My gut feeling is that this reverses the right order of things; that first-order evidence really trumps second-order considerations. But set that aside. It’s possible that E contains a lot of non-testimonial second-order evidence for Z’s position. Perhaps E contains a carefully worked out epistemological theory that entails that E supports p, and Z believes that theory for the right reasons. Presumably RL does not believe that theory, but that will just show (by Z’s lights) that he turns out not to be a very good epistemologist. So even if second-order evidence is always a trump, Z can still (more or less) ignore a peer, as long as she has other second-order evidence.

So that’s where I think things stand. For some kinds of evidence E, the crude equal weight view is obviously wrong. That is, there are some pairs ⟨E, p⟩ such that E supports p, and this support is immune to defeat by peer disagreement. Those are cases where E already contains a lot of peer judgments supporting p. I doubt, for good old-fashioned Quinean reasons, that strength of evidence invariably tracks kind of evidence. So if some such pairs ⟨E, p⟩ exist, there will be other such pairs ⟨E′, p′⟩ where E′ does not contain judgments of other peers, but the support E′ provides for p′ is also immune to defeat by peer disagreement.

I think it’s hard to say just what the contemporary manifestations of the equal weight view are committed to. But I think they must be committed to the rarity of immunity to defeat by peer disagreement. If not, I don’t quite know what the view is. (It can’t be just that defeat by peer disagreement is possible can it? Does anyone disagree with that?) And given these considerations, I don’t know why you’d think this kind of immunity is rare.

1 I’m more and more coming to the opinion that a lot of questions that are discussed in contemporary epistemology are like this one: “A train leaves New York for Chicago travelling west at 75mph. What’s the probability it will snow along the way?” I suspect the Sleeping Beauty puzzle might be like this too. It isn’t obvious that the question has an answer, any more than this question about snow has an answer.

2 Replies to “Being Disagreeable”

  1. I think there are some good things and EWVer can say here:

    (1) As Christensen states it, his position is that in cases of peer disagreement, you can’t favor your own position unless you can explain why the other person would be more likely to be in error by appealing to factors that are “independent” of the disagreement. (The scare quotes indicate the obscurity of what counts as independent of the disagreement, but let that pass.) The fact that loads more scientists agree with you is plausibly independent of the first-order considerations at stake, so that’s why you don’t have to split the difference. (What happens when he still disagrees with you when he’s considered the fact that all these scientists are against him? I don’t know, but maybe you shouldn’t think he’s a peer when it comes to evaluating the force of scientific consensus. On plausible views, peerhood is question specific anyway, and you don’t go for the narrow-scope version of EWV).

    (2) EWV is a constraint on rationality, but the narrow-scope conditional “if you think S is a peer, then all things considered, you should split the difference upon disagreement” is false. The rational response to your disagreement with RL is not to split the difference, but that is because it would be irrational to regard him as a peer. It would be irrational to regard him as a peer because so many people who you regard as peers/experts disagree with him. (What if you don’t know about the scientific consensus? That’s another story.)

    (3) EWV should be generalized when there are lots of folks. How? I’m not sure. But you should take some kind of equal weighting of the judgments of all the people disputing over the first-order evidence. Since there are so many people who disagree with RL, no need to split the difference. (Of course, I agree that once you go this way it becomes more obscure what EWV actually says. We know it isn’t averaging if it is going to be not-crazy. But what is it really?)

  2. Three points about this.

    First, there’s nothing in the description of the case that implies that the ‘first-order’ evidence is sufficient to reason to p. Perhaps it is, perhaps it isn’t. Since I’m only trying to prove an existential, we might be best off considering the case where it isn’t.

    Second, it might be now irrational to judge RL to be a peer. But the usual statements of equal weight require the agent to use their prior judgments of peerhood. And antecedently, RL by hypothesis looked just like a peer

    Third, I think this kind of case really puts pressure on the “independence” principle. It helps to think of this as a version of the point I was making about people who disagree about the epistemology of disagreement. Let RL be someone the agent thought was a peer about making judgments on this question in the presence of a large scientific consensus. It turns out that RL breaches all sorts of things Z took to be canons of rational disagreement. RL also breaches things the EWV takes to be canons of rational disagreement in the course of answering this question. But neither of those breaches are independent of what’s at issue. I don’t see how we get the right result in this case without significantly weakening the clause saying peerhood judgments have to be made on independent grounds.

Leave a Reply