#### Neta’s New Paper — 8 Comments

1. Ram–I’m reading your interesting paper, and have a question for you, at least indirectly. On page 16, you attribute the following view to Schaffer:

Kspq iff (a), p, (b) s has evidence e which is such that, given that either p or q is true, S has e only if p is true, and (c) s is certain that p rather than q, on the basis of (b).

In order for the property in clause (b) to be a substantive one (i.e., to add an additional requirement to the simple claims that p is true, and S bases his/her certainty that p on some e), we can’t read the conditionals as material conditionals (since (a) together with the first part of (b) guarantees that the property obtains if the conditionals are material ones).

And strict conditionals would be too strong, so we’re looking for something like a counterfactual conditional here: p or q counterfactually implies that if S were to have e, then p would be true. A first problem here is that this conditional is rendered true on the standard semantics if clause a and the first part of b obtain, since all three constituents are true. So we need the kind of semantics that defenders of safety hope for, a semantics where more than the actual world is implicated when a subjunctive conditional is being evaluated.

So, here’s the idea, I think? Some such worlds will need to be worlds where q is true, for if only worlds where p is true count, then the counterfactual in question always is true no matter what value we use for e such that S has e. But this is going to be a mess, since some contrasts to p will be themselves impossible. For example, considering knowing that Goldbach’s conjecture is true rather than being counterexampled by only the number 28. Or think of knowing that water is H2O rather than CO2.

I’m also perplexed by the account regarding ordinary skeptical alternatives. Let p=I have hands and let q=the demon is deceiving me into thinking that I have hands. Unless Schaffer restricts the values for e, I can know this easily. Just let e=p. Or let e be anything that entails p and is believed by S. So let e=”Today is Saturday, and that claim materially implies my having hands”.

Of course, the basing requirement might still undermine the counterexample, but that clause needs work anyway. Assume for simplicity a causal basing theory. Then clause (c) will require that the truth of clause (b) causes S’s certainty. I would have thought that e, or S’s believing e, should be doing the causing instead. Even with this change, though, one won’t be able to argue, I think, that the last value for e above can’t cause the belief that I have hands, and if it did, I’d have the strong kind of contrastive knowledge Schaffer doesn’t want to claim I have.

2. Hi Jon,

So how to construe condition b? I think that b is supposed to be understood as follows: take all the worlds in which either p or q is true. Now, among all of those worlds, the only worlds in which S has evidence e are worlds in which p is true.

Now, what does this add to conditions a and c? Well, condition a tells us that Kspq only if s is in a p world. And condition c tells us that Kspq only if s bases her certainty that p on a certain kind of evidence e. But wait: what kind of evidence must s base her certainty that p on, in order for Kspq? If we left out condition b, then we would be giving no answer to that question. And we need to give a substantive answer to that question — s’s certainty must be based on the right kind of evidence in order for Kspq. So that’s why condition b is necessary, I think.

But I should let Schaffer correct me on this one if I’ve misunderstood him.

3. Ram, you and I need to find some night life…

Anyway, I agree a clause is needed here, since lots of true beliefs are improperly based. So, take a belief that is based on some e that we can correctly say S has. Then, when is e the right thing to count as knowledge?

That’s fine, except that clause (b) doesn’t get it right, I think. The cases I used show that if we take all the worlds where either p or q are true, that can be just p-worlds. And if we only consider p-worlds, clause (b) puts no additional restriction on whatever e S has on which s/he bases the certainty in p.

4. While you’re distracted by my Schaffer questions(!), let me express some discomfort about the evidence relation in your account. As I understand your view here, there is an evidence-having relation that varies, and evidence itself. Let’s use capitals for the relation and lower case for the evidence. Then if E is the relation, e is the evidence; E’, e’; E”, e”, etc. The relation fixes the value of the evidence, and so the crucial question for you is how the relation is fixed. Correct?

OK, if that’s right, then switch subjects to moral relativism. The moral relativist says that the truth of a moral claim is relative to the culture one is in, let’s say. To assess the view, I’ll want to know what a culture is and, especially, how to individuate them. So suppose the defender of the view says, “here’s my account. See how the truth values of moral claims varies from context to context, and that will tell you how to individuate cultures.” Such a response leaves me unhappy, though it’s hard to say exactly what is displeasing me. I want an independent account of cultures and their identity conditions, and then I can assess whether the relativistic theory is defensible. The explanation I’m given presupposes the relativism in question in order to individuate cultures, since the procedure suggested would leave the truth of the view a trivial consequence of the suggested procedure.

On page 25, I think you’re doing something similar to the imagined moral relativist. You ask how the value of the evidence relation is fixed in a context, and then you cite a claim about evidence: evidence is what you’re entitled to take for granted in a given case of inquiry. I’m inclined to respond as above: I want information about E, and then we can use that information to determine whether it gives us an adequate account of e. But what you give me goes the other way: here’s what e is, so it follows that the relation is E. But that presupposes the relativism in question rather than clarifying the view so that its adequacy can be assessed.

This is not very clear, but maybe it gives you a glimpse of what I’m worried about?

The issue here is that I don’t recall you talking about relativity to the evidence-having relation before, though I do recall the contextualism about evidence and the relationship to what one is entitled to take for granted in inquiry. So I was puzzled early on by the talk of relativity, since it looked new to me. And if it is new, then there should be an argument for it over and above the arguments for contextualism about evidence that you’ve already given. So I think I must be missing something here…

5. Hi Jon,

This is better than night life!

About condition (b) in Schaffer’s account: I’m inclined to agree with you that the condition, as stated, doesn’t impose a tight enough constraint on the evidence that s must have in order for Kspq. As currently stated, e could just be anything that is true in all and only the p-worlds. So the account of knowledge that I attribute to Schaffer needs a further constraint on e. (There are, of course, different ways that Schaffer might wish to specify this constaint. He might choose to say, with Feldman and Conee, that e includes all and only s’s conscious mental states. Or he might choose to say, with Williamson, that e includes all and only what s knows. Or he might choose to say, with Lewis, that e includes all and only s’s perceptual and memory states. Or …)

Now, about evidence-having and moral relativism. I agree with you completely that my procedure in the last section of this paper is like the procedure of the moral relativist that you describe: all I’m doing there is showing how an evidence contextualist like me is going to design a framework to handle (what I take to be) contextual variation in the evidence that our subject can truthfully be attributed.

That’s why I didn’t want to call the paper “developing the case for evidence contextualism” — it’s not a manifesto, containing the primary arguments and motivation for my view, in the way that Schaffer’s “Contrastive Knowledge” is a manifesto. Instead, I called it “undermining the case for contrastivism”, because that’s all I aimed to do: show that Schaffer’s case for contrastivism (as I understand it) doesn’t give us reason to prefer contrastivism to evidence contextualism. (And, in fact, I think it suggests reason to prefer the latter to the former.)

6. Hi Jon and Ram,

1. What Ram says in comment #2 about construing condition (b) is essentially right. To clarify, my official statement of condition (b) is actually: s has proof that p rather than q. I later add that proof requires conclusive evidence, which is evidence that, within the (p-or-q)-worlds), only obtains in the p-worlds. If we follow Lewis in taking a world to be eliminated when it is incompatible with our evidence, then it requires the elimination of the q-worlds.

2. I agree with what Ram says in comment #5, that I need to specify what evidence is. I do say in a footnote that “I have not said what evidence consists in… Though what I say is compatible with Lewis’s (1996) conception of one’s evidence as one’s total experience.” So for better or worse, that is what I had in mind.

3. I agree with what Jon says in comment #4, that Ram needs to specify how it is that context sets the value of E. Here I think contrastivism may have the advantage. The contrastivist argues that the language encodes a contrast-slot, various forms of knowledge ascriptions contain devices (such as “rather than”-clauses, embedded questions, focus operators, and Stalnakerian context sets) for saturating the contrast-slot. I don’t see any reason to think that (i) the language encodes an E-slot, or (ii) that any devices in knowledge ascriptions saturate the E-slot. Indeed, Ram is explicit that fixing the value of the question Q does not fix the value of E (that is the ‘bit of slack between contrasts and evidence’ that he explicitly cites near the end of his paper). But then my arguments that the language is encoding a value for Q cannot be coopted to show that the language is encoding a value for E. So I would make the argument that contrast-relativity is in the language, while evidence-relativity is just a clever invention.

7. Hi Jonathan,

Thanks for the useful responses!

I agree completely with you and Jon that I need to specify how it is that context sets the value of E. Now, I happen to think that the best answer to that question will be delivered by a substantive theory of what it is to have evidence — but that’s a long story…

But I have a very naive question of clarification for you: what is it for the language to “encode” something? Suppose I assert that Bush is President. Does the language encode that there is some particular nation (or club, or organization, etc.?) of which he is president if my assertion is true? What if I think that being the President is a property that only one being can ever actually have — like being the messiah, or being the inventor of bifocals. Am I then ignorant of the language? Or am I simply ignorant of a matter of empirical fact?

8. Hi Ram,
I’m always afraid of questions labeled ‘naive’! But all I mean when I speak of encoding is that you need linguistic evidence that the language actually features semantic relativity to an E argument. You need evidence that this is more than wishful thinking. For better or worse, the contrastivist presents linguistic evidence that the language actually features semantic relativity to a contrast argument, such as: (i) knowledge ascriptions seem to be semantically focus-sensitive, and (ii) focus operators encode contrasts, from which the natural conclusion is (iii) knowledge ascriptions are semantically focus-sensitive because they are semantically contrast-sensitive. This is contentious of course. But it is the sort of thing I think you need.