Sosa’s Latest Safety Condition

Sosa’s latest account of safety is:

S’s belief that p is safe iff S holds it on a basis that this belief would not have without being true.

He offers this new account to avoid the Kripke red-fake/green-barn counterexample. Here’s how it works with that case. S believes that there’s a green barn in the field, and deduces that there’s a barn in the field. In that case, the first belief is safe, having a perceptual basis, and the second is safe as well, have a deductive basis. Had S simply formed the perceptual belief that there’s a barn in the field, however, it would not have been safe.

What’s interesting is Sosa’s explanation of the intuition that we don’t know that the object in the field is a barn. We have that intuition, according to Sosa, because of the way our perceptual phenomenology works: we are appeared to barn-in-the-field-ly, and also are appeared to greenly, so that the basis of the belief that there’s a barn in the field is the first appearance state. Were the phenomenology different, so that we are appeared to green-barn-in-the-field-ly, no violation of safety would occur either for the belief that there’s a green barn in the field or that there’s a barn in the field.

To appreciate Sosa’s proposal, we must first distinguish between what is in our visual field and what we attend to within this field. It is possible for one’s visual field to contain a green barn, and yet one attend only to the barnishness of the experience. But Sosa needs more than this point. He needs there to be a distinction between unified noticing or attending, where one attends to the green-barnishness of the experience, and conjunctive noticing or attending, where one attends to the barnishness of the experience and in a separate attending, to the greenishness of the experience.

I can make sense of sequentially noting the barnishness of an experience and the color of it. I can also make sense of simultaneously noticing two aspects of my experience, where one part has a property and another part has a different property, such as when there are two objects in my visual field. But I’m not convinced that I abstract away from color in a way that would result in a simultaneous, independent noticing of the greenish character of my experience of an object and the barnishness of that same object.

Suppose we let this point go, however. Even so, I don’t see why one phenomenological account over the other should somehow indicate a difference in safety. I know that the account above implies that they are different, but I don’t see why we should accept that. Take two barn-believers, one of whom has the conjunctive abstraction experience Sosa indicates, and the other of whom has a unified experience. I don’t see why that should make a difference as to whether either knows, and I’m pretty strongly inclined to think that neither knows. Either way, however, if safety is necessary for knowledge, we should expect an adequate account of safety to treat them the same.


Sosa’s Latest Safety Condition — 3 Comments

  1. If I understand Sosa’s latest safety condition correctly, it is quite similar to my condition that the belief should be formed by a method that is “reliably error-avoiding”. (I stated this condition in my paper “The Aim of Belief”; actually, it’s a bit cheeky of me to call it “my condition” here, since it is inspired by, and extremely close though not quite identical to, a condition that Peacocke proposed in his 1986 book Thoughts: An Essay on Content.)

    I’m inclined, at least in rough outline, to agree with Sosa’s treatment of Kripke’s barn example. If your belief “That’s a green barn” is based on the lemma “That’s a barn”, then your method is not “reliably error-avoiding” in my sense, because the first part of the method (which leads from your having a barn-type experience to your believing “That’s a barn”) yields false belief in those circumstances in some nearby possible worlds. Moreover, there is in my view no rational way of basing the belief “That’s a green barn” on an experience of this sort — at least so long as this belief involves the normal human concepts ‘green’ and ‘barn’ — except by means of an inference from “That’s a barn” and “That’s green”. Since knowledge requires rationality as well as reliability, Kripke was quite right that this belief isn’t knowledge. (On the other hand, if someone did possess an unstructured concept “grarn”, which was true of all and only green barns, then they could rationally form the belief “That’s a grarn” directly on the basis of their perceptual experience, and that belief would be both rational and reliably-error avoiding.)

    However, there are the following two differences between Sosa’s condition and mine.

    1. In one way Sosa’s condition is weaker than mine. A belief can be safe in Sosa’s sense even if it based on a false lemma. The “Gettier examples with a counterfactual twist” show this. (E.g. suppose that the only reason why Jones went to such lengths to make Smith believe that he owned a Ford was because Brown was in Barcelona; if Brown hadn’t been in Barcelona, then Jones would never have made Smith believe that he owned a Ford.) My condition is stronger because it does rule out beliefs that are based on false lemmas.

    2. In another way, it seems that my condition is weaker, since it is explicitly restricted to “nearby possible worlds”. I’m not completely sure how to read Sosa’s condition, but it sounds rather like Nozick’s idea that if the belief weren’t true, then I wouldn’t have the basis for it that I actually have. So, it seems, your belief will be “unsafe” in Sosa’s sense if in one of the nearest possible worlds where the belief is false, you have the basis that you have in actual world — no matter how remote from the actual world that possible world may be. So on this reading of Sosa’s condition, your belief “I’m not being deceived by a demon into thinking that I have hands” is going to be unsafe: whatever your basis for that belief is, you might have that basis if the belief were not true.

    Personally, I just don’t see how what goes on in those remote far-out possible worlds is relevant to whether or not I know here in the actual world. Hence I think that a condition that does not use counterfactuals at all, but is explicitly restricted to “nearby possible worlds” is clearly preferable.

  2. Ralph, nice explanation of the motivation for the green barn case. There’s a couple of things to note, though. You and Sosa agree that there is a barnish character to experience independent of color, and in some sense, that is surely true: we can experience barns of lots of different colors. In this sense, there is a barnish character to experience independent of size and geometry as well. But in this sense, it should be an empirical question whether inference is involved in believing that there is a green barn in front of one, or a square barn, or a rectangular barn. In fact, if the barn is green, I’m less inclined to think that I’m inferring that there’s a green barn there–red barns are what I expect, so the greenishness would stand out in my experience in a way that reddishness might not. In any case, we should be able to test response times to the two questions “is that a barn?” and “is that a green barn?” and get slower responses to the second question if inference is involved.

    There’s a caveat for this experiment, however, for it might be that the issue here is independent of the phenomenological question of whether there are two separable experiences, one of greenness and one of barnishness (my suspicion is that the issue is independent of the phenomenology). That is, there may be no reason to think that independent experiences go straight to belief rather than in combination causing directly a complex belief (such as “that’s a green barn”). And if the phenomenology goes straight to complex belief, even if the phenomenology is itself a complex of two independent appearances, Sosa’s phenomenological account has to treat the case differently (in terms of whether it is, or can be, a case of knowledge) from the case where a really simple inference occurs. That I don’t yet find plausible.

    On your second difference noted above, I think to honor Sosa’s insistence that safety is supposed to be something like a contrapositive of sensitivity, we need to construe his language in terms of a conditional relationship between a belief with a certain basis and the truth. So we need something like “if it were believed on this basis, it would be true.” You’re exactly right that if we make the antecedent some false claim, then you get the problem of distant worlds coming into play, but I think that interpretation doesn’t fit well with the history of Sosa’s proposals about safety.

  3. Thank you so much, Jon, for clearing up my misunderstanding of Sosa!

    Presumably, if the antecedent of a subjunctive conditional is true, then (if the truth of the conditional depends on what is going on in any non-actual possible worlds at all) its truth will only depend on what is going on in the “nearby” possible worlds. If Sosa’s condition is the contrapositive of sensitivity, then in the cases that we’re interested in, the antecedent will be true: the believer will hold the belief in question on the basis in question. So in these cases, Sosa’s condition will in effect be restricted to nearby possible worlds — making it even closer to my version of the “reliability” condition!

Leave a Reply

Your email address will not be published. Required fields are marked *