Thursday, February 25, 2010

Knowledge Entails Certainty

The idea that knowledge entails certainty is a very intuitive one. It's easy to forget this, because most of us have it drilled into us, early in our epistemological careers, that embracing a certainty requirement on knowledge leads to skepticism, and we're rightly convinced that skepticsm is crazy, so we start getting used to the idea that there can be uncertain knowledge. Someone can know that p without being certain that p. If we say it enough times, it will stop sounding like a contradiction. And most of us have now said it enough times, so that it has stopped sounding like a contradiction.

Faced with a choice between skepticism and uncertain knowledge, we should indeed choose the latter. But we shouldn't forget that it's an intuitive cost. Given the choice to avoid both, we should consider it seriously.

Is this sounding familiar? It's exactly the opening structure of Lewis's "Elusive Knowledge," but discussing certainty instead of fallibility. I think the argument goes through in just the same way. Contextualists about 'knows' are uniquely positioned to vindicate that knowledge is certain knowledge, and to do it without resulting in skepticism.

The basic structure of it is easy. Let 'knows' and 'is certain that' both be context-sensitive, and let it be that for any context c, the property picked out by the former in c entails that picked out by the latter in c. But a typical effect of asking or enquiring about 'certainty' is to induce a more skeptical context. So I might answer differently to the question "do you know that p" than I would to the question "are you certain that p". But once I've answered 'yes' to the first, I face strong pressure not to say 'no' the second -- or if I do, it will feel like a retraction.

So if you're a contextualist about 'knows,' then you can, if you want, think that knowledge requires certainty. And it looks to me as if there's every reason we should want that. "I'm not certain that p but I know that p" sounds crazy.


  1. "I don't know that p, but p" sounds pretty crazy to most people too. Should we conclude that p being true requires that I know it?

  2. Hey Jonathan,

    I think I'm sympathetic to contextualism about "certain" but am officially agnostic as to whether contextualism about "knows" is correct. Following Stanley, I think there's nothing at all strange in saying:
    (1) I know that John's in the basement. Indeed, I'm certain of it.

    Compare (1) with:
    (2) I know that John's in the basement. Indeed, I believe that he is.
    (3) I know that John's in the basement. Indeed, John is in the basement.

    The thought is that you cannot reinforce entailments but can reinforce things that are pragmatically implied. So, a tiny bit of evidence that "certain" isn't among the conditions necessary for knowledge and that the proper treatment of "He knows but isn't certain" or "I know but I'm not certain" won't require positing a contradiction.

  3. Garret Merriam2/25/2010 02:16:00 PM

    "Let ‘knows’ and ‘is certain that’ both be context-sensitive,"

    How can 'is certain that' be context-sensitive? How can you be CERTAIN in one context but not the other? (Assuming here that 'certainty' is a infallible epistemic relation to fact, rather than simply an attitude of confidence). If I'm certain 'that P' then I must be certain in any context. Unless I'm not understanding what sense of the word 'context' you're using here.

  4. Jonathan Weinberg2/25/2010 03:52:00 PM

    A meta question. I know you're mostly just parroting Lewis here when you write this: "Faced with a choice between skepticism and uncertain knowledge, we should indeed choose the latter. But we shouldn’t forget that it’s an intuitive cost. Given the choice to avoid both, we should consider it seriously." But I'm really wondering, just how do you accommodate this sort of move in your own "intuition, schmintuition" take on methodology? If S credits herself with "there is no uncertain knowledge" as a member of her evidence set, then there's not a _cost_ here -- S just has a straight-up _falsification_ of any uncertain-knowledge-tolerating epistemology. But if she doesn't have _something or other_ to that effect in her evidence set, then she doesn't have any way of making it out as a cost for uncertain-knowledge-tolerating epistemologies.

    One appealing way to accommodate this -- and to provide a relevant something-or-other -- is to observe that the proposition itself has some status that (i) is somewhat indicative of its likelihood, but (ii) nonetheless which might not suffice for knowledge, while (iii) is something that we can know the proposition to possess, even in the absence of knowing that the proposition is true. Intuitiveness is an, er, intuitive candidate for such a status. But it's not one that you want to make use of. (Indeed, maybe you don't want to make use of anything like this kind of machinery at all.) So, that's why I'm asking: how do you want to handle the kind of argumentation of which that quoted bit of text is an instance?

    Independent of all that, and fwiw, I find Clayton's Stanlevian argument very compelling. I also think that it's hard to square a knowledge-entails-certainty view with the existence of non-deductive knowledge, but I know you have some views on that matter, and anyway it'd a bit beside the point here, since I take the same worry to apply to infallibilism as well.

  5. Aidan: I didn't just mean that the assertion sounded crazy; the proposition itself does. You can think that p, even though I don't know p -- I don't think you can think I know p but am uncertain of it.

    Clayton: Good point. However, given the pragmatic differences between 'knows' and 'is certain that' suggested in the post, I think we can explain the pattern of data. The line will be that the 'in fact, he's certain' will generate a more skeptical standard midsentence, so the second conjunct ends up literally stronger than the first. Compare: "it is flat. Indeed, it is perfectly flat."

  6. Jonathan Weinberg2/26/2010 07:26:00 AM

    Jonathan, perhaps you should read Clayton's argument as putting pressure on whether or not one can really make that pragmatics move. Why should inquiring after certainty be a context-shifter in the way you suggest, if knowledge entails certainty?

    It seems to me that this might be a place where some of what works for the contextualist infallibilist wouldn't work for the contextualist certaintist. When I ask you, "is there any way you could be wrong?", then that might induce you to think of some possibilities that were hitherto uneliminated-but-appropriately-ignored, but, once thought of, can no longer count as appropriately ignored. But when I ask you, "are you certain?", I don't see why that should induce you to do anything different (on the view in question) than see whether you already count yourself as knowing. Is it just a basic linguistic fact about "certain"-discourse?

    What is the equivalent of the "indeed" utterances for the infallibilist, btw? Would it be, "I know it. Indeed, there's no possible way that I might be wrong."? But that sounds weird for different reasons. It doesn't sound fine like Clayton's (1), or weirdly redundant like his (2) and (3). Rather, for all that it sounds weird to say, "I know it, but I might be wrong", it sounds to my ear bizarrely hubristic to say, "I know it, indeed, there's no possible way that I might be wrong."

  7. "[...] it sounds to my ear bizarrely hubristic to say, 'I know it, indeed, there’s no possible way that I might be wrong.'”

    But what about: I know that I exist; indeed, there's no possible way I might be wrong [i.e. go wrong in believing that I exist].

    Sounds fine.

  8. jonathan weinberg2/26/2010 09:54:00 AM

    well, sure ... but that's obviously turning on the nature of that particular proposition, isn't it? So it's not telling us anything about any general relation between knowledge and fallibility.

  9. Jonathan W.,
    I'm not sure what you mean by 'general'. Anyway, there are lots of other propositions for which the same is true, e.g. 'something exists', '1>0', '0=0', 'P --> P', some would say 'God exists'.

    Will it sound weird to reinforce non-obvious entailments?

    Jonathan I.,
    On your view, someone psychologically incapable of certainty would thereby be incapable of knowledge. Isn't this an unpalatable consequence?

  10. Jonathan Weinberg2/26/2010 01:25:00 PM

    I don't disagree that there are a fair number of propositions about which it seems fine to claim we just can't be wrong about them. That's something interesting about those propositions, and lots of folks have said lots of good things about them & their epistemology. But I just don't see what that has to do with the argument at hand, which is about context-internal entailments from knowledge that p to infallibility about p, and to certainty about p. Such entailments are not,I take it, supposed to turn on any particular values for p. No?

  11. "Clayton,
    Will it sound weird to reinforce non-obvious entailments?"
    My guess is that it would not, especially if they are especially unobvious entailments.

  12. Sorry for the slow responses, all. Continuing the overdue replies:

    @Garret-3: When I say that 'knows' and 'certain' are context-sensitive, I'm making a linguistic claim: I'm suggesting that the truth-conditions for the relevant discourse depend in part on features of the speaker's conversational context. I don't mean, therefore, to be suggesting that someone can be 'certain in one context but not the other'. Like you, I have a hard time understanding what that would even mean. (Unless it's meant to express the uninteresting possibility that I'm certain at some times about some things, but uncertain about other things at other times.)

  13. @Weinberg-4: RE: methodology: I'm an E=K guy. I think it's legitimate to cite something as evidence just in case you know it. So if S does know that knowledge entails certainty, then you're right -- she can just cite that fact, and deductively prove that admitting uncertain knowledge is a mistake. She'd have to be pretty confident in order to make that move, though. (If I'm right, she'd have to be certain!) If she's like most of us mortals, then she has at best a number of considerations that count in favor of the view. I don't mind counting as among the considerations such propositions as "I have the intuition that knowledge requires certainty" or "it is intuitive that knowledge requires certainty." What I'm concerned to deny is that this sort of proposition comprises all, most, or even a very substantial proportion of the relevant evidence. Facts about how I feel about something comprise at best a pretty anemic sort of evidence for philosophical theories.

    What do you think about the response I gave to Clayton? If you're not a perfect circle, then you're not a circle. But the utterance, "it is a circle; indeed, it is a perfect circle" is a legitimate one. I don't think it's that strange for there to be context-shifters like I am suggesting this is. (I think that saying 'know' with a very confident, authoritative voice and adding stress has a similar effect. I know that p; indeed, I KNOW that p.)

    "What is the equivalent of the “indeed” utterances for the infallibilist, btw?" I don't think it'd be the sort of thing you suggest; the modality of that 'possible' is too fragile, and too easily gets read as something other than the epistemic one needed. For the infallibilist of my Lewisean sort, I take it that the relevant additions would be explicit claims to be able to rule out counterpossibilities that had hitherto been ignored. Suppose we're speaking in a nonskeptical context in which the possibility that my lottery ticket will win is legitimately being ignored. Then I may say: "I know that I won't go on safari next year; indeed, even in light of the fact that I might win the lottery, I know that I won't go on safari next year."


    RE: abductive knowledge: There is no abductive knowledge.

  14. Jonathan Weinberg3/02/2010 05:13:00 PM

    "What I’m concerned to deny is that this sort of proposition comprises all, most, or even a very substantial proportion of the relevant evidence. Facts about how I feel about something comprise at best a pretty anemic sort of evidence for philosophical theories." That's all well and good, but then you really need to take seriously how very _revisionary_ a claim it is. Just look back at your main post: the entire brief against uncertain knowledge as presented there, is that giving it up would present "an intuitive cost". There is no invocation of other considerations (and what would they be?) Moreover, philosophers make that kind of argumentative move all the time. If what you said in your last comment was right, then when philosophers do that sort of thing, they're making a pretty big mistake. And maybe they are making a big mistake! But then you should help try to correct these practices in their erroneous ways.

    Now, I know you have this view that you report as "there is no such thing as abductive knowledge". On any way I can see to understand that claim, though, it is either so boringly trivial that nothing of any note comes of it (i.e., if E=K, then if you know something, it is entailed by your evidence, because it is part of your evidence... ), or utterly insane. Here's a question that'll help me figure out which disjunct it is: there is a vast amount of what folks today currently believe, and which we have very good reason to think that they are correct in believing, and which is very commonly referred to as "scientific knowledge", and which has been reached by using abductive forms of inference. Now, is everyone just wrong when they take it to be knowledge? And if they aren't wrong about that, then why isn't scientific knowledge a primo instance of abductive knowledge?

    "If you’re not a perfect circle, then you’re not a circle." That sentence is false. Indeed, it's _obviously_ false. The vast majority of circles (probably all physically existing ones!) aren't perfect circles, some very odd extant philosophical inclinations to the contrary notwithstanding. Even if I'm wrong about this, then the (putative) fact that only perfect circles are circles can't really do any work to make sense of how utterances like "It's a circle; indeed, it's a perfect circle" generally work, since _no one other than some weirdo philosophers_ believes that (putative) fact. (Well, and maybe some mathematicians. But we know that mathematicians are insane.)

    Anyhow, I agree that shifts in emphasis, tone, etc. can be used to shift standards. It's pretty clear that that just is the conventional job of such shifts, and of course it has nothing to do in particular with the predicate "knows". (E.g., I'm going to Memphis soon, and even though around here you can find bbq, or often just "bbq", it'll be nice to score me some _BBQ_!) What's still unclear, on the entailment account, is why "I'm certain that p" should stand to "I know that p" in anything like the way that "I KNOW that p" stands to "I know that p". As I said earlier, maybe the certaintist will have to just say that it's just a fairly basic linguistic fact, like the fact about tone/emphasis? That doesn't strike me as terribly attractive -- a bit ad hoc, even -- but do they have any better options?

  15. Jonathan, I'm not convinced that I'm making a particularly revisionary claim when I say that intuitions don't comprise a very important part of one's evidence. I know that there are a few people who have written on the methodology of philosophy who sign up to that claim -- I think they're wrong -- but I don't see that there's any kind of widespread commitment to a practice that is by my lights erroneous. And I certainly don't see why you think the passage you've quoted in my post commits to such a practice. Are you supposing that when I said that uncertain knowledge was 'counterintuitive', I meant that thing I said to comprise the central evidence counting against uncertain knowledge? I didn't; that was a rhetorical move, designed to remind the reader of the considerations that do count in that direction. (Compare the use of such sentences as "we all agree that p" -- they don't state evidence; they remind us that there is a compelling case for p.)

    On allegedly abductive knowledge: this is a bit afield, and really ought to be discussed more thoroughly elsewhere, but the basic line I'm interested in is closer to your trivial interpretation than to your inane one -- but not, I think, in a way that renders it genuinely trivial. I do think that lots of the stuff that most people think is abductive knowledge is genuine knowledge, but I deny that it's abductive. Not merely because E=K -- my view's a bit stronger than that. E = K says that for all knowledge, there is evidence that entails it; I claim that all knowledge is BASED on entailing evidence. Since not all knowledge is self-evident, I make a stronger and less trivial claim. Basically, I think that there's lots of tacit knowledge of bridging premises, and that this tacit knowledge plays an important role in inference.

    On circles and perfect circles: as Unger cogently observed, pointing out that something isn't a perfect circle is a pretty good way of convincing people that it's not really a circle at all. So the conditional you say is obviously false is -- wait for it -- intuitive. You're pointing out that it looks like it has crazy consequences, like the consequence that the side of a hockey puck isn't a circle. That's like saying that infallibilism leads to skepticism. The solution to both worries is contextualism.

  16. Jonathan, obviously there's quite a lot here -- more than is really apt for this forum -- but let me just say a couple of things. First, I don't agree that Lewis's methodology is irreconcilable with my metaphilosophical outlook. In the passages you cite, he isn't explicitly making claims about evidence, and I don't see why we should interpret him as making implicit ones.

    Just as a matter of contemporary history of epistemology, I don't think of Unger as (merely) defending a rival to contextualism. Rather, I think of Unger as an important precursor to contemporary contextualism -- that's certainly the way DeRose talks about it, and I think there's something to that.

    Maybe you and I agree that my claims about knowledge and certainty are as plausible as Unger's about circles and perfect circles. I'd like that. I do take that data seriously, although I see that you don't agree with it. I agree that it might be interesting to see what the folk think. I suspect they're friendlier to the Unger line than you suggest. (But if I'm wrong about that, I'm not too worried. They're only the folk.)

  17. Jonathan Weinberg3/15/2010 01:51:00 PM

    I don't really understand what you're saying about Unger; I mean, it's right that his view is very closely related in a lot of ways to the contextualist rival he constructs for it, so of course his discussions play an important role in later elaborations of contextualism. But it's still the case (isn't it?) that the contextualist shouldn't want to tie the meaning of her contextualist term of choice to a construction in which all the contextually-sensitive parameters have been set to a _non_-contextually-sensitive maximum setting. And words like "absolutely", "perfectly", etc. do exactly that -- they set that parameter of a gradable predicate to its maximum setting. Which is why it shouldn't make any sense, if you want to be a contextualist about circles, to say "all circles are perfectly round."

    Regarding Lewis, you say, "In the passages you cite, he isn’t explicitly making claims about evidence...." I take that statement to be directly falsified by Lewis' reference to "the explananda phenomena", unless you have a very weird view on which our explananda are not part of our evidence (which seems required by taking inference to the best explanation to be a good, evidence-based form of inference.) So evidence is very much an explicit part of the discussion.

    Looking at the passages in question, he's clearly saying: the badness of "I know that p but I could be wrong" is part of our evidence; and he's also clearly representing himself as having no further evidence for that badness, beyond how he thinks the re-naive ear will hear it.

    Fwiw, responding to your last parenthetical point: I can't think of any reason to _want_ to be a contextualist, if we have free license to just override the folk usages at will. For Heaven's sake, if that's the methodological situation, then let's just all be traditional, boring, non-sensitivist, non-contextualist invariantists, and call it a day. (That is, in fact, the theory of knowledge I would have if held at gunpoint and forced to have a theory of knowledge.)

  18. An important part of the project, Jonathan, is to be contextualist about terms like 'perfectly' and 'absolutely', too. These are gradable adjectives. So I'm denying that these represent an invariantist maximum standard.

    (Notice also that to describe the view as 'contextualist about circles' is to commit a use/mention error.)

    I'm not going to agree with you about Lewis and evidence. There are lots of non-psychologistic bits of evidence that Lewis can happily recognize. You think the only evidence Lewis represents himself as having in favor of the claim that concessive knowledge attributions are untenable is that they sound bad. But how about this obvious fact: they are conjunctions of knowledge attributions with the epistemic possibility of the negation of the object of the knowledge attribution. That a sentence conjoins a claim of knowledge of p with the possibility of not-p is strong evidence that it is false.