I’m now in London, bumming around until I head off to Beirut on the 9th. If there’s anyone out there in London who wants to meet up with me in the meantime, feel free to email me, or communicate by some other means. On that note, I’m also now on twitter, for anyone who hasn’t already spotted me. This is a very quick post in response to Catarina Dulith Novaes’ post on the KK principle (here), because I couldn’t seem to post a comment on it. It’s thus pretty short (by my standards).
I won’t recapitulate Catarina’s post in any detail, as it’s a very short post itself, but the suggestion she makes is that the problem of whether the KK principle (i.e., if someone knows p, then they know that they know p) is true is amenable to empirical resolution to some extent, on the basis of research into metacognition. She also suggests that Kantian approaches to epistemology are incompatible with such analyses, the implication being that this is another reason why they don’t cut the mustard. I’m an unabashed Kantian on these matters. I take the widespread hostility to transcendental approaches to cognition to stem largely from the assumption that they place illegitimate constraints upon, or are downright incompatible with, empirical approaches. However, even Kant is fairly explicit that transcendental psychology is supposed to be a complement to empirical psychology, rather than a substitute for it. Whether or not Kant’s account of it is correct is another matter (though I increasingly suspect that it is less silly than it is often portrayed to be), rather, the issue is whether there are such things as legitimate constraints upon empirical approaches to cognition.
Unsurprisingly, my answer to this is a resounding yes. The reason for this is that one needs to define what knowledge (or cognition) is in a priori terms before one can deal with the specifics of the mechanisms that produce it, lest one localise it to some particular mechanisms (e.g., homo sapien brain structure, neurological structure as such, etc.). This feeds into debates I have been having with both David Roden (e.g., here) and Levi Bryant (e.g., here) for a while now. Just as I think that we must distinguish between metaphysics and the critique of metaphysics (see my ETR), I think that we must distinguish betweenepistemology and the critique of epistemology. This is to say that we can naturalise epistemology in part, but that the condition for this naturalisation is that there is a part of epistemology which cannot be so naturalised. In essence, a moderate Kantianism is not only possible, but necessary, because one can only work out the balance between a priori epistemological considerations and a posteriori ones on an a priori basis. If one rejects that there are anything but a posteriori considerations, then one ultimately has to sever the connection between epistemology, logic, and semantics, which is tantamount to simply changing the subject.
Given all this, I’m going to attempt to take a moderate Kantian approach to Catarina’s suggestion regarding the KK principle. Really, I’m just going to extrapolate the consequences of the Brandomian/Sellarsian position that I endorse. This gives us good a priori reasons to think that the KK principle doesn’t hold, but it also gives a good explanation of why people might think it does, as well as letting us see why there might be a posteriori reasons to think it holds in limited cases.
I think that Brandom’s analysis of knowledge as a hybrid deontic status is pretty much on the money. To repeat the basic elements for those who aren’t aware of it, he thinks that the externalist dimension of knowledge is provided by the perspectival difference between the one who ascribes knowledge and the one to whom it is ascribed. So, for A to know p from the perspective of B is for B to ascribe commitment to p to A (Belief), entitlement to p to A (Justification), and to undertake commitment to p themselves (Truth). His analysis is thus a fairly classical justified true belief (JTB) account reinterpreted in deontic scorekeeping terms, with a few tweaks here and there.
Setting aside pre-emptive defenses against the standard objections to Brandom’s model (e.g., ‘What about objectivity?’, ‘What about cognitive science?’, ‘This leads to a idealised subject, spinning in the void!’, etc.), this lets us see that KK doesn’t hold from the perspective of the ascriber, insofar as they can ascribe entitlement to p to A, without thereby ascribing a scorekeeping commitment to A that they are so entitled. To repeat that, we can hold that someone believes p and is justified in believing p even when they themselves do not think they are justified. However, because the difference between commitment to p and commitment to Kp is, from one’s own perspective, entirely constituted by whether one takes oneself to be entitled to p, if one does take oneself to be entitled then one automatically takes oneself to know. This means that it is only in the fairly odd situation in which one takes someone to be entitled to p, and takes them not to take themselves to be entitled to p (or, in Gettier cases, takes them not to be entitled to their self-entitlement ascription), that KK breaks down.
This fits neatly into Brandom’s appropriation of Sellars’ version of reliabilism, wherein epistemic expertise is to be understood in terms of the reliability of the causal mechanisms that produce commitments (understood in terms of both abstract idealisations and concrete realisations), be they straightforward perceptual mechanisms, practical coping mechanisms, or computational heuristics. This underwrites defeasible inferences from a commitment (or judgment) being produced by a mechanism to its truth, and thus enables individuals to justify their non-inferentially acquired commitments, be they produced by their own mechanisms, or by the mechanisms of others (i.e., deference). These inferences can be understood either in terms of brute statistics (treating the mechanism as a black box), in terms of specific defeasors that undermine them (dealing with the internal structure of the mechanism), or some combination of the two, but are in each case assessed from the perspective of the scorekeeper. The important point is that questions about whether these inferences are good ones are thus susceptible to empirical evidence.
It is thus at this level that questions to do with metacognition are really to be addressed, insofar as we might locate certain cognitive mechanisms that have interesting kinds of reflexivity. This seems to largely be the case with simple perceptual capacities involving our basic sensory modalities (e.g., sight, sound, taste, touch, etc.). On the one hand, the perceptual mechanisms that produce observational commitments tend also to produce scorekeeping commitments to the effect that those commitments are entitled (thus licensing ascription of the latter), and these latter commitments tend to be given a high psychological weighting in respect to others (i.e., we are less likely to give them up in cases where they conflict with other commitments). On the other, it is possible to make an empirical case that when we (or perhaps certain specific individuals) take themselves to be entitled to certain observational commitments, that they tend to be entitled in this way, or, that these reflexive mechanisms are themselves reliable (statistically and/or structurally). Precisely which mechanisms are reflexive, to what extent they are reflexive, and whether this reflexivity is warranted are thus all a posteriori questions, even if what reflexivity is is not. This demonstrates how Sellars provides us with the tools to connect Kantian critical epistemology with Quinean naturalised epistemology.
I’m sure more could be said on this specific epistemological issue, and there’s much more I want to say about the transcendental/empirical divide, but this will have to do for now.
DEONTOLOGISTICS 
The thing that puzzles me about the KK principle is that it seems to be regularly contradicted just by ordinary experience. I often find myself thinking (and sometimes saying) “Huh – I didn’t know that I knew that”. Surely this basically settles the question?
As with most things in philosophy, a little extra thought on the matter can’t hurt. Understanding why you’re basically right in your intuitions can be pretty interesting, insofar as it’ll contain some counter-intuitive consequences.
Oh sure, I’m not objecting to discussion of the issue or elucidation of the implications of how knowledge functions, etc. I’m just puzzled that it’s such a persistently advanced claim.
There’s a real inattentiveness to personal experience – a dissociation between that experience and the process of theory-formation and -discussion – required to render this sort of thing plausible, I think.
Hi Pete. A quickie on Philosopher’s Union rates I hope!
So (if I’m not mistaken) the Brandom/Sellars/Wolfendale line is:
1) Questions of epistemic reliability are a posteriori. For example, a search algorithm which takes into account the cost of getting to a node as well as the distance of the node from the goal is more reliable than one which jumps to the ‘best’ node without estimating the cost of getting to that point. You can discover this by running a series of searches using the different heuristics.
2) Conceptual questions such as what count as knowledge and its structure are a priori since they reflect invariants in the tissue of normative commitments which underwrite content – such as inference from knowing that p to being justified that p and being committed to p (in the third person case, at least). Among, these questions are questions concerning the nature of reflexivity (e.g. what is it for a representational to be about other representations?).
3) Questions about the balance between the a priori and a posteriori epistemology are also a priori.
3) Seems to preclude a situation where we might discover some state of a naturally occurring or artificial system which is the result of an inferential process on representations but is non-propositional and does not involve socially recognizable vehicles of content.
While one can object that excluding such a possibility is just a matter of sticking to the epistemic a priori which makes the questions about what counts as knowledge intelligible in the first place, I must say I can’t see the force of this argument unless we are already inclined to embrace it on independent grounds. With regard to the specific question of reflexivity, however, I think sticking to our a priori guns could actually have a deleterious effect in cognitive science. This is because there are reasons for thinking that certain states of the brain (e.g. conscious states) exhibit reflexivity (because they involve shared access to content fixations in the head) even though the precise nature of that reflexivity in computational or even logical terms is unclear. For example, if there are no sentences or symbols in the head, then it will not involve symbols with the neural equivalent of quotation marks! This seems to be a legitimate case where not merely the mechanisms but the nature of reflexivity is something that will need to be determined empirically.
Hi David,
Let me try to take this stuff in order:-
1. I’m sure there can be a priori aspects to assessments of reliability, insofar as things like search algorithms can be studied in terms of their a priori features. However, I think arguments about reliability will be a posteriori for the most part.
2. I think you’ve misunderstood what I meant by reflexivity. This may be my fault, as I just pulled the term out of the air because it sounded good at the time. The issue is very specifically about mechanisms that produce theoretical commitments (or dispositions to acknowledge such commitments), and those ‘reflexive’ mechanisms that tend to produce commitments about these commitments (i.e., that they are entitled) at the same time as the former. Again, I think you’re using the word ‘representation’ in too broad a way here.
3. Yes. The idea that the balance between the a priori and a posteriori is itself a priori does preclude discovering an inferential process that is not propositional, as long as you agree with me (and Brandom) that there cannot be inference without propositional structure. The point is that inference is not an empirical topic. This should be clear from logic and the philosophy of logic. It’s analogous to the claim that we aren’t going to stumble upon a computer that violates the laws of computation as laid down by computing theory.
The repeated worry is that somehow a priori analysis of the structure of thought is going to place illegitimate constraints on a posteriori inquiry into the ways this structure is and can be instantiated. My response to this is that the constraints will only be illegitimate if the a priori analysis isn’t done properly. This is possible. A priority does not mean that there is some kind of special apodictic intuition involved, but rather that there is *no* intuition involved. As in maths, it’s possible for us to follow the procedure wrong, or to think we’re characterising the general phenomena when we’re only characterising a species of it (e.g., euclidean geometry). None of this means that a posteriori discourse somehow trumps a priori discourse.
Thanks for clarifying your use of ‘reflexivity’, Pete.
I had assumed that ‘ mechanisms that have interesting kinds of reflexivity’ are those which allow some kind of semantic or intentional ascent. That is to say, they allow a shift from talking about X to talking about talk about X or from representing X to representing X’s mode of representation.
It’s plausible that something like this reflexive operation is necessary for sophisticated reasoning and (if Davidson and Brandom are right) for full-blown propositional attitudes.
In the form of intentional or representational ascent, it may also be involved in non-linguistic cognition or consciousness. For example, not every perceptual state or attitude get’s ‘promoted’ to consciousness. We tend to be conscious of stuff that is salient in the light of our current activity and previous representations may get revised or ‘overwritten’ in the light of incoming information. The fact that we have selective attention or representations of the temporal ordering of our experiences suggests that our brains are able to track, evaluate and make available their own activity in a variety of ways.
It’s hard to see how practices of ‘deontic score keeping’ could get underway without these capacities. Even if language allows us to locate these states within ‘the space of reasons’this basic activity would be impossible if we were unable to monitor our own perceptual activity and agency. This suggests that capacities for meta-representation cannot be exclusively linguistic; even if language, by dint of its publicity and explicit compositional structure provides a wholly new kind of reflexivity.
Ok, I finally got my act together and managed to show up here! Lots of things I could say, but let me focus on the main ones.
As per our Twitter discussion, it’s going to be a matter of balancing how much conceptual underpinning is required, and to what extent this very conceptual framework may be revised on empirical grounds. I guess no one will say that it’s all an a priori issue (perhaps the rabid Kantian, but not even Kant himself, as you point out), nor that it’s all an a posteriori issue, because to be able to conceptualize the empirical data themselves some conceptual underpinning is necessary, always. So obviously, before we even get to start thinking about how KK could be investigated empirically, we must have a fairly robust and largely conceptual story of what knowledge is; nobody could deny that. I suspect that people often misunderstand me when I’m talking about ’empirically-informed philosophy’, as if I was proposing to *reduce* all philosophical questions to empirical questions — and even if that was the case, obviously in the empirical sciences themselves, say experimental psychology, a sufficiently developed conceptual framework is indispensable. (And often things go really wrong in psychology because of the wrong conceptual underpinning; I have a short paper on the not-so-positive influence of Kantianism on the psychology of reasoning tradition, if you are interested. It will appear as a commentary in BBS.)
But I’m more of a Neurath-boat kind of person, I think we often keep working on the foundations of an investigative enterprise as we go along. One of my worries about transcendental approaches in general concerns the possibility of revisability. One frequent argument goes: you can’t revise your foundations in light of empirical elements, because to gather these empirical elements themselves you were using the old theory which you now want to change; so this data-gathering is itself not reliable. I’m not attributing this argument to you, but in general this sort of argument strikes me as totally wrong-headed (we can talk about it in more detail if people are interested).
And finally, I remember very well Brandom’s analysis of knowledge in Making it Explicit, and found his idea that I, the score-keeper, may know that somebody knows something, while the person herself doesn’t know that she knows it (while knowing it) very compelling. But I think there *are* empirical elements in this story in the first place, and reliabilists do rely on actual occurrences of such phenomena (e.g. chicken sexers) to support their case.
To say that KK is amenable to empirical scrutiny does not entail that a priori arguments against it will be of no interest. In my original post, Roy Cook offered an argument on why KK cannot hold, using the usual diagonalization-paradox template. I have no objection to any of this (I’m a methodological pluralist!), my proposal is simply that KK and other similar principles could ALSO be studied taking empirical data into account.
(Hey, can you do something about how comments are typed here? The letters are so small, white against grey, it’s really difficult to see what I’m typing! There must be a million typos along the way… And I probably need glasses.)
Yes, this:
And also this:
Totally agree with these points.
Hey, I know this is totally unrelated, but I wrote this sendup of the Speculative Realist/Object-Oriented Ontological Blogosphere which you might find entertaining. It’s kind of mean in parts, but I think if everyone has a good sense of humor about things it’s not so bad.