miércoles, 7 de septiembre de 2011

Sellars, McDowell, Burge -Perception as Non-inferential Knowledge

- Perception as Non-Inferential Knowledge -

I have been following my reading of Sellars in tandem with some pertinent secondary literature, including: deVries and Triplett's reader's guide, Brandom, McDowell, O'Shea's excellent study, and some criticisms following particularly from Tyler Burge in his 'The Origins of Objectivity'. This has a resulted in a cluster of preoccupations, centered around the Sellarsian account of perception as a form of non-inferential knowledge which includes a conceptual (normative) component, and a non-conceptual sensorial (natural) residue, which remains epistemically inert.

1) The first thing that struck me upon reading Tyler Burge's criticism of what he takes to be an over-intellectualizing tendency in Sellars' account, is the claim that perception must involve a form of conceptual/linguistic response as the only possible way to explain the attributions of sense content by cognizing subjects. I quote Burge apropos Sellars' position as expressed in 'Phenomenalism':
"No serious science of perception agrees with Sellars in taking seeing something to be red as the 'culmination' of the sophisticated linguistic practices that he describes. Sellars leaves no room between (a) 'S has a sensation of x', in the sense of 'S is in that state brought about in normal circumstances by the influence of x on the relevant sense organs', and (b) 'S has a [linguistically informed] thought of x'. That is he allows for no perception of entities as having physical properties that is not backed by linguistically informed thought that attributed such properties to such entities. He moves quickly from a non-representational notion of sensing to a propositional sensing-that- again backed by linguistically informed thought- with no room for any type of perception in-between... Sellars assumes that propositional, linguistically informed thought is the only source of objective representation. Objective representation is epistemic representation. Epistemic representation requires a linguistically grounded propositional ability to represent conditions under which objectivity and knowledge are realized." (OB: Pg 137n).

   Burge's entire point is that perception involves a non-linguistic form of individuation for its contents, which is explained without recourse to conceptual training or the capacity to perform linguistic inferences. This is for Burge a property that is to be found not just in sapient humans, but also in other animals. What is interesting about Burge's response at this juncture is that it seems to obviate completely the relationship which Ray Brassier stresses thoroughly in Sellars' account between sentience and sapience. For Sellars does seem to acknowledge that sentient organisms do respond accordingly to sensible stimuli, without this necessitating conception. It is true, however, that as far as perception is concerned, for Sellars there must always be a conceptual counterpart proper to linguistically trained subjects, to individuate sense-contents in judgment. It would thus seem that what Burge likes to call 'perception' is a form of sentient response that ought to be called positively epistemic while being non-conceptual all the same. This leads into my second observation, which takes us to McDowell's recent polemics with Burge himself. This reveals the core issue as I see it with regard to the status of perception, non-inferential knowledge, and the role of conceptuality in Sellars' scientific realism.

    2) Elsewhere in his text, Burge objects that Sellars' intellectualist position commits him to the view that perceptual knowledge in rational subjects is arrived at always by taking an 'inferential step'. Here McDowell rightfully takes issue with Burge, since it is clear that for Sellars perception in fact can operate as a non-inferential faculty in producing knowledge. This is after all the first premise of the famous inconsistent triad which Sellars endorses: "S senses red-sense-content x if and only if S knows non inferentially that x is red."

Now, this is a point of terminological confusion since 'senses' and 'perceives' seem to be here interchangeable. This is confusing because we have established that Sellars wants to claim that sensation is epistemically inert (i.e. requires conceptual judgment to form knowledge), and yet that there can be a non-inferential form of knowledge described in perception, which is not to say that perception is non-conceptual. Indeed, it is perception's conceptual counterpart to the inertia of sensation that provides its positive epistemic status. So I take it that proposition A of the inconsistent triad, if endorsed by Sellars, refers to the non-inferential knowledge acquired in perceptual episodes, which involves necessarily its conceptual component, and not just in 'mere sensation'. In any case, McDowell goes on to defend Sellars' from Burge's attack. I quote the relevant passage, from his 'Perception as a Capacity for Knowledge':

"One of the formulations of what he finds implausibly intellectualistic about the Sellarsian picture is that it “implies that the formation of a perceptual belief is a piece of reasoning — a transition from a reason to what it is a reason for”.

 But a believer can be self-consciously justified in a belief without having formed the belief by a transition to it from whatever she would cite in giving her justification for it. That is so even if the grounds on which the belief counts as knowledgeable are inferential. When I know that my neighbor is at home on the basis that, as I can see, his car is in his driveway, I do not need to have taken an inferential step to the belief that he is at home. It might be perfectly natural co say I can just see, straight off, that he is at home. Even so, my belief that he is at home counts as knowledgeable, if it does, because there is a good enough inference from the fact that his car is in his driveway to the conclusion that he is at home, And my knowledge that he is at home includes self-consciousness about its warrant, so that I can produce a justification, in Burge’s technical sense, for my belief that he is at home. I know not just that my neighbor is at home but that my warrant for believing that he is consists in the goodness of that inference, even if I did not arrive at the belief by inferring it from the knowledge that grounds it for me." (McDowell, PC, Section 5)

Against Burge, then, it seems as if Sellars would be perfectly at ease in accepting a non-inferential form of knowledge, which involves conceptuality in order to provide judgment all the same. Burge would here insist that such a position deviates from what science teaches, insofar as our perceptual capacities remain non-propositional, and are legitimated thus by the externalist requirement that they cause the proper dispositional/behavioral responses in their cognizing subjects. There is much more to be said about Burge's intricate account of perception, but I shall leave that for the moment.

   The notion of non-inferential knowledge in Sellars, however, is a tricky one to disentangle, seeing that it must not fold us back into some form of the Myth of the Given. The crucial qualification in McDowell's account concerns the relationship between Sellars' endorsement of non-inferential or direct knowledge, and at the same time the his rejection of epistemic independence, i.e. that a proposition p could be known independently of knowing other proposition(s) q, y, etc. Directness requires that the subject not infer a proposition from any other, independence requires the stronger claim that in order to know a proposition p the subject need not know any other proposition(s). This is where O'Shea and deVries bring up some interesting remarks.

To begin, the fundamental difference between the two kinds of knowledge seems to be accounted for as follows:

1) Inferential knowledge - S knows p iff S knows q, S knows that q justifies inferring that p, and S infers p from q.
2) Non-inferential/direct knowledge - S knows p iff S knows that p without inferring p from whatever proposition(s) q justify it, and S knows whatever q justifies inferring that p.

   It is crucial to note that in any case, and against Burge, for Sellars knowledge demands the internalist requirement that the knower be in possession of the justification for whatever he knows, but that this needn't mean in every circumstance the subject will deduce knowledge by inferring it from whatever propositions justify it. This seems to ground the possibility of non-inferential knowledge being direct yet epistemically dependent, where we understand epistemic independence as entailing:

Epistemic Independence - A proposition p can have positive epistemic status for S independently of any other proposition q, i.e. S can know p without knowing anything else.

  Thus every occasion where knowledge obtains, whether inferential or non inferential, will be a case of epistemically dependent knowledge, according to Sellars. While Sellars wants to endorse the idea that no proposition is ever self-justifying or justification-free and so that epistemic independence fails, he wants to do this while accepting that non-inferential knowledge is possible. This is a crucial distinction, since it grounds Sellars' rejection of those 'basic' propositions that would bolster a foundationalist account, while keeping a restricted sense of directness operative. This supplements his rejection of Russellian inspired accounts of non-propositional knowledge by acquaintance, as a second iteration of the Myth of the Given that ought to be rejected.

    Non-inferential knowledge is not foundational since it does not constitute justificatory self-sufficient instances of given knowledge, obtained transparently through the intuition of sensory data. Sensation by itself is epistemically inert, and the capacity to sense sense-contents is acquired in that it requires that one knows a cluster of additional propositions resulting from being imbedded in a linguistic community. This is the anti-foundationalist prescription in Sellars that McDowell underlines, against Burge. This results in the strange idea stated above that knowing p non-inferentially requires knowing q, but that one does not arrive at p by inferring it from q, and so not epistemically independently. Furthermore, the strong internalist demand which McDowell seems to be ascribing to Sellars here is the additional requirement that S not only must know q in order to know p, but he must also know that q justifies p.

This follows from the standard tripartite demands for knowledge which Triplett and deVries argue Sellars endorses:

For S to know that p entails:
1) S believes that p.
2) p is true.
3) S must have the appropriate justification for p.

Sellars must thus insist that for any proposition p, p is epistemically dependent on another proposition(s) q for S, i.e. to posses knowledge of p is justified by / counts as knowledge, on condition that one also possesses knowledge of q. Triplett and deVries write on this account, summarizing Sellars' rejection of epistemically independent propositions:

"Suppose [a proposition assumed as basic] is non-inferentially acquired. Any such proposition can have positive epistemic status for a person only if there are other propositions in the person's epistemic system that support it. For example, a person cannot know the truth of the observation report "this is red" if she is merely capable of reliably producing such reports in appropriate  circumstances [anti-externalism / anti-Burge]... In order to have knowledge the person must know that her reports are reliable. But then her knowledge of her reliability epistemically supports her observational knowledge which therefore cannot be epistemically independent."

 This is what, according to McDowell, would entail that even if one knows non-inferentially that the neighbor is in the house by observing the car, one must nevertheless 'be prepared to justify' their belief if examined. It is not sufficient that one believes what one does; one must know that his belief is justified by appeals to other bits of knowledge, i.e. one must know that seeing the car could serve to inferentially justify that the neighbor is in this house. This is the coherentist demand which Burge finds excessively intellectual in Sellars. And it also constitutes what deVries and Triplett call 'a strong internalist requirement' against externalist accounts. We may thus reformulate requirement (3) above in its 'internalist' enhanced version, endorsed by Sellars:

3*) S must have the appropriate justification q for p, and S must know that q justifies p, i.e. S must be capable of explaining the inference of p from q.

This raises a particular elucidation to O'Shea's description of Sellars' "principle of perceptual reliability" [PR]:

[PR] - S's perceptual judgment [P] that x, over there, is red, constitutes a case of perceptual knowledge if and only if there is a generally reliable connection between cases of S's judging that [P] and its being in fact true that there is a red physical object over there [T&D, Pg. 126)
Thus, under standard conditions, S has reliable (yet not infallible) warrant for x iff one can "infer the presence of a green object from the fact that someone makes this report." This is where O'Shea, however, diagnoses quite menacing problem with regards to the possibility of a vicious regress. If S needs to know q in order to know that p, but also that he must know that q justifies p,  then S needs to have knowledge of the meta-principle [PR] which provides S with the capacity to discern reliable occasions of perception from unreliable ones. But since in general [PR] must follow from empirical observation of instances [P], it is not clear whence from it derives its justification, since it is simultaneously meant to serve as the condition for knowledge of [P]. The impending circularity looms transparently here.

   T&D state: "...it seems reasonable to accept epistemic principle [PR] only because it is in some way supported by particular observations [P]; but it is reasonable to accept observation [P] only in virtue of their being known to fall under the perceptual reliability principle [PR]" [Pg. 132]. And Sellars' response will be to disambiguate between a) the naturalistic account which tracks down the causal process of acquisition of the conditions for our epistemic sapient capacities, and b) the transcendental non-empirical register according to which principles such as [PR] (or the principle of causality) must be taken as normative conditions of possibility for any knowledge whatsoever, i.e. they seem to function as epistemic norms. Apparently, the reason why this is not another instance of the Myth of the Given is because these normative principles can be given a resolute empirical causal explanation in naturalistic terms, while preserving their transcendental status methodologically autonomous for the structure of knowledge. Thus there is no ontological gulf between the normative and the natural causal; the dualism is not metaphysical but methodological. And as I see it, this is where Sellars' provides a subtle corrective to the post-Kantian subordination of the empirical to the transcendental; the  methodological autonomy of the logical space of reasons is nevertheless anchored ontologically and generatively on the causal conditions open to empirical natural-causal description.

   This raises an interesting possibility of re-awakening Meillassoux's challenge to the correlationist defense according to which the separation of transcendental and empirical registers remain insurmountable. The corrective Sellars introduces is to say that there is no explanatory gap in the genetic account that science provides between inorganic matter, sentience and sapience; science is fully capable of causally explaining the genesis of thought, and also how the latter comes to be a condition for the very knowledge of itself and the world, i.e. neurobiological-evolutionary accounts expose how the organism comes to acquire those physical capacities to produce knowledge. This causal account is by no means restricted to self-knowledge; and so does not have to run the gauntlet of the circle of correlation. The empirical-transcendental register is thereby reworked by Sellars so that while the physical is causally responsible for the normative, the latter is epistemically responsible for our knowledge of the causal. So while there is an epistemic correlation between the normative, sapient-endowed critters capable of wielding the transcendental apparatus for rational deliberation, this does not solicit any form of bilateral ontological correlation between thought and matter, but it does solicit nevertheless a unilateral ontological correlation of thought on matter.

    All of this is obviating difficult issues raises elsewhere about the precise nature of the physical for Sellars (the two senses distinguished by O'Shea), the subtle distinctions between levels of sense content (sensing-that or propositional epistemic content, and sensing-of non-propositional responses), the dialectical interplay between the manifest and scientific image leading to the incorporation of sensa into the latter, and also the question of fallibilism with regard to perception (raised particularly by McDowell's brilliant remark against Burge that even if our perceptual faculties are fallible it doesn't follow that we can never posses sufficient grounds for infallible occasions of knowledge, i.e. as Sebastian Rod claims: "from the fact that, when I am fooled, I do not know that I am, it does not follow that, when I am not fooled, I do not know that I am not. When I know that p as I perceive it to be the case, then I know that I perceive that p."
3) However, there is one issue pending which I find extremely confusing in Sellars' account, with regard to the distinction between epistemic independence and non-inferential knowledge. My problem is that I simply do not know what it means for a subject to 'know directly and non-inferentially' with epistemic dependence. Let us return to McDowell's example:
"When I know that my neighbor is at home on the basis that, as I can see, his car is in his driveway, I do not need to have taken an inferential step to the belief that he is at home. It might be perfectly natural co say I can just see, straight off, that he is at home."

    Here the expression "seeing straight off" must be a shorthand for: knowing non inferentially that p, where S also knows that q justifies p, even if at that moment one doesn't make the inference.

   I don't know if this convinces me, for how does one 'simply' observe the car and know that the neighbor is there without making an inference, even if unconsciously or very rapidly? How could one make the connection between the observation of the car and the neighbor being at home without inference taking place? Something like a non-inferential immediate connection between one bit of knowledge and the other would have to be at place here; but what this amounts to remains to me utterly mysterious. Notice that this would seem to hold even in cases where perceptual knowledge was established without mediation/non-inferential derivation, i.e. if the subject in question had seen the neighbor directly sitting in his couch (under standard conditions) it would have been a case of 'seeing that' the neighbor is there, but this would require still the appropriate additional knowledge which would allow one to infer this, i.e. the lighting conditions are standard, knowledge about what constitutes the relationship of interiority, what is a couch, and what it means to-be-sitting, etc. A different connection than inference must be at work here for S to be lead from his observation of the car to the belief that the neighbor is there. For S must already know that a parked car justifies the inference to the presence of the neighbor; he must know that q provides inferential warrant for p, even if he does not infer p from q. Furthermore he must know or believe that q ("the car is parked") is the case, as we surmised above in (3*) as Sellars' strong internalist requirement.

But then since he does not explicitly consider the causal connection between p and q at the moment of gaining this non-inferential warrant, what are we to make between this connection of two instances of (propositional) knowledge? Are we to say that a) S non-inferentially knows that q by seeing the car, and simultaneously non-inferentially knows that p? This seems remarkably awkward; and gratuitous without further explanation. But if Sellars wants to say that b) S knows non-inferentially that q first, and then knows than p, then he needs to explain how one passes from one to the other belief without any inference taking place. How does the non-reflexive, memory-stored knowledge that q justifies inferring p operate so that S may come to know that p on the basis of his seeing that q? Perhaps this has to do with the relation between sentience and sapience, where the acquisition of q upon observation could be defined as a kind of sentient response which nevertheless requires that the subject be in possession of certain propositional knowledge as well; but this is utterly speculative. I haven't found anything in the literature, including Sellars, addressing this particular concern.