Keeping Humpty Dumpty on the Wall: Brandom’s Inferentialist Reliabilism
“And how exactly like an egg he is!” she said aloud, standing with her hands ready to catch him, for she was every moment expecting him to fall. “It’s very provoking,” Humpty Dumpty said after a long silence, looking away from Alice as he spoke, “to be called an egg –very!” “I said you looked like an egg, Sir,” Alice gently explained. “And some eggs are very pretty, you know,” she added, hoping to turn her remark into a sort of compliment. “Some people,” said Humpty Dumpty, looking away from her as usual, “have no more sense than a baby!”
— Lewis Carroll from Through the Looking Glass, Ch. VI
In Sellars’s “Empiricism and the Philosophy of Mind,” Sellars outlines the conditions upon which a non-inferential observation report expresses knowledge that an object is green: “not only must it be a symptom or sign of the presence of a green object in standard conditions, but the perceiver must know that tokens of “This is green” are symptoms of the presence of green objects in conditions which are standard for visual perception” (1956: 75). The first condition for knowledge is a type of reliability condition, because
Brandom thinks the reportability condition is not necessary (in Sellars’s form). [I suggest below, however, that there is an implicit reportability condition in Brandom’s view.] Instead, Brandom thinks that one can augment the belief condition of traditional JTB accounts of knowledge such that a reliable tokener of
McDowell thinks that the robust Sellarsian version of the reportability condition is not necessary, but that a modified dispositional version of the reportability condition is necessary. Further, McDowell suggests that Brandom’s inferentialist reliabilism is not sufficient to guarantee knowledge because his position does not guarantee truth. [McDowell’s account might be read as elucidating a possibility Sellars rejects, i.e., that the knowing-that claim in the reportability condition be cashed out in terms of knowing-how, pace Sellars’s suggestion that K-how presupposes K-that (1956: 75).]
In this paper, I hope to achieve four separate ends. First, I will explicate Brandom’s inferentialist reliabilism as a response to Sellars’s internalist reliabilism. Second, I will criticize Brandom’s reliabilism as failing to “Humpty Dumpty” on a wall between two positions: indicator reliabilism and internalist reliabilism. Third, I will diagnosis the problem for the inferentialist reliabilist’s attempt to straddle the wall between these two positions. Fourth, and finally, I will offer an emendation to inferentialist reliabilism through concentrating on the possibility of building the wall up from the side of internalist reliabilism. A notion that is basic to internalist reliabilism— the notion of seeming-to-perceive—will be shown to be necessary to produce a plausible inferentialist reliabilism. Inferentialist reliabilism can be saved if Brandom incorporates a notion of experience along the lines of McDowell’s two-tiered episodes of knowing. But first, I engage in some name-calling.
Sellars’s position is a type of internalist reliabilism, in which for one’s perceptual state of
Brandom thinks that Sellars requirement upon perceptual knowledge “perhaps goes too far” (1997: 157). As Brandom reads Sellars’s view, under internalist reliabilism, a reporter cannot be credited with knowledge unless she can offer an inferential justification of her belief. In opposition to the level-ascent requirement, Brandom claims, “it is enough that the subject of knowledge be reliable to be entitled to a belief (without having to be able to cite that reliability as a reason for it)” (1995: 906). Elsewhere, Brandom asks, “why isn’t it enough that the attributor of knowledge know that the reporter is reliable, that the attributor of knowledge endorse the inference from the reporter’s responsive disposition noninferentially to apply the concept red to the thing’s (probably) being red? Why should the reporter herself have to be able to offer the inferential justification for her noninferential report?” (1997: 159). [It is not clear that Sellars thinks that one has to provide an “inferential justification,” but simply know that they are reliable. An occurrent requirement for an inferential justification may be too strong, as it might be possible to see the transition from reliability to knowledge that one is reliable as evincing a transition from sentient awareness to sapient awareness.] What is Brandom proposing instead?
Brandom’s reliabilism might be best read as applauding two insights of historical epistemological reliabilism. The first insight (The Founding Insight) is that knowing does not require knowing that you know— reliably-formed beliefs qualify as knowledge even though the subject cannot justify that belief herself. The second insight, (Goldman’s Insight) involves the idea that processes are reliable relative to the contexts of belief, i.e., relative to reference classes for the particular process under consideration.
As Brandom (2000: 115) points out, Goldman showed that the justification condition for knowledge cannot either reduce to the causal antecedents of one’s beliefs (e.g., in the barn facade case, being caused by an actual barn is not sufficient to produce knowledge in cases where there are relevant alternatives in the reference class.) nor one’s internal ability to justify one’s inferences based upon one’s perceptual states. For Brandom, a putative knower’s justification is “external to the subject’s beliefs and to their connection to their causal antecedents” (2000: 115). According to Brandom, one’s love of reliabilism, however, should not lead one to either (1) a denial of the need for classical accounts of justification that rely upon reasons (evincing a Conceptual Blindspot) or (2) a reliance upon natural scientific inquiry at the expense of traditional JTB epistemology (evincing Naturalistic Blindspot). [My criticisms of Brandom’s view are not based upon a belief in either of these blindspots. I agree with Brandom that reliabilism should not imply either position.] What can inferential reliabilism do for us then?
Inferential reliabilism enables us to ascribe knowledge to people that Sellars’s model of reliabilism would rule out as episodes of knowing. Brandom uses several different examples: the Aztec pot expert (2000: 99), the blindsight chicken-sexers (2000: 102-6), Monique and the hornbeams (1994: 219), and others. I will focus on the pottery expert. In that example a pottery expert can “reliably though not infallibly” [This is an important admission for Brandom. There are contextual treatments of relative alternatives that are resolutely infallibilist, e.g., David Lewis’s position in “Elusive Knowledge.”] tell the difference between Toltec and Aztec potsherds. She simply “finds herself believing that some of them are Toltec and others Aztec” (2000: 98) and has a great success rate. She also denies that she is reliable and looks for confirmation before reporting her evidence in journals. Brandom thinks that if her colleagues find that her “gut reactions” are reliably tokened, (because they endorse the material inference from particular observations), then it is reasonable to say that she knows that a pot is, for instance, Aztec. Knowing that one knows is not necessary for knowing, because in this case the pottery expert knows, but doesn’t believe that she knows. [Notice that all that Brandom requires of the transition from reliability to knowledge is that accidental true belief or epistemic luck be ruled out.]
Why does Brandom want to say that the pottery expert knows in this case? If a subject has a differential response disposition to token
Language entry transitions are perceptual beliefs that may or may not be justified in the space of reasons. For instance, the claim that the pottery expert makes, i.e., “That’s an Aztec potsherd” is only reliable (and therefore knowledge) if it is justified by similar types of material inferences made in practice with Aztec pottery around. In order for the pottery expert to possess knowledge, three conditions have to obtain. The pottery expert’s colleague must (1) attribute a commitment (a belief), (2) attribute an entitlement (a justification), and (3) undertake a commitment (a belief in her reliability) (1995: 903-4). It is not necessary for the attributee to possess a commitment or undertake an entitlement in order to possess knowledge—coming by the belief by a reliable process is sufficient.
Episodes of knowing by reliable belief-forming processes are credited with their normative status as knowings based upon the material inferences that sapient creatures make in practice. There is an institution of normative statuses that would justify attribution of knowledge to someone who is a reliable tokener of
Someone might ask (in a naturalistic spirit), “Why aren’t such reliable belief-forming mechanisms external to the space of reasons?” Brandom recoils from this suggestion since all sapient beliefs are articulated as states or episodes inside the space of reasons (and/or concepts). Brandom thinks that so long as the states or episodes are conceptually articulated by creatures within the practice of giving and asking for reasons, reliability should be sufficient to guarantee knowledge.
One way to clarify what Brandom has in mind is to show that Brandom thinks a plausible reliabilism about observational knowledge should attempt to avoid two unsatisfactory views: (1) indicator reliabilism and (2) internalist reliabilism.
Indicator reliabilism is basically the thermometer model of knowledge outlined by David Armstrong: “perception is nothing but the acquiring of knowledge of, or, on occasions, the acquiring of an inclination to believe in, particular facts about the physical world, by means of our senses” (1961: 105). [It pays to recognize also that Armstrong takes consciousness of one’s mental states to be a type of higher-order perception. We perceive that we are seeing and that is what constitutes our seeming to see X. This has given rise to self-monitoring models of consciousness.] Brandom does not want his view to reduce to this type of view. Internalist reliabilism is Sellars’s position (and McDowell’s position) that knowing via a reliable tokening of
Brandom’s position might be read as trying to sit comfortably between these two views. Brandom takes reliability from indicator reliabilism, while trying to distinguishing sapient knowings from sentient knowings* [I’ll use an asterisk to mark non-human animal states.]. Brandom elucidates what is distinctive about sapient awareness from internalist reliabilism while rejecting that second-order knowledge is necessary.
The “joint determination” (1994: 235) both causal and conceptual that Brandom discusses in making sense of language entry moves seems to make his position straddle between reliability and minimal acknowledgement of reliability. Assume we are talking about a human being seeing a red patch. Then, perceptual observations may be reliable seeings (and thus possibly non-sapient knowings for Brandom) or reliable sapient seeings, i.e., seemings-to-see. But, if the latter, then why would a conceptualized content fail to be accessible to sapient consciousness? If seeing is sufficient, then what does sapient consciousness matter? If seeming-to-see is necessary, then why is simply seeing an episode of knowing?
I will argue below that Brandom’s attempt to “Humpty Dumpty” between two sides of the wall while perched on the wall of inferentialist reliabilism leads him to fall upon either side. And. rather than attempt to put him back together as an indicator reliabilist or an internalist reliabilist, I will attempt to build a better wall for him to sit upon. The resolution comes through a distinction between seeing, expressing, and sentience, on the one hand, and seeming-to-see, reporting, and sapience on the other. It will be suggested that Brandom falls off the wall because of his lack of a robust theory of experience.
Why might Brandom’s view be construed as a type of indicator reliabilism? It is possible to show that chief problem that indicator reliabilism faces crops up in Brandom’s reliabilism. One problem for indicator reliabilism is that one cannot distinguish a thermometer or a parrot that has knowledge from a human being that has knowledge. The trouble comes in the difference between a sentient Parrot “report”
Brandom’s view provides us with the problem of construing what exactly is tokened by the reliable process in question. Is it a bird-belief* or human-belief? Brandom points out that the claim that putative knowers may be construed as thermometers or parrots is “a bit cavalier” (1995: 896). He asks, “What is the difference between a parrot who is disposed reliably to respond differentially to the presence of red things by saying “Raawk, that’s red” and the human reporter who makes the same noise under the same circumstances?” (1995: 897).
Brandom just points out that what matters is “how one distinguishes concept use from nonconceptual activity” (1995: 896). Since the pottery expert is a sapient creature, i.e., one that says “We,” the difference between her beliefs and the parrots beliefs* is that her belief condition also satisfies an implicitly contained understanding condition. She inhabits the space of reasons and so understands what she is saying. Parrots inhabit the space of nature and their reports* merely happen to them.
This, however, does not resolve Brandom’s dilemma. One cannot merely say, “It’s a special kind of belief” unless the putative sapient belief can in practice be distinguished from the sentient belief. Why should I attribute to the Potsherd expert a sapient belief? In the discussion of the case, there is nothing keeping me from attributing a brute sentient state to the pottery expert except Brandom’s insistence that she is the type of organism that is prone to inhabit states that sapient creatures inhabit. And, if the parrot exhibits the same degree of reliability (and that is sufficient for knowledge), then why could I not attribute knowledge* to the parrot? And if reliability is sufficient for the parrot, then why would I not attribute knowledge* to the pottery expert?
In fact, there seem to be occasions in which it is more reasonable for me to attribute a non-sapient state. Assume that the Potsherd expert were to reliably discriminate between Aztec and Toltec with 100% success throughout her entire career, but continued to “not believe that she is a reliable noninferential reporter of Toltec and Aztec potsherds” (2000: 98). Should we think she is rational and/or sapient in Brandom’s sense? Should we think that she is reporting on her seeing or merely expressing her seeing? The difference is subtle, but extremely important. If the answer to the second question is negative, then our answer to first should be negative. But, if she is merely expressing her perceptual state, then what reason do we have to attribute robust sapient beliefs to her?
Brandom might respond that, as a sapient being, she is in the line of work of giving and asking for reasons. She can use concepts and have beliefs, can make her way about in the space of reasons, even though she cannot give reasons for her beliefs in this particular case. Brandom wants to say that the inferential articulation of those responses, the role they play in reasoning, “makes those responsive dispositions to apply concepts” (1995: 897) reports that constitute knowledge.
How should we understand these abilities to apply concepts? One reading of such a disposition might be the following. One has the ability to apply concepts if one is in the space of reasons. But, then we need to understand in the particular case of the pottery expert whether she is reliable. It seems plausible to say that she could be reliable and have knowledge locally, meaning, there is a local possibility of a putative knower having reliably tokened a concept and not be able to give reasons. But, it seems unlikely that the pottery expert should be globally unable to provide reasons, given that that’s what locates her in the space of reasons.
Brandom argues that if it is likely that my endorsement of the pottery experts belief leads to truth, then my inference from her expression and reliability to knowledge-attribution— “a belief-endorsing policy is reliable just insofar as it is likely to lead to truths” (1995: 901). But, how would I, the attributor, make out such a policy? It could only be by having a mutual understanding of the concepts involved in the perceptual reports. But, as we saw above, there are reasons to see the pottery expert’s tokenings of
One problem with Brandom’s view is that it rules out the possibility that one at times entertains sentient perceptual states. It makes the idea of one’s sapient awareness of states basic to the states as conceptual states. This is clear in Brandom’s presentation “No Experience Necessary”: “the only sense of ‘immediate awareness’ we need in order to understand our perceptual knowledge of the world around us” (2000(online): §2) is the reliable differential responses to apply concepts. [There are theories of perception that involve an inference (conscious or non-conscious) in the transition from the perceptual state to the consciousness of that perceptual state. It is not clear, however, that Brandom considers the possibility of a non-conscious transition, because for Brandom, a non-conscious transition would be a sentient inference, and inferences are what distinguish sapience from sentience. This might keep Brandom from considering an ability that at some anthropologically prior time, had an inferential or expressive role, that came to have a reporting role, i.e., the possible inference from seeing to seeming-to-see.] To be aware of something is just to apply a concept to it, and nothing else. Brandom faces a worry that one finds oneself as a sapient creature and thus possessing full-fledged human knowledge is “a favor from the world” (1995: 878). Simply by becoming trained as a sapient creature, in Brandom’s reliabilism, a subject is guaranteed to inhabit only sapient states. This is an implausible position about sapient awareness, as it denies the idea that being conscious of a perceptual experience is thinking about it in some respect.
Why might Brandom’s view be construed as a type of internal reliabilism? Brandom wants to avoid the implications of internalist reliabilism, i.e., that knowledge implies knowledge that one has knowledge. We saw above, that we need to assume that a report indicates the existence of a sapient commitment; if one can make a full-fledged conceptual claim about one’s perceptual “experience” then one is likely conscious of that experience. Now, the difficulty for Brandom arises. Let’s return to the pottery expert and assume that she has a sapient belief
This should not be taken to suggest that knowledge should be considered to be internal to the perceptual state of the knower, but it also should not rule out the type of critical reflection that is distinctive of concept use. Brandom (2000(online): §1) places his position in line with the type of Davidsonian and Sellarsian position that McDowell lays out in Mind and World. It is a position that only involves a causal constraint rather than a rational constraint from experience. Every perception, i.e., every language entry transition “stands at the end of a whole causal chain of reliably covarying events, including a cascade of neurophysiological ones” (2000: 206n7). This places Brandom alongside Davidson and Sellars spinning in the void in which all that can justify a belief is another belief. For Brandom, what makes a belief true is its role in inferential practice. This raises a serious issue for the attributor of commitments to the pottery expert.
How does the attributor recognize a reliable belief-forming process as reliable? First, the rule that is provided for the attributor cannot fall into the two problems that Brandom faces in specifying rules in Ch. 1 of Making it Explicit. The pottery expert’s belief cannot be construed as a mere regularity, otherwise the rule may be reduced to law. If that it is possible to reduce the belief to a law, then the attributor (as was shown above) may have reason to attribute a sentient belief*. It the rule is a general explicit rule, then Brandom faces the gerrymandering problem, in which the description of the process can make any belief reliable. Also, the description that the attributor provides cannot be too general, such as “A reliable process is one that reliably produces true beliefs.”
What types of commitments should we attribute, i.e., what is the description of the process under which the pottery expert does have knowledge. The difficulty that Brandom faces here arises from his social perspectivalism. This position does not allow for a definite answer and the demerits of social perspectivalism fall on the attributor. The attributor will believe that the pottery expert is utilizing a reliable process and infer that she has knowledge. It is a problem for Brandom if the attributor does not (in principle) have a way to type processes accurately. At this point, relevance to reference classes will only complicate things, because it is a short step to the claim that the attributor should take that process to be reliable that she should take to be reliable specified by her context.
There may be a way to offer an emendation of Brandom’s view to obviate these difficulties. First, it seems correct to say that one at times achieves a normative status of a knower but for which one cannot occurrently give a reason. But, one should be able to give a reason, because one that is in the business of giving and asking for reasons is also ipso facto a semantically self-conscious being (McDowell, 1997: 162). This means that there is a transcendental requirement upon inhabiting conceptually articulated states that one has engaged in some type of semantic critical reflection, according to McDowell. That seems to be all that is required by Sellars’s reportability condition. [One could argue this by showing that knowing-that is a forming of thinking-that, i.e., making a claim about one’s experience that is directly related to the world as facts or states of affairs. The minimal thinking-that in the reportability condition could be construed as “thinking that one is reliable,” being conscious (i.e., as a type of knowing-with) of one’s reliability, where one might recognize that one is in the “in” crowd, i.e., the space of reasons.]
What is important for Sellars and McDowell is that one make a perceptual transition from a seeing to a seeming-to-see. That is what is constitutive of sapient perceptual experience. Brandom makes the mistake of getting rid of the logic of “looks”/“seems”/“appears” for the logic of “inference X is reliable”. Instead, Brandom makes the perceptual tokenings of concepts, e.g.,
Brandom is merely talking in the dark when he places strong stress on a truth-requirement implicit in the concept of sapient belief. Coming to have a perceptual belief cannot guarantee truth with a mere likelihood of being true. This is why McDowell writes the following, “so far from providing a first glimpse of the world-directedness of (empirical) conceptual content, I think Brandom’s treatment of observation reports makes empirical content unintelligible” (160-1).
I agree with McDowell that Brandom’s reliabilism places a too huge burden on the attributor to be able to spell out the reliability of tokened contents of
One thing we can do is keep the conversation with the pottery expert going. The putative knower says something. What does she say? Now, in answering this question, given the resources of Brandom’s reliabilism, we are forced to either turn her into a parrot (indicator reliabilism) or make the pottery expert hyper-aware of the content of what she says (internalist reliabilism). The minimal requirement of her commitment is that she must at least understand what she says. But, then why does the attributor in understanding what she says accrue a further access to the reasons for what she says? It seems clear that the possibility of the attributor and the attributee understanding what is said presupposes that each have access to reasons. To allow such access does not necessarily lead us to an interiorization of the space of reasons, rather it allows us to be aware that we are within such a space, and makes our reflection on our belief-acquisition, specifically empirical beliefs, indebted to that space.
Brandom’s issue with Sellars, we saw, is that the putative knower on Sellars’s view is required to be able to “inferentially justify the non-inferential claim” (1994: 217). First, it is not clear that Sellars internalism requires one to be able to inferentially justify one’s report. There is a way to describe the transition from a seeing to a seeming-to-see that is either non-inferential (direct) or inferential (indirect). If we focus on seeming as a state of one’s assessments of one’s own experience in an indirect i.e., inferential way, then it seems too stringent to expect a perceiver to be able to inferentially justify one’s report. But, if we explain seeming-to-see in terms of a once-inferential-but-now-directly-accessible relationship to one’s perceptual states, then the requirement is not too stringent. It involves attributing states at the level of dispositions to respond (what is expressed non-verbally (e.g., swerving when a car door opens while biking) or verbally (e.g., trained responses of parrots (and children)) and awareness of those states at the level of reports.
This modification signals that reports are not merely indicative of the language entry transition as a set of laws for stimulus-response mechanisms as Brandom suggests. Rather, reports are indicative of a relationship to one’s perceptual states, i.e., a relationship of being conscious of those states. And as such, they are rule-governed transitions, not law-governed transitions. But Brandom could equally respond that his transitions are not meant to be merely causal inputs and equally causal outputs. But, then he is forced to accept that reasons are something different in kind than causes in such a way that forces the pottery expert’s beliefs into the void.
We have seen that notwithstanding Brandom’s suggestions, experience is necessary to account for perceptual knowledge. Maybe we do not need a concept as robust as Sellars’s notion or McDowell’s (minimal?) notion, but at least some notion of semantic self-consciousness is necessary, i.e., how it “seems” (in Sellars’s sense) for one to see an Aztec potsherd. I think the strength of transcendental empiricism is in construing consciousness itself, a seeming-to-see within the space of reasons, just like any claims about experience. Even if dispositions to become conscious of perceptual tokenings have become well-worn in our mental economy, that does not mean we can spend them without reason.
Brandom would probably not rush to accept this emendation, since the emendation implies that a theory of experience is necessary to make his reliabilism successful. As he notes in his note in Articulating Reasons, “‘Experience’ is not one of my words” (2000: 205n7) and that in using the phrase “perceptual experience” he is speaking with the vulgar. But, if his only other options are speaking with parrots or speaking to himself while spinning in the void, then speaking with the vulgar is not so bad. It may be especially charming considering he would be in the company of Kant, Sellars and McDowell. For inferential reliabilism to have objective import, i.e., to engender claims about the world, it should be explicated in a logic that implies that there is a difference between seeing and seeming-to-see, a logic that engages with the semantics of appearance talk. Any logic that incorporates that distinction guarantees that a theory of experience is necessary.
BIBLIOGRAPHY
Armstrong, D. (1961). Perception and the Physical World. London: Routledge.
Brandom, R. (1994). Making it Explicit Cambridge, MA: Harvard University Press.
Brandom, R. (1995). “Knowledge and the Social Articulation of the Space of Reasons” Philosophy and Phenomenological Research. 55.4: 895-908.
Brandom, R. (1997). “Study Guide” In Empiricism and the Philosophy of Mind. Cambridge, MA: Harvard University Press. pp. 119-181.
Brandom, R. (2000). Articulating Reasons: An Introduction to Inferentialism. Cambridge, MA: Harvard University Press.
Brandom, R. (2000(online)). “No Experience Necessary: Empiricism, Non-inferential Knowledge, and Secondary Qualities” [online—www.nyu.edu/gsas/dept/philo/courses/representation/papers/BrandomNEN.pdf]
deVries, W. and Timm Triplett (2000). Knowledge, Mind and the Given: Reading Wilfrid Sellars’s “Empiricism and the Philosophy of Mind” Indianapolis: Hackett Publishing Co.
Goldman, A. (1976). “Discrimination and Perceptual Knowledge” In The Theory of Knowledge Ed. Louis Pojman. Belmont, CA: Wadsworth
McDowell, J. (1999). “Sellars’s Transcendental Empiricism” in Realism, Rationality and Revision Walter de Gruyter, Berlin and New York, 1999, pp. 42-51.
McDowell, J. (1994). Mind and World. Cambridge, MA: Harvard University Press.
McDowell, J. (1995). “Knowledge and the Internal” Philosophy and Phenomenological Research. 55.4: 877-893.
McDowell, J. (1997). “Brandom on Representation and Inference” Philosophy and Phenomenological Research. 57.1: 157-162.
Rosenthal, D. (2005) Consciousness and Mind. Oxford: Oxford University Press
Sellars, W. (1956). “Empiricism and Philosophy of Mind” In Empiricism and the Philosophy of Mind. Cambridge, MA: Harvard University Press.
Sellars, W. (1963). Science, Perception and Reality. New York: Routledge & Kegan Paul
0 comments:
Post a Comment