I add this to supplement the normative and semantic theme simultaneously. Rationality is presented not as a better way to use language but as its beating heart.
Thinking of conceptual content in terms of inferential role, and of understanding correspondingly as practical mastery of such an inferential role—as the ability to sort into good and bad the inferences in which the concept appears in the premises or conclusions—has other advantages as well. It is a powerful corrective to the philosophically unilluminating and pedagogically damaging cartesian picture of the achievement of understanding as the turning on of some kind of inner light, which permits one then to see clearly. This is what the elementary-school kid thinks happens in math class when the girl next to him “gets it”, and he doesn’t. He is waiting for the light to go on in his head, too, so that he’ll understand fractions. In fact, he’s just got to practice making the moves, and distinguishing which ones are OK and which ones are not, until he masters the practical inferential abilities in question. It is not unusual for teachers of technical material to have students who can do all the problem-sets and proofs, can tell what does and doesn’t follow from some situation described using the concepts being taught, but still think that they don’t understand those concepts. A feeling of familiarity and confidence in knowing one’s way around in an inferential network often lags one’s actual mastery of it. The important thing is to realize that the understanding is that practical mastery, and the feeling (the cartesian light) is at best an indicator of it—often an unreliable one.
...
One might ask whether the inferentialist approach does not require overestimating the extent to which we are rational. Are we really very good at telling what is a reason for what? How often do we act for reasons—and in particular, for good reasons? The question betrays a misunderstanding. We are rational creatures in the sense that our claims and aims are always liable to assessment as to our reasons for them. How good we are at satisfying those demands doesn’t change our status as rational. And it must be kept in mind that on this way of thinking about the nature of semantic content, it makes no sense to think of us first having a bunch of sentences expressing definite propositions, which accordingly stand in inferential relations to one another, and only then having there be a question about how many of those inferences we get right. For it is our practices of treating what is expressed by some noises as reasons for what is expressed by other noises that makes those noises express conceptual contents in the first place. Once the enterprise is up and running, we can certainly make mistakes about what follows from the commitments we have undertaken, and what would justify them. But there is no possibility of us massively or globally getting the inferences wrong (for very much the same Quinean reasons that Davidson has emphasized).
I have been arguing that it is better to think in terms of understanding than knowledge, and better to think of meaning-and-understanding (which on this approach are two sides of one coin) in terms of inference than in terms of truth. So far, I have approached this issue largely from the point of view of semantics and the philosophy of language. But there is more at stake here. For this way of thinking about semantic content goes to the heart of the question of what it is to be sapient—to be the kind of creature we most fundamentally are. It says that we are beings that live, and move, and have our being in the space of reasons. We are, at base, creatures who give and ask for reasons—who are sensitive to that “force of the better reason”, persuasive rather than coercive, which so mystified and fascinated the ancient Greek philosophers. Crossing that all-important line from mere sentience to sapience is participating in practices of giving and asking for reasons: practices in which some performances have the pragmatic significance of claims or assertions, which accordingly, as both standing in need of reasons and capable of serving as reasons (that is, of playing the role both of conclusion and as premise in inference) count as expressing propositional semantic content.
This semantic rationalism—which goes with thinking of content in the first instance in terms of inference rather than reference, reason rather than truth—flies in the face of many famous movements in 20th century philosophical thought. The American pragmatists, above all, John Dewey, used the possibility of explaining knowing that in terms of knowing how not only to assimilate our sapient intellectual activity to the skillful doings of merely sapient animals, but at the same time to blur the sharp, bright line I am trying to draw between sapience and sentience. Wittgenstein famously said that language does not have a ‘downtown’: a core set of practices on which the rest depend, and around which they are arrayed, like suburbs. But inferentialism says that practices of giving and asking for reasons are the ‘downtown’ of language. For it is only by incorporating such practices that practices put in play propositional and other conceptual contents at all—and hence count as discursive practices, practices in which it is possible to say anything. The first ‘Sprachspiel’, language game, Wittgenstein introduces in the Philosophical Investigations has a builder issuing sorderss to an assistant. When he says ‘Slab!’ the assistant has been trained to respond by bringing a slab. When he says ‘Block!’ the assistant has been trained to respond by bringing a block. From the inferentialist point of view, this does not qualify as a Sprachspiel at all; it is a vocal, but not a verbal game. For the assistant is just a practical version of the parrot I considered earlier: he has been trained reliably to respond differentially to stimuli. But he grasps no concepts, and if this is the whole game, the builder expresses none. An order or command is not just any signal that is appropriately responded to in one way rather than another. It is something that determines what is an appropriate response by saying what one is to do, by describing it, specifying what concepts are to apply to a doing in order for it to count as obeying the order. Derrida’s crusade against what he calls the ‘logocentrism’ of the Western philosophical tradition has brilliantly and inventively emphasized all the other things one can do with language, besides arguing, inferring, explaining, theorizing, and asserting. Thus we get the playful essays in which the key to his reading of Hegel is that his name in French rhymes with ‘eagle’, his reading of Nietzsche that turns on what Derrida claims is the most important of his philosophical writings (a slip of paper that turned up in his belongings after his death, reading only “I have forgotten my umbrella,”), and the unforgettable meditation on the significance of the width of the margins of the page for the meaning of the text printed there. But if inferentialism is the right way to think about contentfulness, then the game of giving and asking for reasons is privileged among the games we play with words. For it is the one in virtue of which they mean anything at all—the one presupposed and built upon by all the other uses we can then put those meanings to, once they are available. Again, the master-idea of Foucault’s critique of modernity is that reason is just one more historically conditioned form of power, in principle no better (and in its pervasive institutionalization, in many ways worse) than any other form of oppression. But if giving and asking for reasons is the practice that institutes meanings in the first place, then it is does not belong in a box with violence and intimidation, which show up rather in the contrast class precisely insofar as they constrain what we do by something other than reasons. — Brandom