Making It Explicit, chapter 2: Toward an inferential semantics.

We need to make sense of representation: what it means to purport a representation or take up a representation, and how purported representations might succeed or fail

Big B returns to Brentano, distinguishing two senses of “represent,” one intensional and one extensional. We can levy intensional propositions about the King of France that are meaningful, but that cannot succeed as extensional representations because there is no King of France.

B cautions is against two kinds of collapses. First, we shouldn’t collapse all representation into its extensional mode. Like, we shouldn’t require a purported representation to succeed in order to call it a representation. If we do this, we can’t talk intelligibily about errors.

But we also shouldn’t collapse into the intensional mode, effectively allowing all purported representations to succeed.

Big B is nothing like Bob Ross—no “we don’t make mistakes, only happy little accidents” garbage. Who are we if we can’t talk meaningfully about our mistakes?

Here Big B absolutely throwing shade at the explosion of representeds in, e.g., object-oriented ontology

Of course when we are talking about representational uptake, we are talking about use. This can also go wrong. B warns against a “mind as mirror of nature” account:

Citing Rebecca West: it’s hard to see why one would need a copy of the universe, because “One of the damn things is enough.”

It is not enough to find an isomorphism between a representation and its content; we must show how the representation is more intelligible and does more work than what it represents.

It is also important that whatever understanding of representation we end up with is thick enough to accommodate propositional content. We need to say: what it is to express a proposition, and further, to express a proposition that represents something objective

This approach seems to me a rejection of Wittgenstein's theoretical quietism, as well as a rejection of approaches that consider "propositional content" to be essentially illusory (maybe like enactivism). For Big B propositions are both grounded in social practices _and_ real

In understanding propositional content, it's important that we avoid thinking of "representation" as a semantic primitive, or "unexplained explainer." Semantics don't exceed the grasp of pragmatics.

Where do we look historically for a pragmatic grounding of semantics? First, Kant, who on Big B's reading takes that our ability to discern differences or make judgments depends on the space of possible judgments, whose contents are thoroughly propositional

Also Frege and Wittgenstein. W, fine. But B takes Frege's focus on "truth" to have a pragmatic character. But I wonder if this reading is overdetermined by Brandom's own pragmatic assumptions? I'm no Frege expert but this feels like a stretch

Kinda feels like "Let's do some pragmatic readings of the philosophy that inspired me." Which, honestly, very cool, I'm into it, but not convinced that the work-being-read actually has a thoroughgoing pragmatic character.

Big B emphasizes the sentence level, or the smallest "move" in a language game. Takes some non-sentential constructions (like naming expressions) to be essentially parasitic on the possibility of sentences that include them

This is kind of interesting because I think other theories of communication would take a radically different approach, emphasizing the quasi-linguistic role of actions like pointing gestures. Another sign of Big B's commitment to propositional content

One family of such less-promising alternative theories Big B dubs "the designational model," which takes representation to be solely a matter of naming things—signifier/signified relationships "all the way down."

In a paper on Huw Price, Big B goes into detail, identifying the D model with structuralism/post-structuralism and semiotics; most thinking downstream of Saussure. Derrida, who takes signifiers to simply designate more signifiers, is the most-off track.

https://www.nordprag.org/papers/Price-Brandom.pdf

The two sins of the designational model: (1) collapsing sentences and "complex names;" (2) collapsing judging and predicating.

Committing these sins makes us mere classifiers: "this" is "that" is "this" is "that;" hard to see how such a consciousness could get off the ground.

Next, B looks at the philosophical history of classification and inference, particularly the dispute between empiricists and rationalists. One way to look at this conflict: is classification prior to perception?

The empiricist take is: No. We acquire some sense data first; this is perception or apprehension. Then we classify it; this is conceptual comprehension.

But for Kant, classification precedes and is a requirement of perception; everything is perceived _as being something._

Big B insists this Kantian approach must be "pragmatized" or de-representationalized or naturalized, and Hegel offers a path. Desire is key here; "an animal classifies some particular as food when it 'falls to without further ado and eats it up.'"

Reinforcement learning again.

But it is not enough to simply have a disposition to behave a certain way under certain circumstances. We get a nice counterexample: the case of iron rusting. This disposition classifies moisture in a way arguably similar to eating classifying food. So what's the difference?

"What are the salient differences between a measuring instrument, such as a thermometer or spectrophotometer, and an observer who noninferentially acquires beliefs or makes claims [...]?"

Following Sellars, B distinguishes "merely responsive" from conceptual classification on the basis of whether it can play an active role in inferential reasoning.

E.g., a parrot trained to say "red" when it sees something red cannot infer that the object "is colored"

An interesting consequence of this view is that concepts are never atomic; they are always entangled in a holistic web of other concepts. We cannot understand "red" without access to the network of inferential relations it participates in, involving "color," "scarlet," "blood"

We also get a picture of how propositional contents arise from inferential practice. Concepts exist in virtue of their normative, inferential uses. Inferences make use not of explicit representational properties of concepts, but over their (implicit?) propositional contents

But what could constrain such a process? Does anything prevent Big B's take from collapsing into an "anything goes" a la Feyerabend?

We get a Hegelian hint in Big B's discussion of determinate negation, or "material inferential and incompatibility relations" among concepts.

The invocation of "material" suggests to me that determinate negation is where the mind-independent structure of inferential relations constrains the uses of concepts and thus the form of language games. Looking forward to some more detail on this point...

It is this move that allows Brandom to embrace exactly one (1) Romantic impulse: the rejection of representationalism. But one oughtn't go "full Romantic" and also reject the force of reason.

A bit of a digression here. It strikes me there are some interesting convergences and divergences of Brandom's account with the idea of "explainable AI."

There's a similar starting point, which is wanting a decision-making system to be able to give reasons for its decisions.

But when we ask people for their reasons, we're actually asking for something very different from what we want from an "explainable AI."

In the case of a person, it's how their decision fits into a conceptual/inferential network of the kind Brandom identifies

But in the case of AI, it's more like "What's the material basis of your decision? What parts of the stimulus moved what parts inside of you to produce your response?"

It would be absurd to ask a person such a question.

And producing AI that could be answerable to a broader conceptual and inferential network is an entirely different and much more ambitious project than merely making visible the movements in its guts

Next we get a more robust pragmatic reading of Frege. This section was important, and although a lot of the Frege-details were above my pay grade I found it pretty convincing

B reads Frege, especially early Frege (before he sold out), as thinking of conceptual content in inferential terms. A classic example:

Although "the Greeks defeated the Persians at Platea" and "the Persians were defeated by the Greeks at Platea" are a little different in meaning, the sense in which they are deeply the same is their shared conceptual content

Big B unpacks this in terms of a test of substitution (which is attributed to Frege): Can you swap out one sentence for another and still draw the same inferences? Then they share conceptual content by virtue of playing the same inferential role

To the extent that later Frege abandoned this kind of "inference first" approach, his work was "retrograde"

Inferences that depend on such conceptual content have a "material" character. Some interesting examples:

"Pittsburgh is to the West of Philly" → "Philly is to the East of Pittsburgh"

"Today is Wednesday" → "Tomorrow will be Thursday"

"Lightning now" → "Thunder soon"

With these, logical form alone won't get the job done. These inferences require practical mastery of a network of concepts that play by their own rules (West/East, today/tomorrow, lightning/thunder, time)

It's easy to mistakenly slip into logical formalism, and Big B takes Dennett (in Intentional Systems) to task for inappropriately attributing formal, logical understanding to animals where B sees instead a practical culpability to material inference

Of course, Sellars warned us about all of this by laying out, in listicle form!, the continuum from materialpilled (good) to formalismpilled (bad):

Of course the point is not to deny that logical propositions do work, but to cash them out in pragmatic terms that respect material force, rather than trying to cash _everything else_ out purely in terms of formal, logical validity (hopeless)

Interestingly, Big B's Sellars takes the Socratic method to be a way of making explicit commitments that were previously only implicit in practice, thereby opening them to objections that only become visible when they are caught in a wider inferential net

Back to Frege, on the roles of specific operators in making practices explicit.

Before we have the conditional operator, we can treat judgments as having content only by accepting them or not in practice. After the conditional operator, we can _say_ what inferences are allowed

In the case of disagreement, we can resolve objections by explicitly stating inferential commitments and giving and asking for reasons—but only because operators like the conditional can make implicit practices of judgment explicit in language

For Frege, the dream is to develop a "concept script" that could support the formalization of scientific understanding at the level of rigor that was previously only available to mathematicians

(I want to say "representational format" but I'm hesitant to use that language!)

B highlights an interesting problem with irrelevant inferences. You can "or"-operator propositions that have no material relation—e.g., "Hegel and Hölderlin were roommates" or "43 is prime"—and use that conjunction as a basis for some inference. Formally consistent but stupid

This is a problem because of a tendency to confuse necessary and sufficient conditions. The formal inference might be necessary even though in practice, it's not sufficient. The two-valued conditional is in this sense "an alarmingly bad choice" for explicating proper inferences

B claims this was not for lack of ambition, but because this work of Frege's had a sensibly limited scope. Brandom will pick up the slack

[Tweets are gonna be somewhat more compressed for a while because of work deadlines...]

On irrelevant inferences, I really like Brandom's simple, pragmatic approach. I feel like in my interdisciplinary day-to-day, I mostly run into people who think such problems either (a) undermine the very idea of objective truth, or (b) are trivial and unworthy of consideration

I've often felt these extreme positions are both totally unsatisfying and quite costly, and I've been shocked to see whole subcultural ecosystems organized around the energetic demands of actively denying the costs.

Next section doing double-duty, structural and moral.

Structurally, Big B (via Dummett) distinguishes between _circumstances_ of conceptual application and _consequences_ of application. Insists that our theories of meaning not be one-sided; we must take both into account.

Then some interesting moral meat. Big B (also via Dummett) notes that there may be "a lack of 'harmony' between the circumstances and the consequences of application of a concept"—we may find ourselves caught in a conceptual trap where we must change our concepts to escape

As an example, B considers the pejorative "Boche," used to refer to Germans during the first world war. The "Boche" concept understands German-ness as being essentially cruel, and thereby commits the speaker to treating them as cruel

https://en.wikipedia.org/wiki/List_of_terms_used_for_Germans#Boche_(pejorative)

What do we do, when presented with such a trap? Big B recommends refusal, suggesting a fitting role model:

"When the prosecutor at Oscar Wilde's trial asked him to say under oath whether a particular passage in one of his works did or did not constitute blasphemy, Wilde replied, 'Blasphemy is not one of my words.'"

Big B also strongly recommends we make explicit the courage of our convictions:

This reminds me of @lastpositivist's description of one of Carnap's moral positions: That we see the tools of explicit reason as a powerful defense against a fascism that relies on obfuscation, double-entendre, and mysticism to make its case

Big B, again channeling Dummett, presciently discusses how the way language involves material commitments to action that are not derivable from their purely formal properties makes language about personal identity a key battleground

I am compelled by B's diagnosis, but I wonder about the limits of refusal as a strategy. I sense hints that B isn't 100% committed (in one passage he allows that concepts may "evolve"), but it seems like a Haslanger-style politics of conceptual engineering might fit well here