New APPS

Leif Weatherby does not care for Derrida.  At least, in Language Machines (see here for a synopsis/initial take on this important book) he suggests that Derrida’s (mis)reading of Saussure is a significant part of “how the humanities lost language, allowing both cognitive science and NLP to update analytical and technological approaches that literary theory rarely engaged” (73).  In particular, Derrida’s move to the critique of metaphysics and his tendency to lump pretty much everything together under that umbrella risks abstraction – it’s a proposal that “itself floats above the fray” (73).  This gets to the same place Chomsky did, albeit by a different route:

“By sweeping structuralism’s focus on a concrete object to one side in the name of opposition to metaphysics, poststructuralism fumbled the object itself. Where Chomsky avoids external language by excluding it from science, Derrida finds the law not in cognition but rather at a level of abstraction about culture that ends up having the same effect: a lack of a link between the ‘conditions of im/possibiity’ and the expressions so conditioned” (73).

The Derridean critique, in other words, is so abstract that “it is simply not clear that we need Derrida’s revision of structuralism to proceed with a concrete analysis of computational language” (73).  Worse, post-structuralism in its Derridean version doesn’t have much to say about how language “interfaces with other sign-systems … primarily because it has never taken other sign-systems particularly seriously, perhaps especially mathematics” (73).

There’s a lot going on here, and I’m certainly not in a position to defend Derrida’s level of abstraction.  After all, I lean Foucauldian.  In what follows, I want so say something about the abstraction problem, and then something about why I think Derrida nevertheless has something to offer.

Weatherby targets Derrida’s reading of Saussure and argues that Derrida “wrongly accuses [Saussure] of naivety about ‘exteriorization,’ with the effect of minimizing the type of concrete analysis of all levels of language that structuralism practiced. What then came to pass for close attention to language tended to ignore structure, regularity, and distribution, cutting off any insight that might come from knowledge of linguistics” (70).  Thus, “the result has been a balance of attention massively lopsided toward the general economy of putative metaphysics and not toward the restricted economics of semiological systems” (70-1).

I’m not qualified to assess a reading of Saussure, but the Derridean argument should be familiar, because he in fact makes it a lot.  Weatherby is right to notice how it gets in the way of mathematization or thinking about mathematics as a form of writing.  Consider “Pit and the Pyramid,” which is about Hegel’s account of language.  At several places, Derrida dutifully recounts Hegel’s dismissal of Leibniz.  In my foray into “Pit and the Pyramid,” I cited this passage:

“The heart of the thesis is quickly stated: the privilege or excellence of the linguistic system – that is, the phonic system – as concerns any other semiotic system. Therefore, the privilege of speech over writing and of phonetic writing over every other system of inscription, particularly over hieroglyphic or ideographic writing, but equally over mathematical writing, over all formal symbols, algebras, pasigraphies and other projects of the Leibnizian sort – phonetic writing’s privilege over everything which has no need, as Leibniz said, ‘to refer to the voice’ or to the word (vox)” (Margins of Philosophy, 88; references to this edition)

A few pages later, we read that:

“Projects for a universal writing of a nonphonetic type seem to be marked by the abusive pretensions and insufficiencies of all the formalisms denounced by Hegel. The indictment is directed precisely against the risks of dislocating the word and the name. The principal defendant is obviously Leibniz—his intelligence and his naïveté, his speculative naïveté which impelled him to place his confidence in intelligence, that is, in a formalizing understanding bearing death” (Margins, 96-7).

That passage is then the transition to recounting Hegel’s dismissal of hieroglyphics and Chinese writing (“which fascinated Leibniz and led him astray, as Hegel takes every opportunity to recall,” 102).  Perhaps it’s unfair to tag Derrida here with ignoring Leibniz, since the essay is a critique of Hegel, but where is any sustained discussion of Leibniz’s fascination with other forms of writing? 

Leibniz gets critical attention in “Force and Signification” (Writing and Difference, 9-10), but it’s only to be swept up into the theodicy.  Grammatology opens by noting “the history of ( the only) metaphysics, which has, in spite of all differences, not only from Plato to Hegel (even including Leibniz) but also, beyond these apparent limits, from the pre-Socratics to Heidegger, always assigned the origin of truth in general to the logos: the history of truth, of the truth of truth, has always been-except for a metaphysical diversion that we shall have to explain-the debasement of writing, and its repression outside “full” speech” (3).

Although the status of Leibniz’s inclusion in this list should be interrogated – why is it parenthetical? – Derrida nonetheless concludes “the Leibnizian project of a universal characteristic that is not essentially phonetic does not interrupt logocentrism in any way” (Grammatology, 79) because it functions to model a writing system that has been fully abstracted from history. 100% Leibniz was a metaphysician, but the way he does or does not exceed Hegel’s effort to capture him seems important, at least for a history that is helpfully non-reductive.

The preceding should not be confused with a reading of the Grammatology, of course, or even a reading of Leibniz within the grammatology.  It should however suffice to show why the Derridean project could be frustrating to those who want to study other kinds of sign system, even if they ultimately don’t suffice to the task of language models.

So why might Derrida be useful? Weatherby sets up his complaint in the context of Bataille, as the reference to general economy suggests.  Weatherby writes:

“I find it helpful to put this point in terms of what Derrida, with reference to the theorist George Bataille, calls the general economy. A restricted economy is a general term for a system of symbolic exchange like the kindship system of which Claude Lévi-Strauss had famously written, any local set of distinctions – things exchangeable for like and unlike – is nonetheless included in a larger general economy, which, for Derrida, is constituted by this foundational opacity that establishes the logic of the supplement.  The chain of signifiers carries with it this more general logic but is only tractable for analysis in the restricted economy of some particular practice or system (English, perhaps).” (70)

Here it seems to me is the place to start a conversation about Derrida.  As I read him, in the texts I’ve focused on, Derrida is concerned with the way that a general economy (let’s say, signs in general) become a restricted economy (a given system).  The metaphysical trick lies in convincing oneself that the restricted economy can somehow ground the general one.  This is basically what Hegel is doing with totality, and it’s what the move to orality does when orality is made to ground literacy.  At least, I think that’s a plausible reading for what Derrida is doing.

In Bataille, a general economy is marked by excess, and the need to trim that excess. Bataille writes:

“The living organism, in a situation determined by the play of energy on the surface of the globe, ordinarily receives more energy than is necessary for maintaining life; the excess energy (wealth) can be used for the growth of a system (e.g., an organism); if the system can no longer grow, or if the excess cannot be completely absorbed in its growth, it must necessarily be lost without profit; it must be spent, willingly or not, gloriously or catastrophically” (Accursed Share, 21).

The silent trimming of excess – or maybe, in slightly more Derridean terms – the simultaneous trimming of excess accompanied by the disavowal of such trimming – strikes me as a decent way to characterize what Derrida wants to talk about.  That which is excess to voice – writing – is trimmed away while voice is aggrandized with words like “origin.”

And this is why I think Derrida can be useful: a language model isn’t language.  It’s a language model: something that necessarily involves a trimming or a limitation of the phenomenon of language itself.  I’ve spent a lot of posts on topics like what I called the “hidden normativity” of language models, and explored them in terms of Derrida’s critique of Searle, and in the Platonic critique of writing.  All of that is in the service of trying to think about how the restricted economy of a language model is formed out of the general economy of language itself.  I won’t claim that Derrida is the only way to get there, but I do think he’s useful.