The Algorithm as Originary Stack

Algorithmic governance renders transparent the succession of speech forms hypothesized in The Origin of Language: ostensive>imperative>interrogative>declarative. The algorithm turns the sequence of speech forms into a stack. The basic algorithmic form is “if… then.” You start with a sentence representing a state of affairs, which is taken to be a merely possible state of affairs, which in turn implies a field of declaratives with varying probabilities of being the case instead of the one we start with. So, an algorithm designed for airport security, to determine when an alert should be sent to security personnel, might begin with “if an article capable of hiding explosive materials is left unattended then…” So, we start with a sentence: An article capable of hiding explosive materials is left unattended: this is the description of a state of affairs understood to be possible—it’s what we are looking for. If it’s what we’re looking for, then we need to be able to distinguish it from other states of affairs, especially those closest to it. We need, then, to specify the terms involved more precisely: which articles—bags, packages, etc.—are capable of containing explosive materials? Is it a question of size? Shape? Other features? What do we mean by “unattended”—how far away do people need to be to be leaving the item “unattended”? Do we need to see someone with the article and then leave it? For how long? Etc.

In order to choose one from among a series of declaratives representing differing states of affair, you need to have a sensing system—in the case of airport security, presumably a camera, but also, perhaps, heat sensors, x-ray machines, etc. The sensing system needs to be set—a certain degree of some measurable condition or change triggers the selection of the declarative sentence in question, i.e., notifies us that the state of affairs represented by the sentence is now likely “enough” to initiate some sequence. The sensing system is ostensive: it registers what an individual present, adequately trained and equipped, would be able to point to. As is always the case with ostensives, that hypothetical person is looking for something—he’s not aimlessly surveying a field. He’s already following an imperative—look out for… The terms of that ostensive have, in turn, been determined declaratively, which has assessed the likelihood of certain sequences of events—with a “sequence of events” reducible to a series of ostensive-imperative-ostensive articulations. The ostensive registered by the sensory mechanism triggers another imperative, to affirm that the possible declarative has moved from possibility to close enough to reality so as to trigger another imperative: send out the prepared alert. Finally, when the alert should be sent out is itself determined by some rule of selection from a field of declaratives representing possible states of affairs: set the threshold of triggering the alert too high, and various catastrophes can be imagined, with degrees of likelihood we can keep getting better at determining; set it too low, and other unwanted events will occur, with an easier to determine degree of likelihood—panics, harassment of passengers, leading in turn to other states of affairs, such as an increase in tension, loss of trust, opting out of air travel—the same states of affairs that, of course, we could arrive at through another route by setting the threshold too high.

This virtual condition is already inherent in the first act of shared deferral on the originary scene, which could no doubt be gamed out with probabilities determined for all the possible outcomes of the hypothesized mimetic crisis (Gans has always assumed that the “successful” resolution of the event would have been extremely unlikely, which would mean such mimetic crises might have happened hundreds or thousands of times before the sign was actually issued). When you stand before an ordinary decision, the algorithm is a fair enough model of what you’re doing—you’re noticing things, and one thing rather than another, some thing you are ready to notice, tilts you towards doing something, producing a field of things one might do with a vague and yet actionable representation of the likely consequences of doing one or the other… The more we analyze and come to practice our thinking and decision making in this way, the more we make our actions programmable, while at the same time generating a space of the unprogrammed, where beyond the determination of the likelihood of any outcome we find the miracle of the space of deferral that sets the sequence in motion and that we ultimately want to preserve and enhance. The machines will always relegate the miraculous to the negligible—it is incumbent upon humans to input it. Thinking of the speech forms as layers in a stack will help us to do this.

The originary grammatical stack turns the succession of speech forms into a perpetual object of thought. We are always moving from ostensive to imperative, through interrogative to the declarative and from the declarative back again into a field of possible ostensives. The move from one form to the next should always be seen as a result of the attempt to preserve the existing form, threatened by some break in linguistic presence. So, the first transition, from ostensive to imperative, can be seen as a continuing attempt to keep the object in view. You’re looking at something, sharing attention to it (even in solitude our attention is shared, as you can only see what you can imagine helping others see), and it becomes difficult to continue to see the thing you are looking at in the way you are looking at it. The very realization that you are looking at it in a particular way is already a sign of some difficulty: other things are competing for attention, other modes of attending to that same object obscure what you are seeing. Out of this fragility of the ostensive emerge imperatives: ignore that, move that out of the way, zero in more closely on this part of it, as that is what you were “really” interested in, etc. Eventually: create an institution to be peopled by yourself and others interested in sustaining a tradition of seeing this kind of thing in this way.

Next: the prolongation of the imperative into the interrogative. “Give me that” becomes “will you give it to me,” or “do you have it,” or “can you get it,” but also “is that what I think it is,” “does it only look like that,” and so on. An imperative can always be extended or prolonged—even something as direct as “give me that” opens up the possibility of taking the time to retrieve the requested object; even the more precise “give me that now” implies some distinction between a “now” and a “later” that contains some room to maneuver. If the imperative is pressed (no messing around—give it to me within the next 10 seconds or else…), then the possibility of its being prolonged becomes a way of deferring a crisis. Sometimes the “or else…” is activated, but at least sometimes it will turn out to be a bluff, in which case the “or else” must get converted into some kind of “when,” “where,” “why,” and so on. Sometimes the boundary between imperative and interrogative is thin: “what time is it” is merely a politer way of saying the semantically identical “tell me the time.” But it is the “tell me” that introduces the interrogative into the imperative: the prolongation of a failed imperative opens a space in which a new imperative can replace the original one: instead of “do” something, the imperatee is now told to “say” something. The imperative to say something about the conditions of doing something is the space of the interrogative.

The request to say something about the conditions of doing something is the space filled by the declarative. A declarative refuses some request in its undistilled form: I will not do it. But if that’s all it was, it would merely collapse into the crisis of the imperative, forcing the issuer of the imperative to put up or shut up. The declarative must (is subject to the imperative to) imagine and intimate the conditions under which the speaker might fulfill the imperative; or someone else who, under other conditions might fulfill it; or who the imperator would have to become in order to accept it going unfulfilled. And it must do this under the assumption of some estimated imminence of the initiation of the imperative crisis—how long the imperative can be prolonged without snapping, so to speak. What must ultimately be effected is a transfer of attention from the ostensive that set it all in motion to some new ostensive which will be different even if it is the same. A condition of the completeness of meaning of the declarative is that it issues the imperative: repudiate all ostensive-imperative links inconsistent with the ones constituted or “fielded” by this sentence. You haven’t “understood” a sentence until you’ve relinquished the imperative it is refusing: “understanding” a sentence is a conversion to a new reality. Once you’ve relinquished the refused imperative, a new field of ostensives opens up which you are, likewise, commanded to populate with those who will keep clearing the way to keeping those things in view.

To input the miraculous to the grammatical stack is to hypothesize wildly from the slightest sample. You can test the hypothesis and thicken the sample as you proceed, but even that will be an effect of continuing to hypothesize wildly. This is necessary to sustain the grammatical stack—to have your declaratives grained with the ostensives, imperatives and interrogatives upon which they depend. And sustaining the grammatical stack makes the stack of planetary computation one’s friend, and an instrument against liberalism (equalization through centralization) and its anarchist ontologies. The wild hypothesis is firstness, the same “wager” placed, with only the absolutely necessary degree of intention, on the originary scene. The wild hypothesis is the initial setting of an algorithm, an if… then that sets the stacking in motion.

The declarative issues another imperative—to replicate it by continuing to neutralize other imperatives until there is no imperative that can’t be derived from this declarative. This, linguistically, is the origin of liberalism—the utopianism of the declarative. Every word in every sentence can be defined, and the grammatical relations between all the word explicitly stated. This would convert one declarative into a set of declaratives. It’s one way of rendering the “meaning” of the sentence. Imperatives and ostensives resist this treatment, but not completely, and their resistance makes it all the more important to subdue them. With an imperative, you have to supply the subject, which can then be converted to a declarative: “I am commander of this regiment,” for example. You can continue: “I am informing you that you will be in accord with the description of your role as a ______ once you have had your division moved from point A to point B.” You’d have to go further to completely extirpate the imperatives—in fact, you could never do it, because at some point you would need to do more than offering a description of someone’s actions as being in accord with another description of a type of action given elsewhere.

This is why “consent” is so central to liberalism, and so self-defeating. To consent can be to promise to act in a certain way, fulfill certain obligations, meet certain demands, and so on, but that kind of consent only make sense in the context of an existing relation or project, where we know why we would be making such commitments. Foundational consent means repeating a sentence describing yourself as the kind of person who can repeat that sentence imperative-free, without anyone ordering you to. The only kind of sentence that meets these parameters is something like “I am a free, unconstrained individual who will continue to be such an individual along with other such individuals as we engage in free, unconstrained commerce with each other.” Only the pre-social individual can function as the subject of such a sentence. But, of course, once you sign on to such a sentence you start seeing constraints everywhere, which must be rationalized as “really” a result of some series of freely made agreements or a usurpation, because you haven’t abolished the ostensive-imperative world, just futilely tried to conquer it and thereby forced it to interfere surreptitiously. At this point, it’s a 2.500 years long quagmire. “All” we need to do is to show that this doesn’t “compute.”

This doesn’t mean that you should refuse or defy the command implicit in the declarative sentence to “declarativize” more of the ostensive-imperative world. Sheer defiance of an imperative is never quite right—it locks you into a vicious circle. This is especially the case for imperatives deeply built into our language. We can, rather, hypothesize the origins of our declaratives in ostensive-imperative articulations—this is ultimately what the originary hypothesis itself does, and the originary hypothesis is necessarily formulated in declarative terms. Repudiating any ostensive-imperative articulations inconsistent with a declarative (which can also mean, more broadly, a discourse, configured in a particular way) always open various possible fields of ostensivity—a way of “declarativizing” more of the ostensive-imperative world is by translating the fields into each other. In particular, bring the command to declarativize itself into the ostensive-imperative field so as to revoke its metalinguistic privileges. This is the distinction I’ve made between the metalinguistic and (with a hat tip to Bruno Latour) the “infralinguistic.” So, if we take a concept which was always leaning towards liberalism and has long since tipped over into it, that of “rights,” conversations over who has rights, which criteria for determining rights should be invoked, what’s the proper remedy for the violation of rights, etc., much less to enter into some rights exchange or competition (well, what about my rights as a…), it’s better to ask who had the right to grant rights, and how did those right granters come by their rights? We’d quickly be talking about chains of command and the transfer and delegation of authority. We’re talking about the authority to name, to create new public ostensives. After all, even if one thinks rights are natural or God-given, someone must have, at some time, told the people claiming rights that they had them. How did that happen?

The great advantage of primearchy—guidance by the first—within algorithmic governance is that it has no fear of putting forth any hypothesis, however wild, and using that to feed samples as data into the system. This is already and will increasingly be a serious problem for liberalism, which already knows which outcomes and therefore which inputs it doesn’t want. We can also have guiding questions for constructing hypotheses which are pretty resilient: how to bring power and responsibility close to each other in every single existing institution; and how to arrange for eliciting from each according to his ability and distributing to each according to his needs. A systematic study of power and responsibility, and the maximization of abilities and need meeting—we welcome any hypotheses along these lines, and any assignment aimed at testing those hypotheses. The assignment itself, like the algorithm, is a kind of stack, articulating all the speech forms—one is asked to something that must be stated in declarative terms and whose outcome must be a verifiable practice. Even more, the assignment sets the algorithm in motion, insofar as it involves doing something so as to see what happens, with successive approximations to ensure that the happening is nothing more than the revelation of the doing as a practice. A wide range of practices proposing different ways of populating out a discourse are thereby produced, giving us precise outcomes to hypothesize on.