Preterition

What should technologies do, and which aspects of our labor do we want technologies to conserve or extend?

I recently proposed a definition and from it a tentative law: a technology substitutes capital-intensive processes for labor-intensive processes; the labor saved by technology is stored; and questions of who is exempted, who is replaced, and who pays to change that settlement are decided outside the machine. If the labor a technology saves gets stored, two questions follow: where does the stored labor go, and whose time was it?

Capital’s shape and trace

Marx argues that machinery in its capitalist form is “the historical reshaping of the traditional, inherited means of labour into a form adequate to capital” (Marx 1973, 694). The machine arrives shaped to the relation in which it works, which rules out notions of technological neutrality. And yet for Marx “this in no way means that this use value—machinery as such—is capital,” because “Machinery does not lose its use value as soon as it ceases to be capital” (1973, 699), which likewise rules out notions of hard technological determinism. That pair of eliminations lets me price the positions I sorted in Composition, Noun and Verb: I align toward Olson’s soft technological determinism, which posits that “technological changes tend to be unidirectional” (Olson 1987, 186). For Olson, economic competition provides a rationale, since “If one does not utilize the advantage, another will” (188). And David Harvey helps meet the charge of determinism, arguing that Marx’s argument is deterministic “only in the sense that, once the rules of the game under a capitalist mode of production have been set in stone, then its subsequent evolution is predetermined. It is always possible to change the rules of the game” (Harvey 2023, chap. 9).

Changing the rules has a price; the price has an incidence. Andrew Feenberg’s axis asks—yes or no—whether a technology is under human control. Control carries a cost. Intervening in a substitution means paying again, in living labor, for the labor already stored by the substitution. The bill lands on someone: Alex Reid’s “Post Hoc Public AI” describes how the public resistance comes after the investment, once the infrastructure has been bought and the price of refusing has risen. Jacques Ellul denied the refusal was available at any price, defining technique as “the totality of methods rationally arrived at and having absolute efficiency” (Ellul 2014b, 206), and counting among the aspects of its autonomy that it “does not tolerate being halted for a moral reason” (2014a, 438). Substitutions have been halted for moral reasons many times, and every time someone paid.

Marx’s footnote in the machinery chapter says technology “lays bare the process of the production of the social relations of his life, and of the mental conceptions that flow from those relations” (Marx 1993, 493n4), and so provides the second side of the 2005 coin from my last post: the substitution that hides labor makes inequality visible, to whoever holds custody of the record. Justin Joque relates the mechanism by which the decision cloaks itself from sight. When Google’s chief executive responded to charges of bias in search results, Joque notices, he did so by explaining the method rather than defending the results, so that the authority of the calculation seemed to come from a procedure that had to be followed rather than from anyone at the company (Joque 2022, chap. 8). The settlement was decided outside the machine and presented as the method’s. Charlie Moran saw the same condition and used Richard Lanham’s figure to characterize how the field looked “less at and more through technology” (Moran 2003, 345). Looking through a technology offers places for a substitution to hide.

A box of time

Twenty-plus years of hand-tooled writing resides in the weblog archive I rebuilt this summer. I’ve put the archive online, available to be taken up as fixed capital. Marx’s Fragment on Machines draws the line between the two conditions in one stroke: “The hand tool makes the worker independent—posits him as proprietor. Machinery—as fixed capital—posits him as dependent, posits him as appropriated. This effect of machinery holds only in so far as it is cast into the role of fixed capital” (Marx 1973, 702). Several pages later, Marx quotes the worker of Labour Defended, for whom under machinery “There is nothing on which the labourer can seize: this is my produce, this I will keep to myself” (709). The weblog responds to that worker with product a writer can seize. Marx quotes its measure from another pamphleteer of the 1820s, who argued wealth “is not command over surplus labour time … but rather, disposable time” (706). The archive is a structure of disposable time spent writing, consumed by the writer in the writing as an experience good.

Tiziana Terranova characterized such activity in 2000 as free labor, “Simultaneously voluntarily given and unwaged, enjoyed and exploited,” and counted “the activity of building Web sites” among its forms (Terranova 2000, 33). The name caught the affective euphoria of the early web, and Terranova’s account carries with it some of the diffuse character of that time: free labor is a desire of labor immanent to late capitalism, and late capitalism the field that sustains and exhausts it (51). A field and a desire offer no position from which to identify the traces of appropriation. Terranova argues late capitalism “does not appropriate anything” (51), and about a system she’s correct: a system itself appropriates nothing. A human does, at a use. Stephen Resnick and Richard Wolff provide the accounting that can ask of any economic site who performs the surplus labor, who appropriates it, and who gets a share when it’s distributed. A class process, by their definition, is one in which surplus labor is “pumped out of direct producers” (Resnick and Wolff 1987, 115), and “Class is an adjective, not a noun” (159): it describes a human relation at a site of appropriation.

The archive retained the reader who refused that accounting. In the summer of 2003, the same week I wrote about Olson, I was reading Resnick and Wolff’s Knowledge and Class to prepare for a fall seminar, a graduate student in English who’d never got past Econ 101, posting as I went. One grumpy night’s title was “Resnick and Wolff Unhelpful”. I’d charged them with smuggling the economic back in as a hidden essence, while holding an essentialism of my own: I wanted class to be a quality possessed by a person—a gut check—and they were telling me class has “no existence other than as the site of the converging influences exerted by all the nonclass processes” (116). Resnick and Wolff borrow the concept of overdetermination from Althusser, and I now see parallels to this post’s law: stored labor becomes capital only in a relation, at a use. As a reader in 2003, I couldn’t see the relation; today—for whatever value the report of an interior carries, if any—the writer building on those posts can. The archive traces them both.

By Resnick and Wolff’s accounting, then, under what relation was this archive built? Nothing like the one typically associated with “academic capitalism,” I think. I started out as a graduate student paying for a master’s degree on the GI Bill and working a day job, and later moved to teaching twenty-plus hours a week as a TA and studying twenty-plus more: a wage here, a state allocation there, several class positions at once, as Resnick and Wolff would expect of anyone (124). The weblog was paid for by none of them. Its form is what Resnick and Wolff call the ancient class process, the producer who appropriates his own surplus, though Marx’s singer “sells her song for her own account” (133) while I gave mine away. Gibson-Graham’s diverse economy makes room for that combination: independent in its enterprise, self-provisioning in its labor, a gift in its transactions (Gibson-Graham 2006, 71). In their iceberg diagram, it sits below the waterline, under the wages and markets and capitalist firms that typically stand in as metonym for the economy (70).

Self-appropriation brings its own bills to pay. A writer who keeps his surplus spends part of it keeping the process alive, on paper and postage once, on hosting and a domain now, the payments Resnick and Wolff call subsumed (1987, 118). Those payments go to market even when the writing doesn’t.

The archive’s condition changes with use. David Harvey summarizes Marx’s argument in the Fragment and suggests that “Whether or not things are fixed capital depends, therefore, on their use” (Harvey 2023, chap. 9). The Fragment’s sentence about the general intellect describes a use: the knowledge and skill of “the social brain” get “absorbed into capital, as opposed to labour” (694), a progress “which capital appropriates free of charge” (694–95). Matteo Pasquinelli traces the phrase to the English debates of the 1820s, where the general intellect meant literacy before anything else. The London Magazine of 1828 dated the country’s general intellect from a new mechanical system for teaching reading and writing (Pasquinelli 2023, chap. 4). The general intellect is literacy, and a model trained on writing is literacy taken up as fixed capital.

Alien technology

Training a language model on my research and weblog archive, then, would seem to blur the line between what a writer produces for himself and what gets produced for someone else. Alienation hides in that blurred line. Don Ihde describes how alterity relations are “relations to or with a technology” (Ihde 2014, 553), in which “technological otherness is a quasi-otherness, stronger than mere objectness but weaker than the otherness found within the animal kingdom or the human one” (555). He warns against inflating the relation: “To characterize computer ‘intelligence’ as human-like is to fall into a peculiarly contemporary species of anthropomorphism, however sophisticated” (554), he says, and “the tendency to fantasize its quasi-otherness into an authentic otherness is pervasive” (559), the fantasy running to a fear that decisions “will not only be informed by but also made by computers” (559). These days, it seems like humans don’t like the selves we see anthropomorphized in our technologies.

Ihde’s limit—interpretable as Lovelace’s, seen from the other side—holds that “Were the technofact to be genuinely an other, it would both be and not be a technology” (559). Then Ihde drops a sentence I’m not sure he saw as a hinge: “whatever contest emerges, its sources lie opaquely with other humans as well but also with the transformed technofact” (559). That’s a report posted from inside the relation, admitting its source to parties beyond the scope of the report. Marx names the parties: the science that drives the machine “does not exist in the worker’s consciousness, but rather acts upon him through the machine as an alien power, as the power of the machine itself” (Marx 1973, 693). Ihde describes from the side of experience what Marx’s Fragment describes from the side of the relation, and the razor from Composition, Noun and Verb sorts the two: the alien power can be read from the record—who holds the machine and whose knowledge it stores—while alterity gets reported from inside, where the report can’t always see its own sources.

Capital points to the source of the machine’s stored knowledge: “Along with the tool, the skill of the worker in handling it passes over to the machine” (Marx 1993, 545). In a footnote, Marx borrows William Thompson’s 1824 phrase for what knowledge becomes once it has passed over: “an instrument, capable of being detached from labour and opposed to it” (483n44). The detached knowledge is the quasi-other a person meets in the machine. Lovelace’s engine originates nothing; Marx’s machine, “like every other component of constant capital, creates no new value, but yields up its own value to the product it serves to beget” (509). Those are the same limit, inscribed once in logic and once in value. They explain why the values I found in the engine in Composition, Noun and Verb were ours. Kittler’s chip, an “inhuman bureaucracy” of bus, memory and an arithmetic logic unit built as “the combination of Leibniz and Boole” (Kittler 2014, 187), provides de Prony’s three sections in silicon.

A language model trained on the written commons is the quasi-other in its newest form, though the feelings it produces are older than the model. Ihde describes the case of the word processor, where “this quasi-love relationship reveals its quasi-hate underside” at the crash, in “frustration and even rage” (558). Carolyn Dowling’s writers at the same machine reported “a diminished sense of ownership of the text,” which she heard as “Did I write that?” (Dowling 1994, 230). A word processor can’t respond, and it had already loosened a writer’s grip on her own sentences. A model does respond—Nicholas Carlson characterized the model as a “two-player word processor”—and that’s why it feels like an other from the inside. Read from the record, it answers in one’s own stored words, and the contest has its sources with other humans.

The writer’s record provides the basis for building a language model. Other constructs share that basis, but differ in their builds. Training takes what writers wrote, pools it, and keeps the skill in it, as Capital said the machine kept the worker’s: the skill passes over, and the writer and the dates stay behind. A platform takes a different part—what a writer does while writing, the keystrokes and versions and times—and builds a profile. John Cheney-Lippold describes the profile, itself labor as well, stored another way. “I am utterly overdetermined, made and remade every time I make a datafied move” (Cheney-Lippold 2017, 31), he writes, and the conversations about who he is go on without his knowledge: the performance is the labor, and the profile the by-product taken up as capital. The profile keeps a pattern of a writer, a pattern with no memory of how it was made (71), the details of a life lived in time replaced, in a phrase Cheney-Lippold borrows from Alexander Galloway, by “the synchronic homogeneity of code” (10). The archive keeps the opposite. It holds the 2003 reader beside the 2026 writer, in order, with the dates, and it holds them for the writer. Like the model, the profile is storage that discards time; the archive is storage that keeps time. All three are stored labor. The use decides which becomes capital and whose time it keeps.

I argued the ownership half of this claim in 2018, at Computers and Writing, as a students’ right to their own data: the paradata a writing platform generates, the data exhaust, should download to the writer who generated it, with permissions the writer sets. “If you create it and own it in fixed form,” I said then, “it’s yours and copyrighted, and no one can appropriate it from you.” My argument leaned on the work of Jessica Reyman, who had shown that terms of use split a user’s writing, treated as copyrightable content, from a user’s actions, treated as an authorless by-product. The second classification makes the data, “at its inception, free to be appropriated” (Reyman 2013, 527). The classification decides the appropriation, as the drudgery classification did in my previous post. Reyman cautions, however, that property rights in data “could erroneously lend more support for exclusive ownership rights to uncopyrightable facts” (525), with copyright reaching expression and stopping at fact. The profile is classed as fact. Resnick and Wolff take the position that owning a thing is one process and appropriating the surplus made with it is another, so the deed to the trace doesn’t say whose use takes up that trace. Training shows as much from the other side of Reyman’s line: it takes what a writer wrote, the part a deed can cover, and keeps only what’s statistical in it. What survives those qualifications is narrower: a proposal to write the trace as expression, and so move it across the line that governs its appropriation.

Stored labor and capital

I’ve discussed Marx’s concept of capital as stored labor standing in for living labor. The act of standing-in carries the relation. Stored labor isn’t capital by nature. Machinery “does not lose its use value as soon as it ceases to be capital,” and neither does an archive. The archive you’re reading now is stored labor in each of Kline’s four senses—artifact, know-how, system of production, system of use—and it becomes capital in a relation, at the moment a use of it is taken up by someone positioned to take it. Marx and Moran share a threshold concern with how machinery relates to the ownership of human time.

Moran suggested the “perpetual hope for technology is that it will release us from what we deem to be the boring aspects of our work and enable us to do something better with the time that we will have saved” (Moran 2003, 345). Drudgery, again, is a classification. The classification climbs. Emil Roy wanted essay scoring with “no role for human discretion or error,” and Moran saw that “The computer threatened to take over not drudgery, but a major part of what we consider our special province, our means toward larger ends: reading and evaluating student writing” (346). The substitution replaces discretion with discretization. The time discretization releases was once the time of whoever did the discretized part, and that person rarely gets it returned. Marx had refused the hope in its bourgeois form already: “It is therefore a highly absurd bourgeois assertion that the worker shares with the capitalist, because the latter … makes labour easier for him” (Marx 1973, 701), and capital employs machinery “only to the extent that it enables the worker to work a larger part of his time for capital” (701).

Surplus time goes back into the machine. That’s how capital reproduces itself: the surplus labor a substitution frees gets spent on more substitution, which raises productivity and with it the rate at which the remaining time is taken, until the instrument for reducing labor-time “suffers a dialectical inversion and becomes the most unfailing means for turning the whole lifetime of the worker and his family into labour-time” (Marx 1993, 532). The word processor released the retyping and placed the typesetting on the writer’s desk. The savings traveled up and the labor traveled down, as the last post found with Dowling, and the working day lengthened at the desk as Marx saw it lengthen in the mill. The archive answers the question for one case. It’s disposable time, spent writing and owned by the writer for as long as its use is his. Stored labor becomes capital at the moment that time is taken up by someone else’s use.

Babbage’s “economy,” where the last post began, was thrift: what a manufacture saves by buying only the labor a task requires, a saving Marx would read as relative surplus value. For Gibson-Graham, “the economy” is a discursive-material construction, a system built by technologies of analysis and control, an object Timothy Mitchell finds “materialized in socio-technical networks of calculation” since the 1930s and imagined as a “self-contained and dynamic mechanism” (Gibson-Graham 2006, 54). Its naturalization has been “almost total,” according to Gibson-Graham, “a hegemonic move by which representations of economy have slipped from their locations in discourse and landed ‘on the ground,’ in the ‘real,’ not just separate from but outside of society” (53). Julie Graham’s fall 2003 seminar foregrounded the workings of such naturalizing moves, and taught me to read Resnick and Wolff’s claims as simultaneously economic and rhetorical in the modes of analysis they offer for systems of value.

References

Cheney-Lippold, John. 2017. We Are Data: Algorithms and the Making of Our Digital Selves. New York: New York University Press.

Dowling, Carolyn. 1994. “Word Processing and the Ongoing Difficulty of Writing.” Computers and Composition 11 (3): 227–35.

Ellul, Jacques. 2014a. “The ‘Autonomy’ of the Technological Phenomenon.” In Philosophy of Technology: The Technological Condition: An Anthology, 2nd ed., edited by Robert C. Scharff and Val Dusek, 430–41. Malden, MA: John Wiley & Sons.

Ellul, Jacques. 2014b. “On the Aims of a Philosophy of Technology.” In Philosophy of Technology: The Technological Condition: An Anthology, 2nd ed., edited by Robert C. Scharff and Val Dusek, 205–9. Malden, MA: John Wiley & Sons.

Gibson-Graham, J. K. 2006. A Postcapitalist Politics. Minneapolis: University of Minnesota Press.

Harvey, David. 2023. A Companion to Marx’s Grundrisse. London: Verso.

Ihde, Don. (1990) 2014. “A Phenomenology of Technics.” In Philosophy of Technology: The Technological Condition: An Anthology, 2nd ed., edited by Robert C. Scharff and Val Dusek, 539–60. Malden, MA: John Wiley & Sons.

Joque, Justin. 2022. Revolutionary Mathematics: Artificial Intelligence, Statistics and the Logic of Capitalism. New York: Verso.

Kittler, Friedrich A. 2014. The Truth of the Technological World: Essays on the Genealogy of Presence. Translated by Erik Butler. Stanford, CA: Stanford University Press.

Marx, Karl. 1973. Grundrisse: Foundations of the Critique of Political Economy. Translated by Martin Nicolaus. New York: Penguin Classics.

Marx, Karl. 1993. Capital: A Critique of Political Economy. Vol. 1. Translated by Ben Fowkes. Penguin Books.

Moran, Charles. 2003. “Computers and Composition 1983–2002: What We Have Hoped For.” Computers and Composition 20 (4): 343–58.

Olson, C. Paul. 1987. “Who Computes?” In Critical Pedagogy and Cultural Power, edited by David Livingstone, 179–204. South Hadley, MA: Bergin and Garvey.

Pasquinelli, Matteo. 2023. The Eye of the Master: A Social History of Artificial Intelligence. London: Verso.

Resnick, Stephen A., and Richard D. Wolff. 1987. Knowledge and Class: A Marxian Critique of Political Economy. Chicago: University of Chicago Press.

Reyman, Jessica. 2013. “User Data on the Social Web: Authorship, Agency, and Appropriation.” College English 75 (5): 513–33.

Terranova, Tiziana. 2000. “Free Labor: Producing Culture for the Digital Economy.” Social Text 18 (2): 33–58.