Preterition

The economy of a manufacture

For Charles Babbage, the word “economy” in the title On the Economy of Machinery and Manufactures (1832) names the technological dividend a manufacture saves as thrift. As an economist of labor, Babbage designed the material substrate of the engine Lovelace programmed, seeking the cost of a task sliced sufficiently fine for a machine. His notion of thrift served the master’s question: how does one buy only the labor a task requires, and not a bit more?

In 2005, I proposed—following C. Paul Olson—that digital technologies raise the efficiency of distribution “by substituting capital-intensive processes for labor-intensive processes,” and that they make economic inequality visible “via precisely the same means.” The first issue of Computers and Composition opened on whether computers could “solve some of the perennial problems associated with composition programs—staffing, financial support, faculty work-loads, class size, training for graduate students and paraprofessionals” (Kiefer and Selfe 1983, 1, quoted in Moran 2003, 351): capital operating as a substitute for labor. In Marx’s terms, “capital” means stored labor standing in for living labor. My version, from Marx via Olson: technologies substitute capital-intensive processes for labor-intensive processes.

Babbage proposed that “[p]erhaps the most important principle on which the economy of a manufacture depends, is the division of labour” (1832, sec. 217). Smith had the pin factory; Babbage has the price list. The master, “by dividing the work to be executed into different processes, each requiring different degrees of skill or of force, can purchase exactly that precise quantity of both which is necessary for each process” (Babbage sec. 226). Off the shop floor, he adds, the division of labour “can be applied with equal success to mental as to mechanical operations,” and it “ensures in both the same economy of time” (sec. 241).

Babbage’s proof came from France. The Revolution had adopted the decimal system and wanted tables to put it to use, and Gaspard de Prony, given the work, set out to join perfect workmanship with “l’économie de la dépense et du temps” (sec. 242). The tables outran any lifetime he could expect, even with three or four able collaborators, until he stopped at a bookseller’s, opened a fine edition of Smith to the chapter on the pins, and resolved “mettre ses logarithmes en manufacture comme les épingles” (sec. 243). de Prony’s manufactory had three sections: several mathematicians who chose the formulae, a few more educated hands who turned them into the numbers to be computed, and “from sixty to eighty” who did the computing with “nothing more than simple addition and subtraction,” their work “usually found more correct” than that of the educated hands (sec. 244). Unemployed hairdressers, Matteo Pasquinelli finds in the period sources (Pasquinelli 2023, chap. 2).

From the third section’s labour “the first class were entirely exempt,” because “Such labour can always be purchased at an easy rate.” The engine was for the day “when the completion of a calculating engine shall have produced a substitute for the whole of the third section of computers” (Babbage sec. 245), and the plan “requires a large capital to be employed” (sec. 251). Henry Colebrooke, presenting Babbage a medal in 1823, said the invention “substitutes mechanical performance for an intellectual process” (Pasquinelli 2023, chap. 2), the computer it replaced being then a person, often a woman. The engine was a substitution.

Definitional provenance

Here, then, is the progression I see. Babbage plans the substitution. Marx names the substitution as a tendency and traces its path. Olson suggests the computer permits the substitution. I argue that the substitution is technology’s defining feature.

Marx relied on Babbage for much of what he knew of machinery. The Grundrisse’s Fragment on Machines turns the master’s thrift into capital’s law:

“The increase of the productive force of labour and the greatest possible negation of necessary labour is the necessary tendency of capital … The transformation of the means of labour into machinery is the realization of this tendency” (Marx 1973, 693).

Marx’s path begins in part from Babbage’s “division of labour, which gradually transforms the workers’ operations into more and more mechanical ones, so that at a certain point a mechanism can step into their places” (704). In Marx’s terms, capitalists substitute machinery (which is labor already objectified) for living labor, reducing necessary labor time, displacing workers and raising relative surplus value. Manufacture (manus, hand + factum, passive participle of to make), cutting a craft into detail and apportioning each operation to a hand, is the account of how humans systematize labor: the verb cut into parts before any machine arrives. Capital makes the cut a measure. The machinery chapter opens with Mill wondering whether “all the mechanical inventions yet made have lightened the day’s toil of any human being,” and answers that lightening toil was never the point: “The machine is a means for producing surplus-value” (Marx 1993, 492). Surplus requires a count, the other component of digitization implied by the word’s etymological descent from fingers (digiti) and their purposes, and “The productivity of the machine is therefore measured by the human labour-power it replaces” (513). Comparison without description, applied to labor.

As a graduate student in Massachusetts, I read C. Paul Olson’s “Who Computes?” in the summer of 2003 and wrote about it here. His sentence has inflected much of what I’ve written about technology since: “The extent to which the computer allows substitution of capital intensive processes for what were formerly labor-intensive ones (therefore revolutionizing work relations and processes) has meant that the impact of computers has penetrated every level of the social organization of work processes” (Olson 1987, 184). For Olson, the circumstance offers gains in productivity by “substituting capital for labor; or substituting cheap labor for expensive labor” (189), with the work of “secretaries and file clerks” passing to “word processing, automated filing, answering services, etc.” (189). Olson uses the verb “allows,” and the computer permits the substitution, to “the extent” it permits it. I made the step from “the computer allows” to “a technology is” in 2005 without noticing. Pasquinelli came close in 2023, writing somewhat more vaguely that “any technology influences the metrics of abstract labour” (Pasquinelli 2023, chap. 4).

If the substitution of capital for labor is definitional for technology, three consequences follow. A substitution has an incidence, and the incidence is never in the machine. “Labor-saving” hides a store: the labor isn’t saved but stored, and moved to whoever holds the store. Every writing technology becomes legible in the same terms: the pen, the press, the word processor; each artifact of capital standing in for labor, with language models the case where the labor substituted is that of composing itself. So at the risk of presenting here as excessively ambitious, I’ll propose 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.

The initial process of discretization operates at both technical and social levels. Charles Moran, taking stock of twenty years of Computers and Composition (Moran 2003, 345), found where the field’s hope began: “When someone categorizes an activity as drudgery, that act of classification confers low status on the activity” (345–346). Moran suggested that “[w]e understand the drudgery of writing to be copy-editing, revising, and retyping” (346). Pasquinelli finds the same move, observing mental labour “could therefore be automated because it was a task of the working class, and not one to be regarded as ‘thinking’ proper” (Pasquinelli 2023, chap. 2). Classification comes before substitution: the verb cut into named operations, ranked, the lowest offered to the machine, in public and in advance. Marx’s translator split his Analyse into science’s “analysis” and labour’s “dissection,” and rhetorical analysis as I teach it, breaking an exchange into its parts and their relations, is the second sense performed on a text. The analysis distinguishes the student from Babbage’s master: the student dissects to understand, while the master dissects to buy each part at price.

What the substitution saves, and for whom

Marx’s insight about machinery runs against its common representation as a remedy for scarce hands. Machinery “inserts itself to replace labour only where there is an overflow of labour powers”; it comes in “not in order to replace labour power where this is lacking, but rather in order to reduce massively available labour power to its necessary measure” (Marx 1973, 702). The representation matters for language models, which arrived with the largest overflow of writing there has ever been—the public web and its archives—and whose promise, read through Marx, amounts to reducing that overflow to its necessary measure. Whose saving, then?

Babbage’s plan provides the answer. The first section was exempt from the third section’s labor, and the engine operated to spare the accomplished mathematician, not the hairdressers. The saving travels upward, toward whoever employs the capital. Justin Joque suggests the general form, noting that for any objectification, “the question will always be who is burdened, who is unburdened,” and that such unburdening is “the essence of all computing—and all recording and communication technology” (Joque 2022, chap. 3). Marx adds a caution: where a machine costs as much labor to make as it saves in use, nothing has been saved; labor has only been displaced from one site to another (Marx 1993, 513), and the displacement has an incidence of its own.

The particulars make an incidence legible. In 2007 Alice Horning worked out the arithmetic for her own program. Its union contract paid a starting first-year writing instructor about $3,800 a section, and with twenty-two students in a section, the pay came to “about $172 per student.” Set against Richard Haswell’s estimate of 230 hours to teach a course, it “works out to $16.50 an hour,” and she adds that her institution pays better than many (Horning 2007, 18). Those numbers answer the field’s founding question. Computers did solve the staffing problem, in the sense Olson’s euphemism names: cheap labor for expensive. Christine Nowik recently asked the question about another imagined labor-saving technology: “Who bears the cost when we decide everything needs to change?”

The measure of labor can change along with the machine. Pasquinelli writes that the “abstract eye of capital that regulates the labour theory of value employs a specific instrument to measure labour: the clock,” but notes that “clocks are not universal. Machines can impose a metrics of labour other than time” (Pasquinelli 2023, chap. 4). My 2015 talk put the clocks first with Eli Terry’s mass-produced movements, the difference engine built on the mechanical clock’s principles, Catharine Beecher’s “right apportionment of time” for the household. Writing has its own clocks in the credit hour, Haswell’s hours, and the word count I used to mark a milestone. Language models have provided a new clock in the token, priced by the thousand: the clock’s successor for written labor, and a count of the noun that never reaches the verb.

So what did the word processor save us? In 2005 I pointed to Carolyn Dowling’s “Word Processing and the Ongoing Difficulty of Writing” for what many, following Moran, took as the last word on whether computers improve writing. Moran had written that the field’s hope for improvement “should have come to its natural end” with Dowling’s article (Moran 2003, 348). Dowling had been a teacher, an editor and a professional writer before she lectured in computing, and over four years she talked with seven colleagues and students who couldn’t understand why writing at a computer hadn’t gotten any easier (Dowling 1994, 227–28). She found that the relief made its own trouble, in that “the very alleviation of certain problems connected with writing that is not computer-based could be seen as creating new difficulties” (228). The screen was harder to scan than a page, the mouse pulled attention from the sentence, and the typography that had been someone else’s job was “now in the writer’s hands rather than the professional typesetter’s” (230).

Much of the work Olson saw passing from secretaries and file clerks to word processing came to rest on the people for whom they had performed the typing, and on students, who had never had secretaries to type for them. In the terms of J. K. Gibson-Graham’s table of labor and its compensation, work that had drawn a wage became work that draws no wage of its own (Gibson-Graham 2006, 63). The savings traveled up, as Babbage’s plan would predict, and the labor traveled down to the writer, added to the task list on top of the labor of writing itself.

Dowling noticed two more challenges. Her writers had trouble letting a text go, and she wondered whether they were responding to “the assembly-line imagery suggested by the term processing,” the promise that working over raw material “will necessarily produce a value-added product” (Dowling 1994, 232). The commodity form arrived in the software’s name. And she saw that difficulty carries status: the difficulty of writing had “contributed as much to the apparent intellectual rigor of certain disciplines as did the content itself,” so that calling writing easy threatened long-held attitudes and professional standing alike (234). Moran’s drudgery works the bottom of the same scale. The classification that decides what a machine may take ranks writing’s labor from both ends.

The substitution reached what could be discretized—copying, correction, reformatting, circulation, the noun’s side of writing—and the verb’s difficulty remained with the writer, and added new difficulty. Dowling declined to treat that difficulty as a defect to engineer away. Following Barrett and Paradis, she called it “a generative, perhaps even necessary, contributor” to writing that succeeds (227–28), though she declined to judge word processing bad on balance (234). Moran’s reading, that removing the difficulty might remove the quality (Moran 2003, 348), goes past what Dowling claimed. Two years later Mariolina Salvatori gave the same word a classroom method. Her “difficulty paper” asks students to describe what a text made hard for them, and she finds the descriptions almost always name a crucial feature of the text and carry, in nuce, the move needed to read it, “the generative force of difficulty” (Salvatori 1996, 448).

Line up the three words. Techné is where the difficulty lives; technique, in Ellul’s sense, rationalizes it; technology substitutes capital for that which has been sliced sufficiently fine. Language models are the first writing technology sold as a substitute for the difficulty itself, and if the difficulty is where a writer’s experience good lives, the substitution doesn’t save that labor so much as remove the good. Alex Reid suggested this summer that a model “can continue an unfinished poem,” but not his writing of it, “because once the AI takes over, I have ceased the experience.” Experience ends at the machine.

References

Babbage, Charles. 1832. On the Economy of Machinery and Manufactures. 2nd ed. London: Charles Knight.

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

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

Horning, Alice S. 2007. “The Definitive Article on Class Size.” WPA: Writing Program Administration 31 (1–2): 11–34.

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

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.

Salvatori, Mariolina. 1996. “Conversations with Texts: Reading in the Teaching of Composition.” College English 58 (4): 440–54.