Interlude: When the Medium Shifts

Quantification is a technology of distance.
— Theodore Porter, Trust in Numbers (1995)

The fact migrated

The five properties that Part I traced through seven centuries of commercial practice did not stay in the counting house. They migrated, and the route they traveled rearranges the standard history of modern knowledge.

Mary Poovey documented the migration in patient detail(Poovey 1998, pp. 1–28, 29–91, 92–143). The "modern fact," she argued, was not a discovery of the Scientific Revolution. It was an invention of the merchant's ledger. When Luca Pacioli codified double-entry bookkeeping in 1494, he described a system in which every transaction was recorded twice, once as debit, once as credit, and in which the formal balance of the two columns created what Poovey called an "effect of accuracy" that was immediately conflated with moral authority. The merchants "who keep such records are wise, prudent and honest subjects," Pacioli wrote. The claim went further: the ledger was true, and its truth could be inspected by anyone who understood the form.

The "fact" that the ledger produced was a specific kind of object: a claim bound to a documented procedure, subject to the condition that the books balanced, carrying the stake of the merchant's commercial reputation, offering recourse to anyone who could demonstrate that the books had been altered, and composing with other ledger entries to produce a coherent account of the enterprise. The properties were the same properties the bill of exchange carried. But the ledger's truth-claims did not remain in the domain of commerce. They traveled, first to statecraft, then to natural philosophy, then to the sciences of wealth and society that would eventually become political economy and statistics.

Poovey traced a genealogy in which the epistemic authority of the commercial ledger was progressively appropriated by institutions that had nothing to do with trade. William Petty's "political arithmetic" in the 1660s explicitly modeled itself on mercantile bookkeeping: the state's accounts, like the merchant's, could be balanced, audited, and inspected. The Royal Society's experimental reports borrowed the literary conventions of the commercial correspondence that Italian merchants had perfected over the preceding two centuries: detailed, circumstantial, addressed to a reader who was expected to evaluate the evidence on its own terms. The "fact" migrated from the counting house to the laboratory, and it carried its properties with it.

The gentleman and the air-pump

The migration's most consequential episode occurred in Robert Boyle's laboratory in Oxford in the 1660s. Steven Shapin and Simon Schaffer reconstructed it with the precision of institutional historians(Schaffer 1985, pp. 25, 60).

Boyle's air-pump experiments did not establish the experimental method through theoretical argument. They established it through a verification protocol, one that Shapin and Schaffer decomposed into three "technologies of knowledge-production." The first was material: the air-pump itself, a complex and expensive instrument that few natural philosophers could afford to replicate. The second was literary: Boyle's experimental reports, written in extraordinary circumstantial detail, temperatures, pressures, the behavior of the mercury at specific moments, the names of the gentlemen present, designed to produce in the reader's mind "such an image of an experimental scene as obviates the necessity for either its direct witness or its replication." Shapin and Schaffer called this "virtual witnessing": a literary technology that converted a private observation into a public fact by enabling readers who were not present to evaluate the evidence as if they had been.

The third technology was social. The experiments were performed before witnesses, gentlemen whose social standing made their testimony credible and whose reputations would be damaged by false attestation. "An experience that one man alone witnessed," Shapin and Schaffer concluded, "was not adequate to make a matter of fact." A fact required witnesses. The witnesses had to be identifiable, their testimony had to be public, and their standing had to carry consequences. An experiment performed in private, however rigorously, produced a claim, not a fact. A claim without witnesses was, in the epistemology of the early Royal Society, no better than speculation.

Boyle's experiment shares the same institutional logic as the notary's practice, applied to a different domain. His experimental report bound a specific observation to an identifiable observer and his apparatus. The report specified the circumstances in which the observation was made, in sufficient detail that another experimenter could attempt to replicate it. The witness's social standing created stakes: the gentleman who attested to a fraudulent result would lose the reputation that made his testimony valuable. Replication provided recourse: a contested result could be challenged by performing the experiment again, before different witnesses, under documented conditions. And the findings composed: results from Boyle's laboratory in Oxford could be evaluated alongside results from Hooke's demonstrations in London because both followed the same witnessing protocols — the same condition that had made the bill of exchange portable across cities and currencies.

Shapin and Schaffer drew a conclusion that resonates beyond the history of science: "Solutions to the problem of knowledge are solutions to the problem of social order." The verification infrastructure determines the social structure, not the other way around. The merchants of Bruges understood this. The gentlemen of the Royal Society rediscovered it.

The will to willessness

If Boyle's air-pump was the first medium shift, from documentary to experimental witnessing, the nineteenth century brought a second: from human observation to mechanical recording.

Lorraine Daston and Peter Galison traced this shift through three regimes of scientific objectivity that succeeded each other over three centuries(Galison 2007, pp. 17–19, 39). In the eighteenth century, the regime was truth-to-nature: the natural philosopher selected the best specimen, drew the most representative image, and used his trained judgment to distinguish the essential from the accidental. The witness was the expert's eye, and the expert's authority guaranteed the testimony. In the nineteenth century, a new regime emerged that Daston and Galison called mechanical objectivity: "to be objective is to aspire to knowledge that bears no trace of the knower." The photograph replaced the drawing. The self-recording instrument replaced the observer's notebook. The scientist's goal became what Daston and Galison called "the will to willessness": the deliberate suppression of the human observer's judgment in favor of mechanical registration.

The third regime, trained judgment, emerged in the twentieth century: the expert returned, but now disciplined by mechanical infrastructure. The radiologist reads the X-ray, but the X-ray constrains what she can see. The statistician interprets the data, but the data were gathered by instruments whose calibration is independent of her interpretation.

Each regime was a medium shift. Binding and conditions shifted together: the gentleman's word gave way to the camera's exposure, the experimental protocol to photographic specifications that a stranger could evaluate without ever entering the laboratory. When daguerreotypes replaced hand-drawn illustrations in nineteenth-century botanical atlases, the stakes migrated with the medium. No longer the reputation of the artist who selected the ideal specimen, but the calibration of the instrument that recorded whatever the lens happened to capture, blemishes and all. Had the new regime abandoned recourse entirely, the shift would have destroyed scientific knowledge rather than transformed it; instead, gentlemanly dispute yielded to mechanical replication and, eventually, to the formal apparatus of peer review. But "each successive stage," Daston and Galison observed, "presupposes and builds upon, as well as reacts to, the earlier ones."

Quantification as distance

Theodore Porter completed the genealogy. Quantification, he argued, was not a method but a technology, and its purpose was to solve the same problem the bill of exchange had solved: how to make a claim credible to someone who does not know you and has no reason to trust you(Porter 1995, pp. ix, 8, 14).

"Reliance on numbers and quantitative manipulation," Porter wrote, "minimizes the need for intimate knowledge and personal trust." His central case was the French civil engineer, the ingénieur des Ponts et Chaussées, whose professional judgment about where to build a bridge or how to route a canal had, by the mid-nineteenth century, been subordinated to cost-benefit calculations that any bureaucrat in Paris could read. The engineer knew the terrain; the number made his knowledge transferable to someone who did not. A statistical result carries weight not because the reader trusts the statistician but because the methodology, the sampling procedure, the significance test, the confidence interval, specifies the terms on which the result holds. The methodology is the bill of exchange: a portable, inspectable object that carries the conditions under which a claim can be assessed by strangers.

Porter called these mechanisms "technologies of trust," explicitly citing Shapin and Schaffer's earlier work. The chain from commercial witnessing to scientific experiment to statistical quantification was not an analogy but a genealogy. Each medium shift inherited the requirements of the previous one because the underlying problem — making truth inspectable at a distance — did not change when the medium changed. "Quantification is a way of making decisions without seeming to decide," Porter observed, which was what the bill of exchange also accomplished: it made the exchange rate, the date, and the currency explicit so that strangers could evaluate the obligation without knowing the original parties, without trusting the original drawer, and without being present when the agreement was struck.

Chain of custody

The most explicit survival of this pattern in modern institutional life is legal, not scientific: the chain of custody in evidence law.

A piece of physical evidence that travels from a crime scene to a courtroom must carry, at every stage, documentation of who collected it, the circumstances in which it was stored and transported, and the professional liability of each custodian who handled it. The defense retains the right to challenge any link in the chain. Forensic results from different laboratories must be evaluable alongside each other under the same evidentiary standards. An officer who breaks the chain faces disciplinary consequences; a gap, an undocumented transfer, a period of unsecured storage, can render the evidence inadmissible regardless of its content.

Chain of custody is not a metaphor for the notary's practice. It is the notary's practice, formalized into an explicit protocol by a legal system that recognized, through centuries of case law, that the integrity of evidence depends on the circumstances of its production, preservation, and transmission.

The discontinuity objection

The strongest objection to everything this interlude has argued is straightforward: computation is different. Not merely faster, not merely more scalable, but categorically different: a discontinuity, not a medium shift.

It has force. The Stasi's surveillance apparatus required 170,000 human informers to monitor 16 million citizens, at a cost that consumed a significant fraction of East Germany's state budget. A modern platform requires no informers at all: the users generate the surveillance data themselves, voluntarily, continuously, and in formats that are machine-readable from the moment of creation. The air-pump required a gentleman's presence in the laboratory. The algorithm requires no presence at all: it operates at a distance that Boyle could not have conceived, at a speed that makes the Champagne fairs' twenty-day courier route seem geological, and at a scale that dwarfs every verification infrastructure described in this volume. Mayer-Schönberger's reversal of the forgetting default, documented in Chapter 3, has no analogue in the history of parchment or print. Computation transforms the economics of witnessing, making attestation cheap, permanent, and ubiquitous in ways that no prior medium permitted.

Computation introduces properties, speed, scale, permanence, automation, that the bill of exchange, the experimental report, and the statistical table did not possess. The Interlude does not argue that code is "just like" parchment, any more than Daston and Galison argued that the photograph was "just like" the gentleman's testimony.

But every system that has ever made truth inspectable by strangers, whether the medium is clay, parchment, print, photograph, or code, has had to solve the same problems: how to bind a claim to an identifiable source, how to specify the conditions under which the claim holds, how to ensure that attestation carries consequences, how to provide a path of recourse when attestation fails, and how to compose attestations from different sources into a coherent whole. The medieval bill of exchange and the modern digital signature solve these problems with different mechanisms operating at incomparable scales. The problems themselves have not changed, because the problems are not properties of the medium. They are properties of the task, the task of making truth compose across distances, institutions, and the absence of the person who first asserted it. If the task persists into computation, the constraints it imposes should persist as well.

Forward

The question Part II asks is whether computation satisfies the five properties or merely simulates them.

Two regimes now occupy the space where the notary once stood. One proposes: it interpolates, retrieves, recombines, an engine of plausibility that can produce text indistinguishable from testimony but that cannot bind an assertion to the terms on which it may be relied upon. It cannot produce a witness. It cannot tell you whether two outputs that seem to agree actually agree, or merely resemble each other by accident. The other certifies: it enforces schemas with perfect fidelity, rejects malformed inputs, and answers queries with provable consistency, but it cannot invent a new distinction on demand, cannot recognize that two records from different systems describe the same event unless someone has already encoded the relationship, and when the world offers novelty, it does not adapt. It waits.

Call the first the empire of strings. Call the second the empire of tables.

Each is coherent within itself. Each fails at the seam, at the point where the bill of exchange, the experimental report, and the chain of custody succeed: the border between two systems of truth that must compose.

Whether computation can produce a third kind of object, one that, like the notary's bill, neither replaces nor duplicates the systems it bridges, but allows them to compose under inspectable conditions, with binding that holds, stakes that matter, recourse that functions, and composition that scales, is the question the rest of this volume turns on. The alternative is that the speed, scale, and permanence of the new medium finally break what seven centuries of institutional innovation built.