3.1Discussion
Ten invented records, two lines of a shared past, and one auditor's opinions give an answer in two parts. The auditor in these scenarios lets a number about people decide only when it passes two tests together: it holds when it is computed again, and the person it counts can see it and contest it. And the studio's three marks on every ground of a judgment leave the statement that turns a figure into a person's measure exposed on the page, where it can be doubted as openly as any premise.
Read against the sociology of quantification, the first part answers the call that Espeland and Stevens make in A Sociology of Quantification (2008) for an ethics of numbers with one possible shape, though only an invented one. It is a procedural ethics. It does not ask whether a resolution rate is the right measure of an agent, or whether a contribution score fairly reflects an applicant. It asks whether the rate would come out the same next quarter and whether the agent was told it would be used and could appeal. The scenario places the moral question inside a procedure rather than replacing it: the appeal channel lets the person counted bring their own view of what the number should mean, and the reopening line on the one unqualified record sends the criteria back for review if appeals overturn enough cases.
The second part speaks to Espeland and Stevens's earlier account in Commensuration as a Social Process (1998) of commensuration as a mode of power. In the records the act of commensuration is not hidden inside the figure; the studio's form pulls it out and writes it as a sentence. A gain "splits cleanly"; a rate is read "as equivalent to" ability; one kind of dispute is "treated as falling into the same case type" as another. That is the studio's design, and its presence in every record is not a finding. What is a finding is that, across the ten, such statements are never filed as evidence, and that the auditor's opinions worsen where the record leans on them without a test.
Kiviat's account of de-commensuration finds contention where a prediction holds people to account for what is not their fault. This paper reads the dispatch record as staging that situation and answering it, in the scenario, not with a moral argument but with a proposed measurement: compare two couriers who differ only in signup date. This paper reads Fourcade and Healy's classification situations into the shared name "contribution score" across hiring and lending, and the auditor's refusal to accept its correlation as grounds for automatic rejection is one way the scenario resists a score's becoming a fact. Kellogg, Valentine and Christin's contested terrain is present along the whole work line, and the union that hires an auditor is one of its forms. Two lenses fit less well. Espeland and Sauder's reactivity concerns how the measured change in response to measures, and these one-quarter snapshots show little of that; what they show instead is institutions treating a common index as confirmed by its spread. Porter's history of why numbers attract business and government helps explain the task force's wish, but a record set confined to a few quarters cannot show where such an attraction comes from.
3.2Contributions
The paper makes three contributions, each a claim about how a synthetic record set frames decisions and not about the world. First, it describes a two-sided test for numbers about people, reproduction and reach, and shows that in these scenarios the two must hold together: one record meets both and receives the only unqualified opinion; records that partly meet one receive qualified opinions; records that meet neither receive adverse opinions or none. Second, it shows that the basis form separates figures from the equivalences that let them stand for people, and that in the ten records the equivalences are always claims or assumptions, which gives the debate on commensuration a concrete place to look for the moment a number becomes a person's measure. Third, it answers the request of Mennicken and Espeland in What's New with Numbers? (2019) to track quantification across sites, by reading the studio's own threads: two lines along which a number's logic moves from a firm's ledger toward a single person, with the name "contribution score" recurring from hiring into courts and lending.
3.3Limits
The records are synthetic. Their institutions, figures and opinions are the studio's inventions, and the patterns found are patterns in how the studio writes, not evidence about how real auditors, firms or courts behave. Ten records are few, and the selection rule, two axes of the shared past only, leaves out the same axes in both futures and the six other axes of the trunk. The counts are made by matching words in the records, and the heads under which audit questions are sorted are this paper's own; a different sorting could move a question from one head to another. The frame was measured among the ten records only, not across the studio's library. The works are read through their published summaries, and where this paper goes beyond those summaries it says so.
3.4Further research
Three lines of work follow. The first is to read the same two axes after the fork and ask whether the two futures keep both tests, drop one, or replace them, since the map carries both lines into both branches. The second is to read the trunk's other axes for the same separation between figure and equivalence, to see whether it belongs to these two axes or to the studio's form throughout. The third is to set the scenario's procedural ethics of numbers beside the debate's moral ones more fully, asking, with Kiviat's work in view, when a test of reproduction and reach would be enough and when it would leave the question of desert untouched.
The contribution of this paper, stated once more, is that in ASI 2030's shared past a number may decide about a person only when it holds twice and can be seen by the person it counts, and that the record's form shows where the number is made to stand for that person.