Mechanical Objectivity

"Let the numbers speak, keep subjective judgment out of it" sounds today like a professional virtue. Where that virtue came from may surprise you: public life worships numbers not because science proved numbers more accurate, but because people who are not trusted need numbers to protect themselves. The historian of science Theodore Porter proposed this inversion in Trust in Numbers (1995). He named this kind of objectivity, calculating by rule and excluding personal judgment, mechanical objectivity. It is also why the flattening ruler of R03 got built, and the shared foundation of the audit society and the ranking explosion that follow.

Porter distinguishes two kinds of objectivity. Disciplinary objectivity: a trusted community of experts judging by training, experience and tacit professional agreement, as when a senior doctor looks at a scan and says "something's wrong here," and people believe it because they trust that community. Mechanical objectivity: following public rules and excluding personal judgment, so that anyone computing gets the same result, as when a formula sets an insurance premium. The usual story says quantification "trickled down" from the successes of natural science into other fields; Porter went through the histories of accounting, actuarial work and engineering and reversed the direction: the expansion of quantification in public life was driven mainly by administration and politics, for standardisation, for letting information travel across distance, and for the legitimacy that the word "objective" confers. He calls quantitative procedures "technologies of distance": a number can leave the site that produced it and be checked, compared and aggregated by strangers a thousand miles away, and an expert judgment cannot do that.

From which comes his signature insight: numbers are a weapon of the weak. A strong, trusted professional community, such as the engineers of the French Corps des Ponts et Chaussées at the height of their power, or elite physicians, can rely on discretion, and nobody demands they show their arithmetic, because their "in my view" suffices. A weak, distrusted group exposed to outside scrutiny is forced to arm itself with numbers, because numbers look impersonal, checkable and impossible to accuse of favouritism. So Porter says quantification "is a mark of weak power and low status": objectivity is a defensive posture, a submission to external distrust. That also explains mechanical objectivity's other face, de-responsibilisation: a decision computed by rule has no particular person who must stand behind it, because judgment has been handed to a procedure. For the scrutinised it is a shelter; for anyone seeking accountability it is a wall.

This claim is the political-economy key to the whole red branch. Recall R01: the law schools cornered by rankings are a field of professional education under heavy external scrutiny; public schools and hospitals are in the same position. The less autonomy a field has and the more exposed it is to accountability pressure, the more it gets pulled into quantification and its reactive consequences, not because numbers are truer there but because numbers there are better at withstanding challenge. Upstream, the commensuration of R03 is mechanical objectivity's technical precondition: to let numbers travel across distance you first have to press heterogeneous qualities onto one ruler. Downstream, the audit society (R09) is its institutional form, since rituals of verification consume exactly the legitimacy mechanical objectivity produces. The family has older roots too: Simmel wrote in The Philosophy of Money (1900) that money is the oldest and most thorough commensuration device, pressing all qualitative difference into addable quantity; Weber treated "calculability" as the ground tone of modernity. What Porter did was specialise this grand claim and answer a concrete question, who embraces numbers first: not the strongest, but the first to be doubted.

The French statistician Alain Desrosières completed the picture from the other end. He worked at the French national statistics institute (INSEE) and was also a historian of science, and his The Politics of Large Numbers (1993) shows how statistics and the state made each other: statistics both prescribe what the state should manage and measure how well it does, and the act of "counting the population" itself consolidates authority. His central verb is worth remembering: faire tenir ensemble, to hold things together so that they hold up. Unemployment rate, social class, GDP: these categories did not grow out of the ground, they had to be constructed, agreed and stabilised politically and technically before there could be any talk of them "reflecting" reality; and once stabilised they become the framework reality organises itself around, with policy made by GDP and people mobilised by class. This is the national-statistics version of the looping effect: classification changes the world it classifies, and that lesson is R06, though you can follow this without having read it.

Desrosières also left a handy tool: four attitudes people take toward whether a statistic is real. Metrological realism, where the number faithfully reflects an independently existing quantity, like taking a temperature; accounting realism, where the number is a product of bookkeeping rules and its accuracy depends on the rules; proof in use, where you find out whether a number is any good by using it; and constructivism, where the number is a product of social negotiation. The value of this taxonomy shows up in Espeland and Sauder's 2007 ranking study: people at different positions in the ranking ecology hold different ontologies. The producers of rankings and the uninformed applicants are both naive metrological realists, taking the ranking as objective quality regardless of method; while the administrators deeply involved in manufacturing the numbers slide almost without exception into constructivism, one interviewee saying: what we are forced to do is shameful, we are a Catholic school, we teach honesty, and yet we have to count non-legal jobs as employment like everyone else. The rule is cruel: how much you believe in a number depends on how close you are to its production line, and the closer you are, the less you believe.

One more boundary, and the easiest mistake to make when citing Porter: the mechanical objectivity thesis explains why quantification gets adopted, it does not adjudicate its validity. "Numbers withstand challenge better" and "numbers are more accurate" are two independent claims, and Porter's entire point is not to mistake the first for the second; conversely he does not say rule-following is necessarily worse, since excluding judgment does genuinely exclude room for favouritism, and both the cost and the benefit are real. Zoom out globally and the same mechanism runs under another name: Merry's "indicator culture" (2016), where indicators quantify and package subjective information and then present it as objective and reliable; Governance by Indicators (2012) puts it more bluntly, that whoever defines the indicator redistributes global attention, reputation and resources.

One open question: mechanical objectivity presupposes inspectable rules, so anyone can recompute the result by following the formula. Using a deep model as the grader also excludes personal judgment, yet cannot be checked item by item, and nobody can read out its "rules." The D1 report lists this as an open question: is this a new form of mechanical objectivity, or a third thing with neither judgment nor rules, a "black-box commensuration"?

The one-line takeaway: trust in numbers usually reflects not that numbers are trustworthy, but that no cheaper trust between people is left.