Positions, Not People
Positions, Not People
I was listening to a corporate town hall recently – the kind every large organization holds, with dashboards and values statements and a Q&A at the end – when a familiar phrase went by. Leadership was describing a new workplace policy. The goal, they explained, was a formalized standard: clear, consistent, with minimal variability, moving away from individual arrangements.
Nobody in the room would have flagged the sentence as remarkable. It is how responsible administrators talk. But I had been rereading James C. Scott, and through that lens the sentence stopped being administrative boilerplate and became a thesis statement. The individually negotiated arrangement – this person, this manager, this commute, this caregiving schedule, this odd but productive rhythm – was about to be replaced by something a system could see. And the reason had nothing to do with whether the individual arrangements worked. Many of them worked beautifully. The problem was that they were illegible.
The state’s eye
Scott’s Seeing Like a State (1998) is about what happens when a governing center needs to comprehend something vast and organic – a forest, a city, a language, a population – and comprehends it by simplifying it. The state cannot see the peasant’s tangled strip-farming, so it draws a cadastral map. It cannot tax what it cannot name, so surnames are imposed. It cannot administer a medieval city’s accreted alleys, so the new district is a grid.
His most durable example is the scientific forest. Eighteenth-century Prussian foresters, needing predictable timber yields, replaced the messy old-growth forest with the Normalbaum: same species, same age, planted in rows, inventoried to the tree. For one rotation it was a triumph. Then yields collapsed. The mess that had been cleared away – the underbrush, the fungi, the deadwood, the species interactions nobody had thought to record – turned out to be the fertility of the forest. The simplification did not merely describe the forest poorly; imposed on the forest, it destroyed what it could not describe.
Scott gives the discarded remainder a name: metis – the local, practical, relationship-embedded knowledge that cannot be centralized because it does not survive being written down in standard categories.
The cadastral map of the org chart
Enterprise HR systems are cadastral maps of organizations. The unit of account is not the person but the position: a requisition number, a job code, a grade, a cost center, an FTE fraction. Positions are everything people are not – fungible, comparable, aggregable, budgetable. You can sum positions, benchmark them against industry tables, mark them on-site or remote, and eliminate 3.0 of them from a spreadsheet without meeting anyone.
This is not a design flaw. It is the design. An organization of tens of thousands cannot be administered as tens of thousands of individual stories, any more than Prussia could tax a forest it hadn’t inventoried. The position abstraction is what makes central decisions possible – which is precisely Scott’s point about maps. The question is never whether to have the map. The question is whether the people holding it remember that it is one.
Because here is what the position cannot see. It cannot see that one analyst is the person three other teams quietly route their hardest questions through. It cannot see which nurse is the reason a fragile scheduling truce holds. It cannot see whose institutional memory spans the last two system migrations, or which “redundant” role was actually the connective tissue between two departments that only cooperate because those two individuals trust each other. These interrelations are the organization’s underbrush – its actual fertility – and they are invisible in the system of record by construction, because the system records positions, and interrelations are properties of people.
So when decisions run on positions – this role is now on-site, that grade is eliminated, these arrangements are non-standard – the interrelations are ripped out the way the underbrush was, not out of malice but out of the honest inability of the instrument to register their existence. And as in the forest, the first harvest looks fine. The org chart after the reorganization is cleaner than before. The yield collapse comes one or two rotations later, in attrition and slowed projects and decisions that mysteriously take longer – damage the system’s own instruments cannot attribute to its cause, because the cause was never in the data.
A statistician’s confession
I should be careful here, because I am not an innocent bystander. I build legibility instruments for a living. My current work joins census population grids to healthcare facility service areas – estimating the communities around hospitals and clinics from exactly the kind of standardized aggregates Scott writes about. The census is the founding example of state legibility. County population by age band is a Normalbaum of people.
So the lesson I take from Scott is not that grids are evil. Denominators require grids; you cannot ask “is this rural community underserved?” without an estimate of the community, and an estimate means categories, boundaries, aggregation. The lesson is about rank: the grid is an estimate of the territory and must never be promoted to being the territory. In my own field the discipline shows up in small rules – knowing what an aggregate can and cannot re-identify, distrusting a tidy table until you know what was thrown away to make it tidy, treating a model of demand as a hypothesis about a community rather than a description of it. The statistical tradition at its best is a kind of institutionalized humility about maps. High modernism, in Scott’s telling, is what happens when that humility is deleted.
And that yields a testable claim about the HR case, because a Scott-style failure is not just a mood – it has a signature. If the system sees positions and the organization runs on people, then position-level data should systematically underpredict the disruption of specific departures. Two identical job codes; wildly different network centrality; identical severance math; utterly different consequences. Any organization could measure this – exit events joined to the informal network, predicted versus realized disruption. Almost none do. The instrument that would be indicted is the one that would have to fund the study.
Administering without blinding
Scott’s own prescription was never to abolish maps but to keep them answerable to metis – to design systems that leave room for the local knowledge they cannot contain. Translated to the enterprise, that looks less like a software purchase and more like a set of habits:
- Let variability be data, not noise. A hundred individual arrangements are a hundred observations about what actually lets work happen. “Minimal variability” throws the sample away before analysis.
- Give managers standing to represent what the system can’t see. The supervisor who says “the org chart says these roles are equivalent, and I am telling you they are not” is not obstructing the process. She is the process’s only error-correction channel.
- Measure the aftermath, attribute honestly. If a position-based decision is followed by a slow-motion yield collapse, the postmortem should be allowed to name the decision – otherwise the map is unfalsifiable, which is another way of saying it has become a religion.
- Distinguish the two questions. “What does the budget require?” is a positions question. “What will this do to us?” is a people question. Answering the second with the instrument built for the first is the whole error, compressed.
The town hall that started this essay was, to be fair, full of people sincerely trying to do right by a workforce – and in the same meeting, leadership spoke about wanting to genuinely understand the communities around their rural facilities, which is the opposite instinct: a request for exactly the local knowledge the dashboards don’t carry. Both impulses live in every large organization. Scott’s warning is simply about which one wins by default. The grid always wins by default. It is visible, and the underbrush is not – until the second rotation, when the forest quietly stops growing, and every instrument in the building agrees that nothing is wrong.
Further reading
- James C. Scott, Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed (Yale, 1998) – especially chs. 1 (“Nature and Space”) and 9 (“Thin Simplifications and Practical Knowledge: Metis”).
- Friedrich Hayek, “The Use of Knowledge in Society” (1945) – the economist’s version of the same argument about dispersed, local knowledge.
- Elinor Ostrom, Governing the Commons (1990) – what metis-respecting institutions look like when they work.