Borges imagined a map so detailed that it covered the entire territory it described, at a scale of one to one. The map was perfect in the sense that every feature of the territory was represented, every road, every building, every field boundary reproduced at actual size. It was also useless.
A map that is the same size as the territory it describes has abolished the function that maps exist to serve. The function is compression—the reduction of the territory to something that can be held, consulted, carried from place to place and used to navigate the territory without requiring the territory itself to be present. The perfect map defeats this purpose. It is not navigation. It is duplication.
The categorisation of persons faces the same limit from the other direction. The system that seeks to describe a person with increasing precision is moving, with each additional category, toward the Borgesian map—the description that becomes as complex as the thing described, at which point the description has ceased to be a description and has become a parallel version of the original. The parallel version is not useful for the purposes that description exists to serve. It is not manageable at scale. It does not allow the efficient allocation of resources across a population. It is simply a person, represented in categories rather than in flesh, and it carries all the complexity that makes the flesh unmanageable in the first place.
The categories that population management systems use are broad for reasons that are structural rather than negligent. Male, female, diabetic, citizen, aged, employed, housed—these are categories at the level of granularity that administrative systems can process across millions of people with the resources and time available. They are crude in the sense that they capture very little of what the person inside them is actually like. They are functional in the sense that they allow decisions to be made about resource allocation, service provision, and policy design without requiring every decision to begin from scratch with every individual.
The crudeness and the functionality are not in opposition. They are the same thing viewed from different angles. The category is functional because it is crude. Its crudeness is what allows it to scale. A more precise category—one that distinguished not only between diabetics but between Type 1 and Type 2 diabetics, and then between those managing with medication and those managing without, and then between those with complications and those without, and then between those whose condition is stable and those whose condition is variable, and then between those who have the cognitive and practical resources to manage their condition independently and those who do not—would produce a more accurate description of the relevant variation. It would also multiply the administrative load of the categorisation by a factor that eventually makes the system unworkable.
The system therefore chooses the level of granularity at which it can function, and that level is always below the level at which the person is fully described. The gap between the level of functioning and the level of full description is where the person disappears—not eliminated, but unrepresented, their individuality present and unprocessable, accumulating in the space between what the category captures and what the person is.
The extension of the categories is not a theoretical exercise. Every additional variable added to the description of a person interacts with every other variable in ways that multiply the complexity non-linearly. The gay diabetic with a trauma history and an irregular income and a non-standard housing arrangement and a cognitive style that diverges from the institutional default is not simply the sum of five categories. They are a specific configuration of five variables that interact with each other in ways that none of the five categories, applied separately, can capture. The trauma history affects the diabetic management. The irregular income affects the housing. The cognitive style affects how the person navigates all of the above. The interactions are specific to this person and cannot be read from any of the categories that describe their components.
To describe the interactions would require a further set of categories—interaction categories, describing the relationships between the primary categories. The interaction categories would themselves interact in ways requiring further categories. The web of description would grow until it was not a web of categories but a model of the person—a computational representation of sufficient complexity to capture the dynamics of a specific individual rather than the population tendency of a category. The model would be the person, abstracted. At the point where the abstraction is complete enough to be accurate, it is complex enough to be unmanageable. The map has reached the scale of the territory. The description has collapsed into replication.
This is not a hypothetical limit. It is the direction that sufficiently advanced data systems are already moving, without having arrived at the limit and without necessarily recognising that the limit exists. The commercial platform that tracks a user’s behaviour across enough dimensions and over enough time is accumulating, in its data structures, something that begins to resemble a model of the specific person rather than a classification of the standard user. The model predicts what the specific person will click, what they will buy, what they will respond to, with an accuracy that generic category membership cannot approach. The prediction works because the model has enough variables, interacting in enough ways, calibrated against enough observed behaviour, to approximate the specific person’s responses.
The model is not the person. It is a predictive structure built from observed behaviour in specific contexts. It misses everything that the observed behaviour does not capture—the interior life, the context-dependent variation, the change over time that makes a person different at fifty from who they were at thirty. It also misses its own limits, because the model was built to predict behaviour and not to understand the person, and the distinction between predicting behaviour and understanding the person is the distinction between the Borgesian map and the territory.
The commercial interest in the model is the interest in the prediction, not in the understanding. The platform wants to know what you will click. It does not want to know who you are. The two are related but not identical, and the relationship between them is not in the model’s parameters. The model grows more accurate in its predictions without growing more accurate in its understanding, because the understanding would require categories that the platform has no commercial reason to develop.
The governance system that aspired to total categorisation would encounter a version of the same limit in an administrative rather than a computational form. The bureaucracy that attempted to capture every relevant variable for every person under its administration would require a bureaucratic apparatus as complex as the population it was administering. Each variable would require a process for its collection, a format for its storage, a system for its updating as the variable changed over time, and a mechanism for its integration with the other variables in the production of administrative decisions. The bureaucracy would grow until it was not administering the population but replicating it—a parallel structure of the same complexity as the structure it was supposed to manage.
At this point the bureaucracy has ceased to simplify and begun to duplicate. It has not become more useful to the person it is supposed to serve. It has become more expensive to operate, more difficult to navigate, and no more capable of genuine encounter with the individual than the crude categories it replaced. The duplication is not understanding. The model is not the person. The expanded category system is still a category system, only one that has grown large enough to reveal the limit that all category systems approach without ever reaching.
The person always exceeds the system not because the system has failed but because personhood is operating at a higher resolution than population management can sustain. This is not a criticism of systems. It is a description of what systems are for and what they can therefore not be for. Systems are for managing populations. The population is composed of persons. The persons are more complex than the population description. The excess is not error. It is not the residue of a system that has not yet become sophisticated enough to capture it. It is the permanent condition of the relationship between the person and any system that attempts to represent them, because representation requires reduction and persons resist final reduction by the nature of what persons are.
A system can function by reducing the person. Yet the closer a system moves toward full person-specific understanding, the more it sacrifices the efficiency that makes it systemic. These are not quite mutually exclusive—the skilled practitioner within a system can hold both the category and the person simultaneously, can use the system’s resources while exceeding the system’s categories through the exercise of judgment that the system cannot supply. But this is the practitioner exceeding the system, not the system accommodating the person. The system accommodates the category. The person arrives inside the category, exceeds it, and waits to see whether the practitioner will notice.
Some practitioners notice.
The system does not.
The system cannot.
It was not designed for noticing.
It was designed for processing.
The person who required noticing required something the system was not designed to provide, and no quantity of additional categories will change this, because the limit is not in the number of categories.
The limit is in what categories are.
The Vorpel Foundation