Every institution needs simplified representations of the world in order to act. It needs categories that distinguish one situation from another, indicators that compress complex conditions into observable signals, maps that locate responsibilities and relationships, and models that allow decision-makers to reason about realities too large or intricate to encounter directly. None of this is a defect in institutional cognition. Simplification is one of the conditions that makes organised action possible. The danger begins somewhere else: when a representation becomes so familiar, embedded and authoritative that the institution gradually stops experiencing it as a representation at all.

At first, the distinction between model and reality may be obvious. A new classification system is introduced for a particular administrative purpose; an indicator is selected because it provides a useful approximation of an outcome; a risk model is built around explicit assumptions; a policy framework defines a problem in a way that allows action to begin. People involved in creating these devices understand that choices have been made. They know that categories could have been drawn differently, that indicators leave things out and that models simplify relationships which are more complicated in the world itself. The map is useful precisely because everyone still remembers that it is a map.

Time changes this relationship. Representations that prove useful become embedded in institutional routines. Categories migrate into forms and databases. Indicators become part of reporting cycles. Models shape budgets and organisational responsibilities. Definitions appear in guidance and regulation. New employees learn to work through the resulting architecture without necessarily encountering the reasoning through which it was originally constructed. What began as an explicit simplification can become part of the taken-for-granted environment of the institution.

Nothing about this process requires intellectual negligence. Institutionalisation is partly how organisations preserve useful knowledge. If every generation of officials had to reconstruct every category, metric and model from first principles, government would become impossibly expensive and unstable. Mature institutions need cognitive infrastructure that can be inherited. The problem arises when inheritance preserves the representation while losing awareness of the choices, exclusions and assumptions that made the representation what it is.

A category can then begin to look like a property of reality rather than an administrative distinction imposed for a purpose. An indicator can become synonymous with the outcome it was designed to approximate. A modelled relationship can acquire the status of an unquestioned causal fact. The institution continues to operate coherently, but its coherence increasingly depends on treating its own representational architecture as if it were simply the way the world is.

This is reification: not the existence of simplification, but the conversion of a constructed representation into something experienced as natural, complete or self-evident. The distinction matters because institutions cannot avoid simplification, whereas they can remain aware that simplification has occurred. SIMPLIFICATION ≠ REIFICATION. A model becomes cognitively dangerous not because it leaves things out, but because the institution forgets that anything has been left out.

Consider an agency that divides a population into a small number of administrative categories. The categories may have been designed carefully because they correspond to different service pathways. Over time, data collection, staffing arrangements, funding rules and performance measures are all organised around them. Cases that fit the categories are easy to process because the institution has built an architecture capable of recognising them. Cases that cross the boundaries become exceptions. Eventually, the organisation may begin to interpret those difficult cases as unusual properties of the population rather than signals that the categories themselves are simplifying a more continuous and complicated reality.

The map has started to shape perception of the territory.

A similar process occurs when an indicator acquires conceptual authority. Suppose an institution uses employment status as one measure of economic inclusion. Initially, everyone may understand that employment captures only part of the phenomenon. But if funding, evaluation and political reporting repeatedly revolve around that measure, improvements in the indicator can gradually become indistinguishable from improvements in the underlying objective. Questions about job security, income adequacy, working conditions or sustainability of employment may become secondary, not because anyone has decided they are irrelevant, but because the institution’s established representation makes one dimension much easier to see than the others.

This is different from an inaccurate indicator. The employment measure may be calculated perfectly. It is also different from the representational blindness that occurs when important realities simply lie outside the institution’s current field of vision. Reification concerns the relationship the institution develops with the representation itself. The danger appears when the institution ceases to treat the representation as a selective instrument and begins to organise judgement as though the instrument exhausted the object being judged.

Models can acquire this status even when their users know intellectually that they are models. An economist may readily acknowledge that a forecast depends on assumptions, or a policy analyst that an index combines variables through contestable choices. Yet institutional practice can still behave as though the resulting number possesses a solidity that the underlying assumptions do not warrant. Meetings revolve around it, decisions require it and alternatives struggle to gain attention because they do not fit the established analytical format. Formal awareness of uncertainty can coexist with practical dependence on the model as the institution’s dominant reality.

The problem becomes especially persistent when organisational structures themselves reflect the representation. Once responsibilities, budgets and expertise have been distributed according to a particular model of the world, evidence that crosses its boundaries can become difficult to absorb. A problem that does not fit one department’s mandate may be transferred to another, and then another, because the institutional map contains no location corresponding to the phenomenon as it actually exists. What appears operationally as a coordination problem may partly originate in a representational architecture that has divided reality differently from the way the problem behaves.

Artificial intelligence can intensify this risk because computational models can acquire considerable authority through their predictive performance. A system that repeatedly produces useful predictions may become deeply integrated into institutional decision-making, and that integration may be justified. But predictive success does not transform the model into reality. The system still represents the world through available data, selected objectives, historical relationships and particular forms of classification. When its outputs become embedded in workflows, the danger is not simply that the model might be wrong. It is that the institution may gradually reorganise its understanding of the problem around what the model is capable of representing.

This can create feedback between map and territory. If an institution allocates resources according to a model, institutional behaviour begins to respond to the representation. If inspection is concentrated where a risk model predicts problems, more violations may be discovered there, generating new data that appears to confirm the original distribution of risk. If a classification determines which cases receive particular interventions, future administrative records will reflect those classifications. The representation no longer merely describes the institutional environment; it participates in producing the evidence through which that environment will later be interpreted.

Such feedback does not make the model inherently invalid. Institutional action always changes some part of the reality it seeks to govern. The important question is whether the institution can still distinguish evidence about the world from evidence partly generated by its own representational choices. Without that distinction, repeated institutional use can appear to validate a model when it may instead be reproducing some of the conditions through which the model continues to look plausible.

This is why normalisation should not be mistaken for validation. A representation can become universal within an organisation because systems, routines and incentives have converged around it. Familiarity can make alternatives difficult to imagine, and widespread adoption can make the representation appear empirically inevitable. Yet none of these conditions demonstrates that its assumptions remain appropriate. NORMALISATION ≠ VALIDATION. Institutional endurance tells us that a model has become durable; it does not tell us that the model continues to describe the world adequately.

Avoiding reification does not mean placing every institutional model under permanent suspicion. An organisation that endlessly questioned its basic categories would struggle to act consistently. Models need periods of stability if they are to coordinate behaviour and allow knowledge to accumulate. The objective is not constant representational disruption but retained awareness of contingency: the capacity to recognise that an established representation was designed for purposes, contains exclusions and could become inadequate as reality changes.

That awareness becomes particularly important when evidence repeatedly appears at the edges of the model. Persistent exceptions, categories that require increasing numbers of special rules, outcomes that the established framework cannot explain, or recurring disagreements between formal representations and external experience can all indicate that the map deserves renewed examination. None automatically proves that it should be replaced. They are invitations to recover the distinction between the representation and the thing represented.

Institutions can support this distinction by preserving not only models but the reasoning around them: why categories were created, what assumptions indicators depend upon, which exclusions were considered acceptable, where uncertainty remained and what kinds of change would justify reconsideration. Such memory allows future users to inherit cognitive infrastructure without inheriting it as unquestionable reality. The institution can continue to use the map while retaining knowledge of how and why the map was drawn.

This is ultimately a problem of institutional cognition because models organise attention before individual decisions are made. They shape which objects appear comparable, which differences become administratively meaningful, which relationships can be analysed and which questions can be asked easily. When a model becomes reified, those choices disappear behind the apparent neutrality of the institutional environment. People reason intelligently inside the representation while losing sight of the fact that the representation itself is one of the things that might eventually need to be examined.

A cognitively mature institution therefore does not seek freedom from maps. It seeks the ability to use maps without becoming imprisoned by them. It can rely on categories while remembering that categories are constructed, use indicators while remembering that indicators are proxies, and organise action through models while preserving the possibility that reality may eventually require the model to change.

The map becomes the territory when an institution loses awareness of the difference between a useful representation and the reality it was created to make manageable. The problem is not simplification itself, because institutional action depends on simplification. It is reification: the gradual disappearance of assumptions, exclusions and representational choices from institutional awareness until the model begins to look natural, complete and inevitable. Institutional intelligence therefore requires more than good models. It requires the continuing capacity to remember that even the most useful model remains a map — and that reality retains the right to be different from it.