Institutions routinely confront realities that people inside them know exist but that the organisation itself has difficulty seeing. A frontline worker may recognise a recurring problem that has no category in the reporting system. Citizens may experience a burden that disappears when administrative performance is measured only through processing times. Several departments may encounter different fragments of the same social phenomenon without possessing a common representation that allows those fragments to appear as one problem. In each case, knowledge can exist somewhere within or around the institution while the phenomenon remains weakly available to institutional action. The difficulty is not necessarily ignorance in the ordinary sense. It is a problem of representation.
To govern something consistently, an institution must usually be able to make it present within its cognitive and administrative architecture. The form of representation can vary enormously. A phenomenon might enter through statistics, case categories, maps, professional assessments, narratives, risk registers, legal concepts, indicators or other forms of organised description. It does not need to be reduced to a single number, nor does every relevant property need to be captured. But unless some adequate representation exists, the phenomenon struggles to acquire the visibility required for sustained institutional attention.
This is why representability matters for governability. Institutions cannot continuously organise themselves around everything that might be occurring in the world. They need mechanisms that allow particular conditions to enter workflows, deliberations and decisions. A recognised category can trigger a procedure. An indicator can direct attention. A case classification can determine responsibility. A map can reveal spatial concentration. A professional assessment can make a complex situation legible without reducing it to a metric. Representation gives institutional form to something that would otherwise remain difficult to locate within organised action.
Consider an administrative service that records how long applications take to process but does not represent the effort citizens expend completing them. From inside the institution, performance may look strong: cases are processed within target times, backlogs are low and formal error rates are acceptable. Yet citizens may be making repeated calls, submitting the same information several times or navigating requirements that are individually reasonable but collectively exhausting. The burden is real, and staff may even recognise it informally. But if the institutional representation begins only once a completed application enters the system, much of the citizen’s experience remains outside the object being governed.
The problem is not solved merely by collecting more data. Representability and measurability are not the same thing. Some phenomena can be represented meaningfully through qualitative evidence, professional judgement, structured narratives or combinations of different forms of knowledge. An institution can understand that trust is deteriorating without possessing a single definitive trust metric. It can represent community vulnerability through multiple sources rather than one index. It can recognise emerging forms of harm before they are sufficiently stable to support precise measurement. What matters is whether the phenomenon can acquire a sufficiently coherent institutional form to become available for reasoning and action.
Conversely, measurement does not guarantee adequate representation. A service can possess large quantities of data while representing the wrong dimensions of the problem. If every available measure concerns throughput, the institution may become exceptionally capable of seeing throughput while remaining comparatively blind to accessibility, fairness or unintended burden. The data may be accurate. The representational architecture may still be incomplete for the governing task.
This helps explain why institutional blind spots can persist even in information-rich environments. The obstacle is sometimes not the absence of observations but the absence of a structure capable of turning those observations into an institutional object. Individual complaints remain individual complaints. Frontline exceptions remain exceptions. Local anomalies remain local. Without categories, relationships or interpretive structures that connect them, the institution can repeatedly encounter pieces of a phenomenon without becoming able to see the phenomenon itself.
Administrative boundaries can intensify this difficulty. Problems in society rarely respect the divisions through which governments allocate responsibility. A household experiencing poor health, insecure housing, unstable employment and debt may interact with several public systems, each of which represents one part of the situation according to its own mandate. Every organisation may possess an accurate representation of its fragment while no institution possesses a representation of the interaction among them. The problem is visible everywhere in pieces and nowhere as a whole.
The resulting limitation is cognitive before it is operational. If the institution cannot represent the relationship, it becomes difficult to ask questions about it systematically. Responsibility remains fragmented, evidence remains dispersed and interventions are designed around the visible components. The institution may respond competently within each administrative boundary while remaining unable to govern the pattern produced across those boundaries.
Representational absence can also affect emerging problems. Institutions usually inherit categories designed around realities that were already recognised when administrative systems were created. New technologies, changing family structures, novel forms of employment or unfamiliar combinations of social risk may initially fit poorly within existing classifications. People encounter the new phenomenon before the institution possesses a stable way of describing it. During that interval, cases can appear anomalous because the representational architecture has not yet caught up with the world.
This does not mean that every newly observed phenomenon should immediately acquire a formal category. Institutional representations can become overloaded, unstable or excessively complex if every exception produces a new classification. Governing requires simplification as well as recognition. The challenge is to distinguish noise from evidence that the existing representation is systematically excluding something consequential.
Nor does making a phenomenon administratively legible automatically justify intervention. This distinction is essential. An institution may become capable of representing a behaviour without thereby acquiring a legitimate reason to regulate it. Governments could make many dimensions of private life more visible through additional surveillance, data collection or classification, but greater legibility does not establish democratic authority. Administrative legibility is a condition that can support governing capacity; it is not a substitute for legitimacy.
The same applies to individuals and communities. Being represented by an institution does not necessarily mean being represented fairly. Categories can misdescribe the people they contain, obscure differences that matter or impose identities that those affected do not recognise. A population can therefore become highly legible administratively while remaining poorly understood substantively. Representability creates the possibility of institutional attention; it does not guarantee the quality of the representation.
This is why institutions need more than mechanisms for making reality visible. They need ways of questioning whether the representations through which visibility is produced are adequate to the problems they face. Frontline experience, citizen testimony, interdisciplinary analysis and external evidence can all reveal phenomena that formal systems struggle to represent. Their value lies partly in showing where institutional categories and models fail to provide sufficient cognitive access to what is happening.
Artificial intelligence can alter this boundary by making previously difficult patterns more representable. Systems capable of analysing large quantities of text, images or complex multidimensional data may reveal relationships that traditional administrative categories could not easily capture. This can expand institutional visibility. Yet it does not eliminate the representational problem. AI systems themselves operate through representations: data structures, labels, features, objectives and learned relationships determine what becomes visible to the model. A phenomenon that is absent or badly encoded in those structures can remain difficult to govern even when analytical capacity increases dramatically.
Indeed, new analytical capabilities may make representational judgement more important. Institutions must decide which newly visible patterns deserve attention, which reflect artefacts of data or classification, and which can legitimately inform action. The ability to detect something does not establish what it means or what government should do about it. Expanded visibility creates possibilities for governance; it does not settle the governing decision.
The deeper lesson is that institutional action depends on cognitive access. A problem that cannot enter the institution’s representations has difficulty entering its routines of comparison, prioritisation, responsibility and intervention. It may remain visible to individuals while invisible to systems, repeatedly encountered while never accumulated, widely experienced while administratively peripheral. The institution can be surrounded by evidence of a phenomenon without possessing the representational architecture necessary to treat it as a governable object.
This is not an argument for representing everything. No institution can or should make the whole of social reality administratively legible. The same selective simplification that makes institutional models useful also creates boundaries around institutional visibility. Some exclusion is necessary; some is legitimate; some is inconsequential. The cognitive task is to recognise when an exclusion has become consequential because something important to the institution’s purpose remains persistently outside the forms through which it sees.
A capable institution therefore needs to be attentive not only to what its representations contain but also to what they systematically struggle to contain. Recurring exceptions, persistent cross-boundary problems, citizen experiences that do not correspond to administrative categories and phenomena repeatedly described as unusual may all indicate a representational gap. None proves that the institution’s model is wrong. They are reasons to ask whether the current architecture of visibility is sufficient for the reality the institution is trying to govern.
What an institution cannot represent, it struggles to govern because sustained institutional action requires some way of making a phenomenon cognitively and administratively present. Representation does not have to mean measurement, and administrative legibility does not create legitimacy, accuracy or authority by itself. But when important features of reality cannot enter the categories, models, narratives or other structures through which an institution organises attention, they also struggle to enter its decisions. Institutional intelligence therefore depends partly on recognising the boundaries of institutional visibility and noticing when reality is producing consequential phenomena that existing representations cannot adequately hold.