Large institutions rarely suffer from a complete absence of knowledge. More often, they possess many kinds of knowledge that have difficulty working together. Specialists understand different parts of the problem, departments interpret common evidence through different mandates, professional communities use concepts in different ways, and operational units accumulate forms of experience that may be difficult to express in the language of policy or analysis. The institution can therefore contain considerable intelligence without being able to combine it when a decision crosses the boundaries through which that intelligence has been organised.
The intuitive response is often to increase connection. Share more data. Build common platforms. Create cross-departmental meetings. Standardise terminology. Centralise information. Each of these measures can be useful, and each can solve genuine institutional problems. Yet none guarantees that knowledge produced in one part of an institution will become cognitively usable in another. Information can move without meaning moving with it, people can use the same words while attaching different concepts to them, and departments can consult the same evidence while drawing different implications from it. Connection is therefore necessary in many circumstances, but connection alone is not interoperability.
Cognitive interoperability describes a stronger capability: the ability of differentiated forms of institutional knowledge to remain distinct while becoming mutually usable across boundaries. It does not require every department to adopt the same conceptual framework or every profession to interpret evidence identically. On the contrary, its purpose is to preserve the benefits of specialised cognition while preventing specialisation from turning into cognitive isolation.
This distinction matters because the easiest way to create apparent integration is to remove difference. An institution can impose one vocabulary, one reporting format, one data model or one analytical framework across all of its units. Standardisation may improve efficiency where differences serve no useful purpose, and some common structures are essential for coordination. But cognitive integration cannot be measured simply by the amount of variation eliminated. If a common framework strips away distinctions that specialists need in order to understand their domain, the institution may become easier to connect administratively while becoming less capable cognitively.
Interoperability therefore begins with translation rather than uniformity. Translation allows knowledge constructed within one professional or organisational context to become intelligible in another without pretending that the two contexts are identical. This may require explaining concepts, assumptions, evidential standards or the consequences attached to particular findings. The objective is not to replace specialised knowledge with generic language, but to create enough semantic access for another part of the institution to understand what the knowledge means and why it matters.
Effective translation also requires context preservation. Knowledge can lose much of its significance when detached from the conditions under which it was produced. A risk assessment may depend on assumptions about time, population or operating conditions. A professional judgement may be meaningful because of circumstances that cannot be fully represented in a summary score. An evaluation result may apply to one implementation environment but not another. When knowledge crosses an institutional boundary, interoperability requires some way of preserving the contextual information necessary to understand the limits of what is being transferred.
This is one reason why MORE DATA SHARING ≠ BETTER COORDINATION. Giving every department access to every available dataset can improve informational availability while doing little to improve the institution’s ability to interpret what those data mean across different decision environments. In some cases, greater information volume can increase the integration burden. More material circulates, more dashboards become accessible and more reports enter decision processes, while the cognitive work required to connect them remains unresolved.
The problem is not data sharing itself. Information that cannot be accessed cannot easily participate in institutional reasoning, and unnecessary informational barriers can seriously weaken public organisations. The mistake is to treat access as though it completed the cognitive task. DATA SHARING ≠ KNOWLEDGE INTEGRATION. Access creates the possibility of use; interoperability concerns whether meaningful use across boundaries can actually occur.
Comparability is another component of the capability. When two units reach different conclusions, an interoperable institution needs to determine what kind of difference it is observing. Are they using different evidence? Are they attaching different meanings to the same concepts? Are they interpreting the same information through different mandates? Are they answering different questions? Or do they genuinely disagree about what the evidence supports? Without this decomposition, institutional disagreement can remain cognitively opaque. Each side sees another conclusion, but the institution cannot identify the architecture that produced the difference.
Making cognitive products comparable does not imply that one must always be selected as correct. Two interpretations can be simultaneously useful because they expose different properties of a problem. A finance ministry and a social-policy department may analyse the same programme through different questions without either analysis becoming redundant. Cognitive interoperability allows the institution to understand how those perspectives relate, where they conflict and what each contributes to a decision that neither unit could adequately make alone.
The deeper test is whether knowledge produced in one part of the institution can become consequential elsewhere. Information may circulate widely while leaving other units’ reasoning unchanged because it arrives in a form they cannot incorporate. Genuine interoperability means that an insight generated within one specialised domain can alter assumptions, questions or decisions in another when the underlying problem requires it. Knowledge crosses the boundary not merely when it is received, but when it becomes capable of participating in cognition on the other side.
This does not mean that every piece of knowledge should influence every decision. Institutions need filters because indiscriminate integration would overwhelm decision processes and erode specialisation. The objective is selective cognitive permeability: relevant knowledge should be able to cross boundaries without requiring all boundaries to disappear.
The same principle applies to organisational structure. Centralisation can sometimes make integration easier by bringing authority and information together, but CENTRALISATION ≠ COGNITIVE INTEGRATION. A central unit can become its own cognitive silo, receiving reports from across an organisation while interpreting them through a single framework that loses important distinctions. Conversely, a decentralised institution can achieve strong cognitive interoperability if its interfaces allow specialised knowledge to move, translate and modify reasoning across organisational boundaries.
Nor does interoperability require consensus. An institution may successfully integrate several forms of knowledge and still face genuine disagreement about values, priorities or acceptable risk. In fact, better interoperability can reveal disagreements that were previously hidden behind semantic confusion or fragmented information. Once different perspectives become mutually intelligible, the institution may discover that the conflict is real rather than accidental. This is progress. INTEROPERABILITY ≠ CONSENSUS. Its purpose is to make disagreement cognitively tractable, not to make disagreement disappear.
Artificial intelligence could become important in this architecture because it can potentially translate among specialised vocabularies, synthesise large bodies of heterogeneous information and identify relationships that are difficult for individual units to observe. Used carefully, such systems might reduce some of the cognitive costs of institutional differentiation. But they could also create false interoperability if fluent synthesis conceals differences that matter. A system that compresses several professional perspectives into one coherent summary may appear to integrate them while silently removing assumptions, uncertainty or disagreement. The relevant test is therefore not whether AI produces a unified answer, but whether the resulting representation preserves the distinctions necessary for institutional judgement.
This points towards a broader principle. Integration without erasure is harder than integration through standardisation. It requires institutions to tolerate some cognitive plurality while building interfaces capable of making that plurality usable. The architecture may include shared concepts where commonality is valuable, translation where meanings differ, contextual metadata where knowledge depends on conditions, and deliberative mechanisms where several perspectives must jointly shape a decision. No single mechanism constitutes interoperability on its own. The capability lies in their combined ability to connect differentiated cognition.
The distinction is especially important for public institutions because the problems they govern rarely correspond neatly to administrative divisions. Climate adaptation involves infrastructure, health, finance, land use, social vulnerability and emergency management. Artificial intelligence governance involves technology, law, procurement, organisational design, ethics and public accountability. Demographic change touches healthcare, pensions, labour markets, housing and territorial planning. No single professional language or department can absorb these problems without losing something important.
An intelligent response is therefore not to abolish institutional differentiation but to design for cognition across it. Departments should remain capable of developing deep specialised knowledge, but that knowledge should not become trapped inside the architecture that produced it. The institution needs interfaces through which specialised insights can be translated, compared and combined when the problem requires a wider field of understanding.
This also changes how institutional integration should be evaluated. The relevant question is not simply whether systems are connected, meetings occur or information is shared. It is whether differentiated knowledge can become mutually consequential. Can one department understand why another interprets the evidence differently? Can professional assumptions be made visible across a boundary? Can conflicting conclusions be decomposed into factual, semantic, methodological and value differences? Can an insight generated locally alter a wider institutional judgement without first being stripped of the context that made it valid?
Where the answer is yes, the institution possesses something more valuable than information connectivity. It possesses the beginnings of a cognitive architecture capable of thinking across its own internal differentiation.
Cognitive interoperability is the capability that allows different forms of institutional knowledge to remain different without remaining isolated. It does not require uniform vocabulary, identical interpretations, centralised cognition or universal consensus. It requires interfaces through which knowledge can cross boundaries while preserving enough meaning, context and distinction to become usable elsewhere. More data sharing can support this capability, but it cannot substitute for it. An institution becomes cognitively integrated not when all of its parts think alike, but when what one part knows can enter, challenge and improve the reasoning of another without either part first having to stop being specialised.