Governments rarely confront complex public problems using knowledge generated entirely within the public sector. Companies understand technologies and markets that regulators must oversee; universities and research organisations develop specialised expertise; professional communities observe emerging practices; infrastructure operators possess operational knowledge of systems on which public life depends; and civil-society organisations encounter social conditions that may remain difficult to see through administrative data alone. Recognising this distribution of knowledge is relatively straightforward. The more difficult question is what happens when these actors do more than supply information to government and begin to participate in sustained relationships through which knowledge is interpreted, tested and revised.

At that point, a public–private network can become more than a collection of organisations exchanging information. It can begin to function as a cognitive system.

This does not happen merely because public and private actors are connected. A ministry that requests a technical report from a company has accessed external knowledge, but the relationship need not constitute collective cognition. A regulator that consults industry representatives has broadened its informational inputs, but consultation alone does not create a network-level capability. Similarly, a public agency can purchase expertise, commission research or invite external specialists to a meeting while retaining a fundamentally linear cognitive architecture: knowledge moves from an external source into government, where the actual interpretation and reasoning occur.

A cognitive network requires something more reciprocal. Different actors must be able to contribute knowledge that changes how other participants understand the problem, while the resulting interpretations can in turn alter what the network observes, investigates or questions. Information moves across organisational boundaries, but so do problems, interpretations and feedback. EXTERNAL KNOWLEDGE INPUT ≠ CROSS-SECTOR COGNITIVE SYSTEM.

Imagine a public authority trying to understand the emergence of a new technological risk. Companies may possess detailed knowledge of system design and deployment. Researchers may understand technical limitations that are not yet visible in operational practice. Public agencies may hold evidence about social consequences, complaints or regulatory failures. Independent organisations may observe effects on particular communities. None of these perspectives is sufficient on its own. More importantly, simply transmitting them to a central government team may still leave important relationships among them unexplored.

A different capability appears when these actors can interrogate one another’s observations. Operational evidence can change the questions researchers ask; research findings can change what regulators monitor; regulatory concerns can cause firms to reveal or investigate system properties that previously seemed unimportant; experiences reported by affected groups can expose discrepancies between formal performance measures and actual consequences. The network becomes cognitively significant when knowledge does not merely travel towards a decision-maker but circulates through relationships capable of modifying the production and interpretation of further knowledge.

This is why INFORMATION SHARING ≠ NETWORK COGNITION. Information sharing can be one component of a cognitive network, but the defining capability lies in what relationships allow participants to do with different forms of knowledge. They may compare claims generated under different assumptions, identify contradictions between operational and regulatory representations, recognise that different organisations are using the same term to describe different phenomena, or discover that a problem previously assigned to one organisation is actually produced by interactions among several.

The resulting cognition can have a network-level character because no participant controls the entire process from its own organisational position. The public authority may retain the legal responsibility for deciding what should be done, yet its understanding of the problem can emerge through interactions with actors that remain institutionally autonomous. A company does not become part of the state because its technical knowledge changes a regulator’s interpretation. A university does not acquire governmental authority because its research changes public policy. A civil-society organisation does not need formal decision rights for its observations to become cognitively consequential.

This distinction is essential:

COGNITIVE CONTRIBUTION ≠ DECISION RIGHT.

Public–private cognition therefore does not require public–private government. The network can participate in sensing, interpretation, hypothesis formation, error detection or knowledge revision while legitimate public institutions retain responsibility for decisions involving public authority. Indeed, preserving this distinction can make the cognitive architecture stronger because actors contribute from genuinely different institutional positions rather than being absorbed into a fictitious unified decision-maker.

Those differences include legitimacy. Public institutions derive authority from legal and democratic arrangements that private organisations do not share simply because they possess expertise. Businesses may control important infrastructure or technical systems without acquiring the right to determine public values. Researchers may understand a problem exceptionally well without being accountable for balancing competing social priorities. Civil-society organisations may represent important interests without representing society as a whole. A cognitively integrated network can therefore remain politically differentiated.

INTERDEPENDENCE ≠ EQUAL LEGITIMACY.

This matters because discussions of collaborative governance sometimes blur knowledge and authority. If public institutions depend on external expertise, it can appear that authority must follow expertise; alternatively, fear of private influence can encourage governments to keep external actors cognitively distant even when their knowledge is indispensable. Both responses treat cognition and authority as though they had to occupy the same institutional boundary.

They do not. Government can retain authority while participating in a cognitive architecture larger than government itself.

The challenge is to design relationships that permit knowledge to become consequential without allowing epistemic dependence to become unaccountable political dependence. This requires clarity about who contributes what, how claims can be questioned, where conflicts of interest exist, which forms of evidence remain independently verifiable and where final responsibility lies. Cognitive openness does not remove the need for institutional boundaries; it makes understanding those boundaries more important.

Feedback is particularly important. A network becomes more cognitively capable when participants can observe the consequences of interpretations and adjust accordingly. If a regulator issues guidance based partly on external technical knowledge, subsequent operational evidence can reveal whether the assumptions behind that guidance were correct. If firms alter practices, affected communities may observe consequences that neither firms nor regulators anticipated. If researchers then investigate those consequences, new evidence can return to public decision processes. The network is no longer functioning as a one-directional consultation mechanism. It is participating in a continuing cycle of observation and revision.

Yet COORDINATION ≠ COLLECTIVE COGNITION. Organisations can coordinate successfully while continuing to interpret reality separately. They may align timetables, divide responsibilities or share resources without developing any joint capacity to understand a problem. Cognitive integration occurs when interaction changes what participants can perceive or explain, not merely what they agree to do.

Nor does network cognition require consensus. Public and private actors may have fundamentally different interests and may interpret the same evidence differently. In some circumstances, preserving disagreement is cognitively valuable because it exposes assumptions that would disappear inside an artificially unified account. The network becomes more capable when disagreement can be located and examined rather than suppressed in the name of collaboration.

This also explains why trust and contestability must coexist. Without sufficient trust, actors may withhold uncertainty, emerging risks or information that could expose failure. Without sufficient contestability, trusted relationships may harden into shared assumptions that nobody challenges. A cognitively useful network therefore needs relationships strong enough for consequential knowledge to travel but open enough for claims to remain examinable.

Artificial intelligence can intensify both possibilities. Public and private actors may increasingly depend on shared data infrastructures, models or analytical systems that connect knowledge across organisational boundaries. These technologies can help identify patterns no participant could observe independently, but they can also make a private technical architecture an invisible mediator of public cognition. If the assumptions embedded in that infrastructure cannot be examined by public institutions, apparent cognitive integration may conceal a transfer of epistemic control.

The relevant question is therefore not simply whether government uses external AI expertise, but whether the wider cognitive architecture preserves the ability to question, compare and revise the knowledge on which public judgement depends. Technical dependence should not quietly become cognitive dependence without institutional visibility.

Public–private networks can be particularly valuable when the object being governed is itself distributed. Financial systems, digital infrastructures, supply chains, energy networks and technological ecosystems do not exist inside a single institutional jurisdiction or knowledge base. Understanding them may require knowledge held by actors positioned throughout the system. In such circumstances, attempting to recreate all relevant knowledge inside government may be less realistic than building relationships through which distributed knowledge can become available to public reasoning.

But the purpose is not to dissolve organisational boundaries. On the contrary, those boundaries help explain why different knowledge exists in the first place. Companies, regulators, researchers and social organisations see different things partly because they occupy different positions, pursue different functions and remain accountable in different ways. A cognitively capable network connects those differences without pretending they have disappeared.

The resulting architecture can therefore be collective cognitively while remaining differentiated institutionally. Participants can contribute to a shared process of understanding without sharing identical responsibilities, authority or legitimacy. Public institutions can learn through the network without surrendering the obligation to judge what should ultimately be done.

A public–private network becomes a cognitive system when relationships among autonomous actors allow distributed knowledge to be compared, interpreted, challenged and revised in ways that none of those actors could accomplish from its own institutional position. The cognition can be shared without the authority becoming shared. That distinction is fundamental: complex governance may require knowledge to cross institutional boundaries, but public legitimacy, accountability and decision rights do not automatically cross with it.