When different departments reach different conclusions, one of the first explanations institutions tend to consider is that they are working with different information. Perhaps one team has access to data that another has not seen, records have not been shared, reporting systems are disconnected or important evidence has remained trapped within an organisational boundary. These problems are real, and improving information flows can often reduce them. But institutional disagreement has another source that is harder to solve through data sharing: two departments can receive exactly the same information and nevertheless turn it into different knowledge.
The distinction matters because data does not arrive inside an institution as a finished instruction for action. A set of figures, a case record, a forecast or a research finding becomes institutionally useful only when it is connected to questions about significance, responsibility and consequence. Different departments ask different questions because they exist for different purposes. They possess different mandates, manage different risks, work over different time horizons and are accountable for different outcomes. The informational input may therefore remain constant while the institutional knowledge produced from it changes as it crosses an organisational boundary.
Imagine that several departments receive the same forecast showing a substantial increase in the number of older residents in a region over the next fifteen years. A health department may read the forecast as evidence about future demand for chronic-care services and workforce requirements. A transport authority may see implications for accessibility and mobility infrastructure. A housing department may interpret it as evidence about the suitability and location of future housing stock. A finance ministry may focus on expenditure trajectories and the changing relationship between revenue and public-service demand. None has received different demographic data. Yet the same figures become different knowledge because each department connects them to a different field of institutional responsibility.
This does not mean that interpretation is arbitrary. The departments cannot legitimately make the data mean anything they want. Their conclusions remain constrained by the quality of the evidence, the relationships the evidence can support and the wider body of relevant knowledge. But SAME DATA ≠ SAME KNOWLEDGE, because institutional knowledge includes more than the informational object itself. It also includes an understanding of what that information implies for a particular capacity to decide and act.
This distinction becomes especially important when institutions try to improve coordination by building common data infrastructure. Shared platforms can solve significant problems. They can reduce duplication, make records accessible across organisational boundaries and ensure that different units are no longer reasoning from incompatible factual baselines. Yet the creation of a shared informational environment does not automatically create a shared cognitive environment. Departments may look at the same dashboard, query the same database and cite the same evidence while still understanding the institutional significance of what they see differently.
In some circumstances, that difference is exactly what the institution needs. Specialisation exists partly because complex realities benefit from several forms of attention. A public-health specialist should notice implications in epidemiological data that a procurement specialist may not. A lawyer should identify legal consequences that are not immediately visible to an economist. A cybersecurity team and a service-delivery team can examine the same incident and produce different forms of knowledge because each possesses expertise that makes different properties of the event institutionally meaningful. Cognitive differentiation can therefore increase institutional intelligence rather than diminish it.
The problem begins when the institution cannot recognise or integrate the different knowledge produced from the common input. One department’s interpretation may circulate as though it were simply “what the data says”, while another department’s interpretation appears to contradict the evidence. The disagreement is then misdiagnosed as factual when it may actually concern institutional meaning. Both parties return to the same figures, confirm that the numbers are correct and remain puzzled about why agreement does not follow.
Mandates are one reason for this divergence. A regulator may interpret evidence primarily through questions of compliance and systemic risk, while a delivery agency interprets the same evidence through operational feasibility. Neither perspective is merely an opinion layered onto neutral facts. Mandates create legitimate structures of relevance. They determine which consequences an institutional actor is responsible for noticing and which questions it is required to ask.
Time horizons can produce similar differences. Evidence that appears reassuring to a unit responsible for maintaining service continuity over the next six months may appear deeply concerning to a strategic unit considering conditions ten years ahead. Both departments can agree completely about present facts while converting those facts into different knowledge because the decisions for which they are responsible occupy different temporal frames.
Scale matters as well. A national department may interpret an average improvement as evidence that a policy is working overall, while a local authority examining exactly the same dataset may focus on a subgroup or geographical area where outcomes have deteriorated. The national conclusion and the local conclusion can both be evidentially defensible because they answer different questions about the same distribution. What looks like disagreement about reality may partly be disagreement about the level at which institutional significance should be assessed.
Professional standards add another layer. Different communities may require different thresholds before evidence becomes actionable. A research unit might describe an emerging relationship as promising but uncertain, while an operational team facing an immediate decision may regard the same evidence as sufficient to justify precautionary action. A legal team may require a different evidential threshold again before concluding that a particular intervention is defensible. The evidence has not changed as it moves between them; its relationship to institutional action has.
This is why sending more information across a boundary does not necessarily resolve cognitive fragmentation. If two departments already possess the same data, duplicating the transfer cannot reconcile the knowledge they have constructed from it. What is missing is not information transmission but cognitive interoperability: the ability to understand how another part of the institution has transformed information into significance.
That requires more than asking each department for its conclusion. Conclusions hide much of the reasoning that produced them. Institutions need ways of making visible the questions different units asked of the evidence, the assumptions through which they interpreted it, the consequences they considered important and the decision context in which the resulting knowledge became useful. Once those elements are visible, apparent contradiction can sometimes be decomposed into complementary perspectives.
Consider a policy evaluation showing a modest average positive effect but substantial variation among groups. A programme team may conclude that the intervention is effective because the overall effect is positive. An equality unit may conclude that the programme has a serious problem because some groups benefit much less than others. A finance team may conclude that expansion is not justified because the effect is too small relative to cost. These statements can sound mutually incompatible, yet they may all be legitimate transformations of the same evaluation into knowledge relevant to different institutional responsibilities. The coordinating task is not to decide which department has read the data “correctly” before anything else can happen. It is to construct a sufficiently integrated account of what the evidence means across the several decisions the institution must make.
This also clarifies why semantic alignment alone cannot solve the problem. Departments may agree perfectly on the meaning of terms such as effectiveness, cost and inequality and still produce different knowledge from the same evidence. Shared language reduces one barrier to interoperability, but it does not eliminate differences created by mandate, scale, time horizon or decision responsibility. Semantic compatibility makes translation easier; it does not make specialised cognition identical.
Artificial intelligence creates a useful contemporary illustration. Several departments may receive the same output from an analytical model but use it differently. A fraud team may treat a risk score as a trigger for investigation, a service team as information requiring contextual review, and a policy unit as evidence about population-level patterns. Even when the model output is identical, the knowledge produced around it depends on the institutional function into which it enters. Building one common AI system therefore does not automatically harmonise institutional cognition any more than building one common database does.
The objective should not be to eliminate these differences. An institution in which every department transformed information into identical knowledge would probably have sacrificed much of the value of specialisation. The deeper capability is to preserve differentiated expertise while making its cognitive products mutually intelligible. That means recognising when apparently conflicting conclusions originate in different evidence and when they originate in different legitimate transformations of common evidence.
The distinction can substantially improve coordination. If disagreement is factual, the institution may need better data, verification or evidence. If it is semantic, it may need translation between concepts. If it arises because the same information has become different knowledge, the institution must examine the purposes, frames and decision contexts through which that transformation occurred. Treating all three problems as failures of information sharing leads to solutions that may increase the volume of data moving through the institution without increasing its ability to think across boundaries.
Shared data is therefore an important institutional resource, but it is not the endpoint of cognitive integration. Information can cross an organisational boundary much more easily than meaning, relevance or judgement. The fact that everyone can access the same evidence tells us that a transmission problem may have been solved. It does not tell us that the institution has become capable of integrating what different parts legitimately learn from that evidence.
The same data can become different knowledge in different departments because institutional knowledge is produced at the intersection of information and purpose. Mandates, responsibilities, professional frames, scales and time horizons influence which implications become salient and what forms of action the evidence can support. This differentiation is not inherently a failure; it is one of the benefits of institutional specialisation. The failure appears when the institution assumes that shared information must generate shared knowledge, and therefore lacks the mechanisms needed to understand and integrate the different meanings that its own specialised parts construct from the same world.