Two people can look at the same evidence and reach different conclusions without either of them having ignored the evidence. Inside institutions, this happens constantly. A rise in expenditure can be interpreted as evidence of growing demand, inefficient delivery or successful expansion. A decline in complaints can suggest improved service, reduced accessibility to complaint mechanisms or declining public expectations. A programme that misses its target may appear to one team as an implementation problem and to another as evidence that the target itself no longer represents the problem accurately. The information can be shared while the meaning attached to it remains different.
This possibility becomes easier to understand once we recognise that interpretation is not contained inside information itself. Observations acquire significance in relation to context, but context is not the only thing that shapes meaning. Institutions also approach evidence through assumptions about what matters, professional frameworks for understanding problems, previous experience, responsibilities, expectations about risk and different conceptions of what successful action should look like. These frames do not necessarily distort reality. They are among the tools through which complex reality becomes intelligible. Yet because different actors can use different frames, access to the same information does not guarantee a shared interpretation.
Consider a public programme whose costs have increased while the number of people served has remained stable. A finance team may interpret the pattern primarily through efficiency, asking why expenditure per case has risen. Frontline staff may understand it through complexity, observing that cases now require substantially more time and expertise. A policy team may focus on population change and argue that the programme is serving people with different needs from those for whom it was originally designed. All three groups can be working from the same administrative evidence. Their disagreement may arise not from unequal access to facts but from the different questions through which those facts become meaningful.
Professional specialisation makes this particularly common. Institutions need specialists precisely because complex problems cannot be understood through a single body of expertise. Lawyers are trained to notice different properties of a situation from economists, engineers, clinicians, auditors or service designers. Those differences increase institutional cognitive capacity because each perspective makes certain features of reality easier to see. At the same time, specialisation creates interpretive plurality. The same evidence enters different conceptual vocabularies, and those vocabularies can organise its significance differently.
Responsibilities matter as well. Institutional actors do not interpret evidence from nowhere; they occupy positions with particular obligations. A regulator may be especially attentive to systemic risk, an operational manager to feasibility, a finance ministry to fiscal exposure and a local service provider to individual consequences. These responsibilities influence which implications of the evidence appear most consequential. The difference need not indicate bias in the crude sense. It may reflect the fact that the institution has deliberately distributed responsibility for different dimensions of a problem.
Temporal horizons can create another source of divergence. Evidence that appears positive over the next six months may look troubling over ten years. A programme that produces immediate savings may create future liabilities, while an intervention with high initial costs may generate long-term benefits. Actors responsible for annual delivery and those responsible for long-term resilience can therefore interpret the same performance information differently because they are effectively asking what the evidence means over different periods.
Previous experience also enters interpretation. Institutions possess histories, and those histories affect what they recognise in new information. A small anomaly may appear insignificant to a team that has never seen it precede a serious failure, while another unit with different experience may recognise it as an early warning. Past crises, reforms, successes and failures create expectations about what particular signals can mean. Institutional memory can therefore enrich interpretation, but it can also make different parts of an organisation read the present through different pasts.
None of this implies that interpretation is merely subjective. If different frames can produce different meanings, it does not follow that every meaning is equally defensible. An interpretation can ignore relevant evidence, rely on assumptions contradicted by experience, use an inappropriate comparison or fail to explain important observations. Interpretations can be challenged. They can be compared against new evidence, examined for internal consistency, tested against plausible mechanisms and evaluated according to whether they account for the full pattern better than alternatives.
The distinction between interpretive plurality and epistemic relativism is therefore crucial. Institutions should not attempt to eliminate disagreement simply because disagreement is uncomfortable, but neither should they treat every disagreement as permanently irresolvable. Different interpretations can function as competing hypotheses about what the evidence means. Making those interpretations explicit allows the institution to ask what assumptions separate them and what additional evidence might discriminate between them.
This is more cognitively productive than assuming that disagreement must result from incomplete information. When two teams reach different conclusions, the instinctive response is often to gather more data. Sometimes that is appropriate. But if both teams already possess the same relevant information, additional information may leave the disagreement untouched because its source lies in the frames through which the existing evidence is being understood. The institution first needs to discover what exactly the parties are interpreting differently.
A useful question is therefore not only “What does the evidence show?” but “What are we assuming when we say that this is what the evidence shows?” One interpretation may assume that current conditions are comparable with the historical baseline; another may assume that the population has changed. One may treat an observed relationship as stable; another may believe a structural break has occurred. One may interpret rising demand as evidence of worsening conditions, while another sees it as evidence of improved access. Once these assumptions become visible, disagreement becomes more analytically tractable.
This matters because hidden interpretive differences can otherwise masquerade as factual disputes. Meetings become arguments about numbers even when everyone accepts the numbers. Teams repeatedly exchange evidence that the other side has already seen. Reports grow longer without producing convergence. The institution experiences the disagreement as an information problem because the interpretive architecture underneath it remains invisible.
Making interpretation visible does not require institutions to force consensus. Some problems contain genuine uncertainty that supports more than one reasonable interpretation, and premature agreement can be cognitively damaging. What matters is that the institution knows where agreement ends. It may share the observations while disagreeing about significance; agree about significance while disagreeing about causes; or agree about the diagnosis while preferring different responses. Distinguishing these layers prevents one form of disagreement from being mistaken for another.
The same principle matters when quantitative analysis is involved. Statistical techniques can establish that a pattern exists with a particular degree of confidence, but the institutional meaning of the pattern can still depend on assumptions about mechanism, context and relevance. Analysts may agree entirely on the calculation and disagree about what the result implies for policy. Technical agreement at the level of measurement does not eliminate interpretive judgement at the level of institutional meaning.
Artificial intelligence can reproduce this distinction at greater scale. An AI system may make the same information available across an organisation, summarise competing sources or generate analyses using common datasets. This can reduce differences caused by unequal access to information. It does not necessarily remove differences in how people understand the resulting evidence. Indeed, when AI-generated outputs appear authoritative, institutions may need to become even more explicit about the assumptions through which those outputs acquire significance.
Interpretive diversity can also be valuable. When different perspectives illuminate different properties of a problem, their coexistence can prevent the institution from collapsing complexity too early. A legal interpretation may reveal rights implications that an efficiency analysis misses; operational knowledge may expose implementation constraints absent from a strategic model; citizen experience may challenge assumptions embedded in administrative indicators. The objective is not necessarily to replace several interpretations with one as quickly as possible, but to understand what each reveals and where they genuinely conflict.
The cognitive challenge is therefore to make interpretations inspectable. Institutions need ways to distinguish observations from the meanings attached to them, identify the assumptions that connect one to the other and compare competing interpretations without treating disagreement itself as failure. When those processes are absent, shared data can create an illusion of shared understanding. Everyone may be looking at the same dashboard while constructing a different institutional reality from what it shows.
This distinction changes the meaning of evidence-informed decision-making. Evidence does not simply arrive and determine the answer. It enters an interpretive environment populated by prior knowledge, institutional responsibilities, professional frameworks and expectations about how the world works. Good institutional cognition does not pretend these elements can be eliminated. It makes them sufficiently visible that interpretations can be compared, challenged and revised.
The same information can produce different interpretations because shared observation is not the same thing as shared meaning. Institutions contain different expertise, responsibilities, histories and temporal perspectives, all of which can shape what evidence appears to signify. Recognising this does not require accepting that every interpretation is equally valid. It requires something more demanding: the capacity to make interpretive frames visible enough that competing meanings can be examined against evidence rather than hidden behind it. An institution begins to understand disagreement more intelligently when it can tell the difference between people who possess different facts and people who possess the same facts but understand them differently.