Hierarchies allow institutions to organise authority across large numbers of people. Decisions can be escalated, responsibilities assigned and complex operations coordinated through levels of management. But the knowledge moving through those same levels behaves differently from authority. A decision right can be transferred or concentrated relatively cleanly; rich knowledge cannot. As information moves upward through an institution, it often becomes more usable for decision-making by becoming less complete.
This is not necessarily a failure. Senior decision-makers cannot absorb every observation, document, uncertainty and contextual detail generated across an organisation. Knowledge has to be compressed. Reports summarise cases, dashboards aggregate activity, briefings select relevant findings and managers convert local complexity into forms that higher levels can process.
Compression is therefore functionally necessary. But necessary compression is not cognitively neutral.
Every act of simplification determines what survives and what disappears. A report may preserve the headline result while dropping uncertainty. A dashboard may preserve the average while hiding variation. A briefing may communicate the recommended action while omitting the disagreements that surrounded it. A manager may transmit the conclusion of a technical discussion without the assumptions on which the conclusion depends.
None of these transformations requires dishonesty. The information can become less complete through entirely legitimate attempts to make it manageable. This is what makes hierarchical compression difficult to diagnose. When information is deliberately concealed, the governance problem is relatively obvious. Ordinary compression is subtler because everyone involved may be acting responsibly. Each layer receives something complex, identifies what appears important and passes a more compact representation upward.
After several layers, however, the final representation can be substantially different from the knowledge available at the source. The effect is particularly strong for uncertainty. Frontline staff often encounter ambiguous cases, conflicting signals and contextual exceptions. As information moves upward, ambiguity can be difficult to preserve because senior systems frequently demand clearer categories: on track or off track, high risk or low risk, approve or reject.
The institution may therefore become more certain as information travels, even though no new evidence has justified the increase in certainty. Local context is similarly vulnerable. A service manager may know why a performance indicator changed in one location. Once aggregated nationally, the contextual explanation may disappear while the numerical change remains. Senior officials then see a cleaner pattern than the organisation actually experiences.
COMPRESSION ≠ UNDERSTANDING.
This distinction also separates hierarchical transmission from briefing architecture. A final briefing is itself a cognitively consequential representation, but much of the transformation may have occurred before anyone constructs that brief. Knowledge may already have passed through databases, reporting templates, managerial summaries and departmental interfaces.
By the time a formal decision-maker receives it, the cognitive content may have been compressed repeatedly. Digital systems can intensify this process while making transmission more efficient. Dashboards allow leaders to access organisational information instantly, but what appears on the screen has already been selected, categorised and aggregated. The interface can create a powerful sense of direct visibility while actually presenting a highly compressed representation.
Artificial intelligence introduces another layer. Generative systems can summarise hundreds of pages into a short briefing or extract key points from thousands of records. This can be enormously useful where human attention is scarce. But the improvement in accessibility increases the importance of understanding what compression does.
A good AI summary is not simply one that is factually correct. It should preserve the uncertainty, exceptions and qualifications necessary for responsible judgement. The question is therefore not whether institutions should compress knowledge. They have no realistic alternative. The question is what forms of information are systematically vulnerable during compression and whether those losses matter for the decision being made.
Some detail can disappear safely. Other detail is precisely where risk resides. Institutions can design for this distinction. Briefings can separate facts from interpretations, preserve confidence levels, flag significant exceptions and allow decision-makers to inspect underlying material when necessary. Escalation systems can identify information that should not be aggregated away. Dashboards can display distributions alongside averages.
The objective is not maximal detail. Flooding senior officials with raw information merely replaces compression risk with attention overload. Good transmission architecture preserves the epistemic structure needed for judgement while discarding detail that does not materially affect it. This also requires feedback. People at the source should sometimes be able to see how their knowledge appears after travelling upward. They may discover that something they regarded as essential has disappeared, while senior officials may learn that apparently simple information rests on considerable uncertainty.
The institution becomes cognitively stronger when compression itself becomes observable. Hierarchy solves an authority problem remarkably well: it identifies where decisions can be made. It does not automatically solve the different problem of moving knowledge to those locations without changing its meaning.
The most dangerous loss may therefore occur without anyone deliberately withholding anything. Knowledge can arrive at the top on time, in the correct format and through the proper channels while important parts of what was known at the bottom have quietly disappeared along the way.