Computational Understanding

Information gives us observations about an entity. Understanding requires us to determine what those observations mean together. A system may have access to every relevant document and still fail to recognize an important relationship between them, distinguish a temporary signal from a meaningful change, or notice that several independent observations support the same conclusion.

For the Assumptive Understanding Model (AUM), understanding is the maintained interpretation of what the available evidence currently supports about an entity, including what has been inferred, what has been assumed, and what remains uncertain. Representing that interpretation in a form that can be inspected, revised, and computed over is what makes understanding computational.

Three concepts are useful here. Evidence is information that can support a claim. Inference is the reasoning that produces a new proposition from that evidence. An inference is something we perform; its result can become part of the maintained understanding together with its evidence, uncertainty, and dependencies on other propositions.

The information available to a system never provides a complete view of a company. Public information covers only part of its activity. Internal information may reveal more, but it is still partial. Some information is outdated, some is ambiguous, some conflicts with other information, and some relevant information is unavailable altogether.

Understanding therefore starts from an incomplete view. The goal is not to reconstruct a complete and certain version of the company, because that version is usually not available. The goal is to maintain a current interpretation that is supported by the available evidence while preserving the limits of that interpretation.

Incomplete information first creates uncertainty. Sometimes the right outcome is simply that a question remains unresolved. But useful reasoning can also require a proposition that the available evidence does not establish. When that happens, an assumption enters the reasoning.

An assumption is not a weak observation, and it is not a fact with lower confidence. It is a proposition provisionally accepted within a specific line of reasoning even though the available evidence does not establish it. Reasoning about a partially observed system may therefore require assumptions, but those assumptions must remain identifiable.

The presence of assumptions is not itself the problem. In many situations, useful reasoning requires us to work with propositions that the evidence has not yet established. The problem begins when an assumption is presented in the same way as something that was directly observed, or when an inferred conclusion loses its connection to the evidence that produced it.

AUM is designed to keep those distinctions intact. Evidence remains identifiable as evidence. An inference retains the basis from which it was derived. An assumption remains visible as something the reasoning depends on but the evidence has not established. Uncertainty can remain unresolved when the available information does not justify a stronger conclusion.

AUM does not try to remove assumptions from reasoning. It tries to prevent assumptions from disappearing inside the answer. The objective is not certainty where certainty is unavailable. It is clarity about what is known, what has been inferred, what has been assumed, and what remains open.

Understanding is also not fixed. Companies change, evidence changes, and new information can change the interpretation of information that was already available.

A conclusion that was well supported at one point may later become weaker, historical, or no longer relevant to the current state. New evidence may reinforce an existing proposition, contradict it, or require other propositions that depend on it to be reconsidered. A system that only stores information can preserve both the old and the new. A system that maintains understanding also has to update what those pieces of information mean together now.

This gives understanding continuity without permanence. Earlier reasoning does not need to disappear simply because the state has changed. It can remain part of the history while the current interpretation is revised. What matters is that the model can distinguish between what was once supported and what is supported now.

We use computational understanding to describe the structured computational representation and maintenance of this state. The word computational matters because the understanding is not only stored or displayed. Its evidence, inferences, assumptions, uncertainty, temporal scope, and dependencies are represented so that the system can operate on them.

When evidence changes, an assumption is replaced, or a proposition loses support, those dependencies identify what is affected. The relevant reasoning can then be reconsidered and its conclusions or outputs recomputed. This allows a user not only to inspect an interpretation, but to change part of its basis and see what follows.

The term is intentionally narrower than human understanding. It makes no claim about consciousness or subjective comprehension. It describes a computational state that can be maintained, inspected, changed, and recomputed as its evidence and assumptions change.

The Assumptive Understanding Model, or AUM, is ASUME's framework for maintaining this form of computational understanding. By preserving the relationships among evidence, inferences, assumptions, uncertainty, and conclusions, AUM makes an interpretation something that can persist, be challenged, change, and be computed over rather than disappearing into a single answer.