Adding Meaning Without Overwriting the Ground Truth

Utopia’s graph becomes more useful when it can express not only what a document states, but also what a human-authored criterion concludes from several dated readings. The challenge is to add that meaning without turning a derived judgment into an asserted fact, or attaching an edge-level detail to the wrong entity. This chapter follows Utopia’s business rules, computed conclusions, and relation qualifiers to show how premises, intervals, and uncertainty remain visible as the model grows more expressive.

Rules add a conclusion, not a replacement for evidence

Consider a report about a well. The report already contains a hydrocarbon reading and an interpretation category. A person may define a criterion under which the well is “gas-bearing” when the reading exceeds a threshold and the category indicates a gas anomaly. That conclusion is valuable, but it is not the same kind of thing as the reading. The reading is an asserted attribute backed by a source passage. “Gas-bearing” is a classification derived from a rule written by a person.

Utopia keeps that distinction in its data model. A rule can conclude a derived type or a derived attribute, but it does not write into the entity’s asserted type. The well can remain asserted as a Well while a derived GasBearingWell overlay appears when the rule’s conditions hold. If a premise is retracted, corrected, or no longer holds in the relevant interval, the overlay disappears on recomputation. The asserted identity does not need to be reconstructed from the rule that once classified it.

This is the same boundary as the earlier separation between asserted and derived graph edges, applied to literal values. It prevents a business policy from silently rewriting ground truth. A rule can express an organization’s interpretation—“classify this entity when these readings meet these conditions”—without claiming that the source itself named the classification. The difference matters when a person changes the threshold, retires the rule, or asks why an entity carried the label for only part of its history.

The rule’s premises are structured conditions over an entity’s own attribute facts. They can compare a value, test membership in a set, or require a value to be present. The evaluator retains the readings that satisfied the conditions, intersects their validity intervals, and attaches those premises to the conclusion. If a well has a qualifying reading in 2023 but not in 2025, the classification is not one permanent row that flips back and forth. It is derived over the period supported by the evidence. A later run can retire the old conclusion when the premises no longer produce it.

The restriction to structured conditions is central. A rule is not stored as a paragraph that a model interprets differently on every run. Its subject class, conditions, operators, and conclusion are explicit. The ontology workbench can show what the rule says, and the evaluator can execute that same structure. The model may help a person inspect or use the result, but it does not silently invent the criterion.

Utopia allows one carefully bounded form of disjunction. Conditions in the same group join with and; separate groups join with or. A well can therefore qualify through “high hydrocarbon and gas anomaly” or through a separate condition that already identifies a gas-bearing interpretation. The evaluator runs each group separately, so the proof for a fired conclusion names the readings in the group that actually made it true.