Consistency Is Not Coherence: Orientation Search for Certified Alignments Between 4D Defence Upper Ontologies

In the high-stakes world of national defense and intelligence, data is the currency of safety. Governments and military alliances rely on massive digital systems to track people, events, and places, ensuring that a soldier in the field and a strategist in a command center are speaking the same language. To make these systems talk to one another, experts build digital dictionaries called "ontologies." Think of an ontology not as a simple list of words, but as a rigorous map of how a specific community understands reality. It defines what a "person" is, what an "event" is, and how those things relate to one another. When two different organizations, such as the United Kingdom's defense forces and a broader international alliance, try to share information, they must align their maps. If one map says a "person" is a physical object that exists through time, and the other says a "person" is a series of moments in time, the computers trying to merge these maps will crash or, worse, produce silent errors that lead to bad decisions. The goal of this work is to ensure that when these digital maps are joined, the resulting picture is not just consistent, but actually makes sense.

The researchers in this study tackled a difficult problem involving three specific digital maps used in defense: the Information Exchange Standard, which the UK uses to share intelligence; the Higher Quality Data Model, which underpins the nation's digital twin of the built environment; and the Basic Formal Ontology, a standard used by NATO and the United States. Before this work, no one had successfully created a machine-readable bridge between the UK's Information Exchange Standard and the Higher Quality Data Model. The team took a hand-curated list of seventeen connections, created by human experts who believed the two systems were related, and tested what would happen if a computer treated those connections as absolute logical facts. The result was a startling discovery: simply checking if the combined data was "consistent" was not enough.

When the researchers merged the two maps using the expert's connections, the computer passed the standard consistency check. It found a model where everything could theoretically exist. However, a deeper inspection revealed that the merged system was broken in a way the standard check missed. The combination created one hundred new categories of data that were logically impossible to exist. For instance, the merged system declared that certain real-world operational categories, like "arrested person" or "witness," were empty sets—concepts that could never have a single instance. This happened because the target map, the Higher Quality Data Model, already contained thirty-nine broken categories before the merge even began. These broken categories were invisible to the standard checks, hidden deep within the structure of the map. When the researchers mapped the UK's data onto these broken categories, they were essentially trying to pour water into a cup with a hole in the bottom; the connection was technically "true" because nothing could ever fill the empty space, but it was useless for any real-world application.

The paper argues that the usual method of fixing these problems is too blunt. Typically, when a computer finds a conflict, it simply deletes the connection that caused the trouble. The researchers found that this approach throws away valuable information. Instead, they developed a new method called "orientation search." Rather than asking "should we keep this connection or delete it?", their method asks "in which direction does this connection flow?" It treats the relationship between two concepts as a variable that can be adjusted. Sometimes two things are exactly the same; other times, one is a type of the other, but not the reverse. By letting a computer reasoner test every possible direction for each connection, the team was able to repair the bridge without deleting any of the expert's original insights. They found that for most connections, the two systems agreed perfectly. But for the few where they disagreed, the computer could pinpoint exactly why. It would say, "These two concepts cannot be equal because if they were, it would force a specific real-world event to be a type of physical object, which contradicts the rules of the first system." The computer then provided a specific counterexample to prove the point, turning a potential error into a documented difference in how the two standards view the world.

The study also tested how well existing computer systems could solve this problem on their own. They ran two established matching programs and a large language model, similar to the technology behind modern chatbots, against the same data. All of them failed in the same way. They were tricked by a "false friend"—a pair of words that look identical but mean different things. Both systems tried to connect the concept of an "event" in the UK system with an "event" in the other system. In the UK system, an event is a happening that takes time and involves people. In the other system, an event is a single, instantaneous moment with no duration. The computers, relying on the similarity of the names, forced these two different ideas to be the same, which immediately broke the logic of the entire system. The human experts had warned about this trap, but the automated systems walked right into it. Even when the researchers gave the automated systems the expert's list of connections and asked them to fix the errors, the systems either deleted the correct connections or reversed them, losing the precise meaning the experts had intended.

The final result of this work is a certified bridge between the three major defense ontologies. It is a set of twenty-one logical rules that connects the UK's data standards to the international ones without creating any impossible categories or breaking any existing rules. Crucially, this bridge does not hide the disagreements. Where the systems agree, it states they are equal. Where they disagree, it states the relationship in one direction only and attaches a machine-generated explanation for why they cannot be equal. This approach ensures that the data remains usable and that the differences between the standards are recorded as facts rather than errors. The researchers also discovered that connecting the UK's four-dimensional view of the world to the international three-dimensional view required a specific adjustment: a "person" in the UK system is not a physical object in the international system, but rather the history of that object's existence. This subtle shift was the key to making the entire system work.

The paper concludes that the way we build and test these digital bridges needs to change. Relying on simple consistency checks is dangerous because it allows broken maps to pass inspection. Instead, every time two systems are joined, the process must include a check for "coherence," ensuring that no categories become impossible to fill. Furthermore, the direction of the connection matters as much as the connection itself. By treating the direction as a variable to be tested rather than a fixed assumption, researchers can preserve the knowledge of human experts while letting computers find the precise logical path. This work proves that with the right tools, we can build digital bridges that are not just consistent, but truly coherent, allowing different defense systems to share information with confidence and clarity.