Ensuring the Neutrality of Ontology ​

Let's return to modeling and ontologies with new knowledge baggage about sign systems.

As we discussed, when an ontology identifies objects and defines their names, it cannot completely detach itself from a certain role position (which it knows better or with which it started). It is impossible to form an ontology from the observer's role, completely uninvolved in anything described — the distinction of objects already implies someone's view of the subject area, someone's way of distinguishing them for some use, someone's representation of the purpose of these objects. But an ontology must strive to reflect all necessary views, that is, to make sure that all users see the objects they need in it roughly the way they are used to. And for unfamiliar objects (objects of other roles), they should be able to establish correspondences or relationships with familiar ones, see how these objects are matched with each other (when they are the same objects with different names, when they are more general or more specific concepts, when they are different parts of a whole, etc.).

Ontologies suitable for different roles are called neutral. Neutrality allows for achieving compatibility of models for different roles, both at the level of the objects themselves and their names.

Now you can understand why in subject areas where the conceptual space is denser, it is easier to build neutral ontologies and use them more easily. The more accurate the correspondence of objects of the subject area, the ideas embodied in them and the signs naming them, the clearer the distinction of concepts, the easier it is to recognize objects, relate them to categories, correctly form relationships between them, and the easier it is to understand models created by others.

When receiving an ontological model of an area where the conceptual space is sparse, you spend disproportionately large amounts of time on reference — understanding which objects and ideas stand behind the model signs. And a failure at this stage (misunderstanding what the author meant) can cost you dearly in the future when your interpretation turns out to be incompatible with the author's or other users' interpretation of the same model.

Therefore, ontological modeling is relatively successful in natural-scientific or engineering subject areas, where concepts are defined formally and finding a physical embodiment of objects is not a big problem. But in social-cultural-philosophical subject areas, where conceptual spaces are sparse, the quality of models (texts, other types of formal models in these areas are better left unmentioned) is much lower. Here, texts do not provide successful communication, as the parties draw opposite or incompatible conclusions about the object of modeling from the same models (texts).

To ensure at least some neutrality of the ontology, a very active approach to its construction is necessary. It is essential to analyze texts of different types, to walk and talk to various interested parties, representatives of all groups (performers of each important role in this subject area). Then it can be hoped that the ontology user (in real life, these will be readers of reports, performers of regulations, users of computer systems, knowledge bases, or simply people filling in reporting forms) will find their objects there, named in a way familiar to them.