Defined Classes vs Universals and Their Place in Ontological Realism

[Anthony Petosa]

"Building Ontologies With Basic Formal Ontology" (Arp, Smith, Spear) offers the following definitions for "defined class" (p.178) and "universal" (p.186):

Defined class: A collection of individuals that are grouped together by virtue of their exhibiting some combination of characteristics that does not correspond to any universal.

Universal: A mind-independent, repeatable feature of reality that exists only as instantiated in a respective particular (individual thing, instance) and is dependent upon a particular for its existence.

Pages 19-20 of the book speaks of two distinct defined class families:

• Classes defined by general terms abbreviating logical combinations of terms denoting universals, with "defined by selection" and "defined by combination" sub-categories
• Classes defined by general terms abbreviating logical combinations of terms denoting universals with terms denoting particulars

Elsewhere in the book (pp.18-19) "universal" and "class" are contrasted. It is noted that a class is defined as "a maximal collection of particulars falling under a given general term" and that a universal "has a corresponding maximal class as its extension." Where the maximal class extensions exactly overlap is when a "class" is designated a "natural class", because it reflects the maximal instantiation of particulars for a given universal.

The challenge rises when expressing defined classes. There are many practical cases where an ontology model must express defined classes. For example, the W3C PROV-O Ontology defines "Agent" as follows (see: https://www.w3.org/TR/prov-o/#Agent):

"An agent is something that bears some form of responsibility for an activity taking place, for the existence of an entity, or for another agent's activity."

It presents the following subclasses of Agent: Person, Organization, SoftwareAgent

Person, Organization & SoftwareAgent have the following PROV-O Ontology definitions:

Person: Person agents are people.

Organization: An organization is a social or legal institution such as a company, society, etc.

SoftwareAgent: A software agent is running software.

The preceding Agent, Person, Organization, and SoftwareAgent definitions are nebulous at best. However, the intent appears to be that an Agent is something that carries a responsibility to achieve a certain goal(s). If so, then Agent seems to be a defined class. It is the logical disjunction of Person, Organization and SoftwareAgent, that collectively does not exist as a universal in accordance with the criteria established on p.17 of the book.

From on Ontological Realism perspective, would it not be better to ignore "Agent" and instead express an Agency Role, which is a Specifically Dependent Continuant that s-depends on Person, Organization, SoftwareAgent, etc.?

This raises a larger question about the place that defined classes hold in an ontology model respecting Ontological Realism. Should defined classes be completely removed from such ontology models? There are exceptions, such as the "eukaroyte cell" example on p.20. However, this defined class is considered a "natural class" and, therefore, its maximal class extension exactly overlaps with the maximal particulars that are instantiable over a Eukaryote Cell universal.

Would it be best to delegate defined classes to "application ontologies", which I loosely define here as those ontology models that meet subjective contextualized views of terms that matter to targeted information consumers?

In the case of respecting Ontological Realism while concurrently promoting reuse and ease of extensibility, should not a "domain ontology" objectively model terms corresponding with universals and "application ontologies" subjectively model terms corresponding with defined classes, with application ontologies importing domain ontologies as a point of extension and synchronization?