12.6 Designing the Concept Hierarchy 301
12.6.2 Classes vs. Instances
One important design issue is whether a concept should be represented as a
class or an instance. Up to now we have been tacitly assuming that concepts
are always classes and that it is obvious what it means to be an instance of
the class. Unfortunately, in general, there is no clear division between classes
and instances, and a concept can be either one. Choosing between them
represents a design choice that is dependent on the purpose of the ontology.
Consider, for example, the concept of a disease. Acute promyleocytic leu-
kemia is a particular form of leukemia. Should it be regarded as a subclass of
leukemia or an instance of leukemia? The answer depends on the purpose of
the ontology. If there are usage examples in which there are instances of acute
promyleocytic leukemia (such as particular cases of the disease), then acute
promyleocytic leukemia should be a class. It should also be a class if it has
subclasses. However, if there are no subclasses and no instances, then acute
promyleocytic leukemia should be an instance. This would be appropriate
for ontologies that are concerned with understanding the causes, symptoms,
and progression of diseases rather than with observations of occurrences of
diseases.
In general, instances represent the most specific concepts that are being
represented. If a concept has a subclassification, then the concept must nec-
essarily be a class. However, the absence of a subclassification does not mean
that a concept is an instance. The determination in this case relies heavily on
the usage examples.
12.6.3 Ontological Commitment
Ontology development efforts can sometimes be afflicted with an ailment
that frequently occurs in software development efforts. This is the tendency
to expand the scope of the development beyond the original purpose and in-
tent. This problem is known by many names, such as “featuritis” and “scope
creep.” The accumulation of new features is usually gradual, with each ad-
dition being so small that it generally gets overlooked in terms of its impact
on the project. However, the cumulative effect can be considerable. In on-
tology development, the termontological commitmentrefers to the level of
elaboration and detail that has been chosen for each concept.
Consider, for example, the notion of the temperature of a human being.
This is commonly used for diagnostic purposes. There are, however, sev-
eral ways that the temperature can be measured, so one should also specify
