Choosing classes in conceptual modeling
Jeffrey Parsons, Yair Wand · Communications of the ACM · 1997
Despite the pervasiveness of classification, existing approaches offer little advice for either choosing of evaluating sets of classes. Assuming the objective of classification is to reflect how humans form concepts, the authors proposed two requirements for a class to be permissible and two other requirements to constrain the possible collections of classes. These four requirements provide necessary conditions for a set of classes to reflect cognitive economy and inference. The notion of class structure can help limit the choice of classes when modeling a domain. However, identifying additional constraining criteria may be useful. The notion of class structure suggest such practice might impair classification quality by eliminating the ability to infer all necessary properties. Other restrictions on the possible choice of classes may be found by evaluating classifications with respect to context, or the purpose of modeling. Such additional restrictions would be consistent with the cognitive notion that classes do no exit independently of the observer.