Information Technology Reference
In-Depth Information
Aristotle's position provides much of the philosophical basis for the more disciplined approaches to
modern information modeling. The information modeler observes entities in the real world and then
identifies their attributes. Next, based on the commonality of attributes, these entities are grouped
into entity classes and the common attributes become the class attributes. In object-oriented analysis
the commonality of class attributes can be used to define class hierarchies which follow Aristotle's
concept of genus and species. However, one must remember that Aristotle was classifying physical
objects that had been shaped over millennia by the forces of nature and evolution. Yet, the domain of
the information modeler is the artificial world of the organization where conceptual entities such as
Customer, Account, Part and Employee are useful constructs not shaped over millennia by the forces
of nature. A customer is not a customer because of some set of physical characteristics. A person is a
customer because of a relationship that exists between the organization and the person. Another per-
son may be an employee based on a different relationship to the organization. A person may take on
different roles at different times or have multiple roles at any one time. Further, the customer need not
be a person, the customer may be another organization. Hence, it is not possible to define these classes
based on physical characteristics. These classes are defined by the changing needs of the organization
to organize its information about the world. Thus, although the scientific realism of Aristotle provides
a convenience basis for information modeling, it makes assumptions about the domain of investigation
that are probably not valid.
Locke's view of universals provides quite a different perspective for information modeling. Instead
of forming classes based on common attributes, the modeler needs to define a set of modeling objec-
tives to guide the abstraction process. From Locke's perspective, there are many ways to model an
application domain depending on what the modeler is trying to achieve. Most database applications in
organizations fall into this category. Entity classes such as Customer, Part, Employee, or Account are
defined in terms of attributes that are important to the problem at hand while vast numbers of attributes
are simply ignored as unimportant. This creates problems for the information modeler because it sug-
gests that universals such as Student or Customer may exist in the minds of users and in the language,
but may not have any explicit definition. Further, any appropriate and explicit definition may lie in the
domain in the future but not the present.
Thus, if Hume's view is correct then the process of information modeling becomes the process of
language refinement. Database users employ terms like customer, part or account without a precise
understanding of what they mean. The modeler must talk with users and construct useful definitions.
Further, these definitions must be agreed upon so that information modeling also becomes the process
of achieving a social consensus. Finally, if Wittgenstein is correct, then a domain cannot be modeled
without a serious semantic revision to superimpose semantic order upon it.
ex Tend Ing The Pro BLeM of un IVer SALS To Infor MATIon Mode LIng
Thus far we have discussed the Problem of Universals, some philosophical responses to the problem,
and some ways in which these responses can be seen in the practice of information modeling. Next we
turn to the task of refining our discussion of the Problem of Universals to begin providing a foundation
for the practice of information modeling. From this foundation we will begin making observations about
how the practice of information modeling should be refined. In doing so we define four philosophical
positions that apply directly to the construction of information models. These are: class realism, class
conceptualism, attribute realism, and attribute conceptualism.
Search WWH ::
Custom Search