Information Modeling and the Problem of Universals - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 275

a central problem in metaphysics and at a practical level the central problem in information modeling.
This chapter will review the practice of information modeling, discuss some earlier work related to the
Problem of Universals and then more fully explore the Problem of Universals and the key metaphysical
problems at the heart of information modeling and the heart of information systems.
Information Modeling
Information modeling, sometimes referred to as conceptual database design, is the first step in the
database design process. In this initial step, the information modeler studies the domain of interest and
determines the classes of entities that will be represented in the database and the relationship between
those classes. In a university database, for example, one entity class may be Students while another
entity class may be Courses . The relationship between Students and Courses is that Students Take
Courses . As the modeler continues, attributes of interest are identified for each entity class. These at-
tributes represent facts that are common to all instances of a class. If a particular student has additional
facts, those facts are overlooked in order to have a set of facts common to all students. Eventually,
the conceptual database design is represented in an information model which contains entity class
descriptions, attributes of the entity class, relationships between entity classes, and possibly, additional
information about the nature of those relationships such as cardinality and optionality. In the process
of constructing the information model, a variety of philosophical assumptions are made that address
which classes should be represented in the model, where those classes come from, and whether those
classes are discovered or constructed.
A f irst Pass at The Problem of universals
When we use the word 'same' to refer to 'same kind' we are organizing the things of the world into
categories. Categories are useful because they help us organize our knowledge efficiently. When I point
to a tree and call it a tree, I am assigning it to a category. By doing this, I can apply my general knowl-
edge of trees to the specific tree at which I am pointing. The thing at which I am pointing is actually
an instance of a tree, but we do not make that distinction in normal speech. Yet, philosophically, we
do make that distinction. The instance at which I am pointing is called a particular and the category to
which I assign it is called a universal.
The Problem of Universals attempts to address the question - Where do universals come from? Is
a tree a tree because it is a member of the set of trees or is it a member of the set of trees because it is
a tree? This enigmatic question goes to the heart of universal construction. Is a grouping formed from
things of the same kind or are things of the same kind because they are part of the same grouping? There
really isn't an easy answer to this question and philosophers have provided a variety of answers over the
centuries. (Artz, 1997) A less enigmatic view of the Problem of Universals is to ask -- when we create
categories to organize our knowledge where do those categories come from? Are the categories real and
hence discovered, or are categories constructed and if so according to what criteria? The Problem of
Universals is fundamental to information modeling because the process of constructing entity classes
is no more or less than the Problem of Universals. That bears repeating because the central problem in
information modeling is the Problem of Universals. And understanding what has been said about the
Problem of Universals provides great insight into the process of information modeling. The assump-
tions one makes on the issue of whether universals are discovered or constructed are called ontological