Information Technology Reference
In-Depth Information
iMPlica tions for infor Mation Modeling
The different metaphysical positions we have discussed lead to quite different approaches to informa-
tion modeling and different results in the model produced. Class realism is often adopted by naïve in-
formation modelers and often implicit in topics on information modeling that do not explicitly address
the discovery problem. Unfortunately, class realism is not supportable under any circumstances. If the
class realist develops the correct model it is either just luck, or the demands of the application have been
defined after the fact as those information needs that the database can support.
Class conceptualsim is a far better foundation for the practice. This, however, suggests some fairly
radical changes for the practice of information modeling. Currently, we focus on diagramming tech-
niques and automated tools to support diagramming. If class conceptualism holds, which appears to
be the case, then we should be focusing on the question of how classes are constructed to meet various
information needs. Further, we can no longer validate an information model by comparing it to the
real world (since that assumes class realism). We have to define objectives for the model and evaluate
constructs according to how well they meet those objectives.
Attribute realism is, at least, slightly suspect. Entities in the real world do have physical characteris-
tics. However, the purpose of a database and the purpose of a scientific taxonomy are likely to be quite
different. Scientific taxonomies do not consider functional or artificial attributes in their classification.
Yet information models do. Assuming that entities have a limited (and small) number of physical at-
tributes that can be used to group them into classes is hard to justify in practice. Entities have lots of
attributes (physical and artificial), some that they share in common with other entities some that they
do not. Attribute realism does not allow for functional or artificial attributes. Limited attribute realism
does not allow for the fact that entities may not have uniform properties. Hence, attribute realism does
not provide an adequate foundation for the practice.
Attribute conceptualism seems to provide the richest foundation because it acknowledges the exis-
tence of nonphysical attributes. It also allows for the fact that we pick and choose attributes based upon
(possibly implicit) objectives in the modeling process. It also allows for the fact that we may invent
some of the attributes. This leaves the question of whether or not the attributes can be constructed from
the linguistic usage.
If the model must represent usage and the usage is somewhat consistent, then it may be possible
to construct an information model of the domain in question. If the model must represent usage but
usage is not consistent then it will not be possible to construct a coherent model. Finally, if the model
does not have to reflect usage, it can be constructed to meet information objectives. If the objectives
are consistent, then the model can be constructed, otherwise not. Yet the resulting model may not be
consistent with the users concept of what the various constructs mean.
conclusion
The Problem of Universals provides both a metaphysical foundation for information modeling and
substantial insight into the nature of the process. Information modeling research should focus less
on representational techniques and more on the problem of what is to be represented. Since class and
attribute conceptualism provides a much firmer philosophical foundation, then the work of research-
ers in information modeling is to determine how to define modeling objectives and how to compare
Search WWH ::
Custom Search