Wednesday, October 17, 2007

What is redundant?

In the handy Philosopher's Toolkit book[1], there is a section[2] explaining the difference between "analytic" statements and "synthetic" statements, and in another section[3], "a priori" versus "a posteriori". They are both similar in that they are trying to distinguish between what is known about X just from its definition versus things that would have to be thought through or experiments done to discover them. E.G. "all triangles have 3 sides" and "Joe, the bachelor, is unmarried" are obvious just from the definition of triangle and bachelor. Other statements like "water boils at 212 degrees Fahrenheit" and "Joe, the bachelor, has red hair" are not "entailed by" or "contained within" the definitions of water and bachelor.

There are arguments by philosophers, however, about what IS entailed by a concept or contained within a definition. Some would argue that the boiling point of water *is* contained within the concept of water whether a particular person happens to know that or not. Similarly the statement "all triangles have 180 degrees as the sum of their internal angles" is true independently of whether we know about it, (but a lot of people *wouldn't* know that).

So, in general, there will not be agreement from one ontology/model to another about which statements (i.e. data/attributes/relationships) are redundant. Should the "age" of an applicant for a job be stored because it was a field in the application form? Or, should it always be calculated from the birth date and the date of the application? Should a "short cut" relationship between 2 distantly related entities be explicitly supported by a model, or should it always be "calculated" by traversing all the intermediate links through the "six degrees of separation"?

Even though E/R models will normalize data to an objective standard, different modelers will break normalization for "performance reasons" on a subjective (and thus inconsistent) basis. Existential Programming counts this lack of agreement on "what is redundant" as yet another reason to embrace/support multiple ontologies simultaneously.

[1] The Philosopher's Toolkit, Julian Baggini and Peter S. Fosl, Blackwell Publishers, 2003, ISBN: 0631228748
[2] ibid, section 4.3
[3] ibid, section 4.1

No comments:

Post a Comment