Previous Next

Natural kinds and elementary categories

Which things should be represented by elementary categories introduced using newSub and which things should be defined using a name statement? There is no hard and fast rule. Philosophers argue about whether or not there is any such thing as an elementary concept. However, any computer system must start somewhere from a hierarchy of elementary things.

The answer depends partly on the application and partly on the nature of the thing being represented. Philosophers and psychologists speak of many things which are natural kinds - we know what they are when we see one but would be hard put to give a definition. Notions such as dog, child, tree, red are usually thought of as natural kinds. It might be possible to define them in some sense for some purposes, but that definition might not correspond very well with common usage, and we certainly don’t recognise them by checking their definition. On the other hand, ideas such as parent - a person who has children - or car-owner - a person who owns a car - are relatively straightforward to define, at least for the purposes of a convenient model to cover 95% plus of all usage.

Roughly speaking, it is a waste of time and usually counter productive to try to define a representation for natural kind in GRAIL. Natural kinds can be described by necessary statements, but they will almost always be elementary categories. The reverse is not true. There may be many things in a model which could be defined, but whose definitions are irrelevant to any likely application. Alternatively we may just not yet have refined the model that far. A common strategy is to put things in first roughly as elementary categories and then come back and define them more carefully later.

Previous Next
Making the impossible very difficult, ©, All rights reserved