Abstract
Why do we utilize artificial categories in investigating concepts and categories? No clear answer has been given to this question. Then, how can we claim the validity of artificial categories? This paper makes logical, set theoretical, and psychological consideration around this theme. The major conclusions are: 1) Artificial categories (concepts) should be homomorphic images of existent categories (concepts). 2) Some functional (or abstract) concepts cannot be represented by sets of artificial stimuli; this means artificial categories are not all-purpose tools. 3) As for the types of categories and concepts which can be represented by sets, a) intension and extension, b) the number of instances, and c) modes of uncertainty both in intension and extension, are problems of critical importance. In particular, we must make explicit representational formats for the processing of uncertainty and logical structures inherent in concepts and categories. Finally, 4) people can handle conceptual information in completely differently ways: they are sometimes naive statisticians, sometimes naive scientists, according to the task demands and the data available to them.
