Slide 25 of 32
Notes:
A generalization is when classes from a different sets of coordinate concepts are seen as coordinate concepts for a new set of coordinate concepts. In Table 5 finches, ants, and cows each appear in a different coordinate set corresponding to birds, insects, and mammals. However, each of these classes share the same value, plants, on property 2, Source of Food. By sorting on the second property we can identify a new set of coordinate concepts as in Table 7. Note in this case that the first property discriminates on the second set of coordinate concepts while the second property discriminates on the first level of coordinate concepts.
What is the algorithm by which the learner solves problems of generalization. First, if the value on a property, or the values on two or more properties, are the same across different sets of coordinate concepts, then sorting on this property will yield a new set of coordinate classes. Having a complex knowledge structure of one set of concepts allows the learner to manipulate the knowledge components in this manner in an attempt to identify other coordinate concepts. The learner then searches to find the names associated with these new coordinate concepts (Dijkstra & van Merriënboer level 2). If this is a new categorization then the learner may invent names and thus invent new concepts (Dijkstra & van Merriënboer level 3).
