ABSTRACT
The holistic approach to the generalization task can only be tackled considering the generalization problem as a joint human-machine task. First successful and questionable promises of AI and KBS are discussed, then the advantages and pitfalls of KE methods with respect to knowledge elicitation are laid out so that the last section is centred on conceptual thoughts about the application of CBR, NN and GA. The discussion of the suitability and limitations of Al-related techniques is emphasized rather than the elaboration of technical details. In terms of knowledge systems, building a successful solution, in general, means first to define the representation of the generalization knowledge, then second to elicit the knowledge about operators and sequences, and finally to identify and control the processes. The ability of recording interaction logs promises more insight into the map generalization processes.
