Generalized directive models: integrating model development and knowledge acquisition

Abstract

The Generalized directive model (GDM) methodology for knowledge acquisition is introduced. For GDMs to work two assumptions are required: that knowledge acquisition has a cyclic structure interleaving episodes of model development and domain KA, and that increased specification of one part of a model does not affect other parts. The use of GDMs is illustrated with a real-world example from an Airborne Early Warning system, showing the development of a model for one sub-task using the PC-based GDM tool from the commercial workbench PC-PACK. There is also a small example of a GDM analysis extending an already existing model. Finally, GDMs are compared with the decompositional CommonKADS library.