Symbolic Knowledge Processing for the Acquisition of Expert Behavior: A Study in Medicine.
Interim technical rept.,
CARNEGIE-MELLON UNIV PITTSBURGH PA ROBOTICS INST
Pagination or Media Count:
This research is concerned with the simulation of learning by experience to induce the capability for a knowledge-based system to pre-structure the problem before solving it. The model we present is made of different consecutive modules accounting for the tasks of problem solving, building a dynamic memory and extracting expectations, and pre-structuring or pre-solving the problem. The problem-solver yields internal representation of the problems between which symbolic distances may be defined. The latter are then processed to build the dynamic memory. We used the formalization of medical problem-solving as an example, studying how successive evaluations of cases may lead to the acquisition of the capability to generate an accurate set of initial hypotheses an expert behavior. The knowledge base is not modified, neither are the strategies in the present implementation. To the data gathering about the patients complaints is added a concept-driven process by which the system asks for specific data representative of lthe past experience. The results show that such a system, evolving in a coherent reality increases its qualitative behavior by initially focusing on the right hypotheses or goals. This improvement is induced by the exposure to new situations. Moreover, situations once or rarely encountered are efficiently recognized when re-occurring later.