Case based knowledge acquisition and refinement
This thesis reports research undertaken in two rather distinct phases. Firstly, the thesis reports a study of cognitive processes involved in the task of 'concept identification': given sample instances of a concept, the task is to identify the concept. A computer model which successfully reproduces responses similar to those observed in human subjects is described. Secondly, this thesis reports the design of a case-based learning system REFINER. The system is a 'Learning Apprentice System' for differential diagnosis tasks, to aid the transfer of knowledge from a domain expert to a computer. Knowledge is obtained from the expert(s) in the form of cases which have been diagnosed or classified, and not in the traditional form of classification 'rules' which the experts often find hard to specify. The REFINER program is therefore a Knowledge Acquisiton System which helps an expert refine his knowledge in a more 'natural' way than having rules 'extracted'. Further, the system has the ability to point out that two classifications are not distinct, and can then suggest to the user ways in which the inconsistency might be resolved. Although the system has been used most extensively in the medical domain, it is essentially domain independent.