| Summary: | The thesis investigates the development of medical reasoning processes and how student modelling of such processes can be achieved in intelligent tutoring systems. The domain of orthopaedics was chosen for the research. Literature has shown that medical reasoning has been modelled mainly from an expert point of view. The research problem addressed is to model explicitly various levels of medical expertise in terms of reasoning strategies. The thesis reports on a system, DEMEREST (DEvelopment of MEdical REasoning STrategies), a developmental user model component which describes successive stages of medical reasoning and which could ultimately be part of a medical tutor. The system diagnoses physicians' reasoning strategies, determines the level of expertise and produces a plan corresponding to the application of these strategies. As a basis of doing so, a set of seven reasoning strategies was identified in the medical problem solving literature. These strategies are based on generalisation, specialisation, confirmation, elimination, problem refinement, hypothesis generation and anatomy. An empirical study was carried out to examine the development of these strategies. Protocols of ten physicians at various levels of expertise were collected and analysed. A number of interactions of strategies at different levels of expertise was identified in half of these protocols and this information was used to construct a model of changes of strategies over time. Planning in· artificial intelligence was used as a means of decomposing medical problem solving into a set of goals; the goals being associated with the reasoning strategies. By taking this approach, medical reasoning is viewed as a planning process. The remaining protocols from the empirical study were used to evaluate DEMEREST. The system was tested for its ability to determine a level of expertise for each protocol, model the reasoning strategies applied and their interactions, and generate a plan for each protocol. The assessment of the overall performance of the system showed that it was successful. This assessment also helped to identify conceptual as well as implementation constraints of the prototype system. The main result of the research undertaken in this thesis is that the design of the system DEMEREST demonstrates the feasibility of modelling the development of medical reasoning strategies and its usefulness for student modelling.
|
|---|
