The research objective of this proposal is to develop improved methods for explaining the conclusions of medical diagnostic-consultative computer systems.
The specific aims are to develop an understanding of what limits current explanatory systems, what requirements for explanation are needed for these systems to be acceptable to the physician, and how effective systems can be designed and implemented. The investigation will extend the functionality of diagnostic systems by improving the human-computer interaction. The research will emphasize new ways explanations can be generated by computer systems and new approaches to the computer interface for explanatory systems. A new method, HyperExplain, that unites methods from object-oriented programming, frame systems, natural language generation, and Hypertext will be used to generate explanations. The developed methods for explanation will be incorporated int a working diagnostic system for neurophysiologic analysis. The program will be of clinical value to physicians that treat patients with nerve or muscle disorders. The system will be extensively tested and revised based on user interaction and feedback. The ultimate goal of this program is to improve medical computer-aided diagnosis and its acceptance through better facilities for explanation. Through this effort it is anticipated one significant obstacle to deployment of artificial intelligence in medicine (AIM) programs may be reduced.
| Jamieson, P W (1992) Neurology research: a journal survey, 1980-1990. Ann Neurol 32:87-93 |
| Jamieson, P W (1991) A computational model of levodopa pharmacodynamics in Parkinson's disease. Clin Neuropharmacol 14:498-513 |
