MGR: An Architecture for Problem Solving in Unstructured Environments

Model Generative Reasoning (MGR) is a problem solving architecture designed to support AI applications in unstructured task environments, i.e. in task environments in which the statistical quality, completeness, and relevance of the data that will be available to the problem solver at any particular time cannot be predicted reliably (Hartley et al., 1987; Coombs and Hartley, in press a; Coombs et al., submitted). MGR was originally developed to solve problems in decision support for process control and fault diagnosis (Coombs and Hartley, in press b). Current application areas include scenario-based meteorological data integration (Coombs et al., 1988) and automated nucleic-acid and protein sequence data analysis.