Qualitative decision support

This research describes the basic concepts of Qualitative Decision Support (QDS) and illustrates the processes of constructing and using qualitative behavioral models for consequence prediction and explanantion. The efforts of QDS are congruent with those of research in decision support systems (DSS) in their concern for the development of automated decision support aids. QDS is distinguished from DSS by its commitment to the development and implementation of methodologies that support non-quantitative (i.e. qualitative) reasoning processes, model construction, and decision support. The goals of QDS include (1) the development of a methodology that possesses the ability to recognize and exploit the important behavioral distinctions of an application yet, is far simpler than the sophisticated mathematics of a classical quantitative analysis, (2) to construct qualitative models from generalized descriptions of an applications's important features, and (3) to create intuitively plausible predictions and explanations of the behaviors that arise when the important features of an application are subjected to change. QDS is an outgrowth of research initially aimed at the development of a naive physics to support the automation of significant commonsense reasoning capabilities. QDS draws upon recent work in qualitative reasoning about physical systems and the use of a qualitative calculus and qualitative differential equations (i.e. confluences). QDS accepts a reductionistic approach in assuming that the interesting behaviors of an application can be predicted and explained completely in terms of the application's component mechanisms. A model of the application's behaviors is constructed from models of the constituent mechanisms' behaviors. This approach maintains the individuality and integrity of an application's primitive behavioral units and facilitates modeling flexibility and generalization through the interchangeability and manipulation of the component models. QDS specifically addresses four concerns of qualitative reasoning and mental modeling research: (1) the construction and recognition of intuitively valid causal models, (2) reduction (and elimination under certain restrictions) of non-determinacy, (3) concept generalization and the construction hierarchical reasoning models and (4) the providing of a natural connection for the coupling of QDS models and classic quantitative models.