Article ID: | iaor1992358 |
Country: | United States |
Volume: | 7 |
Issue: | 3 |
Start Page Number: | 265 |
End Page Number: | 289 |
Publication Date: | Sep 1990 |
Journal: | Simulation Transactions |
Authors: | Narain S., Rothenberg J. |
Keywords: | artificial intelligence |
Dynamic systems are frequently modeled using mathematical or engineering descriptions. Frequently, however, humans perform important types of reasoning about them using models of a more qualitative nature. Thus, it is important to develop techniques for systematically constructing such models. DMOD is a new modeling technique based upon a novel view of the causality relation. This paper shows how DMOD can be employed for constructing qualitative models and for answering simple types of questions about them. It offers two main advantages over previous techniques. First, it allows expression of new types of intuitions about causality, notably event preemption. Second, the accuracy of models is limited only by our knowledge about the underlying mathematics. In particular, it is not necessary to discretize real-valued spaces.