This research addresses the issue of automated hypothesization of situation models that are needed as contexts for reasoning and decision making. The three different but related aspects that are addressed by this research are: 1. Learning from domain data the qualitative relationships among domain attribute and their associated probabilistic uncertainty knowledge; 2. Integration of uncertainty knowledge and other known causal qualitative relationships of the domain into a single graphical knowledge structure; and 3. Hypothesization of small and interesting situation models from the intermediate representation of domain knowledge acquired in the preceding step. The learning of qualitative relationships from databases is extended to include acquisition of temporal dependencies among attributes, and also to the cases when the available data is in the form of multi- databases. This learning method is also sought to be applied to the problem of incomplete databases. The knowledge based dynamic hypothesization of situation models is sought to be extended to apply to path planning and temporal modeling problems. The issue of specifying a situation model at varying levels of precision and certainty are also examined.//
