| Summary: | Intelligent behaviour relies heavily on the ability to reason in the absence of complete information. Until recently, there has been little work done on developing a formal understanding of how such reasoning can be performed. We focus on two aspects of this problem: default or prototypical reasoning, and closed-world or circumscriptive reasoning.
After surveying the work in the field, we concentrate on Reiter's default logic and the various circumscriptive formalisms developed by McCarthy and others. Taking a largely semantic approach, we develop and/or extend model-theoretic semantics for the formalisms in question. These and other tools are then used to chart the capabilities, limitations, and interrelationships of the various approaches.
It is argued that the formal systems considered, while interesting in their own rights, have an important role as specification/evaluation tools vis-a-vis explicitly computational approaches. An application of these principles is given in the formalization of inheritance networks in the presence of exceptions, using default logic. === Science, Faculty of === Computer Science, Department of === Graduate
|
|---|
