Improved methods for causal reasoning

Causal structures of devices (circuits, mechanisms or systems) are an adequate aid for expert system manipulation, diagonsis and operation. A causal structure is an abstraction of a physical device where the variables and their mutual dependence are identified. The causal structure of a device qualitatively reflects its behavior, providing a commonsense description adequate to system diagnosis, design and analysis. Two methods for automatic causal structure detection have been advanced [1, 5]. Both define the causal structure of a device starting from the set of equations that model its behavior. Feedback systems, however, are out of the possibilities of these methods. In this work we present two approaches to causal structure detection, based on a topological analysis of the set of structural equations, which can be adequately applied to feedback systems.