Principles and algorithms for causal reasoning with uncertainty

This thesis examines representational issues that arise when reasoning about causes and effects given incomplete and uncertain knowledge about the domain. These issues are largely covered by the famous frame and qualification problems. This thesis shows how traditional nonmonotonic logic approaches can be modified to address these problems in a simple, domain-dependent way. This nonmonotonic approach is then generalized to manipulate statistically-founded beliefs, allowing for more consistent and fine-grained representation of causal knowledge. In addition, this thesis presents an algorithmic approach for efficient parallel computation of statistical predictions. This approach involves two heuristics, highest impact first and highest impact remaining, which control the speed of convergence and error estimation for an algorithm that iteratively refines degrees of belief. This algorithm has been implemented and tested by a program called HITEST, which runs on parallel hardware.