The long-term objective of this project is to develop a comprehensive theory of probabilistic and causal reasoning, specific enough for machine interpretation, so as to guide the construction of computer systems capable of planning, perceiving, and learning in uncertain dynamic environments. The approaches described in this proposal are based on the use of probabilistic causal networks as the basic representational scheme underlying the reasoning process. Specific topics of interest are (1) Information fusion, diagnosis and planning under uncertainty using qualitative approximations of probabilities, utilities, causal and counterfactual relationships; (2) Automatic generation of natural language explanations of actions, recommendations, and unexpected eventualities. The shorter term results of this research will be improved reasoning methods in applications such as medical diagnosis, sensor fusion, and automated management and control of manufacturing processes. Another major impact of the proposed research will be improved experimental methods in the social and biological sciences, where causal inference is hindered by model uncertainty.
http://singapore.cs.ucla.edu/jp_home.html
