Uncertainties in the analysis of large-scale systems

A new approach to analyze uncertainties in large-scale systems is proposed. Unlike the approaches which focus on the uncertainty in the output information, this approach focuses on the inferences drawn from the information. The development of this approach involved three important issues which are the identification of an appropriate mathematical framework for representing inferences; the development of a measure for uncertainty associated with the inferences, which incorporates the context and the attitude of the analyst; and the development of a methodology for the aggregation of inferences from subsystems in a manner that captures the propagation of uncertainty. This approach enables representation of the context-dependent nature of uncertainty, and explains the propagation of uncertainties through long cause-effect reasonings.<<ETX>>