Hierarchical Inference in Large Scale Systems

Abstract

Listen

Translate article

Often there exists one or more information items believed relevant but not fully definitive to allow error free diagnosis of the state of the environment. The process of extracting the (maximum amount of) information concerning the state of the environment and the effect of various decisions or actions is called inference. Bayes' theorem yields an optimum procedure for the sequential aggregation of information across the independent samples of information for cases in which it is possible to identify a mutually exclusive and exhaustive set of hypotheses about the state of the environment. An investigation of the use of state variable concepts for sequential inference analysis with dependent multicue information in possible nonstationary, that is changing over time, environments is described. The central object of the research is to develop extensions to the theory of hierarchical inference in large scale decentralized systems that will allow better use of inference analysis in operational planning and decision support settings. To these ends, the concluding section of this paper describes a possible use of hierarchical inference structures as structured protocols for communication, command, and control system design. These protocols allow identification of various forms of cognitive bias, and appropriate debiasing procedures, in the formulation, analysis, and interpretation of various information and judgment inputs to decision making.