Abstract: The paper discusses the applications of influence diagrams to probabilistic inference, and information with the inference. The concept and method of quasi information filtering are proposed. The fusion of information in belief network is eramed and studied by using quasi-in-formation filtering to represent the flexible inference machanism in human's actions. A flexible inference rule in belief networks is designed and quasi-information filtering provides a new approach to studying uncertainty in decision analysis and A.I.

Key words: influence diagram, decision analysis. probabilistic inference, uncertainty in A.I