Types of reasoning permitted in BBN's

When using BBNs we are interested in making predictions about uncertain quantities conditioned upon some evidence. Mathematically the meaning of the “|” in p(A | B) is straightforward but when modelling the real world we must be careful to ensure that conditioning is employed only to model sensible statements about the world. There are a number of types of reasoning that can be considered as valid conditional propositions in a BBN and the purpose of this section is to provide an overview of what the permitted reasoning types are and outline their differences. The different modes of reasoning described are:

• Causal : effect determined by cause, e.g. accident caused by fault introduced by system designer)

• Statistical : chance of event determined by population of possible events and sampling method, e.g. probability of failure determined by number of demands and number of failures observed

• Structural : phenomena determined by overall structure to which it belongs or object/event or determined by properties of class of objects/events to which it belongs. E.g. failure of module X determined by error introduced in module Y or reliability of system A similar to reliability of system B where A and B belong to class of systems C

Finally we explain how these different types of reasoning are handled by the SERENE idioms.