Belief Maintenance in Bayesian Networks

Bayesian Belief Networks (BBNs) are a pow­ erful formalism for reasoning under uncer­ tainty but bear some severe limitations: they require a large amount of information be­ fore any reasoning process can start, they have limited contradiction handling capabil­ ities, and their ability to provide explana­ tions for their conclusion is still controversial. There exists a class of reasoning systems, called 11-uth Maintenance Systems (TMSs), which are able to deal with partially speci­ fied knowledge, to provide well-founded ex­ planation for their conclusions, and to detect and handle contradictions. TMSs incorporat­ ing measure of uncertainty are called Belief Maintenance Systems (BMss). This paper de­ scribes how a BMS based on probabilitistic logic can be applied to BBNs, thus introduc­ ing a new class of BBNs, called Ignorant Be­ lief Networks, able to incrementally deal with partially specified conditional dependencies, to provide explanations, and to detect and handle contradictions.