Modeling uncertainty using partial information

We are concerned here with the issue of representing the available information, which is often only partial, about the value of a variable. We consider the Dempster-Shafer belief structure and describe a number of different semantics that can be associated with it. Among these is a probabilistic interpretation in which some of the information is not available, we do not have all the distributions. This probabilistic interpretation leads us to consider a view involving selecting balls from an urn. Using this urn framework as a representation of our partial knowledge about the value of a variable leads to number of different uncertainty representations depending upon what information is assumed known. (C) 1999 Elsevier Science Inc. All rights reserved.