We provide some objective foundations for a belief revision process in a situation where i)the decision maker's initial probabilistic knowledge is imprecise and characterized by the core of a belief function; ii)expected new data are themselves consistent with a belief function with known focal sets and iii) is based on belief function combination. We study the properties of the information value for such revising in the GILBOA - SCHMEIDLER multi-prior model.
Keywords. revising, information value, belief function
The paper is available in the following formats: