On Iterated Revision in the AGM Framework

While AGM belief revision identifies belief states with sets of formulas, proposals for iterated revision are usually based on more complex belief states. In this paper we investigate within the AGM framework several postulates embodying some aspects of iterated revision. Our main results are negative: when added to the AGM postulates, our postulates force revision to be maxichoice (whenever the new piece of information is inconsistent with the current beliefs the resulting belief set is maximal). We also compare our results to revision operators with memory and we investigate some postulates proposed in this framework.