A comment on work by Booth and co-authors (Q969423)

In their paper ``A unifying semantics for belief change'' [in: R. López de Mántaras et al. (eds.), Proceedings of the 16th European conference on artificial intelligence, ECAI'2004, including Prestigious applicants of intelligent systems, PAIS 2004, Valencia, Spain, August 22--27, 2004. Amsterdam: IOS Press. 793--797 (2004)], \textit{R. Booth}, \textit{S. Chopra}, \textit{T. Meyer} and \textit{A. Ghose} showed that many of the variations on the AGM postulates for belief revision can be modeled by supplementing the familiar relation between states by a second one, chosen as a subrelation of the first. Roughly speaking, the states that satisfy the result of revision of a belief set \(K\) by a proposition \(x\) are defined to be of two kinds: (1) the best states, under the first relation, among those that satisfy \(x\), and (2) those states failing \(x\) that are nevertheless at least as good, under the second relation, as some state of the former kind. Booth et al. established completeness theorems for a range of AGM-like systems in terms of classes of such structures, via suitable representation theorems for the structures themselves. However, their proofs cover only finite languages, and do not generalize readily to the infinite case. In the paper under review, the authors obtain the same results in the infinite case, using a quite different construction of the representing structures, particularly as regards the second of the two relations. As the authors remark, beyond the specific result obtained, the constructional techniques employed in the paper are of interest as tools for those building their own representation theorems in the infinite case for logic-motivated structures.