Blockage contraction (Q2377189)

In the AGM theory of partial meet belief change, as also in several modellings for qualitative uncertain inference, each output is constructed as the intersection of preferred maximal (or otherwise distinguished) items satisfying desired conditions. In his paper [``On why the best should always meet'', Econ. Philos. 16, 287--313 (2000)], \textit{T. Sandqvist} remarked that, from an intuitive point of view, intersection might not preserve preferential status; perhaps the meet of a class of most preferred maximal items might be less favourable than the meet of some other collection of less preferred items. In the paper under review, the author takes up the challenge that this remark presents, by seeking to reconstruct a formal theory of belief contraction (leaving aside, for the present, revision and uncertain inference) without using intersections. This work is done by specifying a range of potential outcomes for any contraction and a `blocking' relation between them that is reminiscent of the `attack' relation of Dung-style argumentation theory, although ending up with quite different properties. The author articulates conditions on the blocking relation that suffice to guarantee a unique finitely-based belief set as the outcome of contracting a formula (or set of formulae) from a given finitely-based belief set. The resulting class of `blockage contraction operations' overlaps with the class of all AGM partial meet operations, without either class being included in the other. A syntactic characterization is provided for the blockage contraction operations, with a corresponding representation theorem, thus creating a quite novel `non-meet' modelling for belief contraction.