Filters-based revision (Q6903889)

The AGM paradigm for the study of the logic of belief change assumes that the formal language concerned contains at least classical propositional logic; in algebraic terms, it works in the context of Boolean algebras. Over the last two decades, several studies have investigated ways of adapting its ideas to apply to logics that do not satisfy that assumption, e.g. Horn clause logics, description logics, and logic programming systems. The paper under review builds on ideas introduced in [\textit{J. P. Delgrande} et al., J. ACM 65, No. 5, Article No. 29, 34 p. (2018; Zbl 1426.68257)], to cover many of these by lifting AGM theory from the context of Boolean algebras to that of semi-lattices, working with a possibly infinite set equipped with a partial order that has an infimum for each pair and a bottom element. A set of elements is called inconsistent iff one its finite subsets has the bottom element as its infimum, and the counterparts of AGM belief sets are taken to be filters that are in a certain sense `canonical'. The authors lift the Katsuno-Mendelzon semantics for AGM revision to this context and prove an appropriate representation theorem. They suggest that ``This provides a flexible and versatile framework that can be applied across a wide range of logical and non-logical domains, further enriching the landscape of AGM-inspired revision, without being confined to any particular logical system.'' Reviewer's comment: readers seeking an intuitive feel for the crucial notion of canonicity (definition 21) may find the following equivalent and rather more direct formulation helpful: a filter F is canonical iff it contains all elements of the semi-lattice that are consistent with every filter that includes F.