A context for belief revision: forward chaining-normal nonmonotonic rule systems (Q1326782)

For some time it has been widely appreciated that those systems of nonmonotonic logic that are defined in the literature using fixed points (notably Reiter style default logic, autoepistemic systems, logic programs with negation, and certain truth-maintenance systems), have a considerable amount of structure in common. In two earlier papers [``Nonmonotonic rule systems. I, II'', Ann. Math. Artif. Intell. 1, 241- 273 (1990) and ibid. 5, 229-263 (1992)], the authors isolated a common abstract form, and showed how it permitted general proof of many theorems recurring for the various sytems. In the paper under review, the authors continue that line of research. Specifically, they take the well-known notion of a ``normal'' rule in Reiter default logic, and generalize it to their abstract framework. This is not a trivial task, as the abstract framework contains neither negation nor the classical notion of consistency. The authors adapt \textit{D. Scott's} abstract notion of a ``consistency property'' [Lect. Notes Comput. Sci. 140, 577-613 (1982; Zbl 0495.68025)] and use it to formulate constraints on rules. Although in the context of Reiter default logic those constraints are a little weaker than normality, they suffice to prove all the main theorems in the literature posed with that condition as hypothesis, and this in a manner that applies also to the other fixed point systems referred to above.