Two methods of constructing contractions and revisions of knowledge systems (Q757343)

The purpose of this paper is to investigate the formal connections between two of the main approaches to the logic of theory change. One of these is the so-called AGM approach in terms of ``partial meet'' contraction functions, whose canonical presentation is in a paper of \textit{C. E. Alchourrón, P. Gärdenfors} and the reviewer [J. Symb. Logic 50, 510-530 (1985; Zbl 0578.03011)]. The other is the so-called EE approach, in terms of relations of ``epistemic entrenchment'' over the set of propositions of the language concerned. Its canonical presentation is in a paper of \textit{P. Gärdenfors} and the reviewer [Theoretical aspects of reasoning about knowledge, Proc. 2nd Conf., Pacific Grove/CA (USA) 1988, 83-95 (1988; Zbl 0711.03009)]. Some indirect connections between these two approaches already arise out of the representation theorems established in the above papers for each approach in terms of a common set of postulates. A direct connection was established, in the finite case only, between the AGM approach and a certain rather unsatisfactory variant of the EE approach, by \textit{P. Gärdenfors} [Foundations of logic and linguistics, Sel. Pap. 7th Int. Congr. Logic, Methodol. Philos. Sci., Salzburg/Austria 1983, 345-367 (1985; Zbl 0622.03003)]. The author reviews this background. He then constructs an almost direct map between the AGM and EE modellings and shows that it ``works''. The map is ``almost'' direct in the sense that it proceeds via a variant of the AGM approach, in which its order relation (between certain sets of propositions) is given a wider range and is put under more constraints. Indeed, there is a small gap in the argument here: the author does not show explicitly that his variant of the AGM approach yields exactly the same contraction functions as does the original AGM formulation. Nevertheless (as he has shown the reviewer in correspondence) the gap can be filled fairly easily using results from the ``canonical'' papers mentioned above together with the author's theorem 4 and lemma 5. The final two sections of the paper add interesting observations. It was remarked by Gärdenfors and the reviewer that their rule for generating a contraction function out of an EE relation is intuitively awkward, containing a puzzling disjunction. The author shows that if we are interested only in revision, defining it directly from the EE relation rather than indirectly via contraction and the Levi identity, then we can get rid of the disjunction and render the rule more transparent. He also initiates a discussion of iterated contractions, emphasizing the need for revising not only belief sets but also their associated EE relations. A first suggestion in this regard is outlined, and its virtues and limitations discussed. The paper is marked by a remarkable mastery of the literature, clarity of exposition, and sensitivity to subtleties.