Reversing the Levi identity (Q1310625)

In the AGM account of belief change, an agent's beliefs are represented by a set of sentences \(K=\text{Cn} (K)\) closed under a consequence operation that includes at least classical logical consequence. To be accurate, in their paper ``On the logic of theory change: partial meet contraction and revision functions'' [J. Symb. Logic 50, 510-530 (1985; Zbl 0578.03011)], \textit{C. E. Alchourron}, \textit{P. Gärdenfors} and \textit{D. Makinson} did not actually exclude nonclosed sets from consideration, but all their principal results took closure as a hypothesis, and subsequent expositions have tended to write closure into the very definition of a belief set. In this paper, the author investigates belief change on sets for which no closure condition is imposed. Contraction is defined in terms of partial meet operations, as in AGM, but applied directly to belief bases, and a syntactic characterization is obtained. This section builds upon earlier work of the same author in his paper ``In defense of base contraction'' [Synthese 91, 239-245 (1992; Zbl 0760.03009)]. In the case of revision, the author observes that it may be defined by the rule \(K*x=(K-\neg x) \cup\{x\}\), paralleling AGM's \(\text{Cn} ((K- \neg x) \cup \{x\})\), but also in the distinct permuted fashion \(K*x=(K \cup\{x\})-\neg x\), and it is this that is meant by the rather cryptic title of the paper, ``reversing the Levi identity''. Of course, the latter option is not open for closed set operations, for the set \(K \cup \{x\}\) is not closed and so cannot be the subject of a contraction, whilst its closure \(\text{Cn} (K \cup \{x\})\) is equal to the set of all sentences of the language in the principal case that \(\neg x \in \text{Cn} (K)\), thus making trivially identical the revision of any two closed belief sets \(K\), \(K'\) by the same sentence inconsistent with both. The author investigates the formal properties of both kinds of revision, calling them respectively ``internal'' and ``external'' base revision, again obtaining syntactic characterizations. He also suggests that external base revision is not just a mathematical possibility but has much to recommend it.