Bootstrap contraction (Q383564)

This paper introduces the notion of ``bootstrap contraction'', a new kind of contraction functions, which the author argues to be more cognitively realistic than most of the other kinds of contractions so far proposed in the literature. Given a belief set \(K\) and a set \(\mathbb{C}\) of operators on \(K\) which take elements of \(\mathcal{I}\) as inputs. The ``overlap'' of \(\mathbb{C}\), denoted by \(\div_\forall\), is the (partial) operation such that, for any \(i \in \mathcal{I}\): (1) If \(K \div i = K \div^\prime i\) for all \(\div, \div^\prime \in \mathbb{C}\), then \(K \div_\forall i = K \div i\) for any \(\div \in \mathbb{C}\). (2) Otherwise, \(K \div_\forall i\) is undefined. An operation \(\div \in \mathbb{C}\) is ``bootstrappable'' with respect to \(\mathbb{C}\) if and only if for all \(i \in \mathcal{I}\) there is some \(i^\prime \in \mathcal{I}\) such that (\(K \div_\forall i^\prime\) is defined and) \(K \div i = K \div_\forall i^\prime\). A function \(f\) from and to \(\mathcal{I}\) is a ``bootstraping selector'' for \(\div\) if and only if it holds for all \(i \in \mathcal{I}\) that \(K \div i = K \div_\forall f(i)\). The set \(\mathbb{C}\) is bootstrappable if and only if all its members are so. The set \(\mathcal{I}\) of inputs can be either the set of all sentences or the set of all sets of sentences, thus giving rise to singleton, respectively, multiple bootstrap contraction. This paper is focused on the multiple contraction case. If all the contraction operators included in \(\mathbb{C}\) have the same ``outcome set'' -- the set composed of all the possible contraction results -- and \(\mathbb{C}\) is bootstrapable, then the ``bootstraping selector'' \(f\) can be seen as a mechanism for assigning to each different input a certain element of the outcome set (which is defined to be the result of the (bootstrap) contraction by that input). Thus, bootstrap contractions constitute a specific kind of ``Repertoire contractions'' -- a more general type of contraction functions (which are based on some mechanism for selecting for each input \(A\) an element of the outcome set to make up \(K \div A\)) that was introduced and analysed in terms of axioms and their interrelations by the author in [J. Logic Lang. Inf. 22, No. 1, 1--21 (2013; Zbl 1305.03016)]. The main result of the paper is the proof that the following classes of multiple contractions are bootstrapable: Direct partial meet contraction [\textit{C. E. Alchourrón} et al., J. Symb. Log. 50, 510--530 (1985; Zbl 0578.03011); \textit{A. Fuhrmann} and \textit{S. O. Hansson}, J. Logic Lang. Inf. 3, No. 1, 37--75 (1994; Zbl 0791.03014)], base-generated partial meet contraction [\textit{S. O. Hansson}, J. Symb. Log. 58, No. 2, 602--625 (1993; Zbl 0785.03012)], base-generated kernel contraction [\textit{S. O. Hansson}, J. Symb. Log. 59, No. 3, 845--859 (1994; Zbl 0810.03017); \textit{E. Fermé} et al., Stud. Log. 73, No. 2, 183--195 (2003; Zbl 1018.03013)], perimaximal contraction [\textit{S. O. Hansson}, Synthese 190, No. 16, 3325--3348 (2013; Zbl 1284.03122)], and blockage contraction [J. Philos. Log. 42, No. 2, 415--442 (2013; Zbl 1272.03086)]. This result is obtained under the assumption that the original belief set is finite-based and the multiple contraction operators under analysis satisfy ``finite-based outcome'', i.e. the following condition: If \(K\) is finite-based then so is \(K \div A\) for all \(A\). In this regard, the author remarks that these assumptions were made because the aim of this paper is to develop operators with a high degree of cognitive realism. Another significant contribution of this paper is a thorough analysis of the cognitive realism of bootstraping. The author firstly remarks that if \(\mathbb{C}\) is a (bootstrappable) set of operations that differ only in terms of some extra-logical element (e.g. selection functions or incision functions, depending on their particular construction), then bootstrapping replaces that element by the bootstrapping selector. Thus, if the eliminated extra-logical element is cognitively less accessible than a selector, then bootstrapping brings in an improvement in terms of cognitive realism. In particular, the author highlights that the bootstrapping selector represents a form of reasoning (``In order to contract by \(\ldots\) I contract by \(\ldots\)'') that can be found in the reasoning of actual epistemic agents. However, throughout the discussion presented, it is also observed that when a bootstrapping function \(f\) assigns an unmanageably large set \(f(A)\) to a small contractee \(A\) the cognitive problem is not solved. Then, the author argues that, since some contractions are cognitively unmanageable, a possible way for obtaining more cognitively realistic models is replacing standard contraction by shielded contraction [\textit{E. L. Fermé} and \textit{S. O. Hansson}, in: Frontiers in belief revision. Dordrecht: Kluwer Academic Publishers. 85--107 (2001; Zbl 1008.03011)], an alternative operation that does not satisfy the success postulate. The view that motivated this proposal is exposed in some detail and, additionally, it is noted that, just like ordinary contraction, shielded contraction can be bootstrapped.