Online research in philosophy

This paper is closely related to investigations of abstract properties of basic logical notions expressible in terms of closure spaces as they were begun by A. Tarski (see [6]). We shall prove many properties of -conjunctive closure spaces (X is -conjunctive provided that for every two elements of X their conjunction in X exists). For example we prove the following theorems:1. For every closed and proper subset of an -conjunctive closure space its interior is empty (i.e. it is a boundary set). 2. If X is an -conjunctive closure space which satisfies the -compactness theorem and [X] is a meet-distributive semilattice (see [3]), then the lattice of all closed subsets in X is a Heyting lattice. 3. A closure space is linear iff it is an -conjunctive and topological space. 4. Every continuous function preserves all conjunctions.

In this paper we study the relations between the fragment L of classical logic having just conjunction and disjunction and the variety D of distributive lattices, within the context of Algebraic Logic. We prove that these relations cannot be fully expressed either with the tools of Blok and Pigozzi's theory of algebraizable logics or with the use of reduced matrices for L. However, these relations can be naturally formulated when we introduce a new notion of model of a sequent calculus. When applied to a certain natural calculus for L, the resulting models are equivalent to a class of abstract logics (in the sense of Brown and Suszko) which we call distributive. Among other results, we prove that D is exactly the class of the algebraic reducts of the reduced models of L, that there is an embedding of the theories of L into the theories of the equational consequence (in the sense of Blok and Pigozzi) relative to D, and that for any algebra A of type (2,2) there is an isomorphism between the D-congruences of A and the models of L over A. In the second part of this paper (which will be published separately) we will also apply some results to give proofs with a logical flavour for several new or well-known lattice-theoretical properties.

In this paper we study the relations between the fragment L of classical logic having just conjunction and disjunction and the variety D of distributive lattices, within the context of Algebraic Logic. We prove that these relations cannot be fully expressed either with the tools of Blok and Pigozzi's theory of algebraizable logics or with the use of reduced matrices for L. However, these relations can be naturally formulated when we introduce a new notion of model of a sequent calculus. When applied to a certain natural calculus for L, the resulting models are equivalent to a class of abstract logics (in the sense of Brown and Suszko) which we call distributive. Among other results, we prove that D is exactly the class of the algebraic reducts of the reduced models of L, that there is an embedding of the theories of L into the theories of the equational consequence (in the sense of Blok and Pigozzi) relative to D, and that for any algebra A of type (2,2) there is an isomorphism between the D-congruences of A and the models of L over A. In the second part of this paper (which will be published separately) we will also apply some results to give proofs with a logical flavour for several new or well-known lattice-theoretical properties.

Metalogical properties that have traditionally been studied in the deductive system context (see, e.g., [21]) and transferred later to the institution context [33], are here formulated in the -institution context. Preservation under deductive equivalence of -institutions is investigated. If a property is known to hold in all algebraic -institutions and is preserved under deductive equivalence, then it follows that it holds in all algebraizable -institutions in the sense of [36].

The notation and terminology of this paper follow [2], and are dual to those of [6] and [7]. If L is a language in the narrow sense, Cn may be any consequence operation on sets of sentences of L that includes classical sentential logic. Henceforth when we talk of the language L we intend to include reference to some fixed, though unspecified, operation Cn. X is a deductive system if X = Cn(X). Sentences x, z that are logically equivalent with respect to Cn – that is x ∈ Cn({z}) and z ∈ Cn({x}) – are identified. If X and Z are systems we often write X x instead of x ∈ X and Z X instead of X ⊆ Z. If X = Cn({x}) for some sentence x, X is (finitely) axiomatizable. The set theoretical intersection of X and Z has the logical force of disjunction, and is written X ∨ Z; Cn(X ∪ Z), the smallest system to include both X and Z, is written XZ. If K is a family of systems, [K] and [K] may be defined in an analogous way. The logically strongest system S is the set of all sentences of L; the weakest system T is defined as Cn(∅). The autocomplement Z of Z is defined to be the strongest system to complement T, namely the system [{Y : Y ∨ Z = T}]. More generally we may define X − Z as [{Y : X Y ∨ Z}]. In terms of this operation to remainder, Z is identical with T − Z. The class of all deductive systems forms a distributive lattice under the operations of concatenation and ∨; and indeed a Brouverian algebra (a relatively authocomplemented lattice with unit) under concatenation, ∨ – and T.

The notation and terminology of this paper follow [2], and are dual to those of [6] and [7]. If L is a language in the narrow sense, Cn may be any consequence operation on sets of sentences of L that includes classical sentential logic. Henceforth when we talk of the language L we intend to include reference to some fixed, though unspecified, operation Cn. X is a deductive system if X = Cn(X). Sentences x, z that are logically equivalent with respect to Cn – that is x ∈ Cn({z}) and z ∈ Cn({x}) – are identified. If X and Z are systems we often write X x instead of x ∈ X and Z X instead of X ⊆ Z. If X = Cn({x}) for some sentence x, X is (finitely) axiomatizable. The set theoretical intersection of X and Z has the logical force of disjunction, and is written X ∨ Z; Cn(X ∪ Z), the smallest system to include both X and Z, is written XZ. If K is a family of systems, [K] and [K] may be defined in an analogous way. The logically strongest system S is the set of all sentences of L; the weakest system T is defined as Cn(∅). The autocomplement Z of Z is defined to be the strongest system to complement T, namely the system [{Y : Y ∨ Z = T}]. More generally we may define X − Z as [{Y : X Y ∨ Z}]. In terms of this operation to remainder, Z is identical with T − Z. The class of all deductive systems forms a distributive lattice under the operations of concatenation and ∨; and indeed a Brouverian algebra (a relatively authocomplemented lattice with unit) under concatenation, ∨ – and T.

There exist important deductive systems, such as the non-normal modal logics, that are not proper subjects of classical algebraic logic in the sense that their metatheory cannot be reduced to the equational metatheory of any particular class of algebras. Nevertheless, most of these systems are amenable to the methods of universal algebra when applied to the matrix models of the system. In the present paper we consider a wide class of deductive systems of this kind called protoalgebraic logics. These include almost all (non-pathological) systems of prepositional logic that have occurred in the literature. The relationship between the metatheory of a protoalgebraic logic and its matrix models is studied. The following results are obtained for any finite matrix model U of a filter-distributive protoalgebraic logic : (I) The extension U of is finitely axiomatized (provided has only finitely many inference rules); (II) U has only finitely many extensions.

In our previous paper Algebraic Logic for Classical Conjunction and Disjunction we studied some relations between the fragmentL of classical logic having just conjunction and disjunction and the varietyD of distributive lattices, within the context of Algebraic Logic. The central tool in that study was a class of closure operators which we calleddistributive, and one of its main results was that for any algebraA of type (2,2) there is an isomorphism between the lattices of allD-congruences ofA and of all distributive closure operators overA. In the present paper we study the lattice structure of this last set, give a description of its finite and infinite operations, and obtain a topological representation. We also apply the mentioned isomorphism and other results to obtain proofs with a logical flavour for several new or well-known lattice-theoretical properties, like Hashimoto's characterization of distributive lattices, and Priestley's topological representation of the congruence lattice of a bounded distributive lattice.

A pair of deductive systems (S,S’) is Leibniz-linked when S’ is an extension of S and on every algebra there is a map sending each filter of S to a filter of S’ with the same Leibniz congruence. We study this generalization to arbitrary deductive systems of the notion of the strong version of a protoalgebraic deductive system, studied in earlier papers, and of some results recently found for particular non-protoalgebraic deductive systems. The necessary examples and counterexamples found in the literature are described.

The notion of an algebraic semantics of a deductive system was proposed in [3], and a preliminary study was begun. The focus of [3] was the definition and investigation of algebraizable deductive systems, i.e., the deductive systems that possess an equivalent algebraic semantics. The present paper explores the more general property of possessing an algebraic semantics. While a deductive system can have at most one equivalent algebraic semantics, it may have numerous different algebraic semantics. All of these give rise to an algebraic completeness theorem for the deductive system, but their algebraic properties, unlike those of equivalent algebraic semantics, need not reflect the metalogical properties of the deductive system. Many deductive systems that don't have an equivalent algebraic semantics do possess an algebraic semantics; examples of these phenomena are provided. It is shown that all extensions of a deductive system that possesses an algebraic semantics themselves possess an algebraic semantics. Necessary conditions for the existence of an algebraic semantics are given, and an example of a protoalgebraic deductive system that does not have an algebraic semantics is provided. The mono-unary deductive systems possessing an algebraic semantics are characterized. Finally, weak conditions on a deductive system are formulated that guarantee the existence of an algebraic semantics. These conditions are used to show that various classes of non-algebraizable deductive systems of modal logic, relevance logic and linear logic do possess an algebraic semantics.