Systematic Construction of Natural Deduction Systems for Many-Valued Logics

Systematic Construction of Natural Deduction Systems for Many-Valued Logics

23rd International Symposium on Multiple Valued Logic. Sacramento, CA, May 1993 Proceedings. (IEEE Press, Los Alamitos, 1993) pp. 208{213 Systematic Construction of Natural Deduction Systems for Many-valued Logics Matthias Baaz∗ Christian G. Ferm¨ullery Richard Zachy Technische Universit¨atWien, Austria Abstract sion.) Each position i corresponds to one of the truth values, vm is the distinguished truth value. The in- A construction principle for natural deduction sys- tended meaning is as follows: Derive that at least tems for arbitrary finitely-many-valued first order log- one formula of Γm takes the value vm under the as- ics is exhibited. These systems are systematically ob- sumption that no formula in Γi takes the value vi tained from sequent calculi, which in turn can be au- (1 i m 1). ≤ ≤ − tomatically extracted from the truth tables of the log- Our starting point for the construction of natural ics under consideration. Soundness and cut-free com- deduction systems are sequent calculi. (A sequent is pleteness of these sequent calculi translate into sound- a tuple Γ1 ::: Γm, defined to be satisfied by an ness, completeness and normal form theorems for the interpretationj iffj for some i 1; : : : ; m at least one 2 f g natural deduction systems. formula in Γi takes the truth value vi.) For each pair of an operator 2 or quantifier Q and a truth value vi 1 Introduction we construct a rule introducing a formula of the form 2(A ;:::;A ) or (Qx)A(x), respectively, at position i The study of natural deduction systems for many- 1 n of a sequent. The resulting calculi are shown to be valued logics can be motivated by the following two is- sound and cut-free complete by Sch¨utte'sreduction sues: (1) Many-valued logics provide a general frame- tree method. work for the investigation of properties of classical (two-valued) sytems. (2) A general construction of Every sequent rule introducing a formula at a non- sound and complete natural deduction calculi leads to distinguished position is converted into an elimina- an adequate syntactical (proof-theoretic) characteri- tion rule; the sequent rule introducing a formula at zation of many-valued logics for which one wants to the distinguished position is transformed into an in- emphasize the r^oleof a particular truth value. (For troduction rule (in the sense of natural deduction). standard logics, such as the families of G¨odelund Any natural deduction derivation can be translated Lukasiewicz logics, one usually considers such distin- into a derivation of the corresponding sequent calcu- guished truth values.) lus. On the other hand, any cut-free sequent calcu- We consider finitely-many-valued first order logics lus proof translates into a normal natural deduction with arbitrary truth-functional connectives and dis- derivation. (Here normal means that for no branch tribution quantifiers (see Definition 2.2). A natural of the proof tree an elimination follows an introduc- deduction derivation for a logic with the truth values tion; this excludes maximal segments in the sense of Prawitz [1971].) Consequently, the natural deduc- v1; : : : ; vm is defined as a derivation f g tion sytems are sound and complete and every deriva- Γ ::: Γ tion can be transformed into a normal derivation. 1 . m 1 j . j − Such derivations consist of \analytical" paths. Γm 2 Preliminaries where Γi (1 i m) are sets of formulas (Γ1 ::: ≤ ≤ j j Γm 1 represents the assumptions, Γm is the conclu- 2.1. Definition A language for a logic L consists − L ∗Technische Universit¨at Wien, Institut f¨ur Algebra und of: (1) free variables, (2) bound variables, (3) predi- Diskrete Mathematik E118.2, Wiedner Hauptstraße 8{10, A- cate symbols, (4) propositional connectives, (5) quan- 1040 Wien, Austria, [email protected] tifiers, and (6) auxiliary symbols: \(", \)", \," yTechnische Universit¨at Wien, Institut f¨ur Computer- sprachen E185.2, Resselgasse 3/1, A-1040 Wien, Austria, We use a, b, c, . to denote free variables; x, y, z, chrisf, zach @logic.tuwien.ac.at . to denote bound variables; P , Q, R, . to denote f g predicate symbols; 2 to denote connectives; and Q to where a is a new free variable, and M(d=a) is de- denote quantifiers, all possibly indexed. fined as the interpretation equal to M, except that M(d=a)a = d. 2.2. Definition A matrix L for a language is given by: L 3 Sequent calculi (1) a nonempty set of truth values V = v1; : : : ; vm of size m, f g Sequent calculi for classical logic were intro- (2) an abstract algebra V with domain V of appro- duced by Gentzen [1934] and were later general- priate type: For every n-place connective 2 of ized to the many-valued case by Schroter¨ [1955], there is an associated truth function 2: V n VL, Rousseau [1967], and others. More recently, equiv- ! (3) for every quantifier Q, an associatede truth func- alent formulations for tableaux calculi were given tion Q: }(V ) V (see, e.g., Carnielli [1987] or Hahnle¨ [1991]). The n f;g ! A languagee and a matrix for it together fully deter- method used here can also be used to obtain calculi mine a logic L. L is said to be m-valued. for transformation into clause form for many-valued resolution (see Baaz and Fermuller¨ [1992]). The intended meaning of a truth function for a propo- sitional connective is obvious and perfectly analogous 3.1. Definition An (m-valued) sequent is an m- to the two-valued case. A truth function for a quanti- tuple of finite sets Γi of formulas, denoted thus: fier is a mapping from nonempty sets of truth values to truth values: given a quantified formula (Qx)F (x), Γ1 Γ2 ::: Γm j j j such a set of truth values describes the situation where For convenience, we will abbreviate Γ ∆ by Γ; ∆; the instances of F take exactly the truth values in this [ set as values under a given interpretation. Γ F by Γ; F ; and sometimes Γ1 ::: Γm by [m f g j j Γl l=1. We say that F stands (or occurs) at place i, 2.3. Example The matrix for the three-valued G¨odel j j if F Γi; vi then is the truth value corresponding to logic G3 consists of: place2i. (1) The set of truth values V = f; ; t (2) The truth functions for the connectives,f ∗ g e.g.: 3.2. Definition An interpretation M is said to sat- isfy a sequent Γ1 ::: Γm, if there is an i (1 i m) t f f t j j ≤ ≤ : _ ∗ ⊃ ∗ and a formula F Γi, s.t. valM(F ) = vi. A sequent is t f t t t t f t t t called valid, if it is2 satisfied under every interpretation. t f t t ∗ ∗ ∗ ∗ ∗ ∗ f t f t f t f t 3.3. Definition An introduction rule for a connec- ∗ ∗ tive 2 at place i in the logic L is a schema of the (3) The truth functions for the quantifiers and form: (generalized and ): 8 9 ^ _ j j j j t = t t = t DΓ1 ; ∆1 ::: Γm; ∆mE 8f g 9f g j j j I et; = et; = t 2 2:i 8f ∗} ∗ 9f ∗} Γ1 ::: Γi; 2(A1;:::;An) ::: Γm e t; f = f e t; f = t j j j j 8f g 9f g j et; ; f = f et; ; f = t where the arity of 2 is n, I is a finite set, Γl = j I Γl , 8f ∗ g 9f ∗ g S 2 e = e = ∆j A ;:::;A and the following condition holds: 8{∗} ∗ 9{∗} ∗ l 1 n e ; f = f e ; f = Let⊆ fM be an interpretation.g Then the following are 8{∗ g 9{∗ g ∗ e f = f e f = f equivalent: 8f g 9f g e e 2.4. Definition An interpretation M is a mapping of (1) 2(A1;:::;An) takes the truth value vi under M. j j all free variables to elements of a given domain D, and (2) For j I, M satisfies the sequents ∆1 ::: ∆m. of predicate symbols to functions of type Dn V . 2 j j ! A valuation valM is a mapping that extends the in- It should be stressed that the introduction rules for a terpretation to formulas via the truth functions given connective at a given place are far from being unique: in the matrix. We only give the precise definition of Let the expression Avl denote the statement \A takes the valuation function for a quantified formula: the truth value vl". Then every introduction rule for 2(A1;:::;An) at place i corresponds to a con- junction of disjunctions of some Avl which is true valM (Qx)G(x) = Q valM(d=a)G(a); [ e d D iff 2(A1;:::;An) takes the truth value vi (namely, 2 m vl j j I l=1 A ∆j A ). Any such conjunctive normal j; w = vl A(t ) vl V j . Again, it should be V 2 W W 2 l 6 g [ f j 2 n g vi form for 2(A1;:::;An) will do. stressed that in general this is not the only possible In particular, the truth table for 2 immediately rule. yields a complete conjunctive normal form, the cor- 3.6. Example Consider the universal quantifier in responding rule is as in Definition 3.3, with: I V n 8 ⊆ three-valued G¨odellogic G3 given in Example 2.3. In- is the set of all n-tuples j = (w1; : : : ; wn) of truth val- j tuitively, ( x)A(x) takes the value f, if A(t) is false for ues such that 2(w1; : : : ; wn) = vi; and ∆ = Ak 8 6 l f j some t; t, if A(a) is true for all a; and , if A(t) takes 1 k n; vl =ewk .

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    6 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us