Abstract
In this paper we provide a simplified semantics for the logic \(KD45(\mathbf {G})\), i.e. the many-valued Gödel counterpart of the classical modal logic KD45. More precisely, we characterize \(KD45(\mathbf {G})\) as the set of valid formulae of the class of possibilistic Gödel Kripke Frames \(\langle W, \pi \rangle \), where W is a non-empty set of worlds and \(\pi : W \rightarrow [0, 1]\) is a normalized possibility distribution on W.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Strictly speaking, the possibility measure is indeed a generalization of the classical one, but the necessity measure is not, since \(x \Rightarrow 0 \ne 1-x\).
- 2.
Remember that \(u \sim _\varphi v \sim _\varphi w\).
- 3.
Note that, in this case, the first member of the union defining \(\varGamma _{\psi ,u}\) is the empty set.
References
Banerjee, M., Dubois, D.: A simple logic for reasoning about incomplete knowledge. Int. J. Approximate Reasoning 55(2), 639–653 (2014)
Bou, F., Esteva, F., Godo, L., Rodríguez, R.: On the minimum many-valued modal logic over a finite residuated lattice. J. Logic Comput. 21(5), 739–790 (2011)
Bou, F., Esteva, F., Godo, L.: On possibilistic modal logics defined over MTL-chains. In: Montagna, F. (ed.) Petr Hájek on Mathematical Fuzzy Logic. Outstanding Contributions to Logic, vol. 6, pp. 225–244. Springer, Switzerland (2015)
Caicedo, X., Rodríguez, R.: Standard Gödel modal logics. Studia Logica 94(2), 189–214 (2010)
Caicedo, X., Rodríguez, R.: Bi-modal Gödel modal logics. J. Logic Comput. 25–1, 37–55 (2015)
Dellunde, P., Godo, L., Marchioni, E.: Extending possibilistic logic over Gödel logic. Int. J. Approx. Reasoning 52(1), 63–75 (2011)
Dubois, D., Lang, J., Prade, H.: Possibilistic logic, In: Gabbay, et al. (ed.) Handbook of Logic in Artificial Intelligence and Logic Programming, Non monotonic Reasoning and Uncertain Reasoning, vol. 3, pp. 439–513. Oxford UP (1994)
Dubois, D., Prade, H.: Possibilistic logic: a retrospective and prospective view. Fuzzy Sets Syst. 144, 3–23 (2004)
Dubois, D., Prade, H., Schockaert, S.: Reasoning about uncertainty and explicit ignorance in generalized possibilistic logic. Proc. ECAI 2014, 261–266 (2014)
Fitting, M.: Many valued modal logics. Fundamenta Informaticae 15, 325–254 (1991)
Hájek, P.: Metamathematics of Fuzzy Logic. Trends in Logic, vol. 4. Kluwer, Dordrecht (1998)
Hájek, P., Harmancová, D., Esteva, F., Garcia, P., Godo, L.: On modal logics for qualitative possibility in a fuzzy setting. In: Proceeding of the 94 Uncertainty in Artificial Intelligence Conference (UAI 1994), pp. 278–285. Morgan Kaufmann (1994)
Pietruszczak, A.: Simplified Kripke style semantics for modal logics K45, KB4 and KD45. Bull. Sect. Logic 38(3/4), 163–171 (2009)
Acknowledgments
The authors are grateful to the anonymous reviewers for their helpful comments. They acknowledge partial support by the H2020-MSCA-RISE-2015 project SYSMICS, the Spanish MINECO/FEDER project RASO (TIN2015-71799-C2-1-P) and the Argentinean project PIP CONICET 11220150100412CO.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix
Appendix
Proof of Lemma 2
Proof
First of all, notice that if \(\nu \) satisfies the condition b, then necessarily \(\alpha < \beta \). Let \(B= \{ \nu (\lambda ) : \lambda \in \varDelta _\varphi , \nu (\lambda )< 1\} \cup \{0 \} = \{ b_0 = 0< b_1 < \ldots b_N \}\). Obviously, \(b_N < 1\). For each \(0 \le i \le N\), pick \(\lambda _i \in \varDelta _\varphi \) such that \(\nu (\lambda _i) = b_i\). Define now a continuous strictly function \(g:[0,1]\mapsto [0,\delta ) \cup \{1\}\) such that
\(g(1)=1\)
\(g(b_i)= v(\lambda _i)\) for every \(0 \le i \le N\)
\(g[(b_N, 1)]= (\alpha , \delta ).\)
Notice that \(\alpha = g(b_N)\). In addition, define another continuous strictly increasing function \(h:[0,1] \mapsto [\delta , 1]\) such that
\(h(0)=\delta \)
\(h[(0,\beta )] =(\delta , \beta )\)
\(h(x) = x\), for \(x \in [\beta , 1].\)
Then we define the valuation \(w: Var \cup X \rightarrow [0,1]\) as follows:
\(w(p) = \left\{ \begin{array}{ll} g(\nu (p)), &{} \text{ if } \nu (p) < 1,\\ h(u(p)), &{} \text{ if } \nu (p) = 1. \end{array} \right. \)
First of all, let us show by induction that this extends to any propositional formula, that is,
\(w(\varphi ) = \left\{ \begin{array}{ll} g(\nu (\varphi )), &{} \text{ if } \nu (\varphi ) < 1,\\ h(u(\varphi )), &{} \text{ if } \nu (p) = 1. \end{array} \right. \)
Note that, since g and h are strictly increasing mappings, \(g\circ \nu \) and \(h \circ u\) are valuations as well. So, in the induction steps below we only need to check that everything is fine when both are used at the same time when evaluating a compound formula. The base case holds by definition.
-
Assume \(\psi = \psi _1 \wedge \psi _2\). We only check the case when \(v(\psi ) < 1\) and \(v(\psi _1) < 1\) and \(v(\psi _2) = 1\). Then \(w(\psi ) = \min (w(\psi _1), w(\psi _2)) = \min (g(\nu (\psi _1)), h(u(\psi _2))) = g(\nu (\psi _1))\), since \(g(\nu (\psi _1)) < \delta \le h(u(\psi _2)\). But, \(g(\nu (\psi _1)) = \min ( g(\nu (\psi _1)), 1) = \min ( g(\nu (\psi _1)), g(\nu (\psi _2))) = g(\nu (\psi _1\wedge \psi _2)) = g(\nu (\psi ))\).
-
Assume \(\psi = \psi _1 \rightarrow \psi _2\), and consider two subcases:
(1) \(v(\psi _1) < 1\) and \(v(\psi _2) = 1\). Then \(v(\psi _1 \rightarrow \psi _2) = 1\) and \(w(\psi ) = w(\psi _1) \Rightarrow w(\psi _2) = g(\nu (\psi _1)) \Rightarrow h(u(\psi _2)) = 1 = h(u(\psi _1)) \Rightarrow h(u(\psi _2)) = h(u(\psi _1 \rightarrow \psi _2)) = h(u(\psi )). \) (2) \(v(\psi _1) = 1\) and \(v(\psi _2) < 1\). Then \(v(\psi _1 \rightarrow \psi _2) = v(\psi _2) < 1\) and \(w(\psi ) = w(\psi _1) \Rightarrow w(\psi _2) = h(u(\psi _1)) \Rightarrow g(\nu (\psi _2)) = g(\nu (\psi _2)) = g(\nu (\psi _1)) \Rightarrow g(\nu (\psi _2)) = g(\nu (\psi _1 \rightarrow \psi _2)) = g(\nu (\psi ))\).
Properties 1 – 6 in the statement of Lemma 2 now directly follow from the above.
Finally, we prove that \(w \in W^v\). By definition of w it is clear that \(w \sim _\varphi v\). It remains to check that w validates all the axioms. The axioms of \(\mathcal {G}\) are evaluated to 1 by any Gödel valuation. As for the specific axioms of \(KD45(\mathbf {G})\), it is an immediate consequence of Property 3 because it implies that if \(\nu (\psi \rightarrow \phi )=1\) then \(w(\psi \rightarrow \phi )=1\). \(\blacksquare \)
Claim 1 from Lemma 3. If \(u(\Box \psi )=\alpha <1\), for every \(\varepsilon > 0\), there exists a valuation \(w\in W\) such that \(\pi ^\varphi (w) > w(\psi )\) and \(w(\psi ) < \alpha + \varepsilon \), and thus \(\pi ^\varphi (w) \Rightarrow w(\psi ) = w(\psi ) < \alpha + \varepsilon \).
Proof
By definition of Gödel’s implication \(\Rightarrow \) in [0,1], to grant the required conditions on w it is enough to find \(w \in W\) such that \(\alpha \le w(\psi )\) and, for any \(\theta \in Sub(\varphi )\), \(u(\Box \theta )\le w(\theta ) \le u(\Diamond \theta ) \le \alpha \). This is achieved in two stages:
-
first producing a valuation \( \nu \in W\) satisfying \(\nu (\psi )<1\) and preserving the relative ordering conditions the values \(w(\theta )\) must satisfy, conditions which may be coded by a theory \(\varGamma _{\psi ,u}\);
-
and then moving the values \(\nu (\theta )\), for \(\theta \in Sub(\varphi )\), to the correct valuation w by composing \(\nu \) with an increasing bijection of [0,1].
Assume \(u(\Box \psi )=\alpha <1\), and define (all formulas involved ranging in \(\mathcal {L}_{\square \Diamond }(Var)\))
Then we have \(u(\square \xi ) > \alpha \) for each \(\xi \in \varGamma _{\psi ,u}\). Indeed, first recall that, by \(U_\Box \) and \(U_\Diamond \) of Proposition 2, for any \(\lambda \in \varDelta _{\varphi }\) we have \(u(\lambda ) = u(\Box \lambda ) = u(\Diamond \lambda )\). We analyse case by case. For the first set of formulas, it is clear by construction. For the second set, we have \(u(\square (\lambda \rightarrow \theta )) \ge u(\Diamond \lambda \rightarrow \square \theta )= u(\Diamond \lambda ) \Rightarrow u(\Box \theta ) = u(\lambda ) \Rightarrow u(\Box \theta ) = 1\), by FS2. For the third, by FS2 and P, we have \(u(\square ((\theta \rightarrow \lambda )\rightarrow \lambda )) \ge u(\Diamond (\theta \rightarrow \lambda ) \rightarrow \Box \lambda ) \ge u((\Box \theta \rightarrow \Diamond \lambda ) \rightarrow \Box \lambda ) = 1\), since \( u(\Box \lambda ) = u(\Diamond \lambda ) = u(\square \theta \rightarrow \Diamond \lambda ) < 1\). The fourth and fifth cases are very similar to the second and third ones respectively.
This implies
otherwise there would exist \(\xi _{1},\ldots ,\xi _{k}\in \varGamma _{\psi ,u}\) such that \(\xi _{1},\ldots ,\xi _{k} \vdash _{KD45(\mathbf {G})} \psi .\) In such a case, we would have \(\Box \xi _{1},\ldots ,\Box \xi _{k} \vdash _{KD45(\mathbf {G})}\Box \psi \) by Nec and \(K_{\square }\). Then \(\Box \xi _{1},\ldots ,\Box \xi _{k},ThKD45(\mathbf {G})\vdash _G\Box \psi \) by Lemma 1 and thus by Proposition 1 (i), and recalling that \(u(ThKD45(\mathbf {G}))=1,\)
a contradiction. Therefore, by Proposition 1 (ii) there exists a valuation \(\nu :Var\cup X\mapsto [0,1]\) such that \(\nu (\varGamma _{\psi ,u} \cup ThKD45(\mathbf {G}))=1\) and \( \nu (\psi )<1\). This implies the following relations between u and \(\nu ,\) that we list for further use. Given \(\lambda \in \varDelta _\varphi , \theta \in \mathcal {L}_{\square \Diamond }(Var) \), we have :
-
#1. If \(u(\lambda )>\alpha \) then \(\nu (\lambda )=1\) (since then \(\lambda \in \varGamma _{\psi ,u})\).
-
#2. If \(u(\lambda )\le u(\square \theta )\) then \(\nu (\lambda )\le \nu (\theta )\) (since then \(\lambda \rightarrow \theta \in \varGamma _{\psi ,u})\). In particular, if \(\lambda _1,\lambda _2 \in \varDelta _\varphi \) and \(u(\lambda _1)\le u(\Box \lambda _2) = u(\lambda _2)\) then \(\nu (\lambda _1)\le \nu (\lambda _2)\). Furthermore,if \(\Box \theta \in \varDelta _\varphi \) then from \(u(\Box \theta ) = u(\Box \theta )\) by #2, \(\nu (\Box \theta ) \le \nu (\theta )\). That means, taking \(\theta = \psi \), \(\nu (\Box \psi ) \le \nu (\psi ) < 1\).
-
#3. If \(u(\lambda )< u(\square \theta )< 1\) then \(\nu (\lambda )< \nu (\theta )\) or \(\nu (\lambda )=1\) (since then \((\theta \rightarrow \lambda )\rightarrow \lambda )\in \varGamma _{\psi ,u}\)). In particular, if \(\lambda _1,\lambda _2 \in \varDelta _\varphi \), \(u(\lambda _1) < u(\lambda _2)\) and \(u(\lambda _2) \le u(\Box \psi )= \alpha \) then \(\nu (\lambda _1)< \nu (\psi ) < 1\) and thus \(\nu (\lambda _1) < \nu (\lambda _2)\). This means that \(\nu \) preserves in a strict sense the order values by u of the formulas \(\lambda \in \varDelta _\varphi \) such that \(u(\lambda ) \le \alpha \).
-
#4. If \( u(\Diamond \theta ) \le u(\lambda )\) then \(\nu (\theta ) \le \nu (\lambda )\) (because \(\theta \rightarrow \lambda \in \varGamma _{\psi ,u}\)). In particular, if \(\Diamond \theta \in \varDelta _\varphi \) then \(\nu (\theta ) \le \nu (\Diamond \theta )\).
-
#5. If \( u(\Diamond \theta )< u(\lambda )< 1\) then \(\nu (\theta ) < \nu (\lambda )\) or \(\nu (\theta ) =1\). In particular, if \(\lambda _1, \lambda _2 \in \varDelta _\varphi \) and \(u(\lambda _1)< u(\lambda _2)\le \alpha = u(\Box \psi )\) then \(\nu (\lambda _1)< \nu (\lambda _2)\). Furthermore, if \(u(\lambda _2)>0\) then \(\nu (\lambda _2)>0\) (making \(\lambda _1:=\Diamond \bot \) since \(u(\bot )=u(\Diamond \bot )=0\)).
According to the properties #1, #2 and #3, it is clear that \(\nu \) satisfies the conditions of Lemma 2. Consequently, for all \(\epsilon > 0\) (such that \( \alpha +\varepsilon <\beta \)), taking \(\delta = \alpha + \varepsilon \) in Lemma 2, there exists a valuation \(w \in W^{v}\) such that \(w(\psi ) < \alpha +\varepsilon = \delta \). Then in order to finish our proof, it remains to show that:
To do so, we will prove that, for any \(\lambda \in sub(\varphi )\), both implications in (3) are greater than \(\delta \).Footnote 2 First we prove it for the first implication by cases:
-
- If \(v(\Box \lambda ) \le \alpha < 1\) then \(\min (v(\Box \lambda ) \Rightarrow w(\lambda ), w(\lambda ) \Rightarrow v(\Diamond \lambda )) = 1\). Indeed, first of all, by #2, from \(u (\Box \lambda ) = v(\Box \lambda ) \le \alpha = u(\Box \psi )\) it follows \(\nu (\Box \lambda ) \le \nu (\psi ) < 1\). Now, since \(u(\Box \lambda ) \le u(\Box \lambda )\), by #2, we have \(1 \ne \nu (\Box \lambda ) \le \nu (\lambda )\), and by 3 of Lemma 2 we have \( v(\Box \lambda ) = w(\Box \lambda ) \le w(\lambda )\). Then \(v(\Box \lambda ) \Rightarrow w(\lambda ) =1\).
-
- If \(v(\Box \lambda ) > \alpha \) then by #1 and #2, \(1=\nu (\Box \lambda )\le \nu (\lambda )\). Therefore, by 1 of Lemma 2, \(w(\lambda )>\delta \) which implies \(v(\Box \lambda ) \Rightarrow w(\lambda ) > \delta \).
For the second implication we also consider two cases:
-
- If \(v(\Diamond \lambda ) = u(\Diamond \lambda ) > \delta \) then it is obvious that \(w(\lambda ) \Rightarrow v(\Diamond \lambda ) > \delta \).
-
- If \(u(\Diamond \lambda ) < \delta \), by definition of \(\delta \) and taking into account that \(\Diamond \lambda \in \varDelta _\varphi \), then \(u(\Diamond \lambda ) < \alpha \). Now from \(u(\Diamond \lambda ) = u(\Diamond \lambda )\) we obtain by #4, that \(\nu (\lambda ) \le \nu (\Diamond \lambda ) < 1\). Then by Lemma 2 we have \(w(\lambda ) \le w(\Diamond \lambda ) = v(\Diamond \lambda )\) and thus \(w(\lambda ) \Rightarrow v(\Diamond \lambda ) = 1\). \(\blacksquare \)
Claim 2 from Lemma 3. If \(u(\Diamond \psi )=\alpha >0\) then, for any \(\varepsilon >0,\) there exists a valuation \(w'\in W\) such that \(w'(\psi )=1\) and \(\pi ^\varphi (w')\ge \alpha -\varepsilon \), and thus \(\min (w'(\psi ), \pi ^\varphi (w')) \ge \alpha -\varepsilon \).
Proof
Assume \(u(\Diamond \psi )=\alpha >0\) and define \(\varGamma _{\psi ,u}\) in the same way that it was defined in the proof of Claim 1. Then we consider two cases:
-
- If \(u(\Diamond \psi )=1\), let \(U_{\psi ,u}^{{}} = \{ \lambda : \ \lambda \in \varDelta _\varphi \text{ and } u(\lambda ) < 1 \}\). We claim that
otherwise we would have \(\theta _1, \ldots , \theta _n \in \varGamma _{\psi ,u}\) such that \(\vdash _{KD45(\mathbf {G})} \psi \rightarrow ((\theta _1\wedge \ldots \wedge \theta _n) \rightarrow \bigvee U_{\psi ,u})\), and then we would also have \(\vdash _{KD45(\mathbf {G})} \Diamond \psi \rightarrow \Diamond ((\theta _1\wedge \ldots \wedge \theta _n) \rightarrow \bigvee U_{\psi ,u})\), that would imply in turn that \(\vdash _{KD45(\mathbf {G})} \Diamond \psi \rightarrow ((\Box \theta _1\wedge \ldots \wedge \Box \theta _n) \rightarrow \Diamond \bigvee U_{\psi ,u})\). In that case, taking the evaluation u it would yield: \( 1= u(\Diamond \psi ) \le u(\Box \theta _1\wedge \ldots \wedge \Box \theta _n ) \Rightarrow u( \Diamond \bigvee U_{\psi ,u})\), a contradiction, since \(u(\Box \theta _1\wedge \ldots \wedge \Box \theta _n) = 1\) and \(u( \Diamond \bigvee U_{\psi ,u})) < 1\).Footnote 3
Therefore, there is a Gödel valuation \(\nu '\) (not necessarily in W) such that \(\nu '(\psi )=\nu '(\varGamma _{\psi ,u})=\nu '(ThKD45(\mathbf {G}))=1\) and \(\nu '(\bigvee U_{\psi ,u} )<1\). By #2 and #3, it follows that for any \(\lambda _{1},\lambda _{2} \in \varDelta _\varphi \) such that \(u(\lambda _{1}), u(\lambda _{2}) \le \alpha \), we have \(u(\lambda _{1})< u(\lambda _{2}) \le \alpha \ \text{ iff } \ \nu '(\lambda _{1}) < \nu '(\lambda _{2}). \) Thus, \(\nu '\) satisfies the conditions of Lemma 2 because it is strictly increasing in \(\varDelta _\varphi \) (i.e. it satisfies condition b of Lemma 2), and \(\nu '(ThKD45(\mathbf {G}))=1\). Therefore, there exists a valuation \(w' \in W\) such that \(w'(\psi )=1\).
It remains to show that \(\pi ^\varphi (w') = 1\). Indeed, by construction, it holds that \(u(\Box \theta ) \le w'(\theta )\le u(\Diamond \theta )\), and hence \(\min (u(\Box \theta ) \Rightarrow w'(\theta ) , w'(\theta )\Rightarrow u(\Diamond \theta )) =1\).
-
- If \(1> u(\Diamond \psi )=\alpha >0\), then we let \(U_{\psi ,u}^{{}} = \ (\Diamond \psi \rightarrow \psi ) \rightarrow \psi .\) We claim that
otherwise there would exist \(\theta _1, \ldots , \theta _n \in \varGamma _{\psi ,u}\) such that \(\vdash _{KD45(\mathbf {G})} \Box \top \rightarrow ((\theta _1\wedge \ldots \wedge \theta _n) \rightarrow U_{\psi ,u})\), and then we would have \(\vdash _{KD45(\mathbf {G})} \Diamond \Box \top \rightarrow \Diamond ((\theta _1\wedge \ldots \wedge \theta _n) \rightarrow U_{\psi ,u})\), which would imply \(\vdash _{KD45(\mathbf {G})} \Box \top \rightarrow ((\Box \theta _1\wedge \ldots \wedge \Box \theta _n) \rightarrow \Diamond U_{\psi ,u})\). In that case, evaluating with u it would yield \( 1= u(\Box \top ) \le u(\Box \theta _1\wedge \ldots \wedge \Box \theta _n ) \Rightarrow u( \Diamond U_{\psi ,u}))\), contradiction, since \(u(\Box \theta _1\wedge \ldots \wedge \Box \theta _n) > \alpha \) and \(u( \Diamond U_{\psi ,u})) \le \alpha \) (because \(u(\Diamond ((\Diamond \psi \rightarrow \psi ) \rightarrow \psi ))\le u(\Box (\Diamond \psi \rightarrow \psi ) \rightarrow \Diamond \psi )\le u(\Diamond \psi )\le \alpha \)).
Therefore, there is an evaluation \(\nu '\) such that \(\nu '(ThKD45(\mathbf {G}) )=\nu '(\varGamma _{\psi ,u})=1\) and \(\nu '(\bigvee U_{\psi ,u} )<1\). Hence, we can conclude that the three pre-conditions \(\mathbf a\), \(\mathbf b\) and \(\mathbf c\) required in Lemma 2 are satisfied. In addition, the following condition is also satisfied:
-
d. \(\nu '(\Diamond \psi ) = \nu '(\psi )\).
At this point, we can now do a proof dual to the one for Claim 1. Again, by Lemma 2 for \(\delta = \frac{\beta - \alpha }{2}\), we obtain from \(\nu '\) an evaluation \(w' \in W^{v}\) such that \(w'(\psi )=\alpha \). It only remains then to show that \(\pi ^\varphi (w)> \alpha \). But in this case, the proof is the same than the one given for Eq. (3) using \(w'\) instead of w. This finishes the proof. \(\blacksquare \)
Rights and permissions
Copyright information
© 2016 Springer International Publishing Switzerland
About this paper
Cite this paper
Bou, F., Esteva, F., Godo, L., Rodriguez, R.O. (2016). Possibilistic Semantics for a Modal KD45 Extension of Gödel Fuzzy Logic. In: Carvalho, J., Lesot, MJ., Kaymak, U., Vieira, S., Bouchon-Meunier, B., Yager, R. (eds) Information Processing and Management of Uncertainty in Knowledge-Based Systems. IPMU 2016. Communications in Computer and Information Science, vol 611. Springer, Cham. https://doi.org/10.1007/978-3-319-40581-0_11
Download citation
DOI: https://doi.org/10.1007/978-3-319-40581-0_11
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-40580-3
Online ISBN: 978-3-319-40581-0
eBook Packages: Computer ScienceComputer Science (R0)