Abstract
The first aim of this paper is to prove a topological completeness theorem for a weak version of Stalnaker’s logic KB of knowledge and belief. The weak version of KB is characterized by the assumption that the axioms and rules of KB have to be satisfied with the exception of the axiom (NI) of negative introspection. The proof of a topological completeness theorem for weak KB is based on the fact that nuclei (as defined in the framework of point-free topology) give rise to a profusion of topological belief operators that are compatible with the familiar topological knowledge operator Int. Thereby a canonical topological model for weak KB can be constructed. For this canonical model a truth lemma for the modal operators K and B holds such that a completeness theorem for weak KB can be proved in the familiar way. The second aim of this paper is to show that the topological interpretation of knowledge Int comes along with a complete Heyting algebra of belief operators N° that all fit the knowledge operator Int in the sense that the pairs (Int, N°) satisfy all axioms of weak KB. This amounts to a pluralistic relation between knowledge and belief: Knowledge does not fully determine belief, rather it designs a conceptual space for belief operators where different (competing) belief operators coexist that can be compared with each other.
Similar content being viewed by others
Notes
In topological terms Stalnaker’s operator NS is defined as the concatenation IntClInt, where, as usual, the interior kernel operator of a topological space (X, OX) is denoted by Int and the topological closure is denoted by Cl. In this paper this (more or less standard) topological terminology is used throughout. In more detail it is explained in Sect. 3.
The topological reason for the failure of (NI) in general spaces may be informally described as the fact that (NI) requires that many clopen (= open and closed) subsets exist. General topological spaces, however, may lack sufficiently many clopen sets. For instance, connected spaces such as the Euclidean line (ℝ, Oℝ) have only Ø and ℝ as clopen subsets.
For various equivalent definitions of a normal modal logic, see Chellas (1980, Theorem 4.3, p. 115).
Elementary examples based on the Euclidean line (ℝ, Oℝ) show that there are models of weak KB logic the B-fragment of which are not KD45 models (cf. Proposition (4.11)).
The necessity of introducing nuclei for defining the semantics of belief operators distinguishes weak KB logic from original KB logic. Since for KB systems the belief modality B can be defined in terms of K, in these systems the excursion into the theory of nuclei can be avoided, since the semantics of B can be defined in terms of the semantics of K. This is not the case for weak KB. Then, the modal operator B need not be definable in terms of K.
The literature on nuclei in point-free topology has reached a high level of technical sophistication. This paper does not aim to give a full-fledged introduction into the theory of nuclei. Instead, we intend to provide the basic definitions and facts so that the reader can understand that this theory has interesting applications regarding the modal theory of belief and knowledge. For a fuller account, the reader may consult Johnstone (1982), Borceux (1994), or Picado and Pultr (2012, 2021) and the extensive bibliographies on point-free topology mentioned there.
The reader should not be confused by this (established) terminology: every open subset A of X (as an element of OX) defines a closed nucleus and an open nucleus, namely, kA and jA, respectively.
Clearly, a nucleus N and its corresponding belief operator N° determine each other uniquely: N = iN°Int and N° = IntNi. Here, i is, of course, the canonical inclusion i: OX→PX
Sublocales are also called ⇒-ideals.
Actually there is an order-reversing bijection between nuclei and sublocales: A nucleus N:OXOX is uniquely determined by its image i:N(OX)OX. Indeed, N(OX) OX is a sublocale. Thus, a nucleus uniquely determines a sublocale. On the other hand, the inclusion map of a sublocale i:SOX has an adjoint frame map j:OXS such that the concatenation ij:OXOX is a nucleus N [cf. Johnstone (2002, Proposition 1.1.3., p. 486), Picado and Pultr (2012, 5.3.2. Proposition. p. 32)]. We don’t need this result, however.
According to Aiello et al., (2003, p. 896) the canonical topology of the canonical space is the intersection of the Kripke topology and the Stone topology. This entails that this space is compact and dense-in-itself.
If the axiom (NI) of negative introspection is assumed to be valid, the bimodal logic KB boils down to a unimodal logic defined by K since then the belief modality B can be uniquely defined in terms of K, namely B = \(\neg\) K \(\neg\) K (cf. Footnote 4).
This section is somewhat technical. Readers who believe that a knowledge operator Int is always accompanied by many belief operators may therefore skip this section on first reading.
TD is a rather weak axiom satisfied by most topological spaces that “occur in nature”. For instance, Euclidean spaces and, more generally, all T2-spaces, and all T0-Alexandroff spaces are TD-spaces.
Already in Macnab (1981) it is proved that for TD -spaces (X, OX) there is a Boolean isomorphism between PX and the Boolean algebra of regular elements of NUC(OX) [cf. Macnab (1981, Theorem (6.5)(5)]. In contrast, even for the Euclidean line (ℝ, Oℝ) the full structure of NUC(Oℝ) is not fully known up to now (as far as I know).
Roughly, the relation between dense subspaces and dense nuclei of a topological space (X, OX) may be compared with the relation between the field of rational numbers ℚ and the field of complex numbers \(\mathbb{C}\) with respect to their algebraic qualities. A very simple aspect of this issue concerns the solvability of polynomial equations. While there are enough complex numbers to solve all polynomial equations in a neat and elegant way, this does not hold for the more restricted domain of rational numbers ℚ. It is quite difficult to say anything general about the solvability of polynomial equations in rational numbers.
With some more effort it can be easily shown that there exist many spatial nuclei N1, N2 in NUC(Oℝ)d such that N1\(\vee\)N2 is non-spatial and different from NS = IntCl.
Zarycki (1930) erroneously claimed that PF is distributive with respect to \(\cup\) for all subsets A, D of X, not only for closed ones. This error was observed by Vaidyanathaswamy (1947) and Oxtoby (1976). Oxtoby proved a more complex formula for all subsets A, D that yields (6.12)(iv) for closed sets. For our purposes it is sufficient that distributivity ((6.14)(iv)) holds for closed subsets of X.
Simmons (1978, 1982) stated (without explicit proof) that (6.14) (iv) holds, i.e., that the operation PF is distributive with respect to \(\cup\) for closed sets. He then went on to show that PF(AC)C is a nucleus. Actually, Oxtoby proved his more general results on PF only for T1-spaces. A closer inspection of his proof, however, reveals that for the distributivity of PF his proof works for all topological spaces.
A Polish space is a separable topological space that is homeomorphic to a complete metric space [cf. Jech (2002, Definition (4.12), p. 44)].
Already Stalnaker (2006) pointed out that on general topological spaces the operator ClInt does not define a (reasonable) belief operator, since it is not a normal operator, i.e., does not satisfy (2.6)(i) [cf. Stalnaker (2006, p. 195)]. An example for this fact is already provided by the Euclidean line (ℝ, Oℝ).
DSO is an acronym for “Derived Sets are Open.”
References
Aiello, M., van Benthem, J. F. A. K., & Bezhanishvili, G. (2003). Reasoning about space: The modal way. Journal of Logic and Computation, 13(6), 889–920.
Aull, C. E., & Thron, W. J. (1963). Separation axioms between T0 and T1. Indagationes Mathematicae, 24, 26–37.
Baltag, A., Bezhanishvili, N., Özgün, A., & Smets, S. (2017). The topology of full and weak Belief. In H. H. Hansen, S. E. Murray, M. Sadrzadeh, & H. Zeevat (Eds.), Logic, language, and computation (pp. 205–228). Springer.
Baltag, A., Bezhanishvili, N., Özgün, A., & Smets, S. (2019). A topological approach to full belief. Journal of Philosophical Logic, 48, 205–244.
Bezhanishvili, N., & van der Hoek, W. (2014). Structures for epistemic logic. In A. Baltag & S. Smets (Eds.), Johan van Benthem on logic and information dynamics (pp. 381–404). Springer.
Blackburn, P., de Rijke, M., & Venema, Y. (2010). Modal logic, 4th printing. Cambridge University Press.
Borceux, F. (1994). Handbook of categorical algebra 3. Cambridge University Press.
Chellas, B. F. (1980). Modal logic. Cambridge University Press.
Jech, T. (2002). Set theory, 3rd millenium edition, revised and expanded. Heidelberg.
Johnstone, P. (1982). Stone spaces. Cambridge University Press.
Johnstone, P. (1991). The art of pointless thinking: a student’s guide to the category of locales. In H. Herrlich & H. E. Porst (Eds.), Category Theory at Work Research and Exposition in Mathematics (pp. 85–107). Heldermann Verlag.
Johnstone, P. (2002). Sketches of an elephant: a topos theory compendium. Clarendon Press.
Kuratowski, K. (1966). Topology I. Academic Press, New York.
Macnab, D. S. (1981). modal operators on heyting algebras. Algebra Universalis, 12, 5–29.
Mormann, T. (2021). Prototypes, poles, and topological tessellations of conceptual spaces. Synthese, 199, 3675–3710.
Mormann, T. (2022). Topological models of columnar vagueness. Erkenntnis, 87(2), 693–716.
Oxtoby, J. F. (1976). The kernel Operation on subsets of a T1-space, Fundamenta Mathematicae 90(3), 275–284.
Parikh, R., Moss, S, Steinsvold, C. (2007). Topology and epistemic logic. In Handbook of Spatial Logic, 299–341.
Picado, J., Pultr, A. (2012). Frames and locales, topology without points, Basel, Birkhäuser.
Picado, J., & Pultr, A. (2021). Separation in point-free topology. Springer.
Rumfitt, I. (2015). The boundary stones of thought: an essay in the philosophy of logic. Oxford University Press.
Simmons, H. (1978). A framework for topology. In Logic Colloquium ’77, Studies in Logic foundations Math. Vol. 96, (pp. 239–251). Amsterdam, New York.
Simmons, H. (1980). Spaces with Boolean ASSEMBLIES. Colloquium Mathematicum, 43, 23–39.
Simmons, H. (1982). An algebraic version of Cantor-Bendixson analysis. In: Lecture notes in Mathematics. (pp. 310–323). Springer, Berlin.
Steinsvold, C. (2006). Topological models of belief, PhD Thesis, CUNY, Graduate Center.
Stalnaker, R. (2006). On logics of knowledge and belief. Philosophical Studies, 128, 169–199.
Steen, L. A., & Seebach, J. A., Jr. (1978). Counterexamples in topology. Springer.
Steen, L. A., Seebach Jr., & J. A. (1982). Counter examples in Topology. Second edition, New York, Heidelberg, Berlin, Springer.
Vaidyanathaswamy, R. (1947). Treatise on Set Topology I. Indian Mathematical Society, Madras
Van Benthem, J., & Bezhanishvili, G. (2007). Modal logics of space. Handbook of spatial logics (pp. 217–298). Springer.
Willard, St. (2004). General topology. Dover Publications.
Zarycki, M. (1930). Über den Kern einer Menge, Jahresbericht der Deutschen Mathematiker-Vereinigung 39, 154–158.
Acknowledgments
I’d like to thank two anonymous referees for extremely useful critical comments that helped me to improve earlier versions of this paper considerably. Of course, I am solely responsible for all remaining shortcomings and deficits the article. Further, I should like to express my sincere gratitude to Nasim Mahoozi for general advice, conversations and support when writing this paper.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Mormann, T. Completeness and Doxastic Plurality for Topological Operators of Knowledge and Belief. Erkenn (2023). https://doi.org/10.1007/s10670-023-00666-7
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10670-023-00666-7