Abstract
In this paper I propose and motivate a logic of the interdefined concepts of making true and control, understood as intensional propositional operators to be indexed to an agent. While bearing a resemblance to earlier logics in the tradition, the motivations, semantics, and object language theory differ on crucial points. Applying this logic to widespread formal theories of agency, I use it as a framework to argue against the ubiquitous assumption that the strongest actions or options available to a given agent must always be pairwise incompatible. The conclusion is that this assumption conflicts with failures of higher order control of agents over their degree or precision of control, failures exhibited by such imperfect agents as ourselves. I discuss models in this setting for understanding such imperfectly self-controlling agents. In an appendix, I prove several relevant results about the logic described, including soundness and completeness both for it and for certain natural extensions.
Similar content being viewed by others
Data Availability
Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
References
Davidson, D. (1967). The logical form of action sentences. In N. Rescher (Ed.) The Logic of Decision and Action. University of Pittsburgh Press, pp. 81–95.
Horty, J., & Pacuit, E. (2017). Action types in stit semantics. Review of Symbolic Logic, 10(4), 617–637.
Jeffrey, R. (1965). The logic of decision. New York: McGraw-Hill.
Joyce, J. (1999). The foundations of causal decision theory. New York: Cambridge University Press.
Lewis, D. (1981). Causal decision theory. Australasian Journal of Philosophy.
Savage, L. (1954). The foundations of statistics. New York: Wiley.
von Neumann, J., & Morgenstern, O. (1947). Theory of games and economic behavior. Princeton University Press.
Lewis, D. (1981). Are we free to break the laws? Theoria, 47(3), 113–21.
Inwagen, P. V. (1983). An essay on free will. Oxford University Press.
Nagel, T. (1979). Mortal questions. Cambridge University Press.
Williams, B. (1981). Moral luck. Cambridge University Press.
Pritchard, D. (2007). Epistemic luck. Theoria, 73(2), 173–178.
Hartman, R. J. (2017). In Defense of Moral Luck: why luck often affects praiseworthiness and blameworthiness. Routledge.
Chellas, B. F. (1992). Time and modality in the logic of agency. Studia Logica, 51(3-4), 485–517.
Kenny, A. (1976). Human abilities and dynamic modalities.
Belnap, N., & Perloff, M. (1988). Seeing to it that: a canonical form for agentives. Theoria, 54(3), 175–199.
Stalnaker, R. (1984). Inquiry. Cambridge University Press.
Lewis, D. K. (1986). On the plurality of worlds blackwell.
Stalnaker, R. (2006). On logics of knowledge and belief. Philosophical Studies, 128 (1), 169–199.
Horty, J. F., & Belnap, N. (1995). The deliberative stit: a study of action, omission, ability, and obligation. Journal of Philosophical Logic, 24(6), 583–644.
Frankfurt, H. G. (1969). Alternate possibilities and moral responsibility. Journal of Philosophy, 66(23), 829.
Fischer, J. M., & Ravizza, M. (1998). Responsibility and control: a theory of moral responsibility. Cambridge University Press.
Vihvelin, K., Causes, Laws, & Will, Free. (2013). Why determinism doesn’t matter oup usa.
Pacuit, E. (2017). Neighborhood semantics for modal logic, ser. Short Textbooks in Logic Springer.
von Wright, G. H. (1963). Norm and action. New York: Humanities.
Chellas, B. F. (1969). The logical form of imperatives. Ph.D. dissertation, Stanford University.
Segerberg, K., & started, Getting (1992). Beginnings in the logic of action. Studia Logica, 51, 347–378.
Nuel Belnap, M. P., & Xu, M. (2001). Facing the future: agents and choices in our indeterminist world. Oxford University Press on demand.
Xu, M. (1995). On the basic logic of stit with a single agent. Journal of Symbolic Logic, 60(2), 459–483.
Xu, M. (1996). An investigation in the logics of seeing-to-it-that. Ph.D. dissertation, University of Pittsburgh.
Xu, M. (1998). Axioms for deliberative stit. Journal of Philosophical Logic, 27(5), 505–552.
Philippe Balbiani, A. H., & Troquard, N. (2008). Alternative axiomatics and complexity for deliberative stit theories. Journal of Philosophical Logic, 37(4), 387–406.
Lorini, E., & Schwarzentruber, F. (2011). A logic for reasoning about counterfactual emotions. Artificial Intelligence, 37(3-4), 814–847.
Gibbard, A., & Harper, W. (1978). Foundations and Applications of Decision Theory, ser. University of Western Ontario Series in Philosophy of Science. Dordrecht: D. Reidel, 1978, no. 13a, ch. Counterfactuals and Two Kinds of Expected Utility.
Pollock, J. L. (2002). Rational choice and action omnipotence. Philosophical Review, 111(1), 1–23.
Hedden, B. (2012). Options and the subjective ought. Philosophical Studies, 158(2), 343–360.
Koon, J. (2020). Options must be external. Philosophical Studies, 177(5), 1175–1189.
Schwarz, W. (2023). Objects of choice Mind.
Lorini, E., Longin, D., & Mayor, E. (2011). A logical analysis of responsibility attribution: emotions, individuals and collectives. Journal of Logic and Computation, 24(6), 1313–1339.
Boudou, J., & Lorini, E. (2018). Concurrent game structures for temporal stit logic. In Proceedings of the Seventeenth International Conference on Autonomous Agents and MultiAgent Systems ACM Press.
Williamson, T. (2001). Knowledge and its limits. Cambridge University Press.
Fara, D. G. (2003). Gap principles, penumbral consequence, and infinitely higherorder vagueness. In J. C. Beall (Ed.) New Essays on the Semantics of Paradox. Oxford University Press.
Bacon, A. (2018). Vagueness and thought. Manuscript, forthcoming.
Ford, A. (2018). The province of human agency. Noû,s, 52(3), 697–720.
Lederman, H., & Holguín, B. (2023). Trying without fail, 2023, unpublished manuscript.
Cresswell, M. J., & Hughes, G. E. (1996). A new introduction to modal logic routledge.
Acknowledgements
I would like to thank Andrew Bacon for extensive discussion and feedback on earlier drafts of this paper, and an anonymous referee at this journal for pressing me on certain crucial points. I would also like to thank Harvey Lederman, Taylor Friesen, and Douglas Vaaler for helpful discussions on these topics.
Funding
The author has no relevant financial or non-financial interests to disclose.
Author information
Authors and Affiliations
Contributions
This paper has no coauthors.
Corresponding author
Ethics declarations
Ethics approval
No ethical approval requirements are relevant to this research.
Competing interests
The author has no competing interests to declare. She would like to thank Andrew Bacon for advice and comments on several drafts of this paper.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: A
Appendix: A
We here prove relevant results for the logics discussed above. Proofs are largely adapted from analogous ones in [46] and [24].
1.1 A.1 The Logic C o n v
Our language \({\mathscr{L}}\) (or equivalently, its set of wff’s) is as usual that generated by a countable set At of sentential variables p,q,r,… and
\(\mathsf {At} | \neg \varphi | \varphi \vee \psi | \varphi \wedge \psi | \varphi \rightarrow \psi | \Box \varphi \)
Our logic itself, \(\mathsf {Conv} \subset {\mathscr{L}}\), i.e. the φ such that ⊩Convφ, is the set of wff’s generated by all instances of the following axiom schemas
All classical tautologies
- T:
-
\(\Box \varphi \rightarrow \varphi \)
- M:
-
\((\Box \varphi \wedge \Box \psi ) \rightarrow \Box (\varphi \wedge \psi )\)
- O:
-
\((\Box \varphi \wedge \Box \psi ) \rightarrow \Box (\varphi \vee \psi )\)
- Conv:
-
\(\Box (\varphi \vee \psi \vee \chi ) \rightarrow \Box \varphi \rightarrow \Box (\varphi \vee \psi )\)
with the following inference rules
- M P :
-
if ⊩Convφ and \(\vdash _{\mathsf {Conv}} \varphi \rightarrow \psi \), ⊩Convψ
- R E :
-
if \(\vdash _{\mathsf {Conv}} \psi \leftrightarrow \phi \), \(\vdash _{\mathsf {Conv}} \Box \psi \leftrightarrow \Box \phi \)
A model \({\mathscr{M}}\) of our logic is a triple 〈W,V, (Smin,Smax)〉, with W some nonempty set, \(V : \mathsf {At} \rightarrow \mathcal {P}(W)\), and the possibly partial \((S_{\mathsf {min}}, S_{\mathsf {max}}) : W \hookrightarrow \mathcal {P}(W) \times \mathcal {P}(W)\) (or just S) such that (where defined) \(\{ w \} \subseteq S_{\mathsf {min}}(w) \subseteq S_{\mathsf {max}}(w)\). The clauses of the interpretation function ⟦⟧ for the atoms and Boolean clauses are as usual, while \(\llbracket \Box \varphi \rrbracket = \{ w \in W : S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \subseteq S_{\mathsf {max}}(w)\}\). (Note that \(w \notin \llbracket \Box \varphi \rrbracket \) when S is undefined on w, and that as components of a single function from W into the cartesian square of \(\mathcal {P}(W)\) Smin and Smax are each defined iff the other is.) φ is valid in a model \({\mathscr{M}}\), i.e. \(\vDash _{{\mathscr{M}}}\varphi \), iff \(\llbracket \varphi \rrbracket _{{\mathscr{M}}} = W_{{\mathscr{M}}}\), and is valid simpliciter, i.e. \(\vDash _{\mathsf {Conv}}\varphi \), iff for all models \({\mathscr{M}}\), \(\vDash _{{\mathscr{M}}}\varphi \).
We now prove a lemma that will be useful in proving completeness.
Lemma A.1
If \(\vdash _{\mathsf {Conv}} (\varphi \wedge \psi ) \rightarrow \chi \) and \(\vdash _{\mathsf {Conv}}(\neg \varphi \wedge \neg \psi ) \rightarrow \neg \chi \), \(\vdash _{\mathsf {Conv}} (\Box \varphi \wedge \Box \psi ) \rightarrow \Box \chi \).
Proof
Suppose the antecedent. Then by classical reasoning, we have \(\vdash _{\mathsf {Conv}} (\varphi \wedge \psi ) \rightarrow \chi \) and \(\vdash _{\mathsf {Conv}} \chi \rightarrow (\varphi \vee \psi )\). Now assume \(\Box \varphi \wedge \Box \psi \). By M this gives us \(\Box (\varphi \wedge \psi )\), and by O \(\Box (\varphi \vee \psi )\). But then by Conv and RE we have \(\Box \chi \), and discharging we have \(\vdash _{\mathsf {Conv}} (\Box \varphi \wedge \Box \psi ) \rightarrow \Box \chi \). □
Corollary A.2
For any finite {φ1,…,φn,ψ}, if \(\vdash _{\mathsf {Conv}} (\varphi _{1} \wedge {\ldots } \wedge \varphi _{n}) \rightarrow \psi \) and \(\vdash _{\mathsf {Conv}} (\neg \varphi _{1} \wedge {\ldots } \wedge \neg \varphi _{n}) \rightarrow \neg \psi \), \(\vdash _{\mathsf {Conv}} (\Box \varphi _{1} \wedge {\ldots } \wedge \Box \varphi _{n}) \rightarrow \Box \psi \).
1.2 A.2 Soundness
It is easy to see that our semantics is sound.
Theorem A.3
If ⊩Convφ, \(\vDash _{\mathsf {Conv}}\varphi \)
Proof
We omit as routine the proofs of the validity of classical tautologies and MP. The validity of T is straightforward from the stipulation that (when defined) \(\{ w \} \subseteq S_{\mathsf {min}}(w)\). The validity of M is straightforward from the fact that for \(X, Y \supseteq S_{\mathsf {min}}(w)\), \(X \cap Y \supseteq S_{\mathsf {min}}(w)\). The validity of O is straightforward from the fact that for \(X, Y \subseteq S_{\mathsf {max}}(w)\), \(X \cup Y \subseteq S_{\mathsf {max}}(w)\). The validity of Conv is straightforward from the fact that, for any X,Y,Z: \(X \subseteq (X \cup Y) \subseteq (X \cup Y \cup Z)\). By induction, for any φ,ψ provably equivalent and for which our theorem holds, in all \({\mathscr{M}}\), ⟦φ⟧ = ⟦ψ⟧, validating RE by the intensionality of the clause for \(\Box \). □
1.3 A.3 Completeness
Let a consistent set Γ be a set of wff’s in \({\mathscr{L}}\) such that for no \(\varphi _{1} {\ldots } \varphi _{n} \in {\mathscr{L}}\) (for finite n) do we have ⊩Conv¬(φ1 ∧… ∧ φn). Let a maximal consistent set Γ be a consistent set such that for all \(\varphi \in {\mathscr{L}}\) either φ ∈Γ or ¬φ ∈Γ. The proof that for any consistent set Γ there is a maximal consistent set \({\Gamma }^{\prime }\) such that \({\Gamma } \subseteq {\Gamma }^{\prime }\) is as usual. We denote the set of maximal consistent sets in \(\mathcal {P}({\mathscr{L}})\) as \({\mathscr{L}}_{\mathsf {M}}\).
We define the canonical model \({\mathscr{M}}_{\mathsf {canon}}\), as per the above, as the triple 〈W,V, (Smin,Smax)〉 such that
-
\(W_{{\mathscr{M}}_{\mathsf {canon}}} = {\mathscr{L}}_{\mathsf {M}}\)
-
For p ∈At, \(V_{{\mathscr{M}}_{\mathsf {canon}}}(p) = \{ w \in W_{{\mathscr{M}}_{\mathsf {canon}}} : p \in w \}\)
-
\(S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}(w^{\prime }) = \{ w \in W_{{\mathscr{M}}_{\mathsf {canon}}} :\) for all \(\Box \varphi \in w^{\prime }, \varphi \in w \}\) unless no \(\Box \varphi \in w^{\prime }\), in which case undefined
-
\(S_{\mathsf {max}{\mathscr{M}}_{\mathsf {canon}}}(w^{\prime }) = \{ w \in W_{{\mathscr{M}}_{\mathsf {canon}}} :\) for some \(\Box \varphi \in w^{\prime }, \varphi \in w\}\) unless no \(\Box \varphi \in w^{\prime }\), in which case undefined
Lemma A.4
Let Γ be some consistent set of wff’s of the form \(\Box \varphi \), \({\Gamma }_{-\Box }\) be the set obtained by exchanging φ for each \(\Box \varphi \in {\Gamma }\), and \(\neg {\Gamma }_{-\Box }\) the set obtained by negating each member of \({\Gamma }_{-\Box }\). Then if, for arbitrary \(\neg \Box \psi \), we have \({\Gamma } \cup \{\neg \Box \psi \}\) consistent, we have either \({\Gamma }_{-\Box } \cup \{\neg \psi \}\) consistent or \(\neg {\Gamma }_{-\Box } \cup \{\psi \}\) consistent.
Proof
Suppose that \({\Gamma } \cup \{\neg \Box \psi \}\) is but neither \({\Gamma }_{-\Box } \cup \{\neg \psi \}\) nor \(\neg {\Gamma }_{-\Box } \cup \{ \psi \}\) also is consistent. Then there must be some finite \(\{ \gamma _{1}, \ldots , \gamma _{n} \} \subseteq {\Gamma }_{-\Box }\) where ⊩Conv¬(γ1 ∧… ∧ γn ∧¬ψ) and some (not necessarily disjoint) \(\{ \gamma _{n+1}, \ldots , \gamma _{m} \} \subseteq {\Gamma }_{-\Box }\) where ⊩Conv¬(¬γn+ 1 ∧… ∧¬γm ∧ ψ). Indeed, by the monotonocity of inconsistency, we may simplify by saying we have both ⊩Conv¬(γ1 ∧… ∧ γm ∧¬ψ) and ⊩Conv¬(¬γ1 ∧… ∧¬γm ∧ ψ).
By classical reasoning and the first conjunct, we have \(\vdash _{\mathsf {Conv}} (\gamma _{1} \wedge {\ldots } \wedge \gamma _{m}) \rightarrow \psi \). By classical reasoning and the second conjunct, we have \(\vdash _{\mathsf {Conv}} (\neg \gamma _{1} \wedge {\ldots } \wedge \neg \gamma _{m}) \rightarrow \neg \psi \). But since by supposition \({\Gamma } \cup \{\neg \Box \psi \}\) is consistent, we have \(\nvdash _{\mathsf {Conv}} (\Box \gamma _{1} \wedge {\ldots } \wedge \Box \gamma _{m}) \rightarrow \Box \psi \). But by Corollary A.2, this is impossible. □
Theorem A.5
\(w \in \llbracket \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}}\) iff φ ∈ w
Proof
The proof proceeds by induction. It is straightforward to see that it holds in the base step for all p ∈At, as the interpretation function ⟦⟧ in that case simply reduces to the assignment function V, which in turn by construction reduces in \({\mathscr{M}}_{\mathsf {canon}}\) to ∈. We omit as routine the induction steps for the Boolean connectives, leaving only our operator \(\Box \).
We first prove right to left. Suppose \(\Box \varphi \in w\). Now by hypothesis we have \(w^{\prime } \in \llbracket \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}}\) iff \(\varphi \in w^{\prime }\); by construction we then have for all \(w^{\prime } \in S_{\mathsf {min}{{\mathscr{M}}_{\mathsf {canon}}}}(w)\) that \(\varphi \in w^{\prime }\), and thus \(\llbracket \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}} \supseteq S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}(w)\). Moreover, by construction of \(S_{\mathsf {max}{\mathscr{M}}_{\mathsf {canon}}}\) and by the inductive hypothesis, \(\llbracket \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}} \subseteq S_{\mathsf {max}{\mathscr{M}}_{\mathsf {canon}}}(w)\). So by construction of ⟦⟧ we have \(w \in \llbracket \Box \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}}\).
Now suppose \(\Box \varphi \notin w\), i.e. (by maximality) \(\neg \Box \varphi \in w\). Either \(S_{{\mathscr{M}}_{\mathsf {canon}}}(w)\) is defined or not. If not, then by stipulation \(w \notin \llbracket \Box \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}}\). Now let Γ be \(\{\Box \psi : \Box \psi \in w\}\); then, if \(S_{{\mathscr{M}}_{\mathsf {canon}}}(w)\) is defined, by Lemma A.4 and consistency of \(w \supset {\Gamma } \cup \{\neg \Box \varphi \}\) we have either \({\Gamma }_{-\Box } \cup \{\neg \varphi \}\) consistent or \(\neg {\Gamma }_{-\Box } \cup \{\varphi \}\) consistent. But then (as the domain \({\mathscr{L}}_{\mathsf {M}}\) includes all maximal consistent sets, and all consistent sets can be extended to some \(v \in {\mathscr{L}}_{\mathsf {M}}\)) there must exist either some \(w^{\prime } \supset {\Gamma }_{-\Box } \cup \{\neg \varphi \}\) or \(w^{\prime \prime } \supset \neg {\Gamma }_{-\Box } \cup \{\varphi \}\). If the former, then by the inductive hypothesis, the consistency of \(w^{\prime }\), and construction of \(S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}\) we have \(w^{\prime } \in (S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}(w) \setminus \llbracket \varphi \rrbracket )\), and thus \(w \notin \llbracket \Box \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}}\). If the latter, then by the inductive hypothesis, the consistency of \(w^{\prime \prime }\), and construction of \(S_{\mathsf {max}{\mathscr{M}}_{\mathsf {canon}}}\) we have \(w^{\prime \prime } \in (\llbracket \varphi \rrbracket \setminus S_{\mathsf {max}{\mathscr{M}}_{\mathsf {canon}}}(w))\), and thus \(w \notin \llbracket \Box \varphi \rrbracket _{{\mathscr{M}}_{\mathsf {canon}}}\). □
Corollary A.6
\(\vDash _{M_{\mathsf {canon}}}\varphi \) iff ⊩Convφ
Proof
Suppose ⊩Convφ. Then we have {¬φ} inconsistent, and thus by maximality of \(w \in W_{{\mathscr{M}}_{\mathsf {canon}}}\) each such w must contain φ, which by the above means \(\vDash _{M_{\mathsf {canon}}}\varphi \).
Suppose \(\nvdash _{\mathsf {Conv}}\varphi \). Then we have {¬φ} consistent, meaning some \(w \in W_{{\mathscr{M}}_{\mathsf {canon}}} = {\mathscr{L}}_{\mathsf {M}}\) must extend it, so by the consistency of \(w \in W_{{\mathscr{M}}_{\mathsf {canon}}}\) this w must not contain φ, which by the above means \(\nvDash _{M_{\mathsf {canon}}}\varphi \). □
By familiar reasoning, proving completeness for our logic now only requires, in light of Corollary A.6, that the frame \(\langle W_{{\mathscr{M}}_{\mathsf {canon}}}, (S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}, S_{\mathsf {max}{\mathscr{M}}_{\mathsf {canon}}}) \rangle \) satisfies the requirements for a model frame given in the first section. (That it satisfies the requirements for its base valuation function \(V_{{\mathscr{M}}_{\mathsf {canon}}}\) is immediate.)
Corollary A.7
If \(\vDash _{\mathsf {Conv}}\varphi \), ⊩Convφ
Proof
The conditions for a frame are just that, for all w with S(w) defined, a) w ∈ Smin(w) and b) \(S_{\mathsf {min}}(w) \subseteq S_{\mathsf {max}}(w)\).
\(S_{{\mathscr{M}}_{\mathsf {canon}}}(w)\), by construction, is defined only for w with some \(\Box \varphi \in w\). Let w be such a world. Now by T and maximality, for all \(\Box \varphi \in w\), also φ ∈ w, and so by construction of \(S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}(w)\) we have \(w \in S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}(w)\).
Since \(S_{{\mathscr{M}}_{\mathsf {canon}}}(w)\) is defined only when there exists such a \(\Box \varphi \in w\), the set of \(w^{\prime } \in W_{{\mathscr{M}}_{\mathsf {canon}}}\) with some \(\psi \in w^{\prime }\) such that \(\Box \psi \in w\) extends the set of those containing all such ψ, i.e. \(S_{\mathsf {min}{\mathscr{M}}_{\mathsf {canon}}}(w) \subseteq S_{\mathsf {max}.{\mathscr{M}}_{\mathsf {canon}}}(w)\). □
Corollary A.8
\(\vDash _{\mathsf {Conv}}\varphi \) iff ⊩Convφ
1.4 The Logics C o n v 4, C o n v5∗, and C o n v 45∗
The logics \(\mathsf {Conv4}, \mathsf {Conv5^{*}}, \mathsf {Conv45^{*}} \subset {\mathscr{L}}\) are those generated by the inference rules and axiom schemas of Conv (mutatis mutandis) plus each of the following and both, respectively.
- 4 :
-
\(\Box \varphi \rightarrow \Box \Box \varphi \)
- 5∗:
-
\((\Box \varphi \wedge \neg \Box \psi ) \rightarrow \Box (\varphi \wedge \neg \Box \psi )\)
The models of Conv4 and Conv5∗ are just like those of Conv, with the added stipulation that, for the first, \(S_{\mathsf {min}}(w^{\prime }) \subseteq S_{\mathsf {min}}(w)\) and \(S_{\mathsf {max}}(w^{\prime }) \supseteq S_{\mathsf {max}}(w)\) for all \(w^{\prime } \in S_{\mathsf {min}}(w)\) where S(w) is defined; for the second, \(S_{\mathsf {min}}(w^{\prime }) \supseteq S_{\mathsf {min}}(w)\) and \(S_{\mathsf {max}}(w^{\prime }) \subseteq S_{\mathsf {max}}(w)\) for all \(w^{\prime } \in S_{\mathsf {min}}(w)\) where \(S(w), S(w^{\prime })\) are defined. Their canonical models are constructed just like that of Conv, save for maximal consistency in the respective logics replacing maximal ⊩Conv-consistency.
Soundness is routine.
Theorem A.9
If ⊩Conv4φ, \(\vDash _{\mathsf {Conv4}} \varphi \)
Proof
To extend our soundness proof for Conv, observe first that \(w \in \llbracket \Box \Box \varphi \rrbracket \) iff for each \(w^{\prime } \in S_{\mathsf {min}}(w)\), \(w^{\prime } \in \llbracket \Box \varphi \rrbracket \), and for no \(w^{\prime \prime } \notin S_{\mathsf {max}}(w)\) is \(w^{\prime \prime } \in \llbracket \Box \varphi \rrbracket \). Now clearly, if \(w \in \llbracket \Box \varphi \rrbracket \), by the at most restriction of \(S_{\mathsf {min}}(w^{\prime })\) and at most expansion of \(S_{\mathsf {max}}(w^{\prime })\) for \(w^{\prime } \in S_{\mathsf {min}}(w)\), \(S_{\mathsf {min}}(w^{\prime }) \subseteq \llbracket \varphi \rrbracket \subseteq S_{\mathsf {max}}(w^{\prime })\) (i.e. \(w^{\prime } \in \llbracket \Box \varphi \rrbracket \)). Moreover, if \(w \in \llbracket \Box \varphi \rrbracket \), \(S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \subseteq S_{\mathsf {max}}(w)\), so for no \(w^{\prime \prime } \notin S_{\mathsf {max}}(w)\) do we have \(w^{\prime \prime } \in \llbracket \varphi \rrbracket \), much less \(w^{\prime \prime } \in \llbracket \Box \varphi \rrbracket \). So we have \(w \in \llbracket \Box \Box \varphi \rrbracket \) if \(w \in \llbracket \Box \varphi \rrbracket \), i.e. we have 4 valid. □
Theorem A.10
If \(\vdash _{\mathsf {Conv5^{*}}} \varphi \), \(\vDash _{\mathsf {Conv5^{*}}} \varphi \)
Proof
To see that 5∗ is valid, we will show that, whenever \(w \in \llbracket \Box \varphi \wedge \neg \Box \psi \rrbracket \), \(w \in \llbracket \Box (\varphi \wedge \neg \Box \psi ) \rrbracket \). Consider that \(w \in \llbracket \Box (\varphi \wedge \neg \Box \psi ) \rrbracket \) iff for all \(w^{\prime } \in S_{\mathsf {min}}\), \(w^{\prime } \in \llbracket \varphi \rrbracket \) but not \(w^{\prime } \in \llbracket \Box \psi \rrbracket \) (i.e. not \(S_{\mathsf {min}}(w^{\prime }) \subseteq \llbracket \psi \rrbracket \subseteq S_{\mathsf {max}}(w^{\prime })\)), and for all \(w^{\prime \prime } \notin S_{\mathsf {max}}(w)\), either \(w^{\prime \prime } \notin \llbracket \varphi \rrbracket \) or \(w^{\prime \prime } \in \llbracket \Box \psi \rrbracket \). But since for such \(w^{\prime } \in S_{\mathsf {min}}\), \(S_{\mathsf {min}}(w^{\prime }) \supseteq S_{\mathsf {min}}(w)\) and \(S_{\mathsf {max}}(w^{\prime }) \subseteq S_{\mathsf {max}}(w)\) (where \(S(w^{\prime })\) is defined), if we have \(S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \subseteq S_{\mathsf {max}}(w)\) but not \(S_{\mathsf {min}}(w) \subseteq \llbracket \psi \rrbracket \subseteq S_{\mathsf {max}}(w)\) (i.e. if \(w \in \llbracket \Box \varphi \wedge \neg \Box \psi \rrbracket \)), we have the latter also for (defined) \(S(w^{\prime })\) (so that \(w^{\prime } \notin \llbracket \Box \psi \rrbracket \)) and we have \(w^{\prime } \in \llbracket \varphi \rrbracket \) trivially, as \(w^{\prime } \in S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \). (Where S is not defined, \(w^{\prime } \notin \llbracket \Box \psi \rrbracket \) trivially.) And, since if \(w \in \llbracket \Box \varphi \wedge \neg \Box \psi \rrbracket \) we have \(\llbracket \varphi \rrbracket \subseteq S_{\mathsf {max}}(w)\), for no such \(w^{\prime \prime } \notin S_{\mathsf {max}}(w)\) do we have \(w^{\prime \prime } \in \llbracket \varphi \rrbracket \). □
Completeness is similarly routine.
Theorem A.11
If \(\vDash _{\mathsf {Conv4}} \varphi \), ⊩Conv4φ
Proof
By our earlier result, it will suffice to show, where \({\mathscr{M}}^{\prime }_{\mathsf {canon}}\) is the canonical model of Conv4, that for \(w \in W_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}\) with \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) defined, \(S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \subseteq S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) and \(S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \supseteq S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) for all \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\).
Suppose there is some \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) and \(v \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }), \notin S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) (i.e. \(S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \nsubseteq S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\)). Then by construction, there is some \(\Box \varphi \in w, \notin w^{\prime }\). But by 4 this is impossible, since for any \(\Box \varphi \in w\), \(\Box \Box \varphi \in w\), and so \(\Box \varphi \in w^{\prime }\). (For this same reason, \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime })\) is always defined when \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) is defined and \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\).)
Suppose next there is some \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) and \(v \in S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w), \notin S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime })\) (i.e. \(S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \nsupseteq S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\)). Then again by construction, there is some \(\Box \varphi \in w, \notin w^{\prime }\), which we have just seen to be impossible. □
Theorem A.12
If \(\vDash _{\mathsf {Conv5^{*}}} \varphi \), \(\vdash _{\mathsf {Conv5^{*}}} \varphi \)
Proof
Again it will suffice to prove that for \(w \in W_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}\) with \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) defined, \(S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \supseteq S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) and \(S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \subseteq S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) for all \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) with some \(\Box \psi \in w^{\prime }\) (where \({\mathscr{M}}^{\prime }_{\mathsf {canon}}\) is now the canonical model of Conv5∗).
Suppose there is some \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) (with \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime })\) defined) and \(v \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w), \notin S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime })\) (i.e. \(S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \nsupseteq S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\)). Then by construction (and definedness of \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime })\)), there is some \(\Box \varphi \in w^{\prime }, \notin w\). Now since \(S_{{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) is defined, there exists \(\Box \psi \in w\), with ψ≠φ. We then by maximality have \(\Box \psi \wedge \neg \Box \varphi \in w\). But then we have by 5∗ that \(\Box (\psi \wedge \neg \Box \varphi ) \in w\), and so by construction and maximality \(\psi \wedge \neg \Box \varphi , \neg \Box \varphi \in w^{\prime }\), which is by consistency of \(w^{\prime }\) impossible.
Suppose next there is some \(w^{\prime } \in S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) and \(v \in S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }), \notin S_{\mathsf {min}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\) (i.e. \(S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w^{\prime }) \nsubseteq S_{\mathsf {max}{\mathscr{M}}^{\prime }_{\mathsf {canon}}}(w)\)). Then again by construction, there is some \(\Box \varphi \in w^{\prime }, \notin w\), which we have just seen to be impossible. □
Corollary A.13
Conv4 and Conv5∗ are independent.
Proof
For a model of Conv5∗ invalidating 4, let W = {0, 1}, Smin(0) = Smax(0) = W, S(1) be undefined. For a model of Conv4 invalidating 5∗, let W = {0, 1}, Smin(0) = Smax(0) = Smax(1) = W, Smin(1) = {1}. For one satisfying both let W = {0}, Smin(0) = Smax(0) = W. □
Corollary A.14
Conv45∗ is sound and complete on the class of frames where for any w with S(w) defined, S(v) = S(u) where v,u ∈ Smin(w).
Proof
This is straightforward from the preceding proofs. □
Corollary A.15
Among models in which, for all w with S(w) defined, \(S(w^{\prime })\) is defined for all \(w^{\prime } \in S_{\mathsf {min}}(w)\), the Conv5∗ models are the Conv45∗ models.
Proof
Take any such model of Conv5∗ and any w with S(w) defined. For any \(w^{\prime } \in S_{\mathsf {min}}(w)\), \(S_{\mathsf {min}}(w^{\prime })\) (which we know to be defined) must contain w, as \(w \in S_{\mathsf {min}}(w) \subseteq S_{\mathsf {min}}(w^{\prime })\). So, by the defining property of Conv5∗ models, \(S_{\mathsf {min}}(w) \subseteq S_{\mathsf {min}}(w^{\prime })\) (as \(w^{\prime } \in S_{\mathsf {min}}(w)\)) and \(S_{\mathsf {min}}(w^{\prime }) \subseteq S_{\mathsf {min}}(w)\) (as \(w \in S_{\mathsf {min}}(w^{\prime })\)), i.e. \(S_{\mathsf {min}}(w) = S_{\mathsf {min}}(w^{\prime })\); similarly, \(S_{\mathsf {max}}(w) \supseteq S_{\mathsf {max}}(w^{\prime })\) and \(S_{\mathsf {max}}(w^{\prime }) \supseteq S_{\mathsf {max}}(w)\), i.e. \(S_{\mathsf {max}}(w) = S_{\mathsf {min}}(w^{\prime })\). Thus, for all \(w^{\prime } \in S_{\mathsf {min}}(w)\), \(S(w) = S(w^{\prime })\).
The reverse direction is trivial. □
We can now prove the correspondence between Conv45∗ and the principle attributed to Lewis in Section 3.1.
Fact A.16
Take an arbitrary sandwich frame 〈W,S〉. Let \(\mathcal {N}(p \subseteq W)\) be \(\{ w \in W : S_{\mathsf {min}}(w) \subseteq p \subseteq S_{\mathsf {max}}(w)\}\). Let Ow be the intersection of all p ∋ w such that for some \(q \subseteq W\) either \(p = \mathcal {N}(q)\) or \(p = W \setminus \mathcal {N}(q)\).
If for all w (with S(w) defined), \(S_{\mathsf {min}}(w) \subseteq O_{w} \subseteq S_{\mathsf {max}}(w)\), then for all v and u ∈ Smin(v), S(v) = S(u), and vice versa.
Proof
Left to right: Clearly, given the antecedent, for all w (with S(w) defined), Ow = Smin(w), as by definition \(O_{w} \subseteq \mathcal {N}(S_{\mathsf {min}}(w)) \subseteq S_{\mathsf {min}}(w)\). It is now easy to see that for any v ∈ Ow, neither is there u ∈ Ov,∉Ow nor \(u^{\prime } \in O_{w}, \notin O_{v}\). For if the former, then , which is prohibited by construction of O as long as v ∈ Ow = Smin(w); and if the latter, either or no \(\mathcal {N}(p) \ni v\) at all, both of which are prohibited by construction of O as long as v ∈ Ow = Smin(w), since \(w \notin \mathcal {N}(p)\) iff \(w \in (W \setminus \mathcal {N}(p))\). Therefore Ow = Ov, and thus Smin(w) = Smin(v).
Moreover, whenever Ow = Smin(w) = Ov = Smin(v), Smax(w) = Smax(v), as otherwise w,v will be distinguished by some p in that \(w \in , v\notin \mathcal {N}(p)\) (or vice versa). But again by definition of O, this is impossible. So S(w) = S(v).
Right to left: Suppose for all w (with S(w) defined) and v ∈ Smin(w), S(w) = S(v). Then \(w \in \mathcal {N}(p)\) iff \(v \in \mathcal {N}(p)\), so Ow = Ov. So by v ∈ Ov for all v ∈ Smin(w), \(S_{\mathsf {min}}(w) \subseteq O_{w}\). But \(O_{w} \subseteq \mathcal {N}(S_{\mathsf {max}}(w)) \subseteq S_{\mathsf {max}}(w)\). So \(S_{\mathsf {min}}(w) \subseteq O_{w} \subseteq S_{\mathsf {max}}(w)\). □
1.5 A.5 The Logic C o n v +
Let the language \({\mathscr{L}}_{\boxdot }\) be \({\mathscr{L}}\) closed under the introduction of formulas \(\ulcorner \boxdot \varphi \urcorner \). Let the logic \(\mathsf {Conv+} \subset {\mathscr{L}}_{\boxdot }\) be that generated by the axiom schemas and inference rules of Conv (mutatis mutandis), together with the additional axiom schemas
- \(\mathsf {K_{\boxdot }}\) :
-
\(\boxdot (\varphi \rightarrow \psi ) \rightarrow (\boxdot \varphi \rightarrow \boxdot \psi )\)
- \(\mathsf {T_{\boxdot }}\) :
-
\(\boxdot \varphi \rightarrow \varphi \)
- P S :
-
\((\boxdot \varphi \wedge \Box \psi ) \rightarrow \Box (\varphi \wedge \psi )\)
and the additional inference rule
- N e c :
-
if ⊩Conv+φ, \(\vdash _{\mathsf {Conv+}} \boxdot \varphi \)
A model \({\mathscr{M}}^{+}\) of Conv+ is just like a model of Conv, together with an extra parameter \(R : W \rightarrow \mathcal {P}(W)\); R obeys the further constraints that for all w ∈ W, \(S_{\mathsf {min}}(w) \subseteq R(w)\) (where defined), and w ∈ R(w). As usual, \(\llbracket \boxdot \varphi \rrbracket _{{\mathscr{M}}}^{+} = \{ w \in W : R(w) \subseteq \llbracket \varphi \rrbracket _{{\mathscr{M}}}^{+} \}\). The canonical model \({\mathscr{M}}^{+}_{\mathsf {canon}}\) of Conv+ is constructed as for Conv, with the obvious extensions for \(R_{{\mathscr{M}}^{+}_{\mathsf {canon}}}\) (particularly that \(w^{\prime } \in R_{{\mathscr{M}}^{+}_{\mathsf {canon}}}(w)\) iff \(\varphi \in w^{\prime }\) for all \(\Box \varphi \in w\)).
It is easy to see the logic is sound.
Theorem A.17
If ⊩Conv+φ, \(\vDash _{\mathsf {Conv+}} \varphi \)
Proof
The only distinctive case to show is for PS. By the stated constraint on R and Smin, whenever \(w \in \llbracket \boxdot \varphi \rrbracket , \in \llbracket \Box \psi \rrbracket \), \(S_{\mathsf {min}}(w) \subseteq \llbracket \psi \rrbracket \subseteq S_{\mathsf {max}}(w)\) (by construction of \(\llbracket \Box \psi \rrbracket \)) and \(S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \) (by the above constraint), and thus \(S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \cap \llbracket \psi \rrbracket = \llbracket \varphi \wedge \psi \rrbracket \subseteq \llbracket \psi \rrbracket \subseteq S_{\mathsf {max}}(w)\), and so by construction of \(\llbracket \Box (\varphi \wedge \psi ) \rrbracket \), \(w \in \llbracket \Box (\varphi \wedge \psi )\rrbracket \). Whence the conclusion follows trivially. □
The bulk of the completeness proof can rely on those given above.
Theorem A.18
If \(\vDash _{\mathsf {Conv+}} \varphi \), ⊩Conv+φ
Proof
It is easy to check that the proofs for Theorem A.5 and Corollary A.6 can be extended to proofs of their analogues for Conv+. Thus all that is required is a proof that our canonical model \({\mathscr{M}}^{+}_{\mathsf {canon}}\) satisfies the criteria for a model.
For the constraints on R by itself the proofs are well known. We thus need only demonstrate that, when defined, \(S_{\mathsf {min} {\mathscr{M}}^{+}_{\mathsf {canon}}}(w) \subseteq R_{{\mathscr{M}}^{+}_{\mathsf {canon}}}(w)\). Suppose there is some \(w^{\prime } \in S_{\mathsf {min} {\mathscr{M}}^{+}_{\mathsf {canon}}}(w), \notin R_{{\mathscr{M}}^{+}_{\mathsf {canon}}}(w)\) (where these are defined). Then by construction there are \(\varphi \in w^{\prime }, \psi \notin w^{\prime }\) (that is, by maximality, \(\varphi , \neg \psi \in w^{\prime }\)) such that \(\Box \varphi , \boxdot \psi \in w\). But then we do not have (by consistency of \(w^{\prime }\)) \(\varphi \wedge \psi \in w^{\prime }\), and so by construction of \(S_{\mathsf {min} {\mathscr{M}}^{+}_{\mathsf {canon}}}\) we do not have \(\Box (\varphi \wedge \psi ) \in w\), which contradicts the hypothesis given PS (and maximality). □
Say that a model is partitional just in case, for all w, for all \(v, v^{\prime } \in R(w)\), Smin(v) and \(S_{\mathsf {min}}(v^{\prime })\) are (when defined) either identical or pairwise disjoint.
Theorem A.19
No formulas define the partitional models.
Proof
It will suffice to show that, for two models \({\mathscr{M}}, {\mathscr{M}}^{\prime }\) with \({\mathscr{M}}\) partitional but not \({\mathscr{M}}^{\prime }\), the φ valid on \({\mathscr{M}}\) are just those valid on \({\mathscr{M}}^{\prime }\).
Let \(W_{{\mathscr{M}}} = \{ w_{0}, w_{1} \}\); \(S_{\mathsf {min} {\mathscr{M}}}(w_{0}) = S_{\mathsf {max} {\mathscr{M}}}(w_{0}) = \{w_{0}\}\); \(S_{{\mathscr{M}}}(w_{1})\) undefined; and for all atomic p, \(V_{{\mathscr{M}}}(p) = \{w_{0}\}\). Let \(W_{{\mathscr{M}}}^{\prime } = \{ v_{0}, v_{1}, v_{2} \}\); \(S_{\mathsf {min} {\mathscr{M}}^{\prime }}(v_{0}) = S_{\mathsf {max} {\mathscr{M}}^{\prime }}(v_{0}) = \{v_{0},v_{1}\}\), \(S_{{\mathscr{M}}}^{\prime }(v_{1}) = (\{v_{1}\}, \{v_{0},v_{1}\})\); \(S_{{\mathscr{M}}}^{\prime }(v_{2})\) undefined; and for all p, \(V_{{\mathscr{M}}}^{\prime }(p) = \{v_{0},v_{1}\}\). For both, R is the universal relation (the constant function to W). Clearly, \({\mathscr{M}}\) but not \({\mathscr{M}}^{\prime }\) is partitional, so it remains to show \(\vDash _{{\mathscr{M}}} \varphi \) iff \(\vDash _{{\mathscr{M}}^{\prime }} \varphi \).
First we prove by induction on complexity that w1 ∈⟦φ⟧ iff v2 ∈⟦φ⟧. For the atomic base case, this is given. For the Boolean connectives, this follows by classicality of the logic. Moreover, as S(w1) and S(v2) are undefined, the step for \(\Box \varphi \) is likewise trivial.
Next, w0 ∈⟦φ⟧ iff v1 ∈⟦φ⟧. Again, the base step and Boolean inductive step are trivial. For \(\Box \varphi \), if w0∉⟦φ⟧ and v1∉⟦φ⟧, it follows by T that \(v_{1} \notin \llbracket \Box \varphi \rrbracket \) and \(w_{0} \notin \llbracket \Box \varphi \rrbracket \); if w0 ∈⟦φ⟧ and v1 ∈⟦φ⟧, either v2 ∈⟦φ⟧ and (thus, by the above) w1 ∈⟦φ⟧ or not. If so, \(v_{1} \notin \llbracket \Box \varphi \rrbracket \) and \(w_{0} \notin \llbracket \Box \varphi \rrbracket \), since v2∉Smax(v1) and w1∉Smax(w0). If not, then \(S_{\mathsf {min}}(v_{1}) \subseteq \llbracket \varphi \rrbracket \subseteq S_{\mathsf {max}}(v_{1})\) (meaning \(v_{1} \in \llbracket \Box \varphi \rrbracket \)) and Smin(w0) = ⟦φ⟧ = Smax(w0) (meaning \(w_{0} \in \llbracket \Box \varphi \rrbracket \)).
Finally, w0 ∈⟦φ⟧ iff v0 ∈⟦φ⟧. Base and Boolean steps are trivial. For \(\Box \varphi \), by the inductive hypothesis we have w0 ∈⟦φ⟧ iff v0 ∈⟦φ⟧. As (by T) the result is trivial if v0∉⟦φ⟧ and w0∉⟦φ⟧, suppose v0 ∈⟦φ⟧ and w0 ∈⟦φ⟧. Again, either v2 ∈⟦φ⟧ (and thus w1 ∈⟦φ⟧) or not. In the former case, the conclusion follows as before. Otherwise, by the preceding paragraph, we have w0 ∈⟦φ⟧ and v1 ∈⟦φ⟧, so v0 ∈⟦φ⟧ and v1 ∈⟦φ⟧, and thus, as Smin(v0) = Smax(v0) = {v0,v1} and v2∉⟦φ⟧, \(v_{0} \in \llbracket \Box \varphi \rrbracket \). But (since we are assuming w1∉⟦φ⟧), as Smin(w0) = Smax(w0) = {w0} = ⟦φ⟧, also \(w_{0} \in \llbracket \Box \varphi \rrbracket \).
So the \(\varphi \in {\mathscr{L}}\) true at w0 are just those true at v0,v1, and the \(\varphi \in {\mathscr{L}}\) true at w1 are just those true at v2 extending this to \(\varphi \in {\mathscr{L}}_{\boxdot }\) is straightforward. The desired result–\(\vDash _{{\mathscr{M}}} \varphi \) iff \(\vDash _{{\mathscr{M}}^{\prime }} \varphi \)–is now immediate. □
We now prove Fact 3.3.
Proof
Suppose our model is partitional and that (3 - 5) and (Ci) are true (at w ∈ W), where i is the least number such that \(\Box (p \wedge q_{i})\) and \(\neg \Box (p \wedge q_{i-1})\) are true. As (by (4)) \(S_{\mathsf {min}}(w) \subseteq \llbracket q_{500,000} \rrbracket \) and \(S_{\mathsf {min}}(w) \subseteq R(w)\), so (by (3) and PS) for all \(w^{\prime } \in S_{\mathsf {min}}(w)\), \(S_{\mathsf {min}}(w^{\prime })\) is defined and \(S_{\mathsf {min}}(w^{\prime }) \subseteq \llbracket p \rrbracket \subseteq S_{\mathsf {max}}(w^{\prime })\). By partitionality, \(S_{\mathsf {min}}(w) = S_{\mathsf {min}}(w^{\prime })\). We thus have for all \(\Box (\varphi \wedge p)\) true at w (i.e. such that \(S_{\mathsf {min}}(w) \subseteq \llbracket \varphi \rrbracket \cap \llbracket p \rrbracket \subseteq \llbracket p \rrbracket \subseteq S_{\mathsf {max}}(w)\), with Smin defined and fixed throughout Smin(w)) that \(\llbracket \Box (\varphi \wedge p) \rrbracket \supseteq S_{\mathsf {min}}(w)\), and thus \(S_{\mathsf {min}}(w) \subseteq \llbracket \Box (p \wedge q_{i}) \rrbracket \subseteq \llbracket p \rrbracket \subseteq S_{\mathsf {max}}(w)\), and therefore we have \(\Box \Box (p \wedge q_{i})\) true (at w).
Since we also have \(\neg \Box (p \wedge q_{i-1})\) true (at w), we by similar reasoning have \(\Box (p \wedge q_{i-1})\) true at no \(w^{\prime } \in S_{\mathsf {min}}(w)\), and thus \(S_{\mathsf {min}}(w) \subseteq \llbracket p \rrbracket \cap \llbracket \neg \Box (p \wedge q_{i-1}) \rrbracket \subseteq \llbracket p \rrbracket \subseteq S_{\mathsf {max}}(w)\), and thus \(\Box (p \wedge \neg \Box (p \wedge q_{i-1}))\) true (at w).
We by these and M therefore have \(\Box (\Box (p \wedge q_{i}) \wedge p \wedge \neg \Box (p \wedge q_{i-1}))\), or equivalently \(\Box (\Box (p \wedge q_{i}) \wedge \neg \Box (p \wedge q_{i-1}))\), true (at w), which contradicts (Ci). □
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Mollica, L. The Logic of Action and Control. J Philos Logic 52, 1237–1268 (2023). https://doi.org/10.1007/s10992-023-09704-9
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10992-023-09704-9