Abstract
A new deontic logic, Action Type Deontic Logic, is presented. To motivate this logic, a number of benchmark cases are shown, representing inferences a deontic logic should validate. Some of the benchmark cases are singled out for further comments and some formal approaches to deontic reasoning are evaluated with respect to the benchmark cases. After that follows an informal introduction to the ideas behind the formal semantics, focussing on the distinction between action types and action tokens. Then the syntax and semantics of Action Type Deontic Logic is presented and it is shown to meet the benchmarks. Finally, possibilities for further research are indicated. In the appendix, decidability of the satisfiability of formulas is proved via a technique known from monadic First Order Logic.
Similar content being viewed by others
Notes
This claim is based on my own linguistic intuitions, informed by discussions with the people present in various contexts where I have presented this work. I have, however, not tested the claim statistically, hence the qualification ‘would probably accept.’ Since my default belief is that natural language should be respected as far as possible, I elevate these cases to normative standards or benchmarks that a good deontic theory should meet. Here I appeal to normal linguistic intuitions as a regulative guideline. My view is not that a logical theory should be founded on empirical observations of natural language phenomena. On the contrary, a good theory gives a plausible theoretical explanation of the inferences it validates. However, a theory should provide a very good theoretical reason, if it diverges from natural language, preferably a reason arising within the theory itself. Philosophical logicians are often too eager to sacrifice natural language to save the logical theory they have become accustomed to working with.
This objection and the following example was presented to me by Frank Veltman.
Presumably, the agent also has available a set of actions which are not acceptable, but since we are only concerned with idealized deontic reasoning here, we have no need to model these non-ideal actions. It is an interesting methodological point that we can do without non-ideal actions when modeling deontic reasoning, given the enormous effort which has been put into modeling these in the history of deontic logic. On the other hand, the logic must be extended if we are to handle violations of obligations and contrary-to-duty obligations.
I would like to thank an anonymous reviewer for pointing this out.
References
Abney, K. (2012). Robotics, ethical theory, and metaethics: A guide for the perplexed. In P. Lina, K. Abney, & G. A. Bekey (Eds.), Robot ethics: The ethical and social implications of robotics (pp. 35–52). Cambridge: MIT Press.
Anderson, A. R. (1966). The formal analysis of normative systems. In N. Rescher (Ed.), The logic of decisions and actions. Pittsburgh: University of Pittsburgh Press.
Arkin, R. (2009). Governing lethal behavior in autonomous robots. Boca Raton: CRC Press.
Asher, N., & Bonevac, D. (2005). Free choice permission is strong permission. Synthese, 145, 303–323.
Belnap, N., Perloff, M., & Xu, M. (2001). Facing the future. Oxford: Oxford University Press.
Bentzen, M. M. (2011). Deontic reasoning with incomplete trust. Logique et Analyse, 54(215), 327–342.
Bräuner, T., & Ghilardi, S. (2007). First-order modal logic. In P. Blackburn, J. van Benthem, & F. Wolter (Eds.), Handbook of modal logic (pp. 549–620). Amsterdam: Elsevier.
Bringsjord, S., & Taylor, J. (2012). The divine-command approach to robot ethics. In P. Lina, K. Abney, & G. A. Bekey (Eds.), Robot ethics: The ethical and social implications of robotics (pp. 85–108). Cambridge: MIT Press.
Chellas, B. F. (1980). Modal logic. Cambridge: Cambridge University Press.
Chisholm, R. M. (1964). Contrary-to-duty imperatives and deontic logic. Analysis, 24, 33–36.
Davidson, D. (1963). Actions, reasons, and causes. The Journal of Philosophy, 60(23), 685–700.
Fitting, M., & Mendelsohn, R. L. (1998). First-order modal logic. Berlin: Kluwer Academic.
Forrester, J. (1984). Gentle murder, or the adverbial samaritan. The Journal of Philosophy, 81, 193–197.
Geach, P. (1982). Whatever happened to deontic logic. Philosophia, 11, 1–12.
Geurts, B. (2005). Entertaining alternatives: Disjunctions as modals. Natural Language Semantics, 13, 383–410.
Goble, L. (1996). ‘Ought’ and extensionality. Noûs, 30, 330–355.
Goldblatt, R. (1992). Logics of time and computation. Lecture Notes, second edn, Center for the Study of Language and Information.
Hage, J., & Brouwer, B. (2000). Action types and act tokens in deontic logic of the ought-to-do type. DEON, 00, 187–201.
Hilpinen, R. (Ed.). (1971). Deontic logic: Introductory and systematic readings. Dordrecht: D. Reidel.
Hilpinen, R. (Ed.). (1981). New studies in deontic logic. Dordrecht: D. Reidel.
Horty, J. F. (2001). Agency and deontic logic. Oxford: Oxford University Press.
Jackson, F. (1985). On the semantics and logic of obligation. Mind, 94, 177–195.
Jeffrey, Boolos, & Burgess, (1997). Logic and computability. Cambridge: Cambridge University Press.
Kamp, H. (1973). Free choice permission. Proceedings of the Aristotelian Society, 74, 57–74.
Lindström, S., & Segerberg, K. (2007). Modal logic and philosophy. In P. Blackburn, J. van Benthem, & F. Wolter (Eds.), Handbook of modal logic (pp. 1150–1214). Amsterdam: Elsevier.
Mastop, R. (2011). Norm performatives and deontic logic. European Journal of Analytic Philosophy, 7(2), 83–105.
McNamara, P. (2006). Deontic logic. In E. N. Zalta (Ed.), The stanford encyclopedia of philosophy (Spring 2006 Edition). http://plato.stanford.edu/archives/spr2006/entries/logicdeontic/.
Meyer, J.-J. (1988). A different approach to deontic logic: Deontic logic viewed as a variant of dynamic logic. Notre Dame Journal of Formal Logic, 29, 109–136.
Meyer, J.-J., & Veltman, F. (2007). Intelligent agents and common-sense reasoning. In P. Blackburn, J. van Benthem, & F. Wolter (Eds.), Handbook of modal logic (pp. 991–1029). Amsterdam: Elsevier.
Nute, D. (Ed.). (1997). Defeasible deontic logic. Berlin: Kluwer Academic.
Portner, P. (2009). Modality. Oxford: Oxford University Press.
Prakken, H. (1993). Logical tools for modelling legal argument. Amsterdam: Vrije Universiteit.
Prakken, H., & Sergot, M. (1997). Dyadic deontic logic and contrary-to-duty obligations. In D. Nute (Ed.), Defeasible deontic logic (pp. 223–262). Berlin: Kluwer Academic.
Prior, A. N. (1954). The paradoxes of derived obligation. Mind, 63, 64–65.
Ross, A. (1941). Imperatives and logic. Theoria, 7, 53–71.
Ross, W. D. (1930). The right and the good. Oxford: Clarendon Press.
Segerberg, K. (2012). D\(\Delta \)L: A dynamic deontic logic. Synthese, 185, 1–17.
van Benthem, J., & Liu, F. (2007). Dynamic logic of preference upgrade. Journal of Applied Non-Classical Logics, 17, 129–155.
van der Meyden, R. (1996). The dynamic logic of permission. Journal of Logic and Computation, 6, 465–479.
van der Torre, L. W. N., & Tan, Y.-H. (1999). An update semantics for deontic reasoning. In P. McNamara & H. Prakken (Eds.), Norms, Logics and information systems (pp. 73–90). IOS Press.
Veltman, F. (1996). Defaults in update semantics. Journal of Philosophical Logic, 25, 221–261.
von Wright, G. H. (1951). Deontic logic. Mind, 60, 1–15.
Yamada, T. (2006). Acts of commanding and changing obligations. In Proceedings of the seventh international workshop on computational logic in multi-agent systems (CLIMA VII), pp. 1–19.
Zimmermann, T. E. (2000). Free choice disjunction and epistemic possibility. Natural language semantics, 8, 255–290.
Acknowledgments
I would like to thank Frank Veltman, Jens Ulrik Hansen, Klaus Frovin Jørgensen, Carola Eschenbach, and Patrick Blackburn for helpful suggestions regarding this work.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
I now state the definitions and lemmas needed to prove theorem 2 stating that Action Type Deontic Logic is decidable.
Definition 9
Let \(M=\langle G, V, \theta \rangle \) and \(M'=\langle G', V', \theta '\rangle \) be models. Define a strong homomorphism from \(M\) to \(M'\) to be a function \(h\) from \(dom(M)\) to \(dom(M')\), such that.
-
1.
For any \(p_i\in P\), \(\theta (p_i)=\theta '(p_i)\).
-
2.
For any \(T_i\in BT\), \(\alpha \in G\) (\(\alpha \in V(T_i)\) iff \(h(\alpha )\in V'(T_i)\)).
For the following lemma to go through we need a surjective strong homomorphism. Let \(M\) and \(M'\) be models and \(h\) a surjective, strong homomorphism from \(M\) to \(M'\).
Lemma 1
-
1.
For any \(T\in ACT\), \(\alpha \in G\) ( \(\alpha \in V(T)\) iff \(h(\alpha )\in V'(T)\)).
-
2.
For any \(\phi \in WFF\), \(M\vDash \phi \) iff \(M'\vDash \phi \)
Part 2 of the above lemma shows that satisfiability is invariant under surjective, strong homomorphisms.
Proof
-
1.
We define the complexity of an action term as the number of action type terms in the action type term. The proof is by induction on the complexity of action terms.
-
1.
Induction basis. Let \(T_i\in BT\), \(\alpha \in G\). We have \(\alpha \in V(T_i)\) iff \(h(\alpha )\in V'(T_i)\) by definition of surjective, strong homomorphism.
-
2.
Suppose the lemma holds for any action type term of complexity lower than \(n\). Now suppose \(T\) is an action type term with complexity \(n\). The proof divides into several cases.
-
(a)
\(T\) is of the form \(-S\). Left to right. If \(\alpha \in V(-S)\), then \(\alpha \in G\setminus V(S)\), and so \(\alpha \not \in V(S)\). Then \(h(\alpha )\not \in V'(S)\) (by induction hypothesis), so \(h(\alpha )\in V'(-S)\). Right to left. If \(h(\alpha )\in V'(-S)\), then \(h(\alpha )\in G'\setminus V'(S)\), and so \(h(\alpha )\not \in V'(S)\). Then \(\alpha \not \in V(S)\) (by induction hypothesis), so \(\alpha \in V(-S)\).
-
(b)
\(T\) is of the form \((S\cap R)\). Left to right. If \(\alpha \in V(S\cap R)\), then \(\alpha \in V(S)\) and \(\alpha \in V(R)\). Hence, by induction hypothesis, \(h(\alpha )\in V'(S)\) and \(h(\alpha )\in V'(R)\). Hence \(h(\alpha )\in V'(S\cap R)\). Right to left. If \(h(\alpha )\in V'(S\cap R)\), then \(h(\alpha )\in V'(S)\) and \(h(\alpha )\in V'(R)\). Hence, by induction hypothesis, \(\alpha \in V(S)\) and \(\alpha \in V(R)\). Hence \(\alpha \in V(S\cap R)\).
-
(c)
\(T\) is of the form \((S\uplus R)\). Left to right If \(\alpha \in V(S\uplus R)\), then 1) \(\alpha \in V(S)\) or 2) \(\alpha \in V(R)\). Assume 1. By induction hypothesis \(h(\alpha )\in V'(S)\). Since \(V(R)\ne \emptyset \) by definition of \(\uplus \), there is \(\beta \in V(R)\). By induction hypothesis \(h(\beta )\in V'(R)\). Hence \(h(\alpha )\in V'(S\uplus R)\). The other case is similar. Right to left. If \(h(\alpha )\in V(S\uplus R)\), then 1) \(h(\alpha )\in V'(S)\) or 2) \(h(\alpha )\in V'(R)\). Assume 1. By induction hypothesis \(\alpha \in V(S)\). \(V'(R)\ne \emptyset \) by definition of \(\uplus \). Let \(\delta \in V'(R)\). Since \(h\) is surjective, there is \(\beta \in G\) such that \(h(\beta )=\delta \). By induction hypothesis \(\beta \in V(R)\). Hence, \(\alpha \in V(S\uplus R)\). The other case is similar.
-
(a)
-
1.
-
2.
We define the complexity of a formula as the number of propositional connectives in the formula. The proof is by induction on the complexity of formulas.
-
1.
There are two basic cases, according to whether \(\phi \) is a propositional variable \(p_i\in P\) or \(\phi \) is of the form \(\mathbf{may}[T]\) for some \(T\in ACT\). \(M\vDash p_i\) iff\( M'\vDash p_i\) is obvious from definition of surjective, strong homomorphism. Assume \(M\vDash \mathbf{may}[T]\). Then there is \(\alpha \in V(T)\). By part 1 of this lemma, \(h(\alpha )\in V'(T)\). Hence \(M'\vDash \mathbf{may}[T]\). Assume \(M'\vDash \mathbf{may}[T]\). Then there is \(\beta \in V'(T)\). Since \(h\) is surjective, there is \(\delta \in G\), such that \(h(\delta )=\beta \). Hence by part 1 of this lemma, \(\delta \in V(T)\). Therefore, \(M\vDash \mathbf{may}[T]\).
-
2.
Suppose the lemma holds for all formulas of complexity lower than \(n\). Now, let \(\phi \) be a formula of complexity \(n\). We consider two cases.
-
(a)
Suppose \(\phi \) is of the form \(\lnot \psi \). Left to right. If \(M\vDash \lnot \psi \), then \(M\nvDash \psi \) and hence \(M'\nvDash \psi \) (by induction hypothesis) and so \(M'\vDash \lnot \psi \), i.e. \(M'\vDash \phi \). Right to left. If \(M'\vDash \lnot \psi \), then \(M'\nvDash \psi \) and hence \(M\nvDash \psi \) (by induction hypothesis) and so \(M\vDash \lnot \psi \).
-
(b)
Suppose \(\phi \) is of the form \(\psi \wedge \chi \). Left to right. If \(M\vDash \psi \wedge \chi \), then \(M\vDash \psi \) and \(M\vDash \chi \). Hence \(M'\vDash \psi \) and \(M'\vDash \chi \) (by induction hypothesis). So, \(M'\vDash \psi \wedge \chi \) i.e. \(M'\vDash \phi \). Right to left. If \(M'\vDash \psi \wedge \chi \), then \(M'\vDash \psi \) and \(M'\vDash \chi \). Hence \(M\vDash \psi \) and \(M\vDash \chi \) (by induction hypothesis). So, \(M\vDash \psi \wedge \chi \).\(\square \)
-
(a)
-
1.
Now we show decidability of satisfiability with a technique used for showing this for monadic First Order Logic, see e.g. Jeffrey et al. (1997). We prove the following lemma.
Lemma 2
If a formula \(\phi \) is satisfiable, then it is satisfiable in a model with a domain of size at most \(2^n\), where \(n\) is the number of basic action type terms occurring in \(\phi \).
Proof
Let \(\phi \) be a formula and \(M=\langle G, V, \theta \rangle \) a model such that \(M\vDash \phi \). Let \(S=\{s_1,\ldots ,s_n\}\) be the basic action types occurring in \(\phi \), and \(Q=\{q_1,\ldots ,q_m \}\) be the propositional variables occurring in \(\phi \). We first derive the model \(M'=\langle G', V', \theta '\rangle \) from \(M\) as follows.
-
1.
\(dom(M')=dom(M).\)
-
2.
For any \(T_i\in BT\), if \(T_i\in S\), then \(V'(T_i)=V(T_i)\). Otherwise \(V'(T_i)=\emptyset \).
-
3.
For any \(p_i\in P\), if \(p_i\in Q\), then \(\theta '(p_i)=T(p_i)\). Otherwise \(\theta '(p_i)=f\).
In a sense, \(M'\) is \(M\) stripped of all information not pertaining to \(\phi \). Since the truth of \(\phi \) only depends on its subformulas, an easy inductive argument shows that \(M'\vDash \phi \).
Let the signature of an \(\alpha \in dom(M')\) be a sequence \((j_1, \ldots , j_n)\) of \(0's\) and \(1's\), such that, for each \(s_k\in S\)
-
1.
\(j_i=1\), if \(\alpha \in V(s_k).\)
-
2.
\(j_i=0\), if \(\alpha \not \in V(s_k)\).\(\square \)
There are at most \(2^n\) signatures. Call \(\alpha \) and \(\beta \) similar, if they have the same signature. Obviously, similarity is an equivalence relation. For an element \(\alpha \in G'\), we use the notation \(|\alpha |\) for the equivalence class \(C\) s.t. \(\alpha \in C\). Now we define \(N= \langle F, V^N, \theta ^N \rangle \) as follows. The domain of \(N\) consists of exactly one element from each of the equivalence classes induced by the similarity relation. Hence \(dom(N)\) is of size at most \(2^n\). Further, we define:
Definition 10
-
1.
For each \(T_i\in BT, \beta \in dom(N)\) ( \(\beta \in V^N(T_i)\) iff \(\beta \in V'(T_i)\)).
-
2.
For each \(p_i\in P, \theta ^N(p_i)=\theta '(p_i)\).
We now define a function \(h\), from \(dom(M')\) to \(dom(N)\) as follows. For each \(\alpha \in G', h(\alpha )= \delta \), where \(\delta \in dom(N)\), and \(|\alpha |=|\delta |\). Clearly, there is exactly one such \(\delta \), hence \(h\) is a surjective function. Further, by Definition 10 \(h\) is a strong homomorphism. Hence by Lemma 1, \(N\vDash \phi \).
Theorem 2 follows directly from Lemma 2.
Rights and permissions
About this article
Cite this article
Bentzen, M.M. Action Type Deontic Logic. J of Log Lang and Inf 23, 397–414 (2014). https://doi.org/10.1007/s10849-014-9205-0
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10849-014-9205-0