Skip to main content
Log in

Transmission Failure, AGM Style

  • Original Article
  • Published:
Erkenntnis Aims and scope Submit manuscript

Abstract

This article provides a discussion of the principle of transmission of evidential support across entailment from the perspective of belief revision theory in the AGM tradition. After outlining and briefly defending a small number of basic principles of belief change, which include a number of belief contraction analogues of the Darwiche–Pearl postulates for iterated revision, a proposal is made concerning the connection between evidential beliefs and belief change policies in rational agents. This proposal is found to be sufficient to establish the truth of a much-discussed intuition regarding transmission failure.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Similar content being viewed by others

Notes

  1. See for instance Wright (1985, 2000, 2002, 2003).

  2. For article-length introductions to this framework, see for example Arló-Costa and Pedersen (forthcoming), Hansson (2011) or Peppas (2008). For book-length treatments, see Hansson (1999b) or Gärdenfors (2008).

  3. These include the so-called ‘AGM’ and ‘Darwiche–Pearl’ postulates; see Gärdenfors (1984) and Darwiche and Pearl (1997).

  4. The occurrence of ‘\(\vartriangleright\)’ within the scope of a connective is not permitted in this model.

  5. Revision or contraction by evidential beliefs is not modeled here.

  6. Readers who are acquainted with the belief revision literature will note that many accounts of belief revision pack a great deal more into the entities that are revised or contracted. Entrenchment relations (Gärdenfors 2008) are a case in point, having way more structure than a mere set of sentences. Belief sets drawn from an extended language that includes ‘Ramsey Test conditionals’ (Gärdenfors 1986) are another. Very strong constraints on transitions between these rich entities are then imposed, strong enough to restrict the set of rationally permissible contraction or revision functions to a singleton: for any given language, there is a single contraction and a single revision function whose domain is a large set of extremely finely individuated states.

    For what are principally notational motivations, I have opted for an alternative way of proceeding, allowing, for each language, a multiplicity of contraction and revision functions whose domain is a small set of relatively coarsely individuated states. What could alternatively be packed into an entrenchment relation or a set of beliefs with Ramsey Test conditionals is here distributed between the agent’s belief set on the one hand, and her particular revision or contraction function on the other.

  7. It is important to note here the restriction to sentences in \(\mathcal{L}.\) Waiving this restriction may yield principles of dubious plausibility. We return to this point below.

  8. As an anonymous referee has remarked, my axiomatization is slightly unorthodox here. Indeed, in standard presentations of the AGM framework, although belief sets are required to be closed under logical consequence, they are not, as they are here, required to be consistent. This allows for (AGM*2) and (AGM*V) to be replaced by a single axiom to the effect that \(A\in K* A,\) for all \(A\in\mathcal{L}\): inconsistent sentences can be handled in the same way as the remainder of the language. To the extent that I have defined \({\mathbb{K}}\) as the set of all rationally permissible belief sets, which I take to be consistent, this option is not available to me.

  9. Perhaps a contraction by \(\{A, A\rightarrow B\}; \) see Spohn (2010) for a recent discussion of so-called ‘multiple contraction’, i.e. contraction by sets of sentences rather than single sentences

  10. It has been suggested to me that one could perhaps interpret Glaister (2000) as offering a response along these lines.

  11. For instance: natural, aka ‘conservative’, contraction, lexicographic contraction and moderate, aka ‘priority’, contraction. See Ramachandran et al. (2011) for a recent exposition and comparison of these three approaches.

  12. For the record, here are the analogues for the remaining two Darwiche and Pearl postulates, namely C1 and C3, respectively. The first of these is implicitly endorsed by Nayak et al. (2006). The second, however, appears to have been overlooked in the literature. Again, these are both satisfied by the main constructive proposals regarding iterated contraction.

    • \(({\bf{I}}\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!{\bf {1}})\) If \(\neg C\in{\rm Cn}(A)\), then \(B\in K*A\) iff \(B\in (K\!\!\!\buildrel\textstyle \centerdot\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\)

    • \(({\bf{I}}\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!{\bf {3}})\) If \(\neg B\in K*A\), then \(\neg B\in (K\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-}} \!B)*A\)

  13. It should be noted that there is a somewhat looser sense of ‘evidence’ that can be found in the philosophical literature, most notably in the Bayesian confirmation-theoretic literature, that does not fit this claim. Here we mean by ‘evidence’ what could perhaps be termed ‘evidence sufficient for rational belief given one’s background assumptions’. And this, in any case, is the sense in which the term is used in the literature on transmission.

  14. This of course simply mirrors the standard ‘Ramsey Test’ proposal for conditionals.

  15. Although it then also follows, given (AGM*V), that arbitrary contradictions are taken to be evidence for whatever one happens to believe. This could be seen as an awkward result.

  16. See for instance Rott (1986), who makes the point in relation to conditionals and the Ramsey Test. I believe that the point was first noted by Gärdenfors (1979).

  17. This is analogous to the variant of the Ramsey Test labeled ‘(R3)’ by Gärdenfors (1987, p. 324) and first proposed by Rott (1986).

    Note that, by the contrapositive of \(({\text{I}}\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!4)\), if \(A\,\vartriangleright\,B\in K\) according to (EV3), then \(A\,\vartriangleright\,B\in K\) according to (EV2).

  18. The somewhat counterintuitive result noted in footnote 16 above is also avoided.

  19. As an anonymous referee has pointed out to me, in the presence of (CON) and (CL), (TRIV) itself entails an even more strongly counterintuitive proposition, namely that there exists no \({K\in\mathbb{K}}\) and pair of atomic sentences \(A, B\in\mathcal{L}\) such that \(A, B\in K\). Indeed, assume that such a K exists. By (CL), we have \(A\vee B, A\rightarrow B, B\rightarrow A\in K.\) By (CON), it follows that the pairwise inconsistent negations of these sentences, namely \(\neg A\wedge\neg B, A\wedge\neg B,\) and \(\neg A\wedge B\) respectively, are not in K, contradicting (TRIV).

  20. Adapted from Gärdenfors (1987). A more explicit version of his original proof, which clearly states the various assumptions involved, is provided in the Appendix.

  21. Note that all that we need here is the weak claim that such a K be rationally permissible. Rational obligation, though perhaps intuitive in the present example, is more than is required.

    Note that Hansson (1992) voices similar concerns regarding (AGM*3+) and (AGM*4+), in relation to a conditional language.

  22. To see the relationship between (CL) and (CCL), note that (CL) can be equivalently rewritten as:

    • (CL) For all \(A\in\mathcal{L}\) and \(\Upgamma\subseteq\mathcal{L}_{E},\) if \(A\in \text{Cn}(\Upgamma),\) then \(A\in K\) for all \({K\in \mathbb{K}}\) such that \(\Upgamma\subseteq K\).

  23. The deathcap, otherwise known as Amanita phalloides, is a particularly virulent poisonous mushroom commonly found in Europe.

  24. Note that, given the principle according to which, for non-contradictory A, if \(B\in {\rm Cn}(A),\) then \(A\,\vartriangleright\,B\in K\) (which, as we have seen, our proposal entails) and (CL), these cases are also counterexamples to transitivity of evidential support, i.e. the principle according to which \(A\,\vartriangleright\,C\in{\rm Cn}(\{A\,\vartriangleright\,B, B\,\vartriangleright\,C\}).\)

  25. This is a slight departure here from standard usage of the term. Cases of transmission failure are typically taken to be triples \(\langle A,B,C\rangle\) of sentences, such that \(C\in {\rm Cn}(B), A\,\vartriangleright\,B\) is true in some world, but \(A\,\vartriangleright\,C\) is false in that world. Of course, given (CL), if K is a case of transmission failure according to the definition offered here, \(\langle A,B,C\rangle\) is a case of transmission failure according to the standard definition.

  26. Proof omitted, but straightforward.

  27. In fact, one can show that given (CL), \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} }\! {{3}})\), (AMG*2), (AMG*4), and (EV3), (TR1) cannot hold on pains of ‘triviality’, i.e. of placing unwarranted constraints on \({\mathbb{K}}\). More precisely, given these assumptions, (TR1) entails the following:

    • (TRIV2) There is no \({K\in \mathbb{K}}\), and \(A,B,C\in\mathcal{L}\) such that \(A\in K, B,\neg B\notin K\) but \(B\,\vartriangleright\,A\notin K\).

    For the proof, we show that, in the presence of the relevant background conditions, the contrapositive holds: if (TRIV2) is false, then so too is (TR1). Indeed, assume that \(B\notin K\). It then follows by (CL) that \(A\wedge B\notin K\). By \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{3}})\) it then follows that \(K\!\!\! \buildrel\textstyle\centerdot\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!A\wedge B = K\), and hence that \((K\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! A\wedge B )*B = K*B\). Assume that \(A\in K\) but that \(\neg B \notin K\). It follows by (AGM*4) that \(A\in K*B\) and hence that \(A\in(K\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! A\wedge B )*B\). Since, by assumption, \(\neg B\notin K\), it follows by (CL) that \(\neg B\notin {\rm Cn}(\varnothing)\). So by (AGM*2), we also have \(B\in(K\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! A\wedge B )*B\). It then follows by (CL) that \(A\wedge B\in(K\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! A\wedge B )*B\). By the right-to-left direction of (EV3), \(B\,\vartriangleright\,(A\wedge B)\in K\). Assume furthermore that \(B\,\vartriangleright\,A\notin K\). Since \(A\in{\rm Cn}(A\wedge B)\), (TR1) must be false. \(\square\)

    And of course, (TRIV2) is a clearly undesirable constraint to impose. Let A stand for the proposition that it is sunny today in Leuven (which I believe) and B stand for the proposition that a parcel has arrived for me this morning (which I suspend judgment on). It seems permissible for me to not hold that B would be evidence for A (indeed: I do not).

  28. This is not quite how Wright phrases it. His precise claim appears to be that the principle (TR1) fails when one would be entitled to infer B from A only given prior entitlement to believe C (Wright 1985, p. 433).

  29. Adapted from Gärdenfors (1987).

References

  • Arló-Costa, H., & Pedersen A. P.(forthcoming). Belief revision. In L. Horsten & R. Pettigrew (Eds.), Continuum companion to philosophical logic. London: Continuum Press.

  • Chandler, J. (2010a). The transmission of support: A Bayesian re-analysis. Synthèse, 176(3), 333–343.

    Article  Google Scholar 

  • Chandler, J. (2010b). The lottery paradox generalized? The British Journal for the Philosophy of Science, 61(3), 667–679.

    Article  Google Scholar 

  • Darwiche, A., & Pearl, J. (1997). On the logic of iterated belief revision. Artificial Intelligence, 89, 1–29.

    Article  Google Scholar 

  • Gärdenfors, P. (1979). Conditionals and changes of belief. In I. Niiniluoto & R. Tuomelo (Eds.), The logic and epistemology of scientific change. Acta Filosofica Fennica, 30, 381–404.

  • Gärdenfors, P. (1984). Epistemic importance and minimal changes of belief. Australasian Journal of Philosophy, 62, 136–157.

    Article  Google Scholar 

  • Gärdenfors, P. (1986). Belief revisions and the Ramsey test for conditionals. The Philosophical Review, 95(1), 81–93.

    Article  Google Scholar 

  • Gärdenfors, P. (1987). Variations on the Ramsey test: More triviality results. Studia Logica, 46(4), 321–327.

    Article  Google Scholar 

  • Gärdenfors, P. (2008). Knowledge in flux. London: College Publications.

    Google Scholar 

  • Glaister, S. M. (2000). Recovery recovered. Journal of Philosophical Logic, 29(2), 171–206.

    Article  Google Scholar 

  • Hansson, S. O. (1991). Belief contraction without recovery. Studia Logica, 50(2), 251–260.

    Article  Google Scholar 

  • Hansson, S. O. (1992). In defense of the Ramsey test. Journal of Philosophy, 89(10), 522–540.

    Article  Google Scholar 

  • Hansson, S. O. (1999a). Recovery and epistemic residue. Journal of Logic, Language, and Information, 8(4), 421–428.

    Article  Google Scholar 

  • Hansson, S. O. (1999b). A textbook of belief dynamics. Theory change and database updating. Dordrecht: Kluwer.

    Book  Google Scholar 

  • Hansson, S.O. (2011). Logic of belief revision. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Fall 2011 Edition).

  • Lindström, S. & Rabinowicz, W. (1995). The Ramsey test revisited. In G. Crocco, L. Fariñas del Cerro & A. Herzig (Eds.), Conditionals: From philosophy to computer science. Oxford: Clarendon Press.

    Google Scholar 

  • Nayak, A., Goebel, R., Orgun, M., & Pham, T. (2006). ‘Taking Levi identity seriously: A plea for iterated belief contraction’. In Knowledge science, engineering and management: First international conference (KSEM’06), (pp. 305–317). Dordrecht: Springer.

  • Okasha, S. (2004). Wright on the transmission of support: A Bayesian analysis. Analysis, 64(282), 139–146.

    Article  Google Scholar 

  • Peppas, P. (2008). Belief revision. In F. van Harmelen, V., Lifschitz & B. Porte, (Eds.), Handbook of knowledge representation (Vol .3, pp. 317–359). San Diego: Elsevier Science.

    Chapter  Google Scholar 

  • Rott, H. (1986). Ifs, though, and because. Erkenntnis, 25(3), 345–370.

    Article  Google Scholar 

  • Rott, H. (2004). A counterexample to six fundamental principles of belief formation. Synthèse, 139(2), 225–240.

    Article  Google Scholar 

  • Ramachandran, R., Nayak, A. C., & Orgun, M. A. (2011). Three approaches to iterated belief contraction. Journal of Philosophical Logic. Advance online publication. doi:10.1007/s10992-011-9201-7.

  • Spohn, W. (2010). Multiple contraction revisited. In M. Suàrez, M., Dorato & M. Rédei (Eds.), EPSA epistemology and methodology of science (Vol. 1, pp. 279–288). Dordrecht: Springer.

  • Stalnaker, R. (2009). Iterated belief revision. Erkenntnis, 70(2), 189–209.

    Article  Google Scholar 

  • Wright, C. (1985). Facts and certainty. Proceedings of the British Academy, 71, 429–472.

    Google Scholar 

  • Wright, C. (2000). Cogency and question-begging: Some reflections on McKinsey’s paradox and Putnam’s proof. Noûs, 34(s1), 140–163.

    Article  Google Scholar 

  • Wright, C. (2002). (Anti-)Sceptics simple and subtle: G. E. Moore and John McDowell. Philosophy and Phenomenological Research, 65(2), 330–348.

    Article  Google Scholar 

  • Wright, C. (2003). Some reflections on the acquisition of warrant by inference. In S. Nuccetelli, (Ed.), New essays on semantic externalism and self-knowledge (pp. 57–77). Cambridge, MA: MIT.

    Google Scholar 

Download references

Acknowledgments

The research for this article was funded by a Research Foundation—Flanders (FWO) postdoctoral research grant. Special thanks are due to Peter Fritz, for providing me with the result reported in footnote 28, as well as spotting an important mistake in an early incarnation of the manuscript. I am also extremely grateful to two anonymous referees for this journal for their incisive and helpful feedback. The level of detail of the comments provided by one of these referees, in particular, went way beyond the call of duty. Finally, I would like to thank Peter Brössel, Franz Huber, James Joyce, Abhaya Nayak, Jim Pryor and Hans Rott for helpful discussions at earlier stages of the project.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jake Chandler.

Appendix

Appendix

Proof of Observation 1

Assume, for conditional proof, that \(A\in K\). By (CL), (SUP) and (CON) it follows that \(\neg A\notin K\). Now assume, again for conditional proof, that \(B\in K\) and hence that \(B \in{\rm Cn}(K\cup\{A\})\). It follows by (AGM*4) that \(B\in K*A\). By the right-to-left half of (EV2), it then follows that \(A\,\vartriangleright\,B\in K\). \(\square\)

Proof of Observation 2

Assume, for conditional proof, that \(B \notin K\). By \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{3}})\), it follows that \(K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B=K\). So \(B\in K*A\) iff \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B)*A\), and by (EV3), iff \(A\,\vartriangleright\,B\in K\). \(\square\)

Proof of Observation 3

Given \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{3}})\), (EV3) entails (EV1) (see Observation 2). Furthermore, (EV1) entails:

  • (WM) If \(K\subseteq K', A\vee C\notin K'\) and \(C\in K*A\), then \(C\in K'*A\)

Indeed, assume that \(K\subseteq K'\) and \(A\vee C\notin K'\). It follows that \(A\vee C\notin K\), and hence, by (CL), that \(A,C\notin K\). Assume furthermore that \(C\in K*A\). It follows by (EV1), and the fact that \(K\subseteq K'\), that \(A\,\vartriangleright\,C\in K\), that \(A\,\vartriangleright\,C\in K'\), and therefore, by (EV1), that \(C\in K'*A\).

Now assume for reductio that \({K\in\mathbb{K}}\) is such that, for pairwise inconsistent \(A,B,C\in\mathcal{L}, \neg A, \neg B, \neg C\notin K\).

The general proof strategy is as follows. We first derive (i) \(B\in (K*A)*B\vee C\) and (ii) \(C\in (K*A)*B\vee C\). By (CL), this then entails that \(B\wedge C\in (K*A)*B\vee C\), which violates (CON). Here, we simply prove (i). The proof for (ii) is analogous.

From (AGM*3+), (AGM*4+), (SUP) and (MON), it follows that \(K*A\vee B\subseteq K*A\). Indeed consider an arbitrary \(C\in K*A\vee B\). By (AGM*3+), it follows that \(C\in {\rm Cn}(K\cup\{A\vee B\})\). By (MON), we have \(C\in {\rm Cn}(K\cup\{A\vee B, A\})\). By (SUP), the latter is equal to Cn(K ∪ {A}). Now, by hypothesis, \(\neg A\notin K\), so by (AGM*4+), it follows that \(C\in K*A\).

We now show that \(B\vee C\notin K*A\), and hence, by (SUP) and (CL), that \((B\vee C)\vee B\notin K*A\). Assume for reductio that \(B\vee C\in K*A\). Then by (AGM*3), \(B\vee C\in {\rm Cn}(K\cup\{A\})\). By (DT), \(A\rightarrow(B\vee C)\in K\). Since A is inconsistent both with B and with C, it follows by (CL) that \(\neg A\in K\), contrary to our assumptions.

Next, we prove that \(B\in (K*A\vee B)*B\vee C\). We first show that \(\neg(B\vee C)\notin K*A\vee B\). Assume for reductio that \(\neg(B\vee C)\in K*A\vee B\). Then by (AGM*3), it follows that \(\neg(B\vee C)\in {\rm Cn}(K\cup\{A\vee B\})\). By (DT), we have \((A\vee B)\rightarrow\neg(B\vee C)\in K\). By (SUP), we have \(\neg B\in{\rm Cn}((A\vee B)\rightarrow\neg(B\vee C))\), so by (CL), it follows that \(\neg B\in K\), contrary to our assumptions.

Now, since \(\neg A, \neg B\notin K\), by (CL), \(\neg A, \neg B\notin {\rm Cn}(\varnothing)\), and hence, by (SUP), it follows that \(\neg(A\vee B)\notin {\rm Cn}(\varnothing)\). By (AGM*2), therefore, \(A\vee B\in K*A\vee B\). Since AB and C were assumed to be pairwise inconsistent, it follows from this and (SUP) that \(B\in {\rm Cn}(K*A\vee B\cup\{B\vee C\})\). By (AGM*4), and the fact that \(\neg(B\vee C)\notin K*A\vee B\), it then follows that \(B\in (K*A\vee B)*B\vee C\).

From the fact that \(K*A\vee B\subseteq K*A, (B\vee C)\vee B\notin K*A\), and \(B\in (K*A\vee B)*B\vee C\), which have all now been derived, it then follows by (WM) that (i) \(B\in (K*A)*B\vee C\). \(\square\)

Proof of Observation 4

Assume, for conditional proof, that (i) \(C\in {\rm Cn}(B)\) and (ii) \(A\,\vartriangleright\,B\in K\). By the left-to-right half of (EV2) and (ii), it follows that \(B\in K*A\). From this, by (i) and (CL), we have \(C\in K*A\). By the right-to-left half of (EV2), it then follows that \(A\,\vartriangleright\,C\in K\). \(\square\)

Proof of Observation 5

We divide the proof into three cases: (a) the case in which \(C\in{\rm Cn}(\varnothing)\), (b) the case in which \(\neg A\in {\rm Cn}(\varnothing)\), and (c) the case in which neither \(C\in{\rm Cn}(\varnothing)\), nor \(\neg A\in {\rm Cn}(\varnothing).\)

Regarding (a), the proof is trivial. Assume that \(C\in{\rm Cn}(\varnothing)\). It follows from this, by \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} }\! {{3}})\), that (i) \(K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C = K\). Assume \(A\,\vartriangleright\,B\in K\). From (i), it then follows that \(A\,\vartriangleright\,B\in K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\). It also follows from \(C\in{\rm Cn}(\varnothing)\), this time by (CL), that (ii) \(C\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C )*A\). From (ii), by the right-to-left direction of (EV3), it follows that \(A\,\vartriangleright\,C\in K\). So we have \(A\,\vartriangleright\,C\in K\) and \(A\,\vartriangleright\,B\in K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\), and therefore \(A\,\vartriangleright\,C\notin K\) iff \(A\,\vartriangleright\,B\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\).

Case (b) is equally easy to handle. There are two subcases to consider, namely \(B\notin{\rm Cn}(\varnothing)\) and \(B\in{\rm Cn}(\varnothing): \)

  • Assume \(B\notin{\rm Cn}(\varnothing)\). By \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{4}})\) we have (i) \(B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B)\). Assume \(\neg A\in {\rm Cn}(\varnothing)\). By (AGM * V), we then have (ii) \((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! B)*A=K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B\). It follows from (i) and (ii) that \(B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B)*A\) and hence, by the left-to-right direction of (EV3), that \(A\,\vartriangleright\,B\notin K\). It is then vacuously true that if \( C\in {\rm Cn}(B)\) and \(A\,\vartriangleright\,B\in K\), then \(A\,\vartriangleright\,C\notin K\) iff \(A\,\vartriangleright\,B\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\).

  • Assume \(B\in{\rm Cn}(\varnothing)\). It then follows, by (CL), that \(B\in((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B)*A\), and hence, by the right-to-left direction of (EV3), \(A\,\vartriangleright\,B\in K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\). Furthermore, since \(B\in{\rm Cn}(\varnothing)\) and \( C\in {\rm Cn}(B)\), we have \(C\in{\rm Cn}(\varnothing)\). But then, by (CL), we have \(C\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\) and therefore, by the right-to-left direction of (EV3), \(A\,\vartriangleright\,C\in K\). So \(A\,\vartriangleright\,C\in K\) and \(A\,\vartriangleright\,B\in K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\), and therefore \(A\,\vartriangleright\,C\notin K\) iff \(A\,\vartriangleright\,B\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\).

Regarding (c), we consider each direction of the biconditional separately.

The left-to-right direction of this biconditional, from (1) to (2), is easy enough to establish. Indeed, assume that \(A\,\vartriangleright\,C\notin K\). By the right-to-left direction of (EV3), it follows that \(C\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\). Now assume \(C\in {\rm Cn}(B)\). It then follows from \(C\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\), by (CL), that (i) \(B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\). Assume \(C\notin{\rm Cn}(\varnothing)\). By \(({\text{AGM}}\!\buildrel\textstyle\!\!\!\centerdot\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{4}})\), it then follows that \(C\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\). From this, (CL) and the fact that \( C\in {\rm Cn}(B)\), we obtain \(B\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C\). Therefore, by \(({\text{AGM}}\!\buildrel\textstyle\!\!\!\centerdot\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{3}})\), \((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } B=(K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)\). Given this equality, (i) is equivalent to \(B\notin ((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! B)*A\), from which it follows, by the left-to-right direction of (EV3), that \(A\,\vartriangleright\,B\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C\).

Regarding the other direction of the biconditional, from (2) to (1), things are a little trickier. We first derive

Lemma 1

If \(A\,\vartriangleright\,B\in K\), then \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\wedge C\).

Assume for conditional proof that \(A\,\vartriangleright\,B\in K\). It follows, by the left-to-right direction of (EV3), that (i) \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! B) * A\). From this, and the fact that \(C\in{\rm Cn}(B)\), it follows by (CL) that (ii) \(C\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! B) * A\). From (i), (ii) and (CM), it then follows that \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B) * A\wedge C\). Now assume for reductio that \(B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C) * A\wedge C\). Since \(C\in{\rm Cn}(A\wedge C)\), it follows by \(({\text{I}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} }\! {{2}})\) that \(B\notin K* A\wedge C\). By \(({\text{I}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} }\! {{4}})\) it then follows that \(B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! B) * A\wedge C: \) contradiction. Hence \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\wedge C\).

We now prove

Lemma 2

If \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\wedge C\) and \(A\,\vartriangleright\,B\notin K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } C\), then \(A\,\vartriangleright\,C\notin K\).

Assume for conditional proof both that \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\wedge C\) and that \(A\,\vartriangleright\,B\notin K\!\!\! \buildrel\textstyle\centerdot\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\!C\). From the second assumption, by the right-to-left direction of (EV3), it follows that \(B\notin ((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C) \!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!B)*A\). As noted above, in the proof of the left-to-right direction of the biconditional, given \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} }\! {{4}})\), \(({\text{AGM}}\!\buildrel\textstyle{\!\!\centerdot}\over{\hbox{\vrule height3pt depth0pt width0pt}{-} } \!{{3}})\) and (CL), this is equivalent to \(B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\). Given that \(\neg A\notin {\rm Cn}(\varnothing)\), by (AGM*2), it follows that \(A\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\). But by (SUP), we have \(A\rightarrow B\in {\rm Cn}(\{(A\wedge C)\rightarrow B, A \rightarrow C\})\). So it follows by (CL) that either:

  1. (i)

    \((A\wedge C)\rightarrow B\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\), or

  2. (ii)

    \(A \rightarrow C\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\)

Assume (i) for reductio. But note that we have already assumed that \(B\in (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\wedge C\). Since \(A\in {\rm Cn}(A\wedge C)\), it follows from this, by (I*1), that \(B\in ((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A)*A\wedge C\). By (AGM*3), we then have \(B\in {\rm Cn}((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\!C)*A\cup\{A\wedge C\})\). Finally, by (DT), it follows that \((A\wedge C)\rightarrow B\in {\rm Cn}((K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A)\). Contradiction.

It therefore follows that (ii) \(A\rightarrow C\notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} }\! C)*A\). But since, by (SUP), \(A \rightarrow C\in {\rm Cn}(C)\), it follows by (CL) that \(C \notin (K\!\!\!\buildrel\textstyle{\!\centerdot}\over{\hbox{ \vrule height3pt depth0pt width0pt}{-} } \!C)*A\). Finally, by the left-to-right direction of (EV3), it follows that \(A\,\vartriangleright\,C\notin K\).

The desired result then follows immediately from the conjunction of the two lemmas. \(\square\)

Rights and permissions

Reprints and permissions

About this article

Cite this article

Chandler, J. Transmission Failure, AGM Style. Erkenn 78, 383–398 (2013). https://doi.org/10.1007/s10670-012-9364-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10670-012-9364-9

Keywords

Navigation