2
$\begingroup$

I'm working on a formalization of Lindenbaum's completeness lemma for modal logic systems, but I've been stuck in a property. Namely, when trying to prove that:

$$\forall\Gamma,\forall\phi,\enspace\Gamma\ \textit{is consistent} \implies (\Gamma,\phi)\ \textit{is consistent}\ \lor\ (\Gamma,\neg\phi)\ \textit{is consistent}$$

Following some books, e.g., Blackburn's and Popkorn's, this seems to boil down to a lemma:

$$(\Gamma\vdash\phi)\iff (\Gamma,\neg\phi)\ \textit{is not consistent}$$

However, when trying to prove the only if side by contraposition, I get that, assuming:

  1. $(\Gamma\vdash\phi)\implies\bot$
  2. $\Gamma\ \textit{is consistent}$ (derivable from 1)
  3. $\Gamma,\neg\phi \vdash \psi$
  4. $\Gamma,\neg\phi \vdash \neg\psi$

...should derive a contradiction. I can see how to do that if I assume a deduction theorem, thus I can, from 3 and 4, derive both $\Gamma\vdash\neg\phi\rightarrow\psi$ and $\Gamma\vdash\neg\phi\rightarrow\neg\psi$, from which we derive that $\Gamma\vdash\phi$, which is a contradiction, as expected. However, I cannot see how to do this without the deduction theorem, which may not be available in the modal logic system. I'm not sure if I'm missing something, as I've seem this property even left as an exercise to the reader in a book (sorry, I forgot which one).

So, my questions: (1) can this be proved without using the deduction theorem? (2) If so, how? (3) If not so, does this mean that Lindenbaum's lemma actually depends on the deduction theorem?

$\endgroup$
13
  • 1
    $\begingroup$ Maybe have a look at Hakli and Negri's paper 'Does the deduction theorem fail for modal logic?', doi.org/10.1007/s11229-011-9905-9 $\endgroup$ Commented Mar 7 at 19:46
  • $\begingroup$ Your property implies the deduction theorem: if $\Gamma,\phi\vdash\psi$, then $\Gamma,\neg(\phi\to\psi)$ is inconsistent, because $\neg(\phi\to\psi)\vdash\neg\psi$ and $\neg(\phi\to\psi)\vdash\phi$, whence $\Gamma,\neg(\phi\to\psi)\vdash\psi$ by the assumption. $\endgroup$ Commented Mar 7 at 19:49
  • $\begingroup$ Of course, the whole thing only makes good sense if $\vdash$ denotes the local consequence relation of the modal logic (where you can use theorems of the logic and modus ponens, but you cannot apply the necessitation rule to anything derived from nontautological premises). $\endgroup$ Commented Mar 7 at 19:58
  • $\begingroup$ Really, in the context of Lindenbaum’s lemma, consistency should be defined so that $\Gamma$ is consistent iff there is no finite subset $\Gamma_0\subseteq\Gamma$ such that $\vdash\neg\bigwedge\Gamma_0$. $\endgroup$ Commented Mar 7 at 20:01
  • $\begingroup$ @EmilJeřábek, I've been using a global consequence relation. In the version I'm using, there's no deduction theorem. Do you mean that Lindenbaum's lemma actually depends on it? $\endgroup$ Commented Mar 7 at 20:32

1 Answer 1

3
$\begingroup$

You need classical logic for this. Presumably the following rules for negation are valid in your logic:

  1. Elimination rule: if $\Delta \vdash \psi$ and $\Delta \vdash \neg\psi$ then $\Delta \vdash \bot$.
  2. Introduction rule: if $\Delta, \phi \vdash \bot$ then $\Delta \vdash \neg\phi$.
  3. Classical logic: if $\Delta, \neg\phi \vdash \bot$ then $\Delta \vdash \phi$.

We say that $\Delta$ is inconsistent if $\Delta \vdash \bot$.

Let us prove that $\Gamma \vdash \phi$ if, and only if $\Gamma, \neg\phi \vdash \bot$:

  • Assume $\Gamma \vdash \phi$. Then $\Gamma, \neg\phi \vdash \phi$ by weakening, and $\Gamma, \neg\phi \vdash \neg\phi$ by the hypothesis rule, therefore $\Gamma, \neg\phi \vdash \bot$ by the first rule above.

  • Assume $\Gamma, \neg\phi \vdash \bot$. Then $\Gamma \vdash \phi$ by the third rule above.

If the above rules of negation are not valid in your logic, or if the definition of inconsistency is not the one I am using, then please specify the necessary information.

$\endgroup$
3
  • $\begingroup$ Indeed, if I have point 3 you mention this can be done. The system is indeed classic, as it has (as axioms) that $(\neg\phi\rightarrow\neg\psi)\rightarrow(\psi\rightarrow\phi)$ and that $\neg\neg\phi\rightarrow\phi$. Still, I'm not sure how to go from that into your point 3 without the deduction theorem. That's precisely where I'm stuck. $\endgroup$ Commented Mar 7 at 20:19
  • 1
    $\begingroup$ Is it a Hilbert-style system or natural deduction? $\endgroup$ Commented Mar 7 at 20:52
  • $\begingroup$ Hilbert style. I'm assuming the system has at least ten propositional axioms (including the two classical I mentioned above), plus K and possibility (modal). Those are listed in here: github.com/funcao/LML/blob/completeness/coq/… (1 to 10, K, pos). It has modus poenens, but the deduction theorem is not derivable. $\endgroup$ Commented Mar 7 at 20:58

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .