Converse of Axiom K

516 Views Asked by At

Just for clarity, this is the specific formulation I am referring to.

$K$: $\Box(p\to q)\to(\Box p\to\Box q)$

Naturally, we might consider the converse of this statement:

$K'$: $(\Box p\to\Box q)\to\Box(p\to q)$

The question is simple. Given $K$, does $K'$ follow? An accompanying proof of the answer would be greatly appreciated.

1

There are 1 best solutions below

2
On BEST ANSWER

$K$ does not imply $K'$; indeed, I can't think of a reasonable system which does include $K'$, which doesn't trivialize the modality.

Think of a Kripke frame with three worlds $a, b, c$ where

  • $a$ sees all three worlds,

  • neither $b$ nor$c$ see any worlds, and

  • $b\models p\wedge\neg q$, $c\models q\wedge\neg p$.

$K$ is true in this frame, but $K'$ is not: "$\Box p\rightarrow\Box q$" is true at $a$ (since $\Box p$ is false at $a$), but "$\Box(p\rightarrow q)$" is false at $a$ (since $a$ sees $b$ and $b\models\neg(p\rightarrow q)$).


EDIT: It's not hard to pin down exactly what Kripke frames validate $K'$: they are the frames in which each world sees at most one world (possibly itself).

Showing that such frames verify $K'$ is easy. If a world $a$ sees no worlds, then we trivially have $a\models\Box (p\implies q)$; conversely, if $aRb$ and $a$ sees no worlds besides $b$, then "$a\models \Box t$" is the same as "$b\models t$," so $K'$ holds at $a$ since $(p\implies q)\implies (p\implies q)$ holds at $b$.

Conversely, set $p=\neg q$ in $K'$, and note that $K'$ tells us in particular that if $\neg\Box p$ then $\Box (p\implies q)$ - which in our case translates to $(\Diamond q)\implies(\Box q)$ (since "$\neg\Box \neg$"="$\Diamond$" and "$\neg q\implies q$"="$q$"). So every frame validating $K'$ also validates the "trivialization" principle $$(*)\quad(\Diamond q)\implies (\Box q).$$ But this statement clearly implies that in such a frame, no world sees more than one world (if $aRb, aRc,$ and $b\not=c$, set $q$ true at $b$ so that $a\models\Diamond q$ and set $q$ false at $c$ so that $a\models\neg\Box q$).

Note that we've in fact shown that in Kripke frames, validating $K'$ is the same as validating $(*)$. I don't know off the top of my head what the weakest modal system $X$ is such that $X+(*)$ is equivalent to $X+K'$, but I suspect it's quite weak.