4

I'm stuck on question 2 of these lecture notes on propositional logic:

"2. Propositional Proof. Give a formal proof of the sentence p from the single premise ¬¬p using only Modus Ponens and the standard axiom schemata. Warning: This is surprisingly difficult. Though it takes no more than about ten steps, the proof is non-obvious. This problem illustrates the difficulties of working with proof methods devised more for minimality than ease of use."

As the question states, it is surprisingly difficult, for a newbie like me at least.

Could anyone give me a hint on how to begin? I've tried a few approaches but nothing.

For example:

premise: ¬¬p
1) ¬¬p -> p            to prove    
2) ¬¬p -> (p -> ¬¬p)   II, 1    
3) p -> ¬¬p            premis, 2

Clearly not the right start, and (1) looks obviously wrong :(

Thanks in advance.

mbomb007
  • 139
Student
  • 181
  • Never mind...I see you've attached a link to lecture notes listing the axioms you have to work with. Thanks for doing so! – amWhy Jun 09 '13 at 01:54
  • In this (Mendelson) collection of axiom schemata, the only rule involving negation is CR. Can you prove that $\lnot p$ implies both some $q$ and its negation $\lnot q$? – András Salamon Jun 09 '13 at 11:48
  • The lecture notes are from a dead link. Can someone provide what the axioms being used are? – mbomb007 Apr 03 '18 at 21:20

2 Answers2

3

I'll give only a few hints, as you asked. First, observe that only the contradiction realization schema uses the negation, so we have to use it somehow in our proof. If we used only the other two, we'd have no information about how the negation behaves.

Now, let's have a look at it: $$(\lnot\phi \Rightarrow \psi) \Rightarrow ((\lnot\phi \Rightarrow \lnot\psi) \Rightarrow \phi) $$ we'd like to somehow introduce the double negation. Let's guess and substitute $\lnot\phi$ for $\psi$ to get $$(\lnot\phi \Rightarrow \lnot\phi) \Rightarrow ((\lnot\phi \Rightarrow \lnot\lnot\phi) \Rightarrow \phi) $$ This looks promising. We are able to prove $\lnot\phi \Rightarrow \lnot\phi$ (which is also somewhat nontrivial), after which we can apply MP to deduce $$(\lnot\phi \Rightarrow \lnot\lnot\phi) \Rightarrow \phi$$ Now if we have the assumption $\lnot\lnot\phi$, we can prove $(\lnot\phi \Rightarrow \lnot\lnot\phi)$ using the implication introduction schema. Putting it together, we prove $\phi$ from $\lnot\lnot\phi$.


Update: How to prove $\phi\Rightarrow\phi$: \begin{align} [1] & (\phi\Rightarrow (\underbrace{(\phi\Rightarrow\phi)} \Rightarrow\phi)) ⇒ ((\phi\Rightarrow\underbrace{(\phi\Rightarrow\phi)}) \Rightarrow (\phi\Rightarrow\phi)) & \mbox{substitute $\phi$, $\phi\Rightarrow\phi$ and $\phi$ into ID} \\ [2] & \phi\Rightarrow (\underbrace{(\phi\Rightarrow\phi)} \Rightarrow\phi) & \mbox{substitute $\phi$, $\phi\Rightarrow\phi$ into II} \\ [3] & (\phi\Rightarrow\underbrace{(\phi\Rightarrow\phi)}) \Rightarrow (\phi\Rightarrow\phi) & \mbox{Modus Ponens [1] and [2]} \\ [4] & \phi\Rightarrow (\phi\Rightarrow\phi) & \mbox{substitute $\phi$, $\phi$ into II} \\ [5] & \phi\Rightarrow\phi & \mbox{Modus Ponens [3] and [4]} \\ \end{align}

Petr
  • 2,307
  • 13
  • 21
0

My proof, version 2:

$$ premise: \lnot\lnot p $$ \begin{array}{ll} 1) & \lnot\lnot p & : p_1 \\ 2) & (\lnot \phi \Rightarrow \psi) \Rightarrow ((\lnot\phi \Rightarrow \lnot\psi) \Rightarrow \phi) & : CR \\ 3) & (\lnot p \Rightarrow \lnot p) \Rightarrow ((\lnot p \Rightarrow \lnot\lnot p) \Rightarrow p) & : 2, \phi=p, \psi=\lnot p \\ 4) & (\lnot p \Rightarrow \lnot p) = (\lnot p \lor p) = \top & : valid \\ 5) & (\lnot p \Rightarrow \lnot\lnot p) \Rightarrow p) & : 3,4 \\ 6) & (\phi \Rightarrow (\psi \Rightarrow \phi) & : II \\ 7) & (\lnot\lnot p \Rightarrow (\lnot p \Rightarrow \lnot\lnot p) & : 6, \phi=\lnot\lnot p, \psi=\lnot p \\ 8) & (\lnot p \Rightarrow \lnot\lnot p) & : p_1,7 \\ 9) & p & : 8,5 \\ \therefore \lnot\lnot p \Rightarrow p \end{array}

Is this correct, in terms of steps and explanations on the right?

Thank you.

My proof, version 1 (Incorrect)

p1 (premise): ¬¬p
1) ¬¬pp                                      : p1
2) (¬p => ¬¬p) => ((¬p => ¬(¬¬p)) => p)      : 1, CR
3) (¬p => ¬¬p) => ((¬p => ¬p) => p)          : 2, double negation
4) (¬p => ¬¬p) => p                          : 3, (¬p => ¬p) is valid
5) ¬¬p => (¬p => ¬¬p)                        : II, 4
6) (¬p => ¬¬p)                               : p1, MP, 5
7) p                                         : 6, MP, 6
∴ ¬¬p => p
Student
  • 181
  • You've made several mistakes. The most important is in 2-3 where you eliminate the double negation. This is what you're trying to prove! So you've created a circular proof. Less important: It's true that $a \Rightarrow a$ is valid, but you have to prove it. Also you use the rule $b,, a \Rightarrow (b\Rightarrow c),\vdash, a\Rightarrow c$, which you need to prove first. Also, to derive 5 you don't need 4 (it's an axiom). Derivation of 7 from 6 doesn't make any sense to me. – Petr Jun 10 '13 at 06:07
  • Thank you again, Petr. I added version 2 of my proof. Any comments would be greatly appreciated. Thanks – Student Jun 15 '13 at 03:45
  • 1
    That's much better. It's correct now, except for step 4). In this logic system, you have no notion of $\lor$, and if you had, you have to prove its relation to $\Rightarrow$ as well as that $p\lor\lnot p$ is a tautology. Hint: You can prove $\phi\Rightarrow\phi$ by taking a suitable instance of ID and then using using MP twice, each time using a (different) suitable instance of II. – Petr Jun 15 '13 at 12:23
  • Petr, I've tried several ways to prove that (~p => ~p) is valid. I'm regrettably asking you for the answer, if you don't mind. Thank you. – Student Jun 29 '13 at 03:24
  • No problem, I added it to my answer. – Petr Jun 29 '13 at 06:21
  • Thank you, Petr. At some point I started down a similar path, but it seemed wrong, and it looked like it could prove anything to be valid. I see now how the way you replaced formulas, makes all the difference. Thanks again for your help. – Student Jun 30 '13 at 05:23