3

I am trying to prove by formula that the following formula is a tautology but I am not able to do so, although I have proved it as a tautology with truth table

 (A∨B) ∧ (¬B∨C) → A∨C

¬((A∨B) ∧ (¬B∨C)) ∨  A∨C
 (¬A∧¬B) ∨ (B∧¬C)  ∨  A∨C
¬A∨(B∧¬C) ∧ ¬B∨(B∧¬C) ∨  A∨C              (Distributive Law)
(¬A∨B) ∧ (¬A∨¬C) ∧ (¬B∨B) ∧ (¬B∨¬C ∨ A∨C )  
(¬A∨B) ∧ (¬A∨¬C) ∧  T   ∧  (¬B∨A ∨  T) 
∴(¬A∨B) ∧ (¬A∨¬C) 

By using Truth table, I am able to prove it as a tautology. Am I doing something wrong in the second method?

Far
  • 143

2 Answers2

2

Your mistake is here:

¬A∨(B∧¬C) ∧ ¬B∨(B∧¬C) ∨  A∨C              (Distributive Law)
(¬A∨B) ∧ (¬A∨¬C) ∧ (¬B∨B) ∧ (¬B∨¬C ∨ A∨C )  

The result should be:

¬A∨(B∧¬C) ∧ ¬B∨(B∧¬C) ∨  A∨C              (Distributive Law)
((¬A∨B) ∧ (¬A∨¬C) ∧ (¬B∨B) ∧ (¬B∨¬C)) ∨ A∨C   

So be careful with those parentheses! In fact, I believe your mistake was partly based on the fact that you didn't use parentheses in the statement before. That is, instead of:

¬A∨(B∧¬C) ∧ ¬B∨(B∧¬C) ∨  A∨C              (Distributive Law)

I would use parentheses to make the structure of the statement really clear:

((¬A∨(B∧¬C)) ∧ (¬B∨(B∧¬C))) ∨  A∨C              (Distributive Law)

Doing that makes it less likely to make mistakes.

By the way, to continue to put this in to CNF (you asked about this in the comments):

((¬A∨B) ∧ (¬A∨¬C) ∧ (¬B∨B) ∧ (¬B∨¬C)) ∨ A∨C (Complement)

((¬A∨B) ∧ (¬A∨¬C) ∧ $\top$ ∧ (¬B∨¬C)) ∨ A∨C (Identity)

((¬A∨B) ∧ (¬A∨¬C) ∧ (¬B∨¬C)) ∨ A∨C (Distribution)

((¬A∨B) ∨ (A∨C)) ∧ ((¬A∨¬C) ∨ (A∨C)) ∧ ((¬B∨¬C) ∨ (A∨C)) (Association)

(¬A∨B ∨ A∨C) ∧ (¬A∨¬C ∨ A∨C) ∧ (¬B∨¬C ∨ A∨C) (Complement x 3)

($\top$ ∨B ∨C) ∧ ($\top$ ∨¬C ∨C) ∧ ($\top$ ∨¬B∨ A) (Identity x 3)

$\top$ ∧ $\top$ ∧ $\top$

$\top$

(you're not going to get any 'interesting' clauses exactly because this is a tautology!)

Also, here is an easier way to show that your statement is a tautology. It relies on a little known equivalence called Concensus:

$(P \lor Q) \land (\neg Q \lor R) \Leftrightarrow (P \lor Q) \land (\neg Q \lor R) \land (P \land R)$

(It's basically the Resolution rule, but as an equivalence)

Applied to your statement:

$((A \lor B) \land (\neg B \lor C)) \rightarrow (A \lor C) \Leftrightarrow$ (Concensus)

$((A \lor B) \land (\neg B \lor C) \land (A \lor C)) \rightarrow (A \lor C) \Leftrightarrow $ (Implication)

$\neg ((A \lor B) \land (\neg B \lor C) \land (A \lor C)) \lor (A \lor C) \Leftrightarrow$ (DeMorgan)

$\neg (A \lor B) \lor \neg (\neg B \lor C) \lor \neg (A \lor C) \lor (A \lor C) \Leftrightarrow$ (Complement)

$\neg (A \lor B) \lor \neg (\neg B \lor C) \lor \top \Leftrightarrow$ (Identity)

$\top$

Bram28
  • 100,612
  • 6
  • 70
  • 118
  • 1
    @Far You're welcome! And as far as your question about resolution goes: as I already said in my answer, the Consensus rule is very much related to Resolution. But if you want to put this explicitly into CNF ... let me update my Answer ... – Bram28 Apr 02 '17 at 14:52
1

It looks like you may have forgotten some parentheses here. As a rule of thumb parentheses (even when slightly overused) can make a huge difference in Logic. For example in your approach:

$$¬[(A∨B) ∧ (¬B∨C)] ∨ (A∨C)$$ $$ [(¬A∧¬B) ∨ (B∧¬C)] ∨ (A∨C)$$ $$[¬A∨(B∧¬C) ∧ ¬B∨(B∧¬C)] ∨ (A∨C)$$

$$[(¬A∨B) ∧ (¬A∨¬C) ∧ (¬B∨B) ∧ (¬B∨¬C)] ∨ (A∨C)$$

And this is where the mistake happened.

However, there is an easier approach from step 2 above. Incidentally you can drop/re-arrange, a lot of parentheses because the clauses share the same logical connective. Namely:

$$¬[(A∨B) ∧ (¬B∨C)] ∨ (A∨C)$$ $$ [(¬A∧¬B) ∨ (B∧¬C)] ∨ (A∨C)$$ $$ (¬A∧¬B) ∨ (B∧¬C) ∨ (A) ∨ (C)$$ $$ (¬A∧¬B) ∨ (A) ∨ (B∧¬C) ∨ (C)$$ $$ [(¬A∧¬B) ∨ (A)] ∨ [(B∧¬C) ∨ (C)]$$ $$ [(¬A∨A)∧(¬B∨A)] ∨ [(B∨C) ∧(¬C ∨ C)]$$ $$ [T∧(¬B∨A)] ∨ [(B∨C) ∧ T]$$ $$ (¬B∨A) ∨ (B∨C)$$ $$ (¬B)∨ (A) ∨ (B) ∨(C) $$ $$ (¬B)∨ (B) ∨ (A) ∨(C) $$ $$ (¬B∨B) ∨ (A) ∨(C) $$ $$ T ∨ (A) ∨(C) $$ $$T$$

  • Okay thanks a lot. I have another question, can this be solved also using resolution rule? Since resolution rule requires the formula to be in Conjunctive Normal Form i.e (A∨ ....) (C∨..) (B∧..). means between the formulas, this formula also has to be converted to CNF. I tried but it seems not possible. Can you help in this, if I am able to express my problem correctly. – Far Apr 02 '17 at 14:24
  • Yes but It looks like Bram28 beat me to it. I am glad you found it helpful though, that's all that matters. – Pellenthor Apr 02 '17 at 17:11