About Explosion and Reductio ad Absurdum
A reductio ad absurdum to eliminate a negation is not valid in Intuitionistic logic. But that is not the only place in logic where arguments by reductio are considered invalid. And historically, reductio arguments have been criticized on many grounds since even before Descartes and the Port Royal Logic regarded them with disfavor.
## 1. The principles of _Explosion_ and _Reductio_
There is a close relation between reductio arguments and the principle:
_Explosion._ (p & ~p) implies q.
I will give a similar name to its (putative) dual:
_Implosion._ q implies (p v ~p),
There are two forms of reductio arguments (where **f** is the falsum, which may or may not be a sentence of form (p & ~p)):
(_Reduc-Int_) if X, p implies **f** , then X implies ~p
(_Reduc-Elim_) if X, ~p implies **f** , then X implies p
For example, Intuitionistic logic has both _Explosion_ and _Reduc-Int_ , but Intuitionists sharply reject _Reduc-Elim_. In their honor I tend to refer to the first as the _good Reductio_ and to the latter as the _bad Reductio._ In classical logic where (p& ~p) is the _falsum_ and (p v ~p) the _verum_ , there is not much to discuss. The questions I want to address are about these principles in contexts where very little is assumed about negation.
## 2. Assuming only a very minimal background logic.
What I would like to do here is to start with just the &-v fragment common to these logics. The syntax will be the sentential syntax with connectives &, v, ~, and a propositional constant **f** , but the logic will have no principles governing ~ or **f**. (Our quest is to find out what principles re ~ and/or **f** will be needed to arrive at, eventually, the bad Reductio.)
For semantics I take it that any admissible interpretation will assign to each sentence as semantic value (“the proposition expressed”) and element of a lattice (“the family of propositions expressed”), with &, v, ~ paired with respectively the meet ∧, the join v, and a special unary operation *. But to begin we have no principles pertaining to *, we only have the lattice laws:
**Definition 1.** x ≤ y iff x ∧ y = x
**Definition 2.** x ≤ y iff x v y = y.
**Lattice laws.** x ≤ x v y, y ≤ x v y, x ∧ y ≤ x, x ∧ y ≤ y
if x ≤ z and y ≤ z then x v y ≤ z,
if z ≤ z and z ≤ y then z ≤ (x ∧ y)
The definitions are equivalent, but I will sometimes use one, sometimes the other. The defined relation ≤ on the family of propositions corresponds to _implication_ for the sentences expressing those propositions. On the side of the sentences, the logical principles assumed at the outset are therefore the corresponding principles, with “⊢” for “implies”:
p ⊢ p v q, q ⊢ p v q, p & q ⊢ p, etc.
This logic I will call _lattice logic._
When examining proofs I will write (L) if the move is valid by the lattice logic, or the lattice laws, alone.
**Terminology.** We note the following principles for reference, to fix terminology.
_Disjunctive Syllogism_ (DisjSyl) : ~p & (p v q) implies q
_Material Modus Ponens_(MatMP): p & (~p v q) implies q
In classical logic these are equivalent, but that is due to the principle of _Double Negation_ , which we won’t assume at any point.
Once we begin to focus on the semantic side I will give the corresponding lattice-theoretic formulation of all the above logical principles, and will (without expecting any confusing) give them the same names.
## 3. Disjunctive Syllogism entails Explosion
This is the famous argument by C. I. Lewis.
**[T1]** p & ~p ⊢ q, by DisjSyl
1. p & ~p assume
2. p 1, (L)
3. p v q 2, (L)
4. ~p 1, (L)
5. q from 3, 4 by DisjSyl
How could the last move be justified in lattice logic? We can continue for 4. with
* 5-1 ~p & (p v q) 4, 3, (L)
* 5-2 (~p & p) v (~p & q) 5-1, distribution
* 5-3 ~p & q 5-2, emptiness of p&~p
* 5-4 q 5-3, (L)
It would be interesting enough if we could justify Reductio with just distribution as extra principle. But that is not so. The move to 5.3 appears to rest on principle that
(p & ~p) v q implies q
or in lattice terms (x ∧ x*) v y ≤ y, and since the converse is obvious, that is equivalent to (x ∧ x*) v y = y. But _that_ is by Definition 2 the same as (x ∧ x*) ≤ y, which is exactly the principle of _Explosion_. So the ‘proof’ would assume _Explosion_ , which begs the question.
## 4. Material Modus Ponens entails Explosion
**[T2]** p & ~p ⊢ q, by MatMP
Adaption of the previous argument:
1. p & ~p assume
2. ~p 1, (L)
3. ~p v q 2, (L)
4. p 1, (L)
5. q from 3, 4 by MatMP
## 5. Given distribution, Explosion, DisjSyl, and MatMP are equivalent
The arguments above establish that _Disjunctive Syllogism_ and _Material Modus Ponens_ , in the context of just the basic lattice laws, each entail _Explosion_. The converses can be proved if we add _Distribution_ to the lattice logic.
Now it will be most convenient to focus on the semantic side. We have the following results for distributive lattices:
**[T3]** _Explosion_ implies _Disjunctive Syllogism_
1. x ∧ x* ≤ y Assumption (which includes that the lattice is bounded below, with special element 0 (bottom), and (x ∧ x*) = 0).
2. (x* ∧ x) v y = y Definition 2, (L).
3. (x* v y) ∧ (x v y) = y 2, distribution
4. [x* ∧ (x v y)] v [(y ∧ (x v y)] = y 3, distribution
5. [x* ∧ (x v y)] v y = y (L)
6. [x* ∧ (x v y)] ≤ y 5., Definition 2.
**[T4]** _Explosion_ entails _Material Modus Ponens_ :
1. x ∧ x* ≤ y Assumption (which includes that the lattice is bounded below, with special element 0 (bottom), and (x ∧ x*) = 0).
2. (x* ∧ x) v y = y Definition 2, (L).
3. (x v y) ∧ (x* v y) = y 2, distribution, and (L)
4. [x ∧ (x* v y)] v [(y ∧ (x* v y)] = y 3, distribution
5. [x ∧ (x* v y)] v y = y (L)
6. [x ∧ (x* v y)] ≤ y 5., Definition 2.
**Corollary.__** In distributive lattices, _Disjunctive Syllogism_ is equivalent to _Material Modus Ponens_ (although _Double Negation_ is not valid).
## 6. When the good Reductio is valid
Since the good Reductio is formulated using the propositional constant **f** we need to restrict the interpretations of the syntax to ones that give **f** a semantic value. So from here on we will restrict our attention to lattices that are bounded below, with least element 0 (bottom), that is to say, for all elements x, 0 ≤ x. And the proposition expressed by **f** is 0.
**Definition 3.** Lattice elements x, y are _disjoint_ iff x ∧ y ≤ 0 (equivalently, iff x ∧ y = 0).
The good Reductio is
(_Reduc-Int_) if X, p implies **f** , then X implies ~p
For the lattice of propositions, that means that for all elements x, y:
(_Reduc-Int_) If x ∧ y ≤ 0 then x ≤ y*.
Though we have no special principles yet pertaining to operation *, we know that it is a map of the lattice into itself.
Hence it is clear what _Reduc-Int_ says:
y* is an element which is at least as weak as any element disjoint from y.
It follows from this that the lattice must also be bounded above, with largest (weakest) element 1 (top). For all elements are disjoint from 0, and if the lattice does not have a top then no element is at least as weak as any element disjoint from 0. 1 is disjoint from 0, and is at least as weak as any element whatsoever, hence 1 = 0*. Moreover, only 0 is disjoint from 1, so 1* = 0.
A traditional definition of negation is that ~p is the logically weakest statement such that (p & ~p) ⊢**f**. For lattices in general there need not exist a largest (weakest) element in the family of elements disjoint from a given element (other than 0 or 1).
So the good Reductio does not suffice to identify the operation *. For that we need to add _Explosion_.
**[T5]** If _Explosion_ and _Reduc-Int_ both hold in lattice L then x* is the greatest (weakest) element disjoint from x.
This connects very well with a major topic in lattice theory: the _pseudo-complement_. Its definition amounts to the combination of _Explosion_ and the good _Reductio_ :
**Definition 4.** Unary operation * on a lattice is a _pseudo-complement_ iff, for all elements x, y, (y ∧ y*) = 0 and if x ∧ y ≤ 0 then x ≤ y*.
**Corollary**. If _Explosion_ and _Reduc-Int_ both hold in lattice L then x* is a pseudo-complement.
## 7. When the bad Reductio is valid
The pseudo-complement has many of the familiar features of negation in general, but it specifically lacks Excluded Middle.
Now, look at
(_Reduc-Elim_) if X, ~p implies **f** , then X implies p
also called _Indirect Proof_ , which is, for the lattice of propositions,
(_Reduc-Elim_) If x ∧ y* ≤ 0 then x ≤ y,
equivalently, if x ∧ y* = 0 then x ≤ y.
We note that _Implosion_ entails that the lattice has a top element, call it 1, and that for any element x, 1 = x v x*.
_Implosion_ entails the bad Reductio, given distributivity. Precisely:
**[T6]** If the lattice is distributive, bounded above with top 1, and (x v x*) = 1 for all elements x, then _Reduc-Elim_ holds in that lattice.
To prove: If x ∧ y* = 0 then x ≤ y.
1. x ∧ y* = 0 Assume
2. x = x ∧ 1 (L)
3. x = x ∧ (y v y*) 2, Given (_Implosion_)
4. x = (x ∧ y) v (x ∧y*) (L), Distribution
5. x = (x ∧ y) v 0 1, 4, (L)
6. x = (x ∧ y) 5, (L)
7. x ≤ y. 6, Definition 1.
**Notice** that this argument does not depend on any special properties of *.
_Implosion_ has much to answer for.
### Share this:
* Share on X (Opens in new window) X
* Share on Facebook (Opens in new window) Facebook
*
Like Loading...
### _Related_