You are on page 1of 4

Some thoughts about logical implication if A then B

Lucian VINTAN

The main aim of this short low-level didactical note is to understand just for myself - why
logical material implication A B (if A then B) is defined according to Table 1. One question
is: why it is defined in this manner and not in another one? This is the most debated Boolean
function from the 24 possible binary functions f (A, B).
Its corresponding truth table is the following (V(X) means the logical value of sentence X):
V(A)
0
0
1
1

V(B)
0
1
0
1
Table 1

V(A B)
1 (*)
1 (*)
0
1

For sure it would be a non-sense trying to prove a definition, because it must be just
consistent; this means that the defined concepts context must be not an empty set. According
to Table 1, material logical implication definition is consistent, and, therefore, it needs no
comment! From this point of view, my approach might be considered more psychological than logical. However, the effort of understanding this non-intuitive definition
would be interesting. Empty sets, actual infinite sets (defined by Cantor, in contrast with the
well-known potential infinite concept), non-Euclidian geometries, etc. are also non-intuitive
mathematical concepts, having very non-intuitive consequences (example: Card N = Card Q
= aleph0). The main aim of this pure didactic document is to try to offer an aposteriori
justification of this definition according to my (our?) common sense. (A vague, nonmathematical concept, I know)
Observation: we have to distinguish between the logical connector , that transforms two
sentences A and B in a new one (A B), that is defined in Table 1, and the implication A B
meaning that if A is true then B is true. Any mathematical theorem has (or could have) the
form A(hypothesis) B(conclusion).
The link between
Theorem: A
holds.)

and

implications is given by the following well-known:

B if and only if A B is true. (Thus A

B if and only if implication

always

Proof: if A B we cannot be in Line 3, Table 1 and thus V(A B)=1. Vice versa: we suppose
V(A B)=1. That means we can be in Lines 1, 2, 4 from Table 1. If V(A) =1, we are for sure
in Line 4, Table 1 and therefore V(B)=1. Thus A B, q.e.d. Disambiguation of these two
implications, especially in natural languages, is necessary. Below, Ill try to do it.
Consequence (1): According to the Table 1 implications definition, we can easily write:
A B = (~A)VB, representing the disjunctive normal form (symbol V = logical OR, symbol ~
= logical NOT). (Using one of the De Morgan formula this definition involves ~(A B) =
A&(~B), representing the conjunctive normal form, where symbol &=logical AND).

Another Consequence (2). Based on the previous Consequence it is obvious that: A B is


equivalent to (~B) (~A).
Returning to Table 1, I consider entry 4 as being quite natural or obvious according to
my apriori intuition. In other words, if V(A)=1 and V(A B)=1 then V(B)=1.
Comment: This last rule represents a mathematical model of the modus ponens rule. It can be
written: if ((A=1)&((A B)=1)) then B=1, that is equivalent to (A&(A B)) B,
representing the well-known modus ponens truth-functional tautology. It represents the basic
syllogism of our logical thinking. The ancient Greeks have used it in their traditional logic
and mathematics, obviously, before the mathematical logic was born.
Consequence (3): According to Consequence (2) and to modus ponens truth-functional
tautology we can write (~B&(A B)) ~A, representing the well-known modus tollens
tautology.
Also, inspecting Table 1 it seems to be quite natural - according to my apriori intuition that when A is true and B is false then A B must be false (entry 3). But how would be
friendly explained entries one and, especially entry two? (Friendly means intuitively or
not shocking our - or at least my logical common sense)
Some logicians are not agreeing this implication definition, due to the rows 1 and, especially,
2 belonging to Table 1. These two rows are saying: If V(A)=0 then V(A B)=1. I
understand that it is a consistent definition but whats its rationale in terms of (my)
common sense? Therefore how truth implies truth but false implies anything (truth or
false)? This is known as "ex falso quodlibet" rule, too. Initially, my apriority thinking might
consider that false must imply (always) false. Here I used the English word implies in the
sense offered by Table 1 for the connector . Perhaps this initial feeling might be partially
explainable due to my initial confusion between the concepts of logical implication and
causality. Propositional Calculus has no concept of causality, but it does have a concept of
implication[,] [according to the definition given in Table 1; my addition!]. see Wikipedia at
http://c2.com/cgi/wiki?LogicalImplication In other words, A B cannot be substituted with
A is a cause for B.
For example if we are considering the sentences A and B like A: 1+1=2 (true) and B: Sibiu is
in Romania (true), then A B is true even if between A and B sentences are not a causality
relation. In formal logic V(A B) is based only on V(A) and V(B); thus it assumes no
causality relation (direct or indirect) between sentences A and B.
On the other hand, P: 0 X is true (and it really is, according to the previous definition),
X {0,1}, it might mean that P is possible (valid) rather than necessary true
Discussion:
I imagine the following examples.
A: 2 N. Obviously V(A)=0 (it is false from a mathematical point of view). Anyway,
sentence A mathematically involves ( ) sentence B: 2 * 2 =2 N. B is obviously true.
Thus false might imply truth. But here the word imply means
and not !

Just an observation: however, in this example I applied A sentence - a false mathematical


assumption - two times (if 2 N then 2 * 2 =2 N)! The false assumptions are somehow
compensating (an even number of false assumptions), finally given a true sentence. I believe
this fact might be significant (double mistakes might compensate themselves).
On the other hand, sentence A mathematically implies ( ) the sentence C: 2 *2 N. C is
obviously false. Therefore false might imply false, too. But again, the word imply means
here
and not ! Here the false mathematical assumption ( 2 Q) was used only one time
(odd number) during the inference process.
Another similar example: (-1=1)
(1=1), because if (-1=1) then (-1)2=12 thus 1=1, that is an
evidence (again it seems that I used two times the wrong assumption). On the other hand (1=1)
(0=2). Here I used only one time the bad assumption.
In fact false implies false means that the sentence (-1=1)
true! But this does not mean that 0=2 is true!

(0=2) is true. And it is really

In my opinion, another possible (psycho-logical) confusion regarding to Table 1 definition


might be produced by the following question: why it is missing in Table 1 the case of A=0
and (A B)=0? The answer is quite trivial: because the output A B cannot be 0 if its input
A=0, according to the definition. Here, the confusion consists in considering in Table 1 A and
A B as inputs, and B as the output.
Anyway, if my initial puritanism will persist in the (false) assumption that starting from a
false logical sentence (A=0) and through correct derivation - (A B)=1 it must be
obtained a false conclusion (B=0), Table 1 will become:
V(A)
0
0
1
1

V(B)
V(A B)
0
1 (*)
1 (*)
0
0
0
1
1
Table 2
Actually, in Table 2 line 1 becomes identic with line 2. This might be not acceptable because
A=0 and B=1 is an interdicted input combination and the new function (Table 2) is
incompletely defined. In order to solve this situation Table 2 might be formally replaced
with the following one (taking into account that row 2 is our single non-intuitive inference,
it represents the unique possibility):
V(A)
0
0
1
1

V(B)
V(A B)
0
1 (*)
1
0 (*)
0
0
1
1
Table 3
But in this case (Table 3), line 2 seems to be intuitively not acceptable: starting from a false
logical sentence (A=0) and through incorrect or invalid derivation - (A B)=0 it must be
obtained a true conclusion B (?!) On the other hand, the Boolean function defined in Table

3 represents the well-known and well-accepted bi-conditional function (bi-implication, if


and only if) from the classical logic ( ). (A B) = (A B)&(B A).
As a consequence the only acceptable (possible, reasonable) definition of A B
seems to be that given by Table 1.
Note 1: Some interesting paradoxes (formulas which are truths of classical logic, but which
are
intuitively
problematic)
of
material
implication
are
presented
at
http://en.wikipedia.org/wiki/Paradoxes_of_material_implication
Note 2: In fuzzy logic implication has many definitions. Some of them are the followings:

A B ( x, y ) = max[min( A ( x), B ( y ) ), (1 A ( x) )] - Zadeh implication ( is the membership


function)

A B ( x, y ) = min(1,1 A ( x) + B ( y ) ) - Lukasiewicz implication


A B ( x, y ) = min( A ( x), B ( y ) ) - Mamdani implication
A B ( x, y ) = max((1 A ( x)), B ( y ) ) - this is the fuzzy implication derived from the classical
logic implication definition, considering the conjunction, disjunction and complement
definitions provided in L. A. Zadeh, Fuzzy sets, Information and control, vol. 8, no. 3, pp.
338353, 1965.

You might also like