02.11.2012 Views

Reduction and Elimination in Philosophy and the Sciences

Reduction and Elimination in Philosophy and the Sciences

Reduction and Elimination in Philosophy and the Sciences

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Clearly x = y → x = ext y holds, but <strong>the</strong> converse does not<br />

hold <strong>in</strong> H.<br />

Theorem 1 The axiom of extensionality,<br />

( ∀ , y )[ x = y ≡ x = y]<br />

, does not hold <strong>in</strong> H.<br />

x ext<br />

For <strong>the</strong> proof, see [Hajek 2005]. Here, we <strong>in</strong>troduce <strong>the</strong><br />

outl<strong>in</strong>e: this is by <strong>the</strong> follow<strong>in</strong>g lemma.<br />

Lemma 1 H proves that Leibniz equality is a crisp relation.<br />

We note that formula P(x) is crisp (Crisp(P)) if, for any<br />

object a, <strong>the</strong> truth value of P(a) is ei<strong>the</strong>r 0 or 1. S<strong>in</strong>ce =ext<br />

can have a fuzzy truth value for fuzzy sets <strong>in</strong> H, = <strong>and</strong> =ext<br />

are different.<br />

As for <strong>the</strong> basic law V, it implies that = is not <strong>the</strong> Leibniz<br />

equality. For V def<strong>in</strong>es = as <strong>the</strong> extensional equality, <strong>and</strong> if<br />

= is Leibniz equality <strong>the</strong>n <strong>the</strong> axiom of extensionality holds,<br />

<strong>and</strong> this implies a contradiction <strong>in</strong> H. However, Frege’s<br />

<strong>in</strong>tension seems to def<strong>in</strong>e = as Leibniz equality. In this<br />

sense <strong>the</strong> basic law V does not hold <strong>in</strong> H.<br />

In this proof, we imply a contradiction when x =ext y<br />

has a fuzzy truth value: this is possible when not less than<br />

one of x <strong>and</strong> y is a fuzzy set. As for ano<strong>the</strong>r paradox, it is<br />

known that we can imply a contradiction if we assume <strong>the</strong><br />

empty set φ = { x : ⊥}<br />

satisfies <strong>the</strong> extensionality axiom<br />

[Cant<strong>in</strong>i 2003]. So it has been unknown that <strong>the</strong> follow<strong>in</strong>g<br />

scheme is consistent:<br />

(({ x : P(<br />

x)}<br />

≠ext φ ) ∧Crisp(<br />

P))<br />

∧ (({ x : Q(<br />

x)}<br />

≠ext<br />

φ)<br />

∧Crisp(<br />

Q))<br />

→ [{ x : P(<br />

x)}<br />

= { y : Q(<br />

y)}<br />

≡ ( ∀x)[<br />

P(<br />

x)<br />

≡ Q(<br />

x)]]<br />

for any formula P(x), Q(x). This means that, <strong>the</strong> follow<strong>in</strong>g<br />

law might hold (this is a version of <strong>the</strong> RV [Shapiro 2003]):<br />

∀ P∀Q( Good(<br />

P)<br />

∧Good(<br />

Q))<br />

→ [ ext(<br />

P)<br />

= ext(<br />

Q)<br />

≡ ( ∀x)[<br />

P(<br />

x)<br />

≡ Q(<br />

x)]]<br />

Therefore if this is consistent, we can th<strong>in</strong>k that RV gives<br />

an implicit def<strong>in</strong>ition of crisp sets, badness means fuzz<strong>in</strong>ess,<br />

<strong>and</strong> any fuzzy set can be regarded to represent <strong>in</strong>def<strong>in</strong>itely<br />

extensibility.<br />

3 Circularity <strong>and</strong> arithmetic without <strong>the</strong> <strong>in</strong>duction<br />

scheme<br />

When we mention <strong>the</strong> formalization of arithmetic, we often<br />

come to axiom systems with <strong>the</strong> <strong>in</strong>duction scheme as PA,<br />

but it is not a unique way. We also have type systems<br />

which are widely used <strong>in</strong> computer science. For example,<br />

Gödel’s system T 2 is a simple type <strong>the</strong>ory [Girard et al.<br />

1989]. T has two types, Int (<strong>in</strong>tegers) <strong>and</strong> Bool (booleans).<br />

As for Int, it has two type constructors, 0 (constant symbol)<br />

<strong>and</strong> s : Int → Int (successor function). And T does not<br />

have <strong>the</strong> <strong>in</strong>duction scheme. Instead, it has a recursion<br />

operator R for recursive def<strong>in</strong>ition whose type is<br />

R : U → ( U → Int → U)<br />

→ U → U for any type U. It satisfies<br />

R uv0<br />

= u <strong>and</strong> R uv(<br />

sn)<br />

= v(<br />

Ruvn)<br />

n (if we substitute U<br />

for Int). This operator enables us to have <strong>the</strong> primitive<br />

recursion on <strong>in</strong>teger numbers: for example, <strong>the</strong> addition<br />

Int Int<br />

x + y is def<strong>in</strong>ed by Rx(<br />

λ z . λz'<br />

. sz)<br />

y . For, we can calculate<br />

as follows:<br />

Int Int<br />

x + 0 a Rx(<br />

λ z . λz<br />

. sz)<br />

0 a x<br />

Int Int<br />

x + ( sy<br />

) a Rx(<br />

λz<br />

. λz<br />

. sz)(<br />

x + y ) y a s(<br />

x + y )<br />

Here t a s represents that s is a result of <strong>the</strong> computation<br />

whose <strong>in</strong>put is t. So, <strong>the</strong> above computations show that<br />

x + 0 = x <strong>and</strong> x + ( sy<br />

) = s(<br />

x + y ) hold. In this way, we can<br />

2 We note that Gödel’s orig<strong>in</strong>al system T has <strong>the</strong> <strong>in</strong>duction scheme.<br />

The Comprehension Pr<strong>in</strong>ciple <strong>and</strong> Arithmetic <strong>in</strong> Fuzzy Logic — Shunsuke Yatabe<br />

represent primitive recursive functionals <strong>in</strong> T without us<strong>in</strong>g<br />

<strong>the</strong> <strong>in</strong>duction scheme.<br />

The arithmetic developed <strong>in</strong> H is very similar to <strong>the</strong><br />

system T <strong>in</strong> that <strong>the</strong>y do not have <strong>the</strong> <strong>in</strong>duction scheme. H<br />

allows <strong>the</strong> circular def<strong>in</strong>ition 3 (as <strong>in</strong> section 1), <strong>and</strong> it<br />

enables us to use <strong>the</strong> general form of <strong>the</strong> recursive<br />

def<strong>in</strong>ition as R. Here we <strong>in</strong>troduce three po<strong>in</strong>ts (for more<br />

details, see [Cant<strong>in</strong>i 2003][Hajek 2005]). First we can<br />

def<strong>in</strong>e ord<strong>in</strong>al numbers <strong>in</strong> Zermelo style: 0 = φ <strong>and</strong><br />

s n = { x : x = n}<br />

for any f<strong>in</strong>ite ord<strong>in</strong>al n. It is easy to see, we<br />

can def<strong>in</strong>e Fregean term “<strong>the</strong> number of <strong>the</strong> conception P”<br />

(NxPx) by <strong>the</strong> comprehension pr<strong>in</strong>ciple if P is crisp <strong>and</strong><br />

f<strong>in</strong>ite. Second, <strong>the</strong> set ω of all f<strong>in</strong>ite ord<strong>in</strong>als can be def<strong>in</strong>ed<br />

as<br />

ω = ext { x : x = 0 ∨ ( ∃y<br />

∈ω<br />

)[ x = sy<br />

]}<br />

(Because of <strong>the</strong> luck of extensionality, we can not<br />

require <strong>the</strong> uniqueness of ω). Third, any recursive<br />

function’s graph can be def<strong>in</strong>ed. In o<strong>the</strong>r words, any partial<br />

recursive function is numerically representable <strong>in</strong> H<br />

Let us give an example of <strong>the</strong> generalized recursive<br />

def<strong>in</strong>ition <strong>in</strong> H. For example, we can def<strong>in</strong>e <strong>the</strong> graph P of<br />

<strong>the</strong> addition as follows 4 :<br />

x,<br />

0,<br />

z ∈ P ⇔ x = z,<br />

x,<br />

sy,<br />

sz<br />

∈ P ⇔ x,<br />

y,<br />

z ∈P<br />

Both x + 0 = x <strong>and</strong> x + ( sy<br />

) = s(<br />

x + y)<br />

are guaranteed by<br />

very simple way. However, we do not know P satisfies <strong>the</strong><br />

follow<strong>in</strong>g conditions:<br />

1. P is a crisp relation,<br />

2. P def<strong>in</strong>es a function p(x, y) = z<br />

(i.e. ( ∀ x , y )[ P(<br />

x,<br />

y,<br />

z)<br />

∧ P(<br />

x,<br />

y,<br />

z'<br />

) → z = z']<br />

),<br />

3. p(x, y) is a total function.<br />

If x <strong>and</strong> y are st<strong>and</strong>ard natural numbers, <strong>the</strong>n we can calculate<br />

<strong>the</strong> unique value z satisfy<strong>in</strong>g P(x, y, z). However we<br />

have a trouble when one of x, y is a non-st<strong>and</strong>ard natural<br />

number: P(x, y, z) might be a fuzzy truth value. We nei<strong>the</strong>r<br />

know, whe<strong>the</strong>r ω or any graph def<strong>in</strong>ed by recursion becomes<br />

crisp or not. If ω is fuzzy, <strong>the</strong>n we might th<strong>in</strong>k this is<br />

ano<strong>the</strong>r expression of Dummett’s “ω is <strong>in</strong>def<strong>in</strong>itely extensible”.<br />

H develops a fair degree of arithmetic; however it<br />

can not deduce Peano arithmetic PA. In fact, <strong>the</strong> follow<strong>in</strong>g<br />

<strong>the</strong>orem holds.<br />

Theorem 2 H proves that <strong>the</strong> <strong>in</strong>duction scheme implies<br />

<strong>in</strong>consistency.<br />

This means that H is ω-<strong>in</strong>consistent: <strong>in</strong> any model of H, <strong>the</strong><br />

sentence which can be <strong>in</strong>terpreted as “ω conta<strong>in</strong>s a nonst<strong>and</strong>ard<br />

natural number” is truth-value 1. For more details,<br />

see [Yatabe 2007].<br />

Contrary, Girard showed that <strong>the</strong> weak version of<br />

ma<strong>the</strong>matical <strong>in</strong>duction scheme is provable <strong>in</strong> LAST, a set<br />

<strong>the</strong>ory with <strong>the</strong> comprehension pr<strong>in</strong>ciple <strong>in</strong> light l<strong>in</strong>ear logic<br />

[Girard 1998] [Terui 2004]. In LAST, <strong>the</strong> def<strong>in</strong>ition of<br />

natural numbers is quite different: such def<strong>in</strong>ition seems to<br />

enable to prove <strong>the</strong> weak <strong>in</strong>duction.<br />

3 Meanwhile, <strong>the</strong> comprehension <strong>the</strong> pr<strong>in</strong>ciple can be thought as a special<br />

case of <strong>the</strong> recursive def<strong>in</strong>ition. S<strong>in</strong>ce it is <strong>the</strong> foundation placed <strong>in</strong> <strong>the</strong> calculation,<br />

so we had better to say that it is <strong>the</strong> generalized recursive def<strong>in</strong>ition that<br />

is essential pr<strong>in</strong>ciple <strong>in</strong> this <strong>the</strong>ory.<br />

4 For <strong>the</strong> def<strong>in</strong>ition of <strong>the</strong> ordered pair <strong>in</strong> H , see [Cant<strong>in</strong>i 2003].<br />

403

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!