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.

Interpretability Relations of Weak Theories of Truth<br />

Mart<strong>in</strong> Fischer, Leuven, Belgium<br />

1. Introduction<br />

Axiomatic <strong>the</strong>ories of truth are understood as extensions of<br />

a syntactic base <strong>the</strong>ory which is often taken to be Peano<br />

Arithmetic, PA . One way to measure <strong>the</strong> strength of a<br />

<strong>the</strong>ory of truth is to take <strong>in</strong>to account which formulas of<br />

arithmetic it proofs. Weak <strong>the</strong>ories <strong>in</strong> this respect are <strong>the</strong>ories<br />

that do not prove more than PA itself, <strong>the</strong>ories that are<br />

conservative extensions of PA . The concept of conservativity<br />

has ga<strong>in</strong>ed some <strong>in</strong>terest <strong>in</strong> formaliz<strong>in</strong>g philosophical<br />

criteria. This is also <strong>the</strong> case <strong>in</strong> <strong>the</strong> debate on truth, <strong>in</strong><br />

which conservativity is expected to expla<strong>in</strong> <strong>the</strong> `no substance'<br />

claim of deflationism. For <strong>the</strong>ories of truth conservativity<br />

over PA alone seems to be a very crude measure<br />

s<strong>in</strong>ce it does not differentiate between different conservative<br />

<strong>the</strong>ories of truth which have quite different properties<br />

<strong>and</strong> prove different formulas conta<strong>in</strong><strong>in</strong>g <strong>the</strong> truth predicate.<br />

A comparison of <strong>the</strong> truth-<strong>the</strong>oretic strength of<br />

<strong>the</strong>ories of truth is desirable. A direct comparison of <strong>the</strong><br />

truth-<strong>the</strong>oretic strength is <strong>the</strong> subset relation but it is only<br />

a partial order so that not all <strong>the</strong>ories can be compared.<br />

Ano<strong>the</strong>r measure of <strong>the</strong> strength of a <strong>the</strong>ory of truth would<br />

be <strong>the</strong>ir <strong>in</strong>terpretability relations to o<strong>the</strong>r <strong>the</strong>ories<br />

especially <strong>the</strong>ir <strong>in</strong>terpretability or non<strong>in</strong>terpretability <strong>in</strong> PA .<br />

The most famous of <strong>the</strong>se <strong>in</strong>terpretability relations is<br />

relative <strong>in</strong>terpretability, <strong>in</strong>troduced <strong>in</strong> (Tarski et al. 1953),<br />

<strong>and</strong> it is a good measure for PA as base <strong>the</strong>ory. On <strong>the</strong><br />

one h<strong>and</strong> <strong>the</strong> less restricted version of local <strong>in</strong>terpretability<br />

collapses <strong>in</strong> this case <strong>in</strong>to relative <strong>in</strong>terpretability. On <strong>the</strong><br />

o<strong>the</strong>r h<strong>and</strong> Tarskis <strong>the</strong>orem of undef<strong>in</strong>ability of truth shows<br />

that <strong>the</strong>re is no def<strong>in</strong>itional extension of PA by a one place<br />

predicate τ , PA (τ ), so that PA (τ ) proves τ( ϕ)<br />

↔ ϕ for<br />

all sentences ϕ of <strong>the</strong> language of arithmetic.<br />

2. Axiomatic <strong>the</strong>ories<br />

LA is <strong>the</strong> language of arithmetic <strong>and</strong> Lτ : = A {τ ∪ . The<br />

arithmetical <strong>the</strong>ories are as usual. For <strong>the</strong> <strong>in</strong>terpretability<br />

considerations take <strong>the</strong> arithmetic <strong>the</strong>ories to be formulated<br />

with predicate- <strong>in</strong>stead of functionsymbols.<br />

96<br />

L }<br />

( Ind P ) P( 0)<br />

∧ ∀x(<br />

P(<br />

x)<br />

→ P(<br />

x + 1))<br />

→ ∀x(<br />

P(<br />

x))<br />

Q is Rob<strong>in</strong>son Arithmetic <strong>and</strong> PA is Peano Arithmetic,<br />

that is P A L Ind Q ∪ ) ⋅ ( , where P A L Ind ⋅ ) ( is <strong>the</strong> set of sentences<br />

that result from replac<strong>in</strong>g P <strong>in</strong> ( Ind P ) by a formula<br />

of L A with at least one free variable. Accord<strong>in</strong>gly,<br />

IΣ k = Q ∪ ( IndP<br />

) ⋅ Σk<br />

.<br />

Assume that L A conta<strong>in</strong>s <strong>the</strong> relevant syntactical<br />

vocabulary: ‘ Ct ’ for closed term of L A , ‘ Sent ’ for<br />

sentence of L A , ‘ Form 1’<br />

for formula of LA with one free<br />

variable, <strong>and</strong> so on, such that PA proves <strong>the</strong> relevant<br />

syntactical <strong>the</strong>orems. Especially if m is <strong>the</strong> gödelnumber<br />

of a formula ϕ (x)<br />

with one free variable x <strong>and</strong> n of a<br />

term t , <strong>the</strong>n m (n)<br />

is <strong>the</strong> gödelnumber of <strong>the</strong> substitution<br />

τ<br />

of <strong>the</strong> free variable x <strong>in</strong> ϕ (x)<br />

by <strong>the</strong> numeral of t . PA is<br />

PA formulated <strong>in</strong> <strong>the</strong> language L τ . A <strong>the</strong>ory of truth T is<br />

τ<br />

a Lτ -<strong>the</strong>ory with PA ⊆ T .<br />

tot ( x)<br />

: ⇔ Form1(<br />

x)<br />

∧ ∀y(<br />

τ ( x(<br />

y))<br />

∨τ<br />

( ¬ & x(<br />

y )))<br />

Disquotational <strong>the</strong>ories of truth are formulated with a<br />

scheme of T-biconditionals:<br />

( TB P )<br />

τ ( P) ↔ P<br />

( UTB P ) ∀x(<br />

τ ( P(<br />

x&<br />

) ) ↔ P(<br />

x)).<br />

Compositional axioms are <strong>the</strong> universally quantified versions<br />

of <strong>the</strong> follow<strong>in</strong>g formulas:<br />

(C 1)<br />

Ct ( x)<br />

∧ Ct(<br />

y ) → ( τ ( x = & y)<br />

↔ val(<br />

x)<br />

= val(<br />

y)).<br />

(C 2)<br />

Ct( x)<br />

∧ Ct(<br />

y)<br />

→ ( τ ( x ≠&<br />

y)<br />

↔ val(<br />

x)<br />

≠ val(<br />

y)).<br />

(C 3)<br />

Sent( x)<br />

∧ Sent(<br />

y)<br />

→ ( τ( x ∧&<br />

y ) ↔ τ(<br />

x)<br />

∧τ<br />

( y)).<br />

(C 4)<br />

Sent ( x)<br />

∧ Sent(<br />

y)<br />

→ ( τ ( ¬ & x ∧&<br />

y)<br />

↔ τ(<br />

¬ & x)<br />

∨ τ ( ¬ & y)).<br />

(C 5)<br />

Sent( ∀&<br />

yx)<br />

→ ( τ( ∀&<br />

yx)<br />

↔ ∀z(<br />

τ ( x(<br />

z)))).<br />

(C 6)<br />

Sent ( ¬ & ∀&<br />

yx)<br />

→ ( τ ( ¬ & ∀&<br />

yx)<br />

↔ ∃z(<br />

τ ( ¬ & x(<br />

z)))).<br />

(C 7)<br />

Sent( x)<br />

→ ( τ ( ¬ & ¬ & x)<br />

↔ τ ( x)).<br />

(C 8)<br />

Sent( x)<br />

→ ( τ ( ¬ & x)<br />

↔ ¬ τ ( x)).<br />

The axiom of <strong>in</strong>ternal <strong>in</strong>duction for total formulas is:<br />

( It I)<br />

∀ x( tot(<br />

x)<br />

∧τ<br />

( x(<br />

0))<br />

∧ ∀y(<br />

τ(<br />

x(<br />

y )) → τ(<br />

x(<br />

y + 1)))<br />

→ ∀y(<br />

τ(<br />

x(<br />

y )))).<br />

The relevant <strong>the</strong>ories are:<br />

TB : = Q ∪ ( TBP<br />

) ⋅ LA<br />

∪ ( IndP<br />

) ⋅ Lτ<br />

UTB : = Q ∪ ( UTBP<br />

) ⋅ LA<br />

∪ ( IndP<br />

) ⋅ Lτ<br />

PT : PA ( C1)<br />

( C7)<br />

r<br />

= ∪ −<br />

−<br />

PT : = PA ∪ ( C1)<br />

− ( C7)<br />

∪ ( ItI<br />

)<br />

PT : = PA ∪ ( C1)<br />

− ( C7)<br />

∪ ( IndP<br />

) ⋅ Lτ<br />

TC : PA ( C1)<br />

( C8)<br />

r<br />

= ∪ −<br />

−<br />

TC : = PA ∪ ( C1)<br />

− ( C8)<br />

∪ ( ItI<br />

)<br />

r<br />

TC is also known as PA (S)<br />

<strong>and</strong> TC as T (PA)<br />

.<br />

3. Interpretability<br />

Some basic results:<br />

(i) TB ⊂ UTB ⊂ PT = TC<br />

−<br />

(ii) IΣ1 ∪ ( C1),<br />

( C3),<br />

( C5),<br />

( C8)<br />

∪ ( It<br />

I)<br />

= TC<br />

−<br />

(iii) IΣ1 ∪ ( C1)<br />

− ( C7)<br />

∪ ( It<br />

I)<br />

= PT<br />

− −<br />

(iv) PT , TC are f<strong>in</strong>itely axiomatizable.<br />

r r<br />

(v) TB,<br />

UTB,<br />

PT , TC , PT,<br />

TC are not f<strong>in</strong>itely axiomatizable.<br />

r r −<br />

(vi) TB,<br />

UTB,<br />

PT , TC , PT are conservative extensions of PA .<br />

−<br />

(vii) TC , PT,<br />

TC are nonconservative extensions of PA .<br />

Def<strong>in</strong>ition<br />

Let S, T be <strong>the</strong>ories formulated <strong>in</strong> L S, LT<br />

. Then<br />

T is a pure extension of S iff T is an extension of S <strong>and</strong> L T = LS<br />

.<br />

T is reflexive iff T proves Con Δ for all f<strong>in</strong>ite Δ ⊆ T.<br />

T is essentially reflexive iff all pure extensions of T are reflexive.<br />

T has full <strong>in</strong>duction iff for all formulas ϕ of L T : T proves<br />

ϕ( 0)<br />

∧ ∀x(<br />

ϕ(<br />

x)<br />

→ ϕ(<br />

x + 1))<br />

→ ∀xϕ(<br />

x)<br />

.<br />

Full <strong>in</strong>duction <strong>and</strong> reflexivity are connected <strong>in</strong> <strong>the</strong> follow<strong>in</strong>g<br />

way, as shown for example <strong>in</strong> (Hájek/Pudlák 1993, p.189):<br />

Lemma 1<br />

If PA ⊆ T <strong>and</strong> T has full <strong>in</strong>duction, <strong>the</strong>n T is reflexive.

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

Saved successfully!

Ooh no, something went wrong!