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.

62<br />

The Knower Paradox <strong>and</strong> <strong>the</strong> Quantified Logic of Proofs — Walter Dean / Hidenori Kurokawa<br />

arithmetic, 10) is already suggestive of quantification over<br />

a doma<strong>in</strong> of <strong>in</strong>formal proofs — cf. (Tait 2001).<br />

In order to reconstruct <strong>the</strong> Knower <strong>in</strong> a system of<br />

explicit modal logic, we need a version of LP which<br />

conta<strong>in</strong>s quantifiers rang<strong>in</strong>g over proofs. Such a system is<br />

presented <strong>in</strong> (Fitt<strong>in</strong>g 2004) under <strong>the</strong> name QLP. The<br />

language of QLP is given by first specify<strong>in</strong>g a class of<br />

proof terms<br />

TermQLP = c | x | !t | t1 ⋅ t2 | t1 + t2 | <br />

The class of formulas of LP is <strong>the</strong>n specified as follows:<br />

FormLP = P | t: φ | ¬φ | φ → ψ | (∀x)φ | (∃x)φ<br />

The axioms of QLP are as follows:<br />

LP1 all tautologies of classical propositional logic<br />

LP2 t:( φ → ψ) → (s: φ → t·s:ψ)<br />

LP3 t: φ → φ<br />

LP4 t: φ →!t:t: φ<br />

LP5 t: φ → t+s: φ <strong>and</strong> s: φ → t+s: φ<br />

QLP1 (∀x) φ (x) → φ (t)<br />

QLP2 (∀x)(ψ → φ(x)) → (ψ → (∀x) φ (x))<br />

QLP3 φ(t) → (∃x)φ(x)<br />

QLP4 (∀x)(φ(x) → ψ) → ((∃x)φ(x) → ψ)<br />

UBF (∀x)t:φ(x) → :(∀x)φ(x), x ∉ FV(t)<br />

Axioms LP1-LP5 correspond to versions of <strong>the</strong> S4 axioms<br />

where<strong>in</strong> <strong>in</strong>stances of � have been “realized” by proof<br />

terms. Axioms QLP1-QLP4 correspond to a set of axioms<br />

adequate for classical predicate calculus <strong>and</strong> to which <strong>the</strong><br />

usual free variable restrictions apply. UBF is an explicit<br />

form of <strong>the</strong> Barcan formula <strong>and</strong> is justified on <strong>the</strong> basis of<br />

<strong>the</strong> observation that if we possess a proof term t which<br />

serves to uniformly verify φ(x) for all x, <strong>the</strong>n <strong>the</strong>re should<br />

be a proof (denoted by <strong>the</strong> complex proof term )<br />

which serves to justify (∀x) φ(x). The rules of QLP consist<br />

of modus ponens <strong>and</strong> universal generalization toge<strong>the</strong>r<br />

with a rule known as axiom necessitation. This rule says<br />

that if φ is an axiom of QLP, <strong>the</strong>n we may <strong>in</strong>troduce c:φ<br />

where c is a so-called proof constant — i.e. an unstructured<br />

proof term <strong>in</strong>troduced as an atomic justification for φ.<br />

Before reconstruct<strong>in</strong>g <strong>the</strong> derivation of <strong>the</strong> Knower<br />

<strong>in</strong> QLP, it will be useful to record <strong>the</strong> follow<strong>in</strong>g technical<br />

result:<br />

Theorem (Lift<strong>in</strong>g) [Artemov/Fitt<strong>in</strong>g]<br />

If QLP ⊢ φ, <strong>the</strong>n for some proof term t, QLP ⊢ t:φ.<br />

The Lift<strong>in</strong>g Theorem reports that if a statement φ is derivable<br />

<strong>in</strong> QLP, its derivation may be <strong>in</strong>ternalized with<strong>in</strong> <strong>the</strong> system<br />

so as to yield a proof term t which exhibits its structure. As<br />

such, <strong>the</strong> Lift<strong>in</strong>g Theorem serves as a sort of explicit counterpart<br />

to <strong>the</strong> S4 necessitation rule (i.e. ⊢ F / ⊢ �F) which<br />

itself is an implicit form of <strong>the</strong> rule Nec used to justify <strong>the</strong><br />

step 5)-6).<br />

The f<strong>in</strong>al step which we must undertaken before<br />

reconstruct<strong>in</strong>g <strong>the</strong> Knower is to <strong>in</strong>troduce some means of<br />

<strong>in</strong>troduc<strong>in</strong>g an explicit analog of a self referential statement<br />

which mirrors (*). The most straightforward way to proceed<br />

is to simply consider <strong>the</strong> result of adjo<strong>in</strong><strong>in</strong>g a statement of<br />

<strong>the</strong> form<br />

12) d:(¬(∃x)x:D ↔ D)<br />

which formalizes “d is a proof of ‘<strong>the</strong>re does not exist a<br />

proof of D iff D’.” Reason<strong>in</strong>g <strong>in</strong> QLP from 12) as a premise,<br />

we may now derive a contradiction as follows:<br />

13) ¬(∃x)x:D → D left to right direction of 11)<br />

14) (∃x)x:D → ¬D right to left direction of 11)<br />

15) (∃x)x:D → D derivable <strong>in</strong> QLP<br />

16) ¬(∃x)x:D propositional logic<br />

17) D<br />

18) t : D for some term t obta<strong>in</strong>able via Lift<strong>in</strong>g<br />

19) t : D → (∃x)x:D QLP3<br />

20) (∃x)x:D<br />

21) ⊥<br />

The step 17)-18) is analogous to <strong>the</strong> step 5)-6) <strong>in</strong> <strong>the</strong> orig<strong>in</strong>al<br />

derivation. In <strong>the</strong> case of QLP, however, this step is<br />

elliptical <strong>in</strong> <strong>the</strong> sense that although we know a term t exists<br />

via <strong>the</strong> Lift<strong>in</strong>g Theorem, such a term must be explicitly<br />

constructed by <strong>in</strong>ternaliz<strong>in</strong>g steps 13)-18). Construct<strong>in</strong>g t<br />

requires not only constants d1 <strong>and</strong> d2 such that<br />

21) d1:(¬(∃x)x:D → D)<br />

22) d2:((∃x)x:D → ¬D)<br />

(which may be constructed from d <strong>in</strong> 12)) but also a proof<br />

term which serves as a verification of 16). Note that while<br />

this statement is a explicit analog of an <strong>in</strong>stance of <strong>the</strong><br />

reflection axiom T, it is not an axiom of QLP. Not only must<br />

this statement be derived <strong>in</strong> QLP, but to construct t, its<br />

proof must also be lifted. This may be done as follows:<br />

24) x:D →D LP3<br />

25) r:(x:D → D) axiom necessitation<br />

26) (∀x)r:(x:D → D) universal generalization<br />

27) (∀x)r:(x:D → D) → :∀x:(x:D → D) UBF<br />

28) :(∀x):(x:D → D)<br />

29) (∀x)r:(x:D → D) → ((∃x):D → D) QLP4<br />

30) q:[(∀x)r : (x:D → D) → ((∃x):D → D)] axiom necessitation<br />

31) q ·:((∃x):D → D) LP2<br />

With this derivation <strong>in</strong> h<strong>and</strong>, it is <strong>the</strong>n easy to see that we<br />

may take<br />

31) t ≡ d1 · ((a · (q · )) ·d2)<br />

where a is a proof constant for <strong>the</strong> tautology (ϕ → ψ) →<br />

((ϕ → ¬ ψ) → ¬ ϕ).<br />

The <strong>in</strong>sight which we th<strong>in</strong>k QLP provides <strong>in</strong>to <strong>the</strong><br />

Knower can now be framed by consider<strong>in</strong>g <strong>the</strong> role which<br />

UBF plays <strong>in</strong> <strong>the</strong> forego<strong>in</strong>g derivation. For note that QLP<br />

<strong>in</strong>cludes nei<strong>the</strong>r a general necessitation rule analogous to<br />

Nec, nor even a local <strong>in</strong>stance of this pr<strong>in</strong>ciple ak<strong>in</strong> to U.<br />

As we have just seen, however, <strong>the</strong> explicit forms of both<br />

pr<strong>in</strong>ciples — i.e.<br />

32) (∃x):D → D <strong>and</strong><br />

33) q · :((∃x):D → D)<br />

— are derivable <strong>in</strong> QLP. Both of <strong>the</strong>se pr<strong>in</strong>ciples are required<br />

<strong>in</strong> order for <strong>the</strong> derivation 13)-21) to go through.<br />

However UBF turns out to essential to <strong>the</strong> derivation of 33)<br />

as it may be shown , without UBF, no statement of <strong>the</strong><br />

form t:((∃x):D → D) is derivable <strong>in</strong> QLP. 2<br />

This is significant for diagnos<strong>in</strong>g how <strong>the</strong> pr<strong>in</strong>ciples<br />

<strong>in</strong>volved <strong>in</strong> <strong>the</strong> orig<strong>in</strong>al derivation of <strong>the</strong> Knower conflict.<br />

Several recent commentators have proposed that <strong>the</strong><br />

paradox should be resolved by reject<strong>in</strong>g U. 3 However, <strong>the</strong><br />

2 This follows from <strong>the</strong> fact that UBF is not conservative over <strong>the</strong> QLP-UBF for<br />

statements not conta<strong>in</strong><strong>in</strong>g terms of <strong>the</strong> of . In particular, it may be<br />

shown that for no φ, do we have QLP−UBF ⊢ (∃y)y:((∃x): φ → φ).<br />

3 More specifically, among <strong>the</strong> three pr<strong>in</strong>ciples employed <strong>in</strong> <strong>the</strong> orig<strong>in</strong>al M&K<br />

derivation (i.e. T, U <strong>and</strong> I), <strong>the</strong> consensus among recent commentators has

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

Saved successfully!

Ooh no, something went wrong!