Fundamental Knowledge-Part 2: Models

The next task is to absorb the traditional area of mathematical logic.  One key missing ingredient is a model.  Let us recall the traditional setup (taken from [1]).

Definition 2.1.  Let S be a set (of symbols).  An S-structure is a pair \mathfrak{A}=(A,\mathfrak{a}) where A is a nonempty set, called a universe, and \mathfrak{a} is a map sending symbols to elements, functions, and relations of A.  An assignment of an S-structure (A,\mathfrak{a}) is a map \beta:S\to A.  An S-interpretation is a pair \mathfrak{I}=(\mathfrak{A},\beta) where \mathfrak{A} is an S-structure and \beta is an assignment in \mathfrak{A}.

For shorthand notation, the convention (with some of my modifications) is to write:  c^\mathfrak{A}=\beta(c), (f(t_1,...,t_n))^\mathfrak{A}=\mathfrak{a}(f)(\beta(t_1),...,\beta(t_n)), and (xRy)^\mathfrak{A}=\beta(x)\mathfrak{a}(R)\beta(y).  These are the terms.  Formulas are then built from the terms using traditional (although this can be generalized) logical connectives.

The notion of a model is then defined via induction on formulas.

Definition 2.2.  Let \mathfrak{I}=(\mathfrak{A},\beta) be an S-interpretation.  We say that \mathfrak{I} satisfies a formula \phi (or is a model of \phi), denoted \mathfrak{I}\vDash\phi, if \phi^\mathfrak{A} holds, where \phi^\mathfrak{A} is defined via its components and \beta and \mathfrak{a} where necessary.

Formal languages in convention are built up from the formulas mentioned above, which are nothing more than special cases of Alt Definition 1.3.  A model for a language is hence nothing more than an A-interpretation into a structure, where A is an alphabet (provided it is equipped with a logic system).  This is precisely what I have constructed in Part 1;  the symbols of W\subset A^* are mapped to the universe \mathcal{L}_{F,T,W}.  The next thing to establish is that every model is a language model.  This is trivial since a model by definition satisfies a set of formulas as well as compounds of them (i.e. it must satisfy a language).  Hence we have no need to trouble ourselves with interpretations and may simply stick to the algebra of Part 1.

While we have absorbed model theory, there are a few more critical topics to absorb from mathematical logic. We return to the language of Part 1 (no pun).  Let X be a theory of \mathcal{L}_{F,T,W} and \varphi:F[X]\to V be a binary logic system.  A formula \phi\in\mathcal{L}_{F,T,W} is derivable in X if it is a proposition (i.e. is in F[X]).  We may write X\vdash\phi.  This definition is in complete agreement with the traditional definition (namely, there being a derivation, or finite number of steps, that begin with axioms and use inference rules);  it is nothing more than saying it is in F[X].  Similarly \phi\in\mathcal{L}_{F,T,W} is valid if \varnothing\vdash\phi, or equivalently, it is derivable in any theory.  In our setup this would imply \phi\in F[\varnothing]=\varnothing.  Hence no formula is valid.

Let F have a unary operation \lnot and \varphi:F[X]\to V be a logic system on a theory X.

If we assume \lnot to be idempotent (\lnot\lnot\phi=\phi), then since \varphi is a homomorphism, we have \varphi(\phi)=\varphi(\lnot\lnot\phi)=\lnot\lnot\varphi(\phi).  That is, the corresponding unary operation in V must also be idempotent on ran(\varphi).

Definition 2.3.  A unary operation \lnot (not necessarily idempotent) is consistent in \varphi if for all \phi\in F[X], \varphi(\phi)\neq\varphi(\lnot\phi).

If we assume \lnot is consistent in \varphi and that \varphi is a binary logic system, then the corresponding \lnot in V is idempotent since

\varphi(\phi)=0\Rightarrow\varphi(\lnot\phi)=1\Rightarrow\varphi(\lnot\lnot\phi)=\lnot\lnot\varphi(\phi)=0.

Again, proofs in a binary system are independent of the choice of valence.   If we assume consistency and idempotency, then we have a nonidentity negation which is idempotent on the range.  The case for assuming binary system and idempotency yields either a trivial mapping of propositions (all to 0 or all to 1), or that \lnot is consistent and idempotent on V.  And lastly if we assume all three (idempotency and consistency of \lnot together on a binary system), we obtain a surjective assignment with idempotent negation in V.

Let \varphi:F[X]\to V be a binary logic system where V is a boolean algebra.  Then the completeness and compactness theorems are trivial.  Recall these statements:

Completeness Theorem.  For all formulas \phi and models \mathfrak{I},

\mathfrak{I}\vDash\phi\Rightarrow X\vdash\phi

where \mathfrak{I}\vDash X.

Compactness Theorem.  For all formulas \phi and models \mathfrak{I},

X\vdash\phi\Rightarrow\mathfrak{I}\vDash\phi

where \mathfrak{I}\vDash X.

Traditionally these apply to, what we would call, a binary logic system \varphi:F[X]\to V where V is a boolean algebra (hence F has a consistent, idempotent negation) under traditional operations, and in particular this fixes the operational/relational structures of F, T, and W , but X is arbitrary.  In this setup, all “formulas” (or what we would hence call propositions since they are generated by a theory) are trivially satisfiable since they have a language model.  Hence Compactness is true.  Moreover since they are propositions in a binary logic system, they are in some F[X] for a theory X and are hence derivable; so we have Completeness.

Lastly we wish to address Godel’s Second Incompleteness Theorem;  recall its statement:

Godel’s Second Incompleteness Theorem.  A theory contains a statement of its own consistency if and only if it is inconsistent.

We have only defined what it means for a unary operation in a logical system to be consistent.  Hence we can say that a binary logic system with a unary operation is consistent if its unary operation is consistent.  But all of these traditional theorems of mathematical logic are assuming a binary logic system where V is a boolean algebra , \lnot is idempotent, and the map \varphi:F[X]\to V is surjective.  Hence \lnot is consistent (from above discussion), and the consequence in the theorem is false.

The weakest possible violation of the antecedent of Godel’s theorem is to use a structure to create itself (i.e. that it is self-swallowing), which makes no sense, let alone using it to create a larger structure within which is a statement about the initial structure.  That a binary logic system with unary operation could contain a statement of its own consistency is itself a contradiction, since the theory itself, together with the statement \phi, are in a metalanguage.  It is like saying that one need only the English language to describe the algebraic structure of the English language.  As we previously said at the end of Part 1, one can get arbitrarily close to doing this–using English to construct some degenerate form of English, but you can never have multiple instances of a single language in a language loop.  Another example would be having the class of all sets, then attempting to prove, using only the sets and operations of them, that there is a class containing them.

Hence the antecedent is also false.  So both implications are true.

[1]  Ebbinghaus, H.-D., J. Flum, and W. Thomas.  Mathematical Logic.  Second Edition.  Undergraduate Texts in Mathematics.  New York: Springer-Verlag.  1994.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: