From: Jan Burse on
Nam Nguyen schrieb:

> Thanks for the link. As mentioned with JB, the context here
> is "FOL proof". My intention is to claim (A xor ~A) _signify_
> the impossibility to syntactically prove the consistency of
> a consistent system. But I'm not so sure the formula would
> be a correct characterization of such.

The problem here is that you're changing the formula, the
formula where the notion "tautology" is applied to. But I guess
for your venture you probably need to change the notion. Go
more down to the bowels of truth respectively derivation.

One way to do it, would be to use a modual logic of provability.
There you would have available inside the logic what is usually
only available out side. The modal operator can refer to
interpretations, like we do in the notion of tautology. Alternatively
the modal operator could also refer to derivations, like it
is done in the notion of valid consequence.

Lets try the later. Let []A denote that there is a derivation
for A. Then consistency is simply ~[]f, see below. Then the
imposibility to prove this would be ~[]~[]f. Well this is
just a shoot from the hip. No clue whether this would really
work out to the end. The problem that now remains is a precise
definition of your provability logic, and of course a derivation
of a all the necessary theorems, if possible.

> The scodnary intention is to go from there and find a replacement
> for CON(T) that doesn't depend on the notion of the natural numbers.
> (A very long shot though, imho).

CON(T) should be the same as T |/- f. So falsum is not derivable
from the theory T. By the deduction theorem this is the same as:

|/- T -> f

Which is the same as:

|/- ~T

So the negation of the theory T is not derivable. This of course
only works when T is finitely representable. I.e. when it is
not an infinite set of formulae. When it is a finite set of formulae
we can take the conjunction and we will stay in FOL.





From: Jan Burse on
Jan Burse schrieb:
> Nam Nguyen schrieb:
>
>> Thanks for the link. As mentioned with JB, the context here
>> is "FOL proof". My intention is to claim (A xor ~A) _signify_
>> the impossibility to syntactically prove the consistency of
>> a consistent system. But I'm not so sure the formula would
>> be a correct characterization of such.

> Lets try the later. Let []A denote that there is a derivation
> for A. Then consistency is simply ~[]f, see below. Then the
> imposibility to prove this would be ~[]~[]f. Well this is
> just a shoot from the hip. No clue whether this would really
> work out to the end. The problem that now remains is a precise
> definition of your provability logic, and of course a derivation
> of a all the necessary theorems, if possible.

Looks like ~[]~[]f does not really say what you wanted to say.
First of all I was using the wrong modal operator. It seems
that <>A is better suited to denote derivation, since <>
usually indicates a possibility(*).

Then we would need to distinguish between two modal operators.
One for the semantical truthness in all interpretations [1]
and one for the syntactic derivability <2>. Then I guess
what you want to say would be, where T is your theory you
are examine:

~[1]~T |/- ~<2>~T

The outer |/- is again a modal operator. So maybe the whole
thing would boil down to:

~<2>(~[1]~T -> ~<2>~T)

Bye

(*)
But in literature they neverthelss use the box instead the diamond.
http://en.wikipedia.org/wiki/Provability_logic
From: Nam Nguyen on
Jan Burse wrote:
> Nam Nguyen schrieb:
>> Please note the word "proof" in my "As far as FOL proof is
>> concerned". Are we then saying in _all_ formal systems
>> there's a _proof_ for (A xor ~A)?
>
> Your saying "As far as FOL proof is concerned (A \/ ~A)
> signifies a tautology" doesn't make much sence.

Let's not fight with terminology where we don't have to.
We could have coined a new term "syntactical tautology",
"logical theorem-hood", etc... The point is from the
syntactical point of view it'd signify something, just
as (A /\ ~A) would signify the consistent theory of the
underlying language.

> Since
> the notion tautology is not related to derivability
> in a proof system, it is related to a notion of truth.

One could say tautology definition is based on the notion
of truth values, but as far as FOL provability is concerned
I'd not say it's not "related" to provability. There's such
a thing as the (meta) Tautology Theorem, which concerns
about provability of some formulas.

>
> A tautology(*) is a formula that is true under all
> interpretations. A proof system does not primarly
> define a notion of interpretation. A proof system
> usually defines a notion of derivation.

Right.

>
> But what should your saying "As far as FOL is concerned
> (A /\ ~A) signifies a contradiction" mean? The difficulty
> here is the same.

It's not the same. (A /\ ~A) represents an inconsistent
theory _where contradictions can be proven_ !

> The notion of contradiction, as
> a property of a formula, is not related to derivability
> in a proof system, it is related to a notion of truth.

That's simply not true: semantically, (A /\ ~A) means
a contradiction!

>
> A contradiction(*) is a formula that is false under all
> interpretations.

Be that as it may, a contraction could also be defined
as a formula of the form (F /\ ~F) for some formula F.

All this still doesn't make it wrong to ask what (A xor ~A)
would syntactically signify within the context provability
w.r.t. to a particular T. You yourself said "A xor ~A is also
a tautology" so apparently it does signify a "logical-theoremhood",
as far as FOL proof is concerned, right?

--
-----------------------------------------------------------
Normally, we do not so much look at things as overlook them.
Zen Quotes by Alan Watt
-----------------------------------------------------------
From: Nam Nguyen on
Nam Nguyen wrote:
>
> The [secondary] intention is to go from there and find a replacement
> for CON(T) that doesn't depend on the notion of the natural numbers.
> (A very long shot though, imho).

It still seems a long short but let me attempt. The meta statement
I think we'd like to make would be something like

(*) Let T be a formal system satisfying certain "conditions", then
if T is consistent there is a formula of L(T), named say KONT(T),
that is true but not provable in T.

Where "conditions" would be anything except "as strong as arithmetic".
Tentatively I have a suggestion: "expressing non-finiteness". So the
statement would now be:

(**) Let T be a consistent formal system that doesn't express finiteness,
there's a formula of L(T), named say KONT(T), that's true but not
provable in T.

For example, the extension G' of the basic group theory G with the
additional axiom A[x=y] would be a system that would express finiteness.
Otoh, Q and PA would express non-finiteness.

***

Now then we do have a trivial meta theorem about a general consistent T:

(*') If T is a consistent formal system, then there's a formula of L(T)
that if true in T is unprovable.

I think it's debatable how close (*') is to (**) but seems to be in
the right direction. Any suggestion so far though?

--
-----------------------------------------------------------
Normally, we do not so much look at things as overlook them.
Zen Quotes by Alan Watt
-----------------------------------------------------------
From: Nam Nguyen on
Nam Nguyen wrote:
> Nam Nguyen wrote:
>>
>> The [secondary] intention is to go from there and find a replacement
>> for CON(T) that doesn't depend on the notion of the natural numbers.
>> (A very long shot though, imho).
>
> It still seems a long short but let me attempt. The meta statement
> I think we'd like to make would be something like
>
> (*) Let T be a formal system satisfying certain "conditions", then
> if T is consistent there is a formula of L(T), named say KONT(T),
> that is true but not provable in T.
>
> Where "conditions" would be anything except "as strong as arithmetic".
> Tentatively I have a suggestion: "expressing non-finiteness". So the
> statement would now be:
>
> (**) Let T be a consistent formal system that doesn't express finiteness,
> there's a formula of L(T), named say KONT(T), that's true but not
> provable in T.
>
> For example, the extension G' of the basic group theory G with the
> additional axiom A[x=y] would be a system that would express finiteness.
> Otoh, Q and PA would express non-finiteness.

After some reflections I'd like to alter the intended meta statement.
The revised statement would now be:

(*) If T is a formal system that expresses non-finiteness and that has
a model M, then there's a formula of L(T), say KONT(T), that if true
in M is not provable in T.

Also, the following isn't a full (meta) proof and is only a take-it-at-
own-risk sketch. But here it is:

(a) We'd make use of Cantor theorem that there's no 1-1 mapping
between U (of M) and P(U) [which is the Power set of U].

(b) We'd make use the fact that we could "re-define" L(T) so that
there be more non-logical symbols available to us that L(T)
might have.

Would you think the revised (*), steps (a) and (b) are in the right
direction?


--
-----------------------------------------------------------
Normally, we do not so much look at things as overlook them.
Zen Quotes by Alan Watt
-----------------------------------------------------------