George, you can do that indeed, but then you are particularizing
things. This can be helpful from a pedagogical point of view, but the
advantage of the axiomatic approach (to a knowledge theory) is that
once you agree on the axioms and rules, then you agree on the
consequences independently of the particular instantiation you think
about. Word like "machine", "access", "memory", "world", data, are,
fundamentally harder than the simple idea of knowledge the modal S4
axioms convey. Using machines, for example, could seem as a
computationalist restriction, when the axioms S4 remains completely
neutral, etc. Also, acceding a memory is more "opinion" than knowledge
because we can have false memory for example. (And then what are the
inference rules of your system?).
S4 is a normal modal logic with natural Kripke referentials
(transitive, reflexive accessibility relations).
A bit more problematic is your identification of "true" with "exist".
This hangs on possible but highly debatable and complex relations
between truth and reality. This is interesting per se, but imo a bit
out of topics, or premature (in current thread). Perhaps we will have
opportunity to debate on this, but I want make sure that what I am
explaining now does not depend on those possible relations (between
truth and reality).
Bruno
Le 24-nov.-07, à 21:23, George Levy a écrit :
> Bruno thank you for this elaborate reply. I would like these three
> statements to make use of cybernetic language, that is to be more
> explicit in terms of the machine or entity to which they refer. Would
> it be correct to rephrase the statements in the active tense, using
> the machine as the subject, replacing proposition p by the term data
> and replacing "true" by "exist"? The statements would then be:
>
> In a world W there is a machine M, data p and data q such that
> 1) If M has access to p (possibly in its memory), then p exists in W.
> 2) If M has access to p, then M has access to the access point to p.
> 3) If M has access to the information relating or linking p to q then
> if M has access to p, it also has access to q.
>
> I assumed that the term "has access" means "in its memory"... but it
> does not have to.
> I also assumed in statements 3 that the multiple uses of M refers to
> the same machine. I guess there may be cases where multiple machines
> can have access to the dame data.
> Same with statement 4
>
> George
>
> Bruno Marchal wrote:
> Le 22-nov.-07, à 20:50, George Levy a écrit :
> Hi Bruno,
> I am reopening an old thread ( more than a year old)
> which I found very intriguing. It leads to some startling conclusions.
> Le 05-août-06, à 02:07, George Levy a écrit :
> Bruno Marchal wrote:I think that if you want to
>>>> make the first person primitive, given that neither you nor me can
>>>> really define it, you will need at least to axiomatize it in some
>>>> way.
>>>> Here is my question. Do you agree that a first person is a knower,
>>>> and
>>>> in that case, are you willing to accept the traditional axioms for
>>>> knowing. That is:
>>>>
>>>> 1) If p is knowable then p is true;
>>>> 2) If p is knowable then it is knowable that p is knowable;
>>>> 3) if it is knowable that p entails q, then if p is knowable then
>>>> q is
>>>> knowable
>>>>
>>>> (+ some logical rules).
>>>>
>>>
>>
>>> Bruno, what or who do you mean by "it" in statements 2) and 3).
>>
>>
>>
>>
>> The same as in "it is raining". I could have written 1. and 2. like
>>
>> 1) knowable(p) -> p
>> 2) knowable(p) -> knowable(knowable(p))
>>
>> In this way we can avoid using words like "it", or even like "true".
>> "p" is a variable, and is implicitly universally quantified over.
>> "knowable(p) -> p" really means that whatever is the proposition p,
>> if it is knowable then it is true. The false is unknowable (although
>> it could be conceivable, believable, even provable (in inconsistent
>> theory), etc. The "p" in 1. 2. and 3. is really like the "x" in the
>> formula (sin(x))^2 + (cos(x))^2 = 1.
>>
>> "knowable(p) -> p" really means that we cannot know something false.
>> This is coherent with the natural language use of know, which I
>> illustrate often by remarking that we never say "Alfred knew the
>> earth is flat, but the he realized he was wrong". We say instead
>> "Alfred believed that earth is flat, but then ...". . The axiom 1. is
>> the incorrigibility axiom: we can know only the truth. Of course we
>> can believe we know something until we know better.
>> The axiom 2. is added when we want to axiomatize a notion of
>> knowledge from the part of sufficiently introspective subject. It
>> means that if some proposition is knowable, then the knowability of
>> that proposition is itself knowable. It means that when the subject
>> knows some proposition then the subject will know that he knows that
>> proposition. The subject can know that he knows.
>>
>>
>>
>>
>>
>>> In addition, what do you mean by "is knowable", "is true" and
>>> "entails"?
>>
>>
>>
>> All the point in axiomatizing some notion, consists in giving a way
>> to reason about that notion without ever defining it. We just try to
>> agree on some principles, like 1.,2., 3., and then derives things
>> from those principles. Nuance can be added by adding new axioms if
>> necessary.
>> Of course axioms like above are not enough, we have to use deduction
>> rules. In case of the S4 theory, which I will rewrite with modal
>> notation (hoping you recognize it). I write Bp for B(p) to avoid
>> heaviness in the notation, likewize, I write BBp for B(B(p)).
>>
>> 1) Bp -> p (incorrigibility)
>> 2) Bp -> BBp (introspective knowledge)
>> 3) B(p->q) -> (Bp -> Bq) (weak omniscience, = knowability of the
>> consequences of knowable propositions).
>>
>> Now with such axioms you can derive no theorems (except the axiom
>> themselves). So you need some principles which give you a way to
>> deduce theorems from axioms. The usual deduction rule of S4 are the
>> substitution rule, the modus ponens rule and the necessitation rule.
>> The substitution rule say that you can substitute p by any
>> proposition (as far as you avoid clash of variable, etc.). The modus
>> ponens rule say that if you have already derived some formula A, and
>> some formula A -> B, then you can derive B. The necessitation rule
>> says that if you have already derive A, then you can derive BA.
>>
>>
>>
>>> Are "is knowable", "is true" and "entails" absolute or do they have
>>> meaning only with respect to a particular observer?
>>
>>
>> The abstract S4 theory is strictly neutral on this. But abstract
>> theory can have more concrete models or interpretations. In our
>> lobian setting, it will happen that "formal provability by a machine"
>> does not obey the incorrigibility axiom (as Godel notices in his 1933
>> paper). This means that formal provability by a machine cannot be
>> used to modelize the knowability of the machine. It is a bit
>> counterintuitive, but formal provability by a machine modelizes only
>> a form of "opinion" by the machine, so that to get a knowability
>> notion from the provability notion, we have to explicitly define
>> knowability(p) by "provability(p) and p is true". (Cf Platos's
>> Theaetetus).
>> Here provability and knowability is always relative to an (ideal)
>> machine.
>> I will come back on this in my explanation to David later. But don't
>> hesitate to ask question before.
>>
>>
>>
>>
>>> Can these terms be relative to an observer? If they can, how would
>>> you rephrase these statements?
>>
>>
>> An observer ? I guess you mean a subject. Observability could obeys
>> quite different axioms that knowability (as it is the case for
>> machine with comp).
>> Just interpret "knowable(p)" by "p is knowable by M".
>> "M" denotes some machine or entity belonging to some class of
>> machine/entity in which we are interested.
>>
>>
>>
>>>
>>
>>
>> "p" has to refer to a proposition. Of course in english (at least in
>> french) we often use similar word with different denotation or
>> meaning. For example, you can say "I know Paul". And Paul, a priori
>> is not a poposition. But such a "know" is not the same as in "I know
>> that Paul is a good guy".
>> S4 is a good candidate for the second "know". The "know" (in "I know
>> Paul") has a quite different meaning, somehow out of topic (to be
>> short). Actually "I know Paul" really means humans variate and
>> pragmatic things like "I met Paul before", or "I know Paul is not the
>> right guy to hire for this job", etc.
>>
>> With the epistemological sense of "knowing", we cannot know a
>> knower, nor an observer. We can only know propositions. Those
>> proposition copuld bear on a knower: like in I know that Paul know
>> that 17 is prime. Sure.
>> Of course we can observe an observer. This illustarte already that
>> observations and knowledge obeys different logics; hopefully related,
>> of course, as it is with the arithmetical hypostases).
>>
>> Oops, I must already go. Have a good week-end, George, and all of
>> you,
>>
>>
>> Bruno
>>
>> PS Marc, Thorgny: I will comment your post Monday or Tuesday.
>>
>>
>>
>>
>> http://iridia.ulb.ac.be/~marchal/
>>
>>
>>
>
>
> >
>
http://iridia.ulb.ac.be/~marchal/
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list.domain.name.hidden
To unsubscribe from this group, send email to everything-list-unsubscribe.domain.name.hidden
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---
Received on Mon Nov 26 2007 - 09:01:59 PST