ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Distributed Knowledge?

To: Len Yabloko <lenya@xxxxxxxxxxxxx>
Cc: ontolog-forum <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Wed, 15 Oct 2008 23:20:58 -0500
Message-id: <5F145445-1003-4EE2-8AAC-497F81686DBB@xxxxxxx>

On Oct 15, 2008, at 5:56 PM, Len Yabloko wrote:    (01)

>>> Pat,
>>>
>>> You answer makes clear how Semantic Web is supposed to operate.
>>> However, it seems to leave very important practical consideration
>>> outside of its scope.
>>
>> Can you elaborate?
>
> Sure. The most important to me consideration and the one I was  
> asking you about is how to negotiate assumptions in order to  
> complete required inferences.    (02)

What kind of negotiation would be required? If you can point to a real  
example I might be able to make sense of your concern, Right now it  
seems to be a worry about nothing.    (03)

> Did I miss your answer to that questions?    (04)

I havnt understood the question yet :-)    (05)

>
>
>>
>>> SB>>     Alternatively, one might hypothesise that individuals may
>>> only
>>>>> pick up on the knowledge they need - use only the terms without
>>>>> further inference. That means that they have, to some extent,
>>>>> misunderstood the term.
>>>
>>> I strongly agree. There are implicit assumptions behind any
>>> inference which are not part of any specification.
>>
>> No. That is the whole point of the semantic web, and why it is based
>> on a monotonic system of logics. When you publish some SWeb content,
>> you take on the responsibility for supplying enough information to
>> enable a reader to draw the appropriate conclusions. If some
>> assumptions are left 'implicit', then you, the publisher, have not
>> done your job properly.
>
> I think some assumption are always left implicit.    (06)

Why do you think so? It would follow, if you are correct, that all  
communication was incomplete and inadequate. But communication does  
happen, and jobs do get done.    (07)

> How can anyone do that job properly? Do you?    (08)

Of course. Isn't this exactly what ontology engineering is all about?    (09)

> However, you are not required to publish the
>> conclusions themselves: the process of drawing the conclusions is up
>> to the reader. And a reader may, of course, input content from many
>> sources and hence be able to draw other conclusions which you had not
>> anticipated.
>
> I was asking not about unanticipated conclusions, but about  
> assumptions. Nobody need to anticipate all conclusions, but some  
> assumptions must not be contradicted or should be negotiated (ie  
> accepted or rejected).    (010)

I do not need to know your assumptions, only what it is you are  
asserting. How you derived or came to the propositions you are  
asserting is not my concern.    (011)

>>> If one wants to apply Semantic Web to a particular problem, then
>>> some of these assumptions must be evaluated with respect to the
>>> problem domain. Or are you suggesting that assumptions implicit in
>>> SW core meta-theory
>>
>> I am not sure what you mean by 'core meta-theory'. The nearest
>> interpretation which makes sense woul dbe the specifications of the
>> actual logics involved (RDF, RDFS, OWL, etc.) and their semantics. If
>> this is what you mean, then yes, indeed...
>>
>>> are never in conflict with any problem domain,
>>
>> ... that is being assumed. The logics themselves are indeed 'domain-
>> neutral'. They can be used to describe any topic under the sun.
>>
>>> as long as the problem defined in terms of the same meta-theory? In
>>> that case we are forced into Closed World Assumption which may not
>>> be practical for distributed knowledge.
>>
>> No, we are not forced into a closed world assumption. I have no idea
>> how you came to this conclusion.
>>
>
> Very simple: agent interaction in most cases are conducted on basis  
> of incomplete information.    (012)

All information is incomplete in some sense, but ....    (013)

> Therefore in the absence of any contradiction to the initial  
> assumptions agents indeed treat those as true.    (014)

Nonsense. If I do assert a FOAF RDF page about myself and my contacts,  
this says nothing about, say, my weight, whether or not I have had  
cancer, or the phases of the moon. Nevertheless, it would be insane to  
draw any conclusion about their truth, from the fact that no  
information is provided. Your talk of 'initial assumptions' begs the  
question: there are no 'initial assumptions' other than what is  
actually asserted.    (015)

> For example if you pay with credit card then bank is assuming that  
> you will pay back to the bank.    (016)

No, the bank isn't assuming anything: it is partaking in a system of  
contractual obligations which give it a guarantee of payment,  
underwritten by insurance schemes which are paid for the fees  
retailers pay to the credit companies in return for their increased  
volume of business. Mechanical support of such contractual dealings is  
exactly one of the primary commercial use cases for SWeb technology.  
But the actual reasonings and communications involved are far more  
elaborate than simplistic assume-unless-you-hear-otherwise inferences.  
When written out in detail, none of them are in fact non-monotonic.  
Some local information sources use such reasoning internally, and  
thereby take a risk of error which they (or their owners) are willing  
to bear; but published Web content has to always be treated as a  
simple monotonic assertion.    (017)

> Isn't financial system then based on Closed World Assumption?    (018)

No.    (019)

> If Semantic Web does not offer any mechanism for negotiating  
> assumptions, then every transaction is based on CWA.    (020)

Even with your faulty reasoning, this conclusion would not follow.    (021)

>
>
>>>
>>> PH>No. They can do as much inference or as little as they like. But
>>> their
>>>> not doing so does not imply that they are MISunderstanding  
>>>> anything.
>>>> They may need only a very limited set of valid conclusions to do
>>>> their
>>>> job. That is not misunderstanding.
>>>
>>>
>>> But what if one can not draw necessary conclusions to do the job -
>>> where do we go from here?
>>
>> Then the agent in question is not able to use the semantic web, ie it
>> is in the same position as all agents were before the SWeb was  
>> created.
>>
>
> Which basicaly renders SW useless for most practical purposes  
> (unless augmented with negotiating mechanism)    (022)

Im afraid I disagree, as do the now rather large number of actual SWeb  
users. Think: your reasoning applies without change to the non- 
semantic WSeb, the one we all use. Evidently there must be something  
wrong with your argument.    (023)

>>> Does SW offer any means for negotiating additional required
>>> knowledge to make further inferences?
>>
>> Any agent, even yourself, can of course go looking on the Web for
>> additional knowledge. That is pretty much what the Web is all about,
>> after all. There are aids to such a search, such as Swoogle.  Most
>> interesting SWeb applications use knowledge from multiple sources in
>> this way.  It is of course much more efficient if you already have
>> some idea where to look.
>>
>
> May be that is what SW is all about, but certainly not the Web.  
> Today it is about doing business online, - not about posting  
> ontologies.    (024)

Ontologies are used to do business with. Every time you use a credit  
card, reasoners are consulting large ontologies to judge whether the  
transaction is plausible, given your spending history.    (025)

>>> SB>> This raises two questions; firstly, how can one quantify the
>>> level
>>>>> of misunderstanding? and secondly, how can one then determine if  
>>>>> the
>>>>> level of misunderstanding is significant?
>>>>>
>>>
>>> I think that is the question of great practical significance. The
>>> "level of misunderstanding" can be replaced with more formal
>>> criteria based on specific conclusions that could not be made. Then
>>> some of initial assumptions can be revised and resulting "damage"
>>> assessed to answer the second question.
>>
>> Welcome to the semantic web. I suggest reading some of the now quite
>> extensive technical literature before expressing any more views in
>> public.
>
> Thank you for welcoming me to (your) semantic web, but I am not sure  
> I want to come in. Pehaps you are not doing your job properly (which  
> is by the way not censorship of this forum). And may I suggest that  
> you read more technical literature about the Web applications before  
> expressing any more views in public.    (026)

Thanks for the advice. I'll make sure I read as much as I can before  
writing any more Web specification documents.    (027)

Pat    (028)


> ------------------------------------------------------------    (029)

IHMC                                     (850)434 8903 or (650)494 3973
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (030)






_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (031)

<Prev in Thread] Current Thread [Next in Thread>