|From:||"RK Stamper" <stamper.measur@xxxxxxxxx>|
|Date:||Fri, 10 Oct 2008 15:01:42 +0100|
At last an opportunity to respond to comments on this thread. Please forgive the length.
Mike Bennett asked:,
Ronald: Is the approach you describe related to the system of "Norms and
Affordances" described by James Backhouse of the London School of
Economics? He and some colleagues have built a semantics system called
"NORMA" which, to my initial casual reading of it, has a similar
approach to what you describe.
Regrettably it is. I supervised Backhouse's PhD until I moved with my research programme to U. of Twente in the Netherlands. He preferred to stay at the LSE but I continued to support him as his new supervisor had no familiarity with our research. Backhouse avoided showing me his thesis, disregarding our established practice of seeking expert critical comment on all our work. When I read a library copy I understood why. The trouble is evident in the paper you mention,
"Searching for Meaning – Performatives and Obligations in Public Key
Infrastructures" (Tseng and Backhouse 2000),
which includes a semantic schema that was Backhouse's personal contribution. I find it useful for teaching: such a schema must obey a few very strict constraints and he has furnished an excellent case study of how to get the analysis thoroughly wrong.
I looked at Backhouse's work a while ago, but I could not see how to
relate it to other work in AI and logic, such as the works often
referenced in here.
You are right to say that the "approach is very different" but when it is garbled it is even more difficult to compare with the work often cited in Ontolog discussions.
I would be intrigued as to how to relate these two bodies of work to each other.
I'll do my best to help with these comparisons. Let me pick up a remark by Pat Cassidy:
Ronald: You raise interesting issues. Do you have an ontology or otherwise based computational system that demonstrates how you would address them? Or can you point to one?
We have not been aiming to construct an ontology (Ontolog sense) nor primarily to develop a computational system, so that should be the first point of comparison. Our focus has been organized social behavior governed by norms using legal norms as a major source of empirical material.
Nevertheless, to impose on ourselves a strict scientific methodology, we formulated our hypotheses as computable formalisms, claiming them to be capable of accommodating any fragment of legislation. These hypotheses were eminently refutable. (I succumbed to Karl Popper's influence immediately I arrived at the LSE.)
So we built a series of hypotheses and implemented each in a computer system to handle legal norms. Unsurprisingly, each one was quickly refuted, pointing us towards an improved hypothesis.
We now have a computer system that can implement most kinds of the sophisticated administrative norms.
Much earlier however, we discovered that huge improvements in computer applications can be achieved by basing them on ontological dependency schemas. One such system was the administrative system for a university built under Yasser Ades. On this, the first attempt to take the method from analysis to implementation, it far outperformed a highly sophisticated package that had had the benefit of 200 previous implementations.
Already we can certainly contribute to practical business computing even without using computer support for our analysis methods. The software now under development will automatically generate a computer application to support a system specified in terms of semantic schema plus the rules that people should obey. Moreover the resulting system is robust in the face of changing organizational requirements.
Building such a computer system does not provide a satisfactory formal basis for our work. We acknowledge that we have much more work to do on the formal aspects of the research.
Now that brings me to Pat Cassidy again:
My current interest is purely practical - to help find a way to enable semantic interoperability among computer systems. That means that computers will be able to transmit information among them and automatically perform useful actions with the transmitted information, without human intervention.
I have been itching to attack a problem of this kind because I'm convinced that our methods are well suited to it. I am also convinced that human intervention cannot be avoided.
In order to do this I believe that we need to encode that information in a format at least as expressive as OWL-full for simple applications, and at least FOL for the general case. So any knowledge representation system will be relevant to this task only when it has been implemented with some form of reasoner. I take it from your explanation that your ontology has not yet been formalized in that way.
Our current formalization goes quite a way in the direction you indicate.
But formality is less important than arriving at insights that are truly worth formalizing.
I believe that we have something new and worthy of formalization but it does not seem to fit comfortably into the establish mould. For example, from my limited acquaintance with it, OWL allows one to model all kinds of semantic confusion but, my God, it certainly looks formal. Our semantic schemas rule out many models acceptable to OWL.
For another comparison, note that the system we are constructing CANNOT be formalized completely. Each affordance exists as an invariant between a start and a finish with an authority associated with each of those events. As those authorities are always rely upon responsible human beings, our solutions can never be confined to logical formulations either on paper or in computers.
Far from diminishing the role of formalization, we seek to maximize it, subject to keeping it within its sphere of competence. For example, no logical system can supply information with any meaning or intention.
I believe that the best prospect for achieving semantic interoperability . . . is for the users to agree on using the same common foundation ontology that is open to inclusion of elements from any source, mediated by a technical committee that is also open for membership. The structure of that ontology, and the content of the most basic ('primitive') concept representations will have to be agreed to by the users.
We support your contention. Human beings have already arrived at a huge corpus of shared perceptions, which they label in different languages, dialects and jargons. We contend that, although each culture fine-tunes those perceptions, a kernel is invariant across cultures and languages. That kernel is represented by a canonical ontological dependency schema.
Thus, in the little illustration I gave earlier Society——————person——————marriage the agents responsible for the start and finish of the marriage determine the differences between Protestant, Catholic, Jewish, Islamic, Buddhist, Parsee, Hindu and other forms of marriage.
The law, in its own domain, often spells out the important refinements of meaning and specifies the people whose authority we should accept when assigning operational meanings (the judge, a qualified medical practitioner, a certified engineer of some discipline etc).
Similarly designers of different computer systems, supposedly handling perceptions in overlapping domains of discourse, may well have presupposed different associated authorities. I conjecture that our kind of semantic schema provides a useful bridge between systems but cannot eliminate the finer details of meaning, which will have to be negotiated. The negotiations may yield a precise compromise or only a partial one with a clear understanding of the remaining semantic fuzziness.
The COSMO ontology is open to inclusion of concept representations of interest to anyone, provided that they are not logically contradictory to other elements in the ontology. . . . If a new language is developed that is shown to have some superior practical utility, the COSMO might be ported to that.
But this problem can never be reduced to one in OWL, FOL, CL and solved on paper. I'd like to understand the role COSMO would play.
People cannot be eliminated from this work and you rightly insist on their involvement. We take that to the point of insisting on no jargon in our schemas. One must employ the words of the users to make sure that they understand what we are trying to do on their behalf. Try showing them a page of OWL!
Jeff Schiffel made many valuable points and I'd like to comment on two.
References to norms, affordances, and other elements of organizational
semiotics can be a bit hard to dig out.
Too true! And, when you find them, their meanings are far from fixed. Because all meanings are made by people through discussion and critical examination of each other's work, this can take a long time. Internationally the process is expensive and error prone. I truly welcome the opportunity to obtain from Ontolog participants critical reactions to my ideas.
The key to computability is in semantic normal forms. When natural
language texts (along with the affordances) can be stated in SNF, then
the resulting well-structured format may be stored in your favorite
data- or knowledge base. Then you can do what you like.
Absolutely! Our experience suggest that basing a system design on a schema in the Semantic Normal Form will yield huge benefits whatever techniques you employ beyond that point.
Pat Hayes provided excellent illustrations of Gibson's notion of an affordance and concluded:
As you can see, these have to do with perception, and they have nothing at all to do with culture or behavior. This is typical: Gibson was primarily a perceptual psychologist, after all.
Jeff Schiffel and Pat Hayes discussed the rights and wrongs of adopting Gibson's affordances without making careful reference to its earlier meaning.
I feel no obligation to work in the field of perceptual psychology, having adopted Gibson's idea and perhaps bent it to my purpose. I see little distortion anyway apart from the fact that Gibson presumes an objective reality whereas I wish to understand the mechanisms whereby we arrive at a shared belief in an objective reality.
I'm an engineer and I regard what I have done (hopefully) to solve a difficult problem as no worse than bending a piece of metal from one machine to serve my purpose in another. The test is: Does it work? HCI people are welcome to their interpretation (Paola).
Rob Freeman was supportive:
I quite like your "affordance" based approach to ontology. To me it
emphasizes the subjectivity of category, whether that subjectivity be
personal or social.
Were it possible to find a complete objective ontology this personal
or social aspect would not be so important. That a personal or social
aspect is so important strikes me as further evidence no complete,
objective way of categorizing the world exists. . . . because it should
be clear social affordances might be very numerous . . .
Categories fit into our analysis in cognitive norms. The belief that whales are a sub-category of fish was discarded a while ago. That and other generic/specific structures are quite different from the perceptual norms that hold our ontological knowledge.
The law creates perceptual norms when it adds to the furniture of our social world by starting and finishing the existence of things such as copyrights, for example. Cognitive, evaluative and behavioral norms rest on the foundation of our perceptual norms, whether directly experienced or socially negotiated.
Norms themselves count among the social object we recognize; they are started and finished, in the UK, by the Queen in Parliament. Norms of all kinds remain invariant between their start and finish events – that allows their component sub-norms to be added or deleted.
Then the John Sowa and Patrick Cassidy continue the discussion of interoperability about which I shall say no more for the moment.
Earlier, Ravi Sharma had raised a point about cyclic time appearing to accept the possibility of cosmic time as a cyclic and repeating processes. I should conclude with a note about time in our view.
Our insistence on accepting only the present time rules that out of our models. Indeed, with only the present moment available to us, we can vist neither the start or the finish of any realized affordance. With a person, we can use their name and we can also visit them and shake them by the hand but starts and finishes of things we can only name. Hence we only know times by their names, which we assign using calendars and chronometers and structure using relationships of simultaneity and sequence at a specified location.
Start and finish EVENTS we do not confuse with beginning and ending PROCESSES that we can experience and learn to recognize as potentially leading to certain events as we formulate cognitive norms about causation.
Cyclic times are legion. Institutions define for their own purposes all kinds of taxation periods, charging periocs, semesters, terms. Our schemas are often littered with them.
On Wed, Oct 8, 2008 at 1:18 PM, Conklin, Don <don.conklin@xxxxxxxx> wrote:
_________________________________________________________________ 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 (01)
|<Prev in Thread]||Current Thread||[Next in Thread>|
|Previous by Date:||Re: [ontolog-forum] Axiomatic ontology, Patrick Cassidy|
|Next by Date:||Re: [ontolog-forum] Axiomatic ontology, Christopher Menzel|
|Previous by Thread:||Re: [ontolog-forum] Looking forward at the past, Conklin, Don|
|Next by Thread:||Re: [ontolog-forum] Looking forward at the past, Pat Hayes|
|Indexes:||[Date] [Thread] [Top] [All Lists]|