Re: No universal things Re: comparing XML and RDF data models

Hi Alan

[I suggest we keep this off-line. I think this goes far off the Semantic 
Web list focus.]

Thanks for the lesson :-) . Not that I agree on everything. Some 
attempts to clarify below.

Alan Ruttenberg a écrit :
>> I for one assume very strongly indeed that 'the description of the 
>> thing causes the thing to exist', and my hunch is that Bijan will 
>> follow me on this (at least I hope so). What we deal with in our 
>> information systems are only descriptions (signs). A 'thing' exists 
>> insofar as we have made a description which somehow cuts out the 
>> world representation in some arbitrary way between 'this thing' and 
>> the rest.
>
> There are two things you are saying and they contradict. It is one 
> thing to say that "the description of the thing causes the thing to 
> exist". This evokes the idea that there is an empty vacuum, a thought 
> comes in to your head, and presto the space is now occupied by the 
> thing you thought of.
Sorry if it gives this impression. What I mean actually is that a thing 
does not exist *as a distinct entity* before someone makes the 
distinction. For example when you name a mountain, you arbitrarly wave 
at a region of the world and call it "Monte Perdido" or "Pic sans nom". 
This region was not separated from the rest of the world before you name 
it. I can reasonably bet it existed somehow, but not with an 
'individual' status. Your language or mental activity does not make 
mountains appear from thin air, of course, but they make appear a 
distinction between a mountain and another one. Distinction is not in 
the world, but in description.
>
> It is another thing to say that the world exists but that there is no 
> objective way to decided where the boundary between things are. Rocks, 
> in this view, don't pop in to empty space - it's just that it may be 
> hard for us to see where the boundary of one (existing) rock ends, and 
> the next begins.
Well, yes or no. In what you write here, more or less, do you assume 
that individual rocks exist (have distinct identity) before you care to 
pay attention and speak about them. You can reasonably say that for two 
rocks  you hold in each of your hands, but it's based on the implicit 
notion that identity has something to do with spatial connectivity, and 
space-time boundaries.
>
> Note that the word "thing" is used in two ways in the above two 
> paragraphs. In the first it is meant as a mental structure. Certainly, 
> if you think of something there is something about your your mind that 
> enables that thought to be carried.
>
> In the second paragraph it is used to denote some portion (albeit 
> indeterminate) portion of material.
Well, I see your point but am not sure I agree completelely with this 
distinction. In other words, not sure about the 'reality' of the second 
referent.
>
> It is a common, and damaging, confusion in knowledge representation to 
> unknowingly mix these these two sense up. Ideas pop in to reality 
> while thinking. Rocks don't.
Well, yes and no. Or let's say I'm agnostic about the independent 
existence of rocks when it comes to implementing knowledge in systems. 
You don't implement rocks, only representations (or more excatly formal 
descriptions of representations)
>
> Let's call things in the first sense "concepts" and things in the 
> second sense "individuals".
Let's assume that. But to make sure, how do you fit this vocabulary 
together with owl:Thing and owl:Individual?
>> The world is a continuum, it's not naturally divided into things 
>> which would be given, and  which we could describe as accurately as 
>> possible.
>
> things = "individuals"
Indeed. So let's say in don't believe in pre-defined individuals.
>
>> Things emerge from descriptions, conversation about descriptions
>
> things = "concepts"
Indeed. Since I don't care about individuals, let's say I care only 
about concepts. I assume that.
>
>> , and so do identity and difference.
>
> There is a difference between "perceived difference" and "perceived 
> identity" and "actual different" and "actual identity". All the 
> biologists I know of try to get at the "actual difference" rather than 
> the "perceived difference". Commonly there is initially no "perceived 
> difference" but in experiments we are able to determine "actual 
> difference".
Ha. Here we are at the core of our disagreement, I think. I know a bit 
about what science and experimentation is about, but *you* are the 
scientist. :-)
>
> The difference between an atom of gold and an atom of hydrogen has 
> nothing to do with what you think. Rather the other way around.
Here, do you consider "an atom of gold" as an individual or a concept? 
Given the success of physics, people take for granted that there exists 
such individuals as atoms of gold and hydrogen, so I can't prevent you 
from thinking that in this case. But most "things" we are dealing with 
in our information systems have a much more fuzzy status than those 
"atomic things". I'm working today on a vocabulary for multimedia 
description and indexing. What about "smiling", that I have in this 
vocabulary? What is the "actual difference" between a photograph of 
someone "smiling", and someone "not smiling"? And most things we deal 
with are closer to smiles than atoms.
>> We might at some level of granularity agree we are describing the 
>> 'same' thing, based on the mutual logical consistency of 
>> descriptions, and a mutual agreement on which set of property/value 
>> pairs we consider as "identifying".
>
> It is more than consistency and agreement. It is easy to be 
> consistent: Call everything "blah" and say all things are related by 
> the relation "flah". If this is your conceptual universe, how do you 
> arrive at an inconsistency? Similarly, you and I may identify gold 
> atoms and hydrogen atoms as the same thing. That would make us idiots, 
> not adepts in the process of identifying.
Of course, YES. But we don't start from nothing, right? We don't build 
ontologies from scratch. We have the legacy of unformal, semi-formal, 
more or less formal representations (called generally "data") and try to 
make sense of data. ("Making Sense of Content" has been the tagline of 
Mondeca since its very beginning in 2000. I'm very happy with this 
tagline, all the more so I am absolutely for nothing in it).
>> Then further down the road, drilling down more precisely our 
>> respective descriptions, we can come to a point were those 
>> descriptions are no more consistent, and we then agree we were 
>> speaking about different things indeed. This should not be an issue, 
>> but the basis of day-to-day practice.
>
> The point is that the consistency that we search for, at least in 
> science, is not between ourselves, but between what we predict and 
> what happens in reality. Science is about doing our best to remove the 
> bias that our senses and preconception put in the way of us 
> understanding reality, and instead understanding the mind-independent 
> truth. Are scientists successful at it, always? No. However there is 
> something to fail *at*, and that this is the case is in opposition to 
> the idea that we construct all of reality in our heads.
Indeed, I agree with all that ... except on the "what happens in 
reality". We do not test science models against reality at large, but 
against *interpretation of results of experiments (on reality)*. You 
know how those things are difficult. At some point you have to pass 
through some interagreement about the interpretation of experiments. You 
know pretty well that some currently agreed upon models rely on a small 
number of critical experiments, very difficult to set and to interpret. 
Look at cosmology and particle physics. If next year people at the LHC 
say they have discovered the Higgs boson, who will be able to verify 
that elsewhere?
>
>> I had this viewpoint even before beginning to work in those 
>> technologies, and was quite happy to find out that ten years or so of 
>> work in very various industries and projects had brought me a huge 
>> pile of examples to illustrate and confort it. Not only I don't know 
>> what a universal thing is, but I'm pretty much convinced that this is 
>> a void and useless concept.
>
> Really? So the periodic table is a void and useless concept?
>
No, of course, the periodic table is a very smart model, and a very 
productive one, both in organization of existing knowledge and in its 
predictive power, but it's not what I call a universal thing, in the 
sense Peter seemed to indicate anyway. Like some universal notion of 
"smile" or "document" or "person" ...
>> It's been counter-productive in science for centuries. Physics had to 
>> go over the notion of universal thing to understand that light is 
>> neither a wave, nor a particle.
>
> This is a separate issue. Physicists understand that there are 
> *different* things than waves and particles, namely things that have 
> properties of both. This is just as universal - just corrected from 
> previous understanding in the light of new knowledge. And while you 
> may lament the centuries that passed until this insight was achieved 
> you fail to note the millions of people and tasks that are daily 
> successfully accomplished not because people are good at agreeing at 
> arbitrary ideas, but because the understanding that science has 
> achieved make predictions possible. Those predictions are not enabled 
> by agreements between people, they are enabled by agreement between 
> what is thought and what is.
OK (modulo the above remark about the fact that we test theories by 
interpretation of partial experiments etc ...). Let me make a point 
here, to go back to the original debate, which is about models for 
implementation in our systems. There is a slight difference between an 
ontology as an engineering artefact built up to be the backbone of an 
information system, and a theoretical model of the world, like quantum 
mechanics or molecular biology. Despite my scientific background, I'm 
more used those days to deal with the former.
>
>> Biology to go over the notion of taxa as rigid concepts based on 
>> phenotypes to understand genetics etc.
>
> There is much more to the evolution of biology, and to the difference 
> between understanding based on taxa and understanding based on taxa, 
> than a stubborn insistence of the notion of "universal thing". I hope 
> you understand how simplistic this comes across.
Indeed. Bad exemple. Writing it, I was sure someone would point at it as 
simplistic. :-)
>
>> Many examples can be found in all science domains.
>> My day-to-day experience in ontology building, listening to domain 
>> experts, is indeed not that 'there are things that people are trying 
>> to describe',
>
> Things = "individuals". The results of that description effort: Things 
> = "concepts".
OK
>> but that 'there are descriptions people take for granted they 
>> represent things before you ask, but really don't know exactly what 
>> those things are when you make them look closely'. When you play the 
>> game, simply asking questions such as "in which way X is different of 
>> Y", or "what does this mean exactly", you always bring domain experts 
>> to acknowledge this.
>
> What are they acknowledging exactly? What you are observing is not 
> something about the nature of reality, it is something about the 
> nature of how people think, what they bring to the forefront in 
> consciousness, etc. Do you think you are teaching scientists something 
> about science by asking them questions? Or revealing more clearly what 
> they know?
Well, maybe, sometimes. Scientists are "domain experts" as others. And 
often specialists which have the nose so close to their domain they 
happen to take a lot of concepts for granted. Like the "nature of 
reality" :-)
>
>> And all our job is not to figure out 'what the domain things really 
>> are' and 'describe them properly', but to bring people to acknowledge 
>> the arbitrary nature of their descriptions, the fundamental 
>> impossibility to capture their referent in an exhaustive way, hence 
>> the necessity of consensus at a certain level of complexity and 
>> granularity, on which to build systems that work.
>
> That systems (or at least, shall we say, biological experiments) work 
> is not fundamentally due us agreeing about something. Biological 
> experiments work when there is something out there - the cells that 
> work in certain specific ways, for example, and we achieve a way of 
> exposing that mechanism
Well, I'm not sure I follow you there. You are a biologist, right? 
Strangely enough, the 20th century has shaken the basis of "realism" in 
physics with singularly quantum mechanics. But this very important 
conceptual revolution has not (yet) been transferred to other sciences. 
Singularly biologists seem to deal with cells and cellular components as 
physicists dealt with atoms in the pre-quantic area ...
>
>> The world does not need to be consistent and logically organized, but 
>> our systems need to be.
>
> The world does not need to be anything. It just is. It is not the sort 
> of thing for which the predicates "consistent" or "logically 
> organized" apply. 
I vehemently agree here!
> Those predicates apply to what you call descriptions, or others call 
> concepts, or ideas. 
Indeed. This is *exactly* my point!
> In order to do ontology in the life sciences I have found that 
> understanding this distinction at the outset helps avoid a lot of 
> churn and poor results. For one, it help avoid type errors such as the 
> ones you are making here.
Hmm. I thought that what I said was meant to stress this distinction. 
Hope I clarified a bit ... :-(

Bernard

<http://mondeca.wordpress.com/>

Received on Thursday, 3 July 2008 13:58:27 UTC