[seek-kr-sms] Re: Thoughts on GUIDs - ontologies

[Nico M. Franz]

Hi Shawn:

    that's a nice little paper for an outsider to look at. I'm writing this 
mainly to give the "ontologists" in SEEK a bit of an idea about biological 
systematics. It's tempting to think about how taxonomic concepts ought to 
look like so that computers can understand a lot about them.

    Though I must say I started stumbling already on page 1, when the 
authors started talking about more, or less rigid essences.

    I suppose you can look at most comprehensive and internally consistent 
biological classifications as ontologies. In the Linnean system, these tend 
to be hierarchical (ranks), and higher taxa subsume lower ones. A genus 
includes (necessarily) at least one species. The hierarchy is based on 
perceived similarities and differences in traits. These can range from DNA 
base pairs to bird songs. So far so good.

    Systematists could observe all kinds of similarities, e.g. whether an 
ant was collected by Shawn Bowers on the roof on the SDSC (check: yes), or 
not (check: no). What they strive to observe specifically to figure out the 
natural (evolutionary, phylogenetic) relationships among organisms, is 
called special similarity, or homology. Homology is similarity due to 
common ancestry. It's usually (i.e. in systematics) applied to traits fixed 
among species and at higher levels, not among populations whose trait 
frequencies still vary due to continuous interbreeding. Example of a good 
homology: the wings of bats and birds are homologous AS tetrapod (these are 
land vertebrates) fore arms.

    Now, it seems to me that the essence/rigidity notions are (not yet?) 
sophisticated enough to describe what systematists do. First of all, 
homologies are inherently non-rigid. Nearly all traits vary somewhat among 
individuals pertaining to a species. They'll quite often vary considerably 
among species pertaining to a genus that is nevertheless said to have a 
particular trait. This is a feature of traits like wings or feathers being 
brought forth by thousands of DNA bases and being controlled by multiple 
genes, all of which are potentially subject to modification. So the 
"essential non-rigidity" (pun intended) goes all the way down. We have lots 
of reasons to believe it's real. At a larger time scale, we can talk of 
birds being dinosaurs and the birds' feathers being homologous to the 
dinosaurs' reptile-like scales. Meaning that there's a historical, 
evolutionary identity that one might reasonably propose, in spite of the 
fact that there has been so much transformation that a lot of the DNA 
responsible for producing bird feathers no longer looks like that which 
once brought forth dinosaur scales. So the "is homologous to" can be a 
fairly sophisticated way of saying "is a". And scientists can have 
alternative, almost equally plausible solutions to this, backed up by 
varying amounts of relevant yet conflicting observations.

    Which leads me to the second point. Even though the properties of a 
taxon, in the evolutionary (homologous) sense, are in some way necessary, 
they're not necessarily obvious at the moment that taxon is first named and 
classified. That can be the case regardless of whether one gets the 
classification right or wrong. So one could correctly recognize and name 
what ultimately turns out to be a valid genus of ants, yet get most of the 
features that make it a natural evolutionary entity wrong. Or just not 
mention them in a 1758 Latin publication. In modern days, many systematists 
probably consider the naming/classifying business as a final, almost 
trivial step. The discussions really revolve around what's the right/best 
kind of evidence that a taxon is natural or not. For example: flies have a 
set of highly modified hind wings (called halteres) and so do the fore 
wings (!) of an enigmatic lineage of parasitic insects (Strepsiptera in 
Italian). Whether these are homologous or not depends on whether flies and 
these parasites are one or two independent lineages. Lots of genes are 
being sequenced to figure this out, and the morphology is being reanalyzed 
too. There's still no convincing solution. People also try to see whether 
developmentally it's possible to have a mutation that produces halteres on 
another body segment (fore to hind wings). Etc. Once it's all said and 
done, those parasites may truly have "halteres" (not just things that look 
like them), and also be classified in a new way. So essences can be 
necessary (must be there always) and contingent (we won't see them until 
very late) at the same time.

    I wrote this partly because so far I've had this hunch that CS/AI 
ontologists (what's a proper name? OntoClean sounds like a mouthwash, I'm 
sure it'll be synonymized soon) draw most of their examples from 
classification where humans STIPULATE properties. In systematics, *we* 
slowly try to work them out. Essences in taxonomy are ontologically soft 
(due to evolution), and in any case hard to figure out. Yet they're the 
backbone for building the hierarchies and naming taxa. I don't think that's 
impossible to incorporate in ontologies, though *I* haven't seen it yet. 
Has it been done? One way might be to represent properties 
probabilistically. "Is a, with a chance of 80%, according to study X of 
person Y." Maybe this makes no sense (yet)...

    BTW, various of the other topics dealt with in the paper come up in 
systematics too. And philosophers of science have wondered for 50 years 
whether species are "classes or individuals". The really seem to be both.

Cheers,

Nico



At 12:38 PM 5/25/2004 -0700, Shawn Bowers wrote:

>Beach, James H wrote:
snip

>>If the key itself has information then you will inevitably run into a 
>>situation where the key will need to be changed because something about 
>>the information represented by the key value has changed or is in doubt 
>>or is a matter of interpretation, (thus losing the temporal uniqueness of 
>>the GUID).
>
>Again, then the information used as the key isn't really "identifying" 
>information, and you have a problem anyway.
>
>There is a very interesting article that people may want to read 
>concerning properties of things and classification, including identity and 
>unity, that may be relevant to what taxon is trying to accomplish with 
>concepts.
>
>The paper can be found here, and was published in the Communications of 
>the ACM in 2002.  There are longer, more detailed versions available, but 
>this is a good primer.
>
>http://www.loa-cnr.it/Papers/CACM2002.pdf
>
>snip 2
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mercury.nceas.ucsb.edu/ecoinformatics/pipermail/seek-kr-sms/attachments/20040602/0608e8fb/attachment.htm