Dear Bob:

I think that your "metapoint" as I understand it, i.e. (1) that different levels of semantic resolution will be necessary and/or sufficient for a particular task, and (2) that considerations of specific use cases should take into account how much resolution is needed, is solid. The argument from example can work the other way as well, however:

http://onlinelibrary.wiley.com/doi/10.1046/j.1523-1739.1999.013002427.x/pdf

I also think that your anecdote touches on another subject that might need more attention. Here's a set of related (and similarly self-exploring) quotes from my "Letter to Linnaeus":

"We’re at a juncture in systematics when more precise phylogenetic estimates are published at an increasing rate. There is a concomitant trend to archive the results in networked repositories intended to serve as the primary ‘hubs’ for systematic information. Both the systematic and the computer science community seem to have bought into this vision. However it is likely that each community underestimates just how much we need to adjust our linguistic habits in order to achieve long-term integration of systematic products. Computer scientists use a formal language (description logic) to build highly structured networks (ontologies) that may include classes, instances, parts, properties, relationships, and other components and qualifiers. Once the structure is in place then powerful algorithms can ‘reason’ about the constituent elements, connect them to other ontologies created for related subject areas, and so on.

As computer scientists learn about systematics they must initially see a strong match between an ontology and a published taxonomy. However, as we’ve seen, a classification is never entirely comprehensible in isolation, and instead represents a complex mosaic of previous and new elements with implicit identities and relationships to each other. Too often such expert-made classifications are only comprehensible to other expert speakers, i.e., persons who share an intimate understanding of the contextuality of the new system and are thus able to make explicit the implicit semantic links to previous systems." [...]

"Why have systematists relied so much on painstakingly acquired, implicit assumptions about the taxonomic history of particular groups when presenting their new classifications? I believe the reason is neither some form of elitism (“take that, users!”) nor a lack of self-esteem (“who wants to read about all these subtle similarities and differences?”). More likely, it’s simply human habit—we make things just as explicit as we think is needed at the moment—paired with the similarly human notion that the latest perspective is really the one that’s going to last for a long time, in spite of all historical evidence to the contrary. And so we pass the burden of full semantic resolution, both looking backward and forward, on to future specialists." [...]

"However, the Linnaean system is not capable of capturing the entirety of semantic adjustments that occur when a previous classification is revised in light of new evidence. [...] Instead of abandoning the Linnaean system, this observation should lead us to express more clearly and more consistently what we mean when presenting a new classification. [...] At the human level, this requires that we routinely acknowledge the ephemerality of our latest insights, spend more time comparing our perspective to a previous one that we no longer think holds true, and generally pay more attention to the context in which we use taxonomic names. [...] If we supplement the Linnaean system with these conventions, there will be more linguistic transparency and less mistaken urgency to purge the idiosyncrasies of the past or legislate a wrong consensus."

http://academic.uprm.edu/~franz/publications/LetterLinnaeus.pdf

---------------

Few additional comments:

The view that taxonomic concepts represent hypotheses about how certain names, types, and descriptions relate to perceived entities in nature is by no means incompatible with your point that people understand each other sufficiently well in many particular situations using even crude shorthands for names and concepts (euphorb versus cactus). The reliability of the notion that those two lineages are phylogenetically distinct is in the ballpark of that of the law of gravity (well, actually I am rather foolishly relying on the veracity of your judgment of Dr. Thiele's expertise...making all kinds of ancillary assumptions that undermine deductive reasoning, sorry Prof. Popper). So yes, we are advancing.

But, isn't the point of some of the most critical use cases (e..g., the EEA one), not just to properly spell names, but to load up the "system" (ontologies, databases, metadata annotations, what have you) with some degree of specific taxonomic insight? If and when so then we shouldn't assume that matters of contextuality are going to be largely insignificant, and instead at some level will have to "teach" the system that contextuality.

Binomials and informal names are shorthands that can hold the water in most casual conversations among humans, especially if and when the involved speakers share a similar scientific and even taxon-specific training (in addition to all the other semantic and inferential expertise they share just by having been born into and raised in society; see Quine). I do feel, however, that our reluctance (if there actually is one) to go deeper with ontological representations, is neither necessarily due to an obvious limitation of computers - that remains to be shown - nor is it the most prudent way to move ahead.

Nico Franz

On 11/20/2010 5:45 PM, Bob Morris wrote:

What puzzles me about the highly taxonomically technical parts of these threads is not that the codes of nomenclature seem difficult to parse in the sense of formal languages---that's true of lots of human-produced legislation. It is that in 15 years of hanging out with biologists, I have rarely heard them use anything other than binomials in conversation about anything other than whether binomials are adequate. Why, I wonder, are they not utterly confused during all those other conversations, and if they are, does that mean that conversations about biological topics can not advance biology? (This seems unlikely to me, else why do they keep doing it?). Does it mean that "only" hypotheses can come out of these discussions, but that support for hypotheses can only come from data that is rigorously tied delicate name formalisms? It is hard to believe that only hypotheses can be the subject of these conversations, except for the position that everything in science is "only" hypothesis. But maybe when the amateurs leave the room, they suddenly start talking in more code-compliant names.

There are plenty of use cases--and successful information systems---that don't depend on rigorous names. Some aspects of morphology form a simple example. For some uses, it is not a problem to illustrate what a sepal is with several images of different taxa which are either not named, inadequately named, or even incorrectly named. Furthermore, this wouldn't change if those images were fetched from a database in which it is impossible to decide which of those name defects is in play, e.g. one in which there is nothing other than binomials as names.

Another example I was personally party to was this conversation, from memory, that I was party to in Morocco a few years ago:

Bob Morris: Ooh, that's a beautiful cactus. Kevin Thiele: It's a Euphorb, not a cactus. There are no cacti here. Bob: Why does it look like a cactus? Kevin: It's pretty much the most successful way to deal with very dry environments. But they are pretty distant from an phylogenetic point of view.

Since most of the listeners were biologists, I imagine I was the only one this was news to. But what I don't believe is that some of the party had a radically different understanding of the conversation than I did.

So, the importance of code-compliant names not withstanding, I would find it very interesting to see a resource devoted to use cases and competency questions that are independent of them, along with accompanying "not fit for use X" annotations. Sort of like warnings on pharmaceutals.

Bob Morris

[...]