I love to run when it's forty-five degrees and rainy. It is my absolute favorite weather. I love it because it's just wonderfully silent and rhythmic--I never see anyone on these days-- it's just me, the rain, and the sound of my breathing-- slow and methodic. I thought this morning: I am so blessed; I am going to spend the next four years of my life in a place where it will be like this EVERY DAY (+ mountains).
But that's not what I am really going to talk about. You see, as I was enjoying this weather, I was thinking about Clemson. One of the things I really love about the program here is how intuitively we are trained to work the forest. Labs often involve going out into the woods and looking at the trees and ground and trying to tell the forest's "story." What did it look like in 1600? In 1840? 1920? If you can't tell the story of the forest, you can't work with it effectively. You have to know the forest like you know your family-- you know its history, its composition, it's behavior. It's funny-- the best advice I got from a professor here about how to improve my forestry was "walk around in the woods more." What a tough life :) . This morning, I wondered-- with all the precision agriculture techniques and algorithms driving them out there, would it ever be possible to simulate the absolute intuition of an experienced forester? It's sort of like Gary Kasparov's situation. But, I am not so sure that a real forester's experience could ever be trumped by a model (I would say "a constrained optimization" but that implies that there is a set of definite goals, which is not necessarily true in real situations). I needed an analogous way to assess this situation.
This made me think of literary criticism methods.
In the early 20th century, literary critics (and art critics) began to really push the limit of "what is a work?" They sought to break the boundary between something that is art and something that is real. To break that boundary, they had to define it. And they found that definition was significantly more complicated than expected. (If you read "Pale Fire" or "House of Leaves," you'll see this in full force). Of course, people have been looking at this for a long time, but there are three critical movements that I think really fostered ways we think about "what is a work?"
The first is the deconstructivist movement. Descartes. The idea behind deconstructive criticism is to look at juxtapositions of literary constructs in a work and examine symbolic fallacies. Diction, for example, you might look at certain noun-noun or noun-verb or verb-adjective (they don't have to just be dualistic combinations, but that's the easiest to describe) and think of all the possible historic and social symbolic references for each term (for example, blood might invoke "red," and "war," and "love," and "water," etc.) and then you look at the references for another term and you compare them and try to find combinations that "fall apart," which can basically be described as combinations that make you say "why would anyone write about that?" You go through the whole "work" this way. You can also deconstruct syntax, plot, character specifications, etc. The idea is that by deconstructing the literature you find the fallacies in the literature that prevent it from being "real."
The second movement is the... and I am somewhat hesitant to use this term loosely, but I will, "constructivist" movement. I am going to allude to an author here who is not exactly a typical constructivist, but I think he does a good job of describing the idea. Donald Davidson's poem "The Ninth Part of Speech" describes how a literary work is composed of eight parts of speech combined in such a way that different actions are "run" within the work. In a way, it is sort like thinking of literature as a computer program. "Yes! (You) get the blue ball quickly" combines interjection + subject (noun) + verb (act) + adjective, adjective (specification) + adverb (specification to verb). That sentence "constructs" a very specific scenario. I cannot go get the blue ball. You cannot go get the red ball. You cannot go get the blue cow. You cannot go get the blue ball slowly. Etc. Every construct destroys all the other possible constructs. However, that construction still does not exactly mimic reality. Davidson suggests that a "ninth part of speech," something that can be akin to, I guess, an "intuition" or a "soul"-- something greater than the sum of constructs in the context of all the impossible constructs-- cannot be written, and therefore no matter what constructions are employed, no literary work can ever have the dimensionality of reality.
The third movement is the vorticist movement. I don't think many literary professors would be happy that I mention this here, but I think it is the most superior method for trying to unite a work and reality. A vorticist constructs a "web" of meaning. I actually thought of vorticists yesterday when I looked at graph theory-- just like coloring the vertices of graphs in certain manners can create an impression for the whole graph (and there is a distinct theory behind how the vertices should be colored in order to avoid repetition), a vorticist's web (or vortex) of meaning combines parts of speech / their relevant symbols in such a way that the literary work creates on overall meaning greater than the sum of its parts. It is therefore considered to be a "vortex" with symbols, words, images, etc. on the periphery that sucks in the reader or viewer. Because the reader or viewer is required to participate in the full work in order to elucidate this deeper meaning, it is therefore a powerful method of breaking the barrier between work and reality.
So, what does this have to do with foresters? I think, in a sense, that the deconstructive method is very similar to a forester in a forest. He or she has the intuition to look at problems on a microscopic scale and construct generalizations about the forest, and then to go back and assess these generalizations on the microscopic scale. I think of precision agriculture as being similar to constructivist method, where different combinations and constraints can be put together in order to seek the optimal "solution" given landowner objectives. I wonder how the idea of the vorticists could be used as an analogy to forestry... is there some way that an intuitive, evolving model could be constructed... a model that is greater than the sum of its parts?... a model so good that, dare I say it, it might be a perfect representation of reality?
I'm not sure how I feel about this. Ethically, it feels unstable. Maybe it's just me, but I like thinking that the "woo-whee'ers" of the world will always have just that little edge on what I can do with my models.
That is an interesting analogy. I didn't realize that was the point of deconstruction! So, if you put a nonfiction story through this kind of analysis, would this "look for combinations that fall apart" technique allow you to conclude that the story was a real account? That could be pretty useful, to verify that "nonfiction" books are actually nonfiction... (I remember there being some ruckus over "A Million Little Pieces," although they debunked that account based on checking for police reports of crimes the author claimed to have committed.)
ReplyDeleteHave you ever learned about neural networks? I wonder if that's not a bit similar to the "vortex" you mention. The idea is that, rather than trying to program pattern-recognition into a computer by a brute-force, top-down approach (an endless series of "if" statements), you construct several layers of connected "software neurons." The layers (which can be written as vectors) have weighted connections between them (which can be written as a "connectivity matrix") that determine whether the neuron will "fire," which in turn determines the network's output. To train the network to recognize patterns, define a cost function, which allows the neural net to adjust its internal weights, until the network's output is close enough to the actual answer. (If you're interested, I wrote sort of a bare-bones C++ neural net, and posted the code in this post.)
Hmmm, I have never tried this with non-fiction. That is an interesting idea, and I remember the scandal with "Million Little Pieces"-- that was the one about the guy who was in jail, only he wasn't in jail, or something? I am not sure if anyone has deconstructed non-fiction before, but it seems sort of similar to "profiling;" you know, how they know someone is lying if the body language does not agree with the verbal language. Good thought!
ReplyDeleteI have never learned about neural networks. That is very similar to the idea that I have, only I do not have a knowledge of what the actual answer would be. Looking at your system, I think that the same principle could ultimately work for a controlled forest situation and actually would probably be pretty useful if you were looking for something like dynamic optimization from a mixed hardwood stand (something that is very hard to model currently because of hardwood interaction patterns). That is actually a pretty big need in our field right now; appraisers up in the NE are always on their toes trying to value mixed hw's; with mixed hw's there is always the question of what is more optimal: standing timber (aesthetic value) or wood timber, and although appraisers can normally tell just by looking at it which is going to be more valueable for a site, having it in math would be very cool.
I still wonder, though... if you pitted an awesome system against a silviculturalist on two separate plots with the goal of "maximize value"-- would the system win?