|
Author
|
Topic: Function, Specificity, and Multiple Realizability
|
Micah Sparacio
Member
Member # 6
|
posted 10. May 2002 16:51
I was recently asked a question regarding the association of specificity and function. The questioner indicated that he understood the independent pattern involved in the bacterial flagellum. Humans had invented a motility device in much the same way decades before this nanomachine was discovered. So the detachable pattern was the human designed motility system (rotory motor). However, this person said that he certainly hopes that the detachable pattern involved in determining the specificty of function would not always demand such a nice sequence of events (humans design an object which is later found out to already exist in the molecular world).
Here are some of my thoughts:
First of all, function does not entail a design inference. A rock that holds open a door performs a function but does not warrant my inference to design. Obviously, the functioning system must also be complex (both Dembski and Behe address this).
When we come accross a functioning biochemical system we'll always need consider both its function and those parts involved in performing the function. The function, we can say, is multiply realizable. In other words, function is not determined by the material in which it is realized. I think that in this sense, function can be said to be detachable. It is not dependent on the particular system in which it has been situated. The concept of multiple realizability is what drives the Artificial Intelligence community. After all, if functional intelligence was limited to grey matter and could not be realizable in silico, then the AI enterprise would be doomed. Then why do we persist? Because there are features of intelligence which we believe are not dependent on grey matter (though some features are believed to be dependent).
So, my point is that if funciton is multiply realizable, then it can be detached from any particular implementation.
After function has been determined as the specification, we must then consider how the function is implemented. This is where complexity comes in. If the implementation of the function is a complex system of interacting parts that meet a certain level of complexity measurement, then we can be warranted in questioning whether the localized(blind) watchmaker had much, if anything, to do with constructing the current implementation.
To consider the complexity of the implementation, I've heard several proposals: you can start at the level of nucleic acid sequences, move up to amino acid sequences, and finally consider protein parts. From what I know, research is being done at each of these levels.
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 11. May 2002 09:15
Micah:
A couple of real quick points.
First, I think that function does necessarily imply intelligence (or "design," if you prefer). Your example of the rock holding the door is no function at all, unless some human being has placed the rock there on purpose.
Second, you assume multiple realizability is the correct analysis of function (along with almost everybody else, of course), but I disagree. Why do people continue to pursue AI? Well, why do people continue to pursue Darwinism? People take time to get the message---what can I say?
AI IS doomed, by the way. Even some of its most fervent supporters (e.g., Jaron Lanier) have now admitted as much. True, the emphasis has moved from the symbol-processing system theory to the dynamical systems theory, which is a huge step in the right direction, but that, too, is a doomed effort, IMHO.
What do I mean by that? I certainly do not deny that we can obtain valuable insights by running AL simulations and tinkering with Rodney Brooks's robots. But Kismet is a simulation of intelligence, not the real thing, for one simple reason: the matter out of which Kismet is made has no INTRINSIC tendency to maintain itself in its functional state. It does so purely because of boundary conditions we have imposed. In other words, there is no functional state so far as Kismet itself is concerned. All the functionality is with respect to us.
Organisms are different. They have their own intrinsic, spontaneous functionality in the sense of cognitively guided, purposive, goal-seeking behavior. Organisms actively strive toward their goals (both instrumental and ultimate---staying alive). The only way to explain this is by looking to the kind of matter out of which they are made, I am convinved. If I am right, then the multiple realizability thesis is simply mistaken.
I would go further and claim that the multiple realizability thesis is at the root of all our troubles in our struggle to understand life and mind. It is really just the assumption that organisms are machines, in disguise. But organisms are not machines, and so, while multiple realizability does apply to artificial systems that are only functional in relation to our purposes, it does not apply to organisms that have purposes of their own.
[ 11 May 2002, 09:18: Message edited by: James A. Barham ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 11. May 2002 11:00
Dear James,
I was not trying to imply that organisms as a whole be considered multiply realizable. I was simply trying to argue that function, especially the function of individual enzymes and various molecular machines, is detachable in the sense that Dembski's specificity requires: the function of the bacterial flagella can be considered and defined independently of the particular instantiation in bacteria. Motility can be considered indpendently of the paricular implementation. The particular implementation is what one considers when determining complexity.
In terms of AI, notice I intended to reflect that fact that there are certain components of intelligence (computation, rule following, etc.) that do seem conducive to duplicating via AI. For example, summing the number 1 through 100 can be performed both by the grey matter in our brains and in silico by a computer. This particular function is multiply realizable. I also indicated that there are certain elements of human intelligence that do seem to depend on their biological medium.
So, let me clarify my position. First, function can be considered independently of its implementation. Whether there is only one instantiation of a particular function in the known universe is irrelevant. The function of hammering nails is independent of the object you use to hammer nails. The function of motility is detachable from any system that performs this function. If we can consider a possible world in which function is implemented in multiple ways, then I think we can say the function is independent of the physical implementation.
Second, function is not alone sufficient for a design inference (I think James disagrees with me here). If a rock rolls down a hill and rolls on top of a leaf, the rock is performing a function. It is acting as a sort of paper weight on the leaf. Function is not necessarily tied to purpose or intention.
Finally, my use of multiple realizability is only to clarify the point that function is not tied to its implementation. Perhaps there is only one way to implement a certain function in our universe. My point is only that there are some functions that can be implemented in multiple ways. Therefore, the function is not dependent on any one implementation. I think this can be extrapolated to all function.
Consider the mousetrap. We can have both a single component mousetrap (a glue trap) and a multipe component mousetrap (the traditional model). The general function performed by either mousetrap can be considered to be the same. The function of catching mice is multiply realizable and therefor independent of either realization.
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 11. May 2002 22:44
Micah:
Thanks for your detailed response.
I think we have a very important disagreement that is absolutely fundamental to everything we are about here at ISCID, so I would like to pursue it a little further. However, I am not quite sure how to proceed to gain clarity in this difficult and vexed area of inquiry.
One idea that occurs to me that may be underlying our difference of opinion is the distinction between a function and a simulation. I certainly agree that many aspects of concrete biological systems can be profitably modeled in various ways. A model operates by emphasizing specific elements of a system, while ignoring others---i.e., by abstraction. In a sense, all science works by simplification via abstraction, and any future science of teleology will have to do the same, in a broad sense.
But it is one thing to abstract mechanistic elements out of the operation of a teleological system; it is something else to try to model the teleological aspect in itself. The latter simply cannot be done using a linear, mechanistic model or algorithm, it seems to me, but can only be done via a nonlinear or "qualitative" model, I believe. Any exact or analytical or integrable model is going to inevitably lose the telic aspect, however insightful in may be in other respects.
Now, when you speak of the multiple realizability of functions in organisms, I think maybe you have in mind something like different ways of analyzing, modeling, or simulating them.But, while important, this must be distinguished from the claim that a concrete teleological function can be physically instantiated in fundamentally different kinds of material systems. The latter claim is what most philosophers and cognitive scientists have in mind when they speak of "multiple realizability," if I am not mistaken. I agree, of course, about the scientific utility of simulations, but disagree that this utility implies true multiple realizability in the second, ontological sense.
Also, I must ask you: Why would you want to say that the case of the rock rolling against a door or a leaf or whatever constitutes a "function"? This would not be true, either in common speech, or in any philosophical or scientific meaning of the term "function" that I am familiar with. Could you explain a little more what you mean? If such a case, considered intrinsically and objectively, without regard to any exernal agent, constitutes a "function," then any physical process whatsoever must be a "function," it seems to me. In that case, the term has lost all meaning.
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 11. May 2002 23:32
I have to agree with James about rocks and function. I don't see how a rolling rock that happens to land on a leaf is performing a function.
On the other hand, I think the concept of a molecular machine itself is detachable in the sense that Dembski's specificity requires.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 12. May 2002 02:23
Perhaps I'm misguided, but my working definition of function is this:
playing a causal role in a particular domain or context.
Does the rock on top of the leaf not play a role in keeping the leaf from blowing around? And could not some other object play the same role?
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 12. May 2002 21:28
Micah:
Karl Popper always used to say that we shouldn't quarrel about words. On the other hand, we must reach a basic level of agreement about the meanings of words if we are not to simply talk past each other.
My Webster's New Collegiate Dictionary, 9th ed., states as the first def. of "function" a professional position or occupation. I think we can agree that that is not what we are talking about. The second def. is "the action for which a person or thing is specially fitted or used or for which a thing exists: Purpose." I believe that this is what is commonly intended in all biological, social scientific, and philosophical contexts. My Penguin Dictionary of Biology states that "the function of a component in an organism is the contribution it makes to that organism's fitness."
In short, the notion of appropriateness or suitability in relation to a value, goal, or purpose is the crux of "fucntion." To use the word for your leaf example would only make sense if an external cognitive agent wanted the leaf to be held down for some reason.
Now, the Darwinians who wrote the def. in the Penguin Dictionary of Biology felt called upon to drag in "fitness" precisely because they are embarrassed by the teleological nature of the concept of "function," and so want to try to "reduce" it to mechanism by means of the theory of natural selection. I personally believe that "fitness" is either tautological or question-begging, depending on whether it is given a population-biological or an ecological-engineering interpretation, and so this gambit fails. But regardless of where one stands on the success of Darwinism in reducing the teleological aspect of functions to mechanism, no one---not even a Darwinian, I think---would seriously dispute the fact that functions are essentially attributes of living things (or perhaps of some complex, cybernetic feedback systems), but not of ordinary inanimate systems or processes.
I think you will find that this is the way the term "function" is used, for example, throughout the excellent collection "Nature's Purposes: Analyses of Funcion and Design in Biology," ed. by Colin Allen, Marc Bekoff, and George Lauder, MIT Press, 1998.
IP: Logged
|
|
|
Printer-friendly view of this topic