|
Author
|
Topic: Joshua A. Smart: On the Application of Irreducible Complexity
|
GP
Member
Member # 570
|
posted 13. June 2003 02:12
quote:
Why this fundamental difference?
Mike,
This will be my last post, per forum regulation. But it will be short, since your question is answered in my post above. Think about it this way -- what else can the flagellar components evolve? If your answer is nothing, then there is a huge difference. If not, I'd like to hear what they can evolve.
<I deleted a rather intemperate response. My frustration has once again temporarily surpassed my curiosity. Apologies.>
[ 13. June 2003, 02:24: Message edited by: GP ]
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 13. June 2003 02:44
I'm afraid I'm going to plead laziness with Nelson's responses. There are so many interrelated errors that I'd have to write a paragraph for each line. I realize that this is a bad way to conduct a discussion, but given how pressed I am for time, the signal-to-noise ratio's gotten too low for me to invest time in it.
Mike points out many differences between Avida and biology. I can point out many more. However, in so doing he seems to overlook the key similarities between the Avida results and claims about IC in biology--although maybe he is granting those points?
In any case, Avida instructions are identical and do not contain localization information, while protein subunits exist in much greater diversity and do contain localization information. As such, noting that flagellar components are not re-used all over the flagellum is uninformative except to note that there is a difference between Avida and biology, which we knew already. The key question is: what is the evidence that a system with large numbers of cross-reactive self-localizing components will behave differently under selection that a system with small numbers of brittle externally-localized components?
I don't have the answer to that question. We do have an answer to the question of whether complex systems of interacting parts can develop IC components under selection: the answer is "yes". Selection is massively helpful. If IC is to hold for biology, then either (1) selection must not be helpful for biological systems, or (2) selection may be as helpful, but the problem is so difficult that the help still isn't enough to get anything.
All the stuff about 3D vs. computer code is interesting, but how are we to evaluate whether perturbing 3D structures is more or less likely to yield something adaptive or co-optable than perturbing computer code? 3D structures have the advantage of being more tolerant of small errors than assembly language programs; but assembly language programs have the advantage of having powerful general capabilities implemented with few components, whereas 3D structures have much less general computational power. Certainly, 3D structures must be fairly well matched at interfaces; but so must subsections of a program, since without the proper state output by one subsection of the program, the next subsection will perform the wrong calculation.
There are various other probable misconceptions arising from the experimental conditions in Avida vs. this planet (e.g. when comparing the results of multiple Avida runs to one run through Earth's history), but I'll leave those for someone else.
IP: Logged
|
|
|
|
Jerry D. Bauer
Member
Member # 756
|
posted 13. June 2003 15:02
"With the EQU, we have 26 different parts sitting in front of us. You can arrange them into at least 23 different assemblies to get the same function. With the flagellum, we have 20 different parts sitting in front of us. Can we rearrange those parts into a few dozen different assemblies to get the same function? No."
Jerry: There's the difference and the reason I believe that complexity and specificity must be calculated separately. A flagellum may be as complex as the EQU; but its specificity is light-years higher.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 13. June 2003 18:00
"specificity" isn't well-defined. Do you mean specificational resources, or the inverse thereof? Does "higher specificity" mean that something is easier or harder to specify?
IP: Logged
|
|
Jerry D. Bauer
Member
Member # 756
|
posted 13. June 2003 22:45
Well, I think you’re right Rex, that specificity needs more refinement in definition. But there are two separate bananas in CSI the way I see it and Mike just pointed out the problem.
Complexity is numbers and specificity is the way those numbers are arranged. In my view one can be high and one can be low. So this is the way I understand it and someone can correct me if I’m off base.
D. E. Berlyne defines complexity as “a pattern can be considered more complex the larger the number of independently selected elements it contains.”
So is a big bag of marbles more complex than a small bag of marbles? It is if we want to look at a formula like Cs = log2(W). But is a larger number of marbles more specified? I don’t think so because specificity is a different deal altogether.
I think specificity is really ‘design’ IMHO. Consider this analogy I picked up somewhere, I believe from Dembski’s older stuff. Anyhow I embellished it a bit:
If we took an archer and made him stand a hundred yards from a huge wall, say the wall of a football stadium and blindfolded him; then asked him to hit the wall with an arrow—we wouldn’t be surprised if he did, because the wall is so large the odds are good he will hit it.
Next I’m going to paint the walls of this stadium in small black and white squares in a ‘checkerboard’ pattern; blind fold him and ask him to hit a white square. I still wouldn’t be surprised if he did because, providing he hits the wall, he has a 50/50 chance of hitting a white square.
Now suppose I begin to paint this wall into an ever increasing number of colored squares. I will do a 4 squares wall, an 8 squares wall, a 32 squares wall, etc. with each square being a different color.
I will eventually come to a point that the odds of him hitting the color I ask him to hit will become so astronomically large against him that it becomes overwhelmingly unlikely he will hit it.
Finally, I’ll try this experiment with the archer. I will blindfold him, find a spot on that huge wall and paint a tiny one inch circle. I will order him to hit it, but he won’t.
If he did hit it, the information (the arrow) is not very complex (just one arrow), but it is so specified that specificity makes up for complexity. In this case, I would have to suspect that he cheated as in the politician in the EF analogy.
I see increasing complexity as more arrows in the wall and more specificity as the increasing likelihood that they won’t hit a specific target.
We need two different calculations for CSI.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 14. June 2003 06:25
So when you claim that the flagellum is more specified, you are claiming that the flagellum is a smaller target.
But there's a problem: how do you know that the flagellum is a target at all, and isn't just something that happened to develop in a certain, arbitrary way?
Dembski's specificational resources takes into account how hard it is to pick out the target. In the case of a pre-painted bullseye, it is easy: "hit the bullseye". But if you find the arrow in the wall first, then it takes a lot of work to say it was between this crack and that spot and the bird dropping there and the spiderweb here, and so forth.
EQU, as a basic logic function, is exceedingly simple to specify; the specificational resources are very low; and hence we expect there to be few chances to randomly stumble into something as easily specified as EQU.
The flagellum seems much more difficult to specify, and therefore the specificational resources are much higher. There may be many things that we could accidentally create that could be described as simply as we can describe a flagellum. Interestingly enough, Dembski never bothers to calculate the specificational resources for a flagellum.
IP: Logged
|
|
Jerry D. Bauer
Member
Member # 756
|
posted 14. June 2003 16:37
I think we can leave the target analogy when we consider the flagellum.
The difference in the arrow analogy and a flagellum is that the latter is an ICS and the former is not.
I think it is in IC systems that we find our specificity. When we shoot 25 arrows into a wall, none of the arrows really care about the other arrows. But in an ICS the parts are so dependent on each other for function that the system is impotent without all of them present and working.
I see specificity manifest to the max as a long assembly line to build, say, automobiles. All workers in that line are dependent on the other workers to do their jobs correctly if the car is to function. If the guy putting in the sparkplugs doesn’t do it, we are left with a system that cannot function. Behe has compared flagella to outboard boat motors. Something certainly seems to have assembled this organelle similarly as General Motors does an outboard motor. And in our assembly line analogy, that guy’s job is very specified. Just put in the sparkplugs—every time—in the same manner—and don’t miss even one car or it will not work.
As to Dembski’s specificational concepts; one cannot reverse the arrow of time when one considers specificity. IOW, we can only consider a pre-painted target to view this concept and anything else doesn’t count.
What are the odds that someone will win the lottery? 100%. It’s going to happen.
Let’s consider a random number generator. What are the odds that the generator will generate a number that is CSI? 100%--because it will generate some random number every time.
But I err in logic when I view the scenario from this perspective because I’m reversing the arrow of time and looking from the present back into the past.
Let’s do this properly. I must look from the present into the future. If I write down a long number and turn on the generator to see how long it will take the generator to cough it up in the future, it will never happen.
And I don’t see flagella difficult to specify mathematically at all. I think it’s high time that IDists get off the philosophy and into the math like other areas of science, don’t you?
The proper way to calculate specificity is to use a scale based on logarithms much like the pH scale is used to consider alkalinity/acidity and the Richter scale used to compare the intensity of earthquakes.
The difference of only a couple of numbers shows a huge increase in specificity and when we consider the many parts that must work together in a flagellic system--that specificity is huge, indeed--Fairly elementary to me. We can calculate the specificity of a flagellum anytime you would care to.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 14. June 2003 19:50
I'm failing to understand how your definition of specificity leads to anything different than a calculation of probability. Dembski's method of specificational resources is a completely different calculation than his computation of probability, and for good reason: he considers every possible target, and picks out those targets which are easy to describe but hard to hit as those which indicate design. To detect this mismatch, at least in theory, he has to both calculate the probability of hitting the target and the ease of describing the target. (In practice, the latter step is usually omitted.)
Perhaps you can clarify the distinction between specificity and probability again, or sketch the calculation for the flagellum (or some sample problem) to make the distinction more apparent. (The toy examples you have been using are too simplistic to be really instructive.)
Added in edit: Incidentally, I usually express probability in negative log units when doing calculations for myself.
Also, keep in mind that one should not attribute design to, for example, a shaken-down box of sticks, where there is good complementarity between the sticks, but no design behind that complementarity (just gravity and volume exclusion).
[ 14. June 2003, 19:58: Message edited by: Rex Kerr ]
IP: Logged
|
|
Jerry D. Bauer
Member
Member # 756
|
posted 15. June 2003 00:27
Specificity is no longer probability once it arrives.
I’m CSI. What are the odds of me being here? 100%. I’m here, aren’t I? We can forget the musings of if I can occur and go with ‘what do I now mathematically do with me?’
Dembski’s math uses the arrow of time to go back in history and calculate the odds of CSI occurring from the past to the present and this is fine, but some of us want to take it further than this.
You ask me to clarify the distinction between specificity and probability. So I will ask you to clarify the distinction between probability and the lottery. One uses the other like a mechanic uses a wrench but the mechanic is not the wrench, is he?
Once we understand that parts are specified—tissue—organs and organelles, we then need up this and take it to IC systems and garner a standardization of the concept to further the hypothesis.
Once the specificity is determined to be there in the sub-systems, then it’s no longer probability and this specificity can be defined macroscopically in the main IC system; and this is where I’m coming from.
I think that system specificity can be defined as: well-matched and specified parts numbering more than one that exhibit intelligence to work together to cause a physically reactive operation.
A physically reactive operation means something must happen, or something might happen as a result of this specified system.
Let’s contemplate a thought experiment considering your ‘sticks in a box’ analogy. Imagine a system of two rocks in the road. Two rocks lying in the road are more than one part. Does my system then have a specificity we can calculate? No--Because these are not two well-matched parts from which a function can be gleaned and the rocks do not work together to cause a physically reactive operation. They are just lying there like your sticks—end of story.
Consider an environmental reaction, such as a tornado that somehow causes one hundred railroad ties to stack in such a manner as to form a crude stairway that I can now use to climb to the top of a bluff where access was previously denied me due to physical limitations.
If we view the railroad ties as a system that now has a function, that function being to serve as a stairwell, can we now calculate the specificity of this system? No, we cannot.
This system is unspecified as it still does not cause a physically reactive operation. The fact that I can now use the system for something isn’t logically followed by “therefore, the system ‘caused’ me to do so.” I can choose either to walk up them or not to walk up them. This is not a specified system.
Flagella can get very complicated when we dissect the parts that come together to form the tissue, but remember we are musing on macroscopic specificity. We can draw the flagellum into a macroscopic format to examine it and when we do, we find that it consists of five major working parts: a rotor, stator, bearing, hook and flagellar filament.
Here we have a good example of 5 parts of an IC system. This is a specified system because it consists of multiple, well-matched parts that cause a physically reactive operation in that the organism now physically moves. We can mathematically define and refine this for use in the lab in comparison to other systems whether specified or not.
System specificity does not increase arithmetically but logarithmically in the math I use.
In the example of a flagellum, we have five IC parts and can see that the system specificity of the flagellum = 5 on our logarithmic scale.
An IC system with only one more part with an SS of 6.0 is 10 times more specified than one of 5.0 and similarly, one of 7.0 is 100 times more specified.
This becomes useful in that now we can compare the flagellum to, say, the cardiovascular IC system in mammals. That system would consist of at least: a heart, veins, arteries, red blood cells, a clotting mechanism, corpuscles, plasma, lungs, one kidney and a brain.
You can do the math.
I also believe we can use this concept for predictions.
IP: Logged
|
|
Joshua A. Smart
Member
Member # 773
|
posted 23. June 2003 14:38
Mike,
You had mentioned something about providing some commentary on my paper. If you have the time, I would still love to hear anything you have to say. Thanks.
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 23. June 2003 18:45
Hi Josh,
Yes, I'm easily distracted. Thanks for reminding me. I'll push ya toward the front of the line and try to craft something. Like I said, I'll probably post blurbs and mostly ignore the critics. Otherwise, I'll meander off into various other tangents. ![[Smile]](smile.gif)
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 23. June 2003 20:52
Josh,
I like your emphasis on getting beyond the exchange of generalities and moving toward application. In biology, air-tight definitions are rare and if biologists sat around coming up with precise definitions that satisfied everyone, they’d never do any experiments. All we need are ‘working definitions’ that approximate the concepts we have in mind. Applications can help us to refine those definitions.
I also enjoyed your discussion on choosing systems for evaluation and attempting to focus on things that exhibit as little hierarchy of structure as possible. In doing so, you target the gene products that comprise biochemical complexes. The gene products not only constitute the structural “parts” of the IC system, but its these “parts” we must account for. You write:
quote:
The farther removed from that genetic level a system is, the more genes must be looked at, and not just more genes, but more complex genetic interactions as well.
Maybe it’s just me, but this would make more sense if you said “structural aspects of a system” rather than “genetic level of a system.” This makes it more clear we’re not talking about regulatory features behind the assembly of a system (if you say “genetic level,” I include regulatory features). However, as it becomes more clear that even regulatory features may approximate a semi-solid state, they too may eventually fall under this umbrella. But then again, such a semi-solid state may depend on the overall architecture of the genome/nucleus/cell and it would be difficult to score any of this as a “part.”
Minimizing hierarchy is what I tried to argue long ago, although you put it much better. For example, Miller argues that the middle ear bones of mammals constitute an IC system, yet fossil evidence shows us how they evolved by cooption. Yet the differences between middle ear bones and a molecular machine are not insignificant. As I explained on ARN (back in 2000):
quote:
The important point, however, is that this type of organismic evolution does not give us reason to think evolution and natural selection worked at the level Behe speaks of. Most developmental biologists would attribute the ear-bone evolution to developmental regulatory changes. That is, no new material was/is employed. Instead, the same old material was simply reshaped. And this could occur because of changes in the timing of expression of certain genes. Yet such regulatory/timing schemes seem largely irrelevant to Behe's focus. Proteins do not change because of the time when they are expressed. To change a protein, you need to change the amino acid sequence. To create a molecular IC system, we need to account for the various parts without the help of a developmental program. Thus, unlike the ear-bones, evolution of the cellular systems involve changing the material and coming up with new material.
I think this is a crucial point. More and more biologists are arguing that morphological evolution is driven by changes in regulatory elements. In fact, some have even proposed that alterations in the patterns of gene regulation have been far more important in evolution than changes in protein function. But what does this mean? It would mean that all of the fossil evidence Stevens can cite is largely the consequence of trivial evolutionary events that have little meaning for the type of things Behe talks about. If most of evolution and the fossil record can be explained by changing the pattern of gene expression, then most of evolution and the fossil record is not relevant to questions about the origin of those genes or the basic process of gene expression itself. Darwin might be vindicated at the level of organismic evolution, but at a very high price. That price being that almost all of the evidence of evolution now becomes irrelevant to the deeper aspects of life.
Let me put it another way. If much/all of morphological evolution can be explained with little/no change in the basic biochemistry and cell plans of organisms, then that morphological evolution, which depends on that basic biochemistry and cell biology, does not apply at this level, but instead exists only because it is built upon this level. This is why the black box that Behe writes about is so important. That biochemistry and cell biology that Stevens waves away as irrelevant is what makes her evolution possible. But while those stories build on what biochemistry and cell biology provide for evolutionary tinkering, they don't explain the origin of that biochemistry and cell biology themselves. To put it one way, the evolution of clothing and fashion depends on the human body, but nothing about fashion-evolution explains the origin of the human body.
Or look at it this way. If we don’t minimize the hierarchy, we end up with a rather odd list of parts for the IC system. Take the middle ear bones. What are the list of parts? Incus, stapes, malleus. Okay. But wait, there is a regulatory gene known that belongs to the RUNX class. Knock it out and no skeleton forms. So our list has expanded to Incus, Stapes, Malleus, and RUNX. RUNX? Any biologist would recognize this as a strange list. It’s a list that ignore hierarchy.
As for knockout experiments, I don’t think that they alone suffice to get us to the IC core. Take the flagellum. In gram negative bacteria, an additional set of rings is needed – the P and L ring. If we were to knock out the flgI gene in E. coli (that codes the P ring component), the flagellum would not function. But E.coli-like flagella work just fine in gram-positive bacteria.
To get to the IC core, I think we need two things. First, we need many different variants of a structure that allows us to compare parts to uncover those that are universal (or near-universal) to the structure. Secondly, we need to understand the logic of the system to understand why the parts are required. The second point helps us to assess the relevance of a missing part among the core if that missing part appears to be derived and restricted to specific lineages. In such a case, the functional role becomes the truly important feature that stands behind the part.
Actually, I realize you discuss this stuff in more detail and precision. In fact, so good is your article that my general response is to say “No need to tinker with it; it’s time to start applying some of these ideas to specific systems.”
[Finally, let me add one correction. The type III components were not discovered by looking for flagellar homologs. People were working on virulence systems and used stanadard genetics to uncover its machinery. When they pulled out the genes, they discovered, somewhat to their surprise, that the type III genes were similar to many flagellar genes. Novel scientific discoveries are often uncovered unintentionally. ]
IP: Logged
|
|
Joshua A. Smart
Member
Member # 773
|
posted 09. July 2003 17:13
Mike,
Thanks for your comments. As for the role of knockouts in determining the ir.core, I certainly agree that situations like the one you mentioned pose a problem. This is why I think it is important to have specific ways of determining the ir.core with well defined approaches, i.e. the six or so approaches that I listed under "Methods of Compilation."
IP: Logged
|
|
|
Printer-friendly view of this topic