|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
charlie d.
Member
Member # 159
|
posted 13. May 2003 17:57
This is just a lot of humming and hawing, definition-shifting, and confusion about levels of analysis.
I agree that this discussion will go nowhere until somebody who disagrees with the paper's conclusions takes on the task of doing an objective, systematic analysis of the EQU functions; it's actually not that complicated. However, I do understand the reluctance.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 13. May 2003 19:29
quote:
This is just a lot of humming and hawing, definition-shifting, and confusion about levels of analysis.
Humming and hawing, huh ![[Razz]](tongue.gif)
Actually, you gotta know how the experiment is relevant before you can critique it. I'm still trying to figure out how (and if) its relevant. As I said at the beginning, my concern is the quick, unskeptical jump to conclusions about *what exactly the experiment achieves.*
I've printed out copies of several organism samples and will be analyzing them tomorrow. However, all this humming and hawing seems perfectly legitimate to me...and actually makes salient the things that need to be analyzed more carefully.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 13. May 2003 21:10
Argon writes,
quote:
I think what RBH is asking for someone to evaluate the final products of the simulation: Are they IC or not? Here's one suggestion for testing: Knock out a few of the instructions used by the coding sequences to generate the EQU function. What happens?
Well, if they are instructions critical to the EQU function, you will no longer have EQU. But that's not sufficient to call it IC. Knock out one of the letters that produced METHINKS IT IS LIKE A WEASEL, and you no longer have that. Like EQU, it either is or isn't, by definition. But like Micah points out, how does this translate to the biological world? Is METHINKS IT ID LIKE A WEASEL good enough to provide some functionality or not? Is a function that matches EQU on a 32 bit string for 31 of the 32 positions good enough to provide a selective advantage? Note that the simulation does indeed reward some of the component functions used to build EQU, and that puts them reasonably "close by" in an evolutionary sense, for some of the reasons John points out.
I don't see the EQU function simulation as producing an IC output.
[ 13. May 2003, 21:33: Message edited by: Roger R ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 13. May 2003 23:38
Roger,
It seems that what is IC and what is not IC becomes quite a mobile target.
Behe: By irreducibly complex I mean a single system composed of several well-matched, interacting parts that contribute to the basic function, wherein the removal of any one of the parts causes the system to effectively cease functioning.
The basic function seems to be quite clearly EQU and the parts are well natched and interaction, where each part contribute to its ultimate function.
Roger: Is a function that matches EQU on a 32 bit string for 31 of the 32 positions good enough to provide a selective advantage? Note that the simulation does indeed reward some of the component functions used to build EQU, and that puts them reasonably "close by" in an evolutionary sense, for some of the reasons John points out.
The simulation did not reward 31 out of 32 positions to be good but it did reward component functions that you claim are reasonably close by. But ICness has nothing to say about 'reasonably close by' merely that it stops functioning when one of its parts is removed.
But how close by in evolutionary sense are these components? The paper may give us some insight here when it states that
quote:
No single mutation in the ancestor can produce even the simplest of these functions. Instead, several mutant instructions must appear in the same lineage, and such that they are coordinately executed, to perform even a simple function.
But all the data are there to do the calculations. But irregardless, it seems to me that EQU matches at least one of the various definitions of IC.
Kirk: A system is not IC if the informational gap is so small that it can be jumped by random processes.
Surely that is not the definition of IC as given by Behe. It seems to me that we are chasing some fast moving goalposts here.
IP: Logged
|
|
Roger R
Member
Member # 200
|
posted 14. May 2003 07:22
Pim van Meurs writes,
quote:
The basic function seems to be quite clearly EQU and the parts are well natched and interaction, where each part contribute to its ultimate function.
I'll reply to the rest later, but I couldn't let this pass without comment. The parts are "well matched" to the function via ID. The designers of the program picked a limited set of components conducive to the goal of EQU. That isn't anything like a bacterial flagellum in the real world.
[ 14. May 2003, 07:23: Message edited by: Roger R ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 14. May 2003 07:45
All this talk of moving goalposts is bull*&^$ as far as I'm concerned.
Truly... has a theory ever been put forth that hasn't been modified to deal with new data sets? Isn't it the practice of science to generalize originally specific theories onto larger domains?
The people moving the goalposts of this entire conversation are the ones who are demanding of IC and Behe something that is not demanded of any other scientific theory in the history of the world. It is exceedingly frustrating.
This is why it is important to remember the context in which Behe proposed his theory. It certainly was not in the context of logic primitives which can be slapped together in a sequential 2 dimensions to achieve new functions.
Personally, I think that those who get wrapped up in this "moving the goalposts" tactic have very little interest in really finding out how relevant this project is to Behe's original conception of IC. Blindly defending any research that comes out which seems to have implications against ID is cheap cognition. As the friendly moderator often says: this is the mindset of people fighting a battle and not of people engaging in dialogue. Unchartiable standards are the signs of closed minds.
Additionally, if people aren't willing to acknowledge that Behe's notion just might need modification to be applicable in this domain, and feel as if the conversation needs to stay static (i.e. Behe gets only one chance to develop his theory), then perhaps these people aren't worth engaging.
Just my thoughts. I won't be responding to people who engage in such debate tactics from this point on. My interest isn't in a static debate but in finding a way to map Behe's notion onto this project. Surely there will be modification of his ideas along the way.
Anyone who thinks otherwise has a naive view of science and intellectual life in general.
Finally, what I'm finding is that if Behe's notion is to be given a larger domain (outside the domain of three dimensional mechanical function) then a better defined "complexity" metric needs to be developed...which I think leads us to Dembski's specified complexity. This fits well with Dembski's claim that IC is specific instance of his more general specified complexity.
[ 14. May 2003, 08:02: Message edited by: Micah Sparacio ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 14. May 2003 08:27
Actually, Micah, that's precisely what I think most of us would think is the appropriate response to this article: an honest recognition that the standard definition of IC, as proposed by Behe originally and still often repeated by ID advocates, is insufficient to make a design inference, because systems that fit that definition have now been clearly shown to evolve.
This doesn't mean however that any ID inference necessarily goes down the drain. If Behe's IC is not a strong argument for design anymore, people are still free to adopt other parameters that they think better fulfill their theoretical needs. This seems to be what you suggest regarding CSI.
As for goalpost moving, that is simply a fact, and arises when people refuse to honestly look at the evidence, draw the (so far still inescapable, as far as I am concerned) conclusions, and instead engage in evasive tactics, that have nothing to do with the question at hand.
Really, claims that "this is not biologically relevant" are plain bogus, when ID theorists have long argued that any object, including, and foremost, man-made ones, can exhibit IC and allow ID inferences. Similarly, the argument that the simulation is not valid because EQU was one of the possible outcomes of the combination of the original simpler functions would seem to require any valid simulation to evolve the unevolvable. Other arguments made here are along similar lines, and equally flawed.
The only purpose of these arguments is to refuse to admit that Behe's IC concept as a hallmark of unevolvability was a theoretical illusion. Alas, until that's done, ID won't be able to move on.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 14. May 2003 08:42
But charlie, was it an illusion in the domain in which it was proposed? I think his argument generalizes relatively well over mechanical function in space which is qualitatively different than function in other senses.
I seriously am not trying (flailing my arms) to "save" Behe's theory here. I'm just pointing out that he developed the idea around the notion of mechanical function (mousetrap, flagella) but didn't qualify this fact in his definition. We should be thoughtful enough to see this and to ask, first of all, whether his notion can be applied in the domain of logic functions (which we are finding may not be the case) and then ask whether the core of his argument can be maintained with an adjusted concept to fit in this new domain. That might lead us to a more general notion of IC that incorporates both mechanical and logical function (perhaps others).
IP: Logged
|
|
YZ2
Member
Member # 91
|
posted 14. May 2003 09:51
I do think this experiment is an interesting one that can give us many insights into biological evolution, possibly due to Darwinian processes.
What the shortest hand-written EQU program describes are applications of NAND function to transformed and partitioned problem subspaces. I could be wrong but if this is true, then it is a trivial case of Darwinian evolution that everybody can agree on that could happen. The new EQU function is a new application of existing function at different parts of the problem subspaces. There is no novel function created.
Unfortunately, it does have the APPEARANCE of a novel function. The so-called new EQU is no more and no less IC or complex than the original NAND function. That is also why EQU can be generated in different paths, by partitioning the problem
subspaces
differently,sometimes sub-optimally.
I do think that biological evolution can be more powerful than that. If indeed no novel function has been created, it is possible that digital simulation may not be a good paradigm to analyze biological evolution.
Comments?
[ 14. May 2003, 09:57: Message edited by: YZ2 ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 14. May 2003 10:34
In case anyone is interested in some mappings:
Y nand (X nand Y) = X or ~Y
This should be obvious, but it is just saying that for the left side to be true, you can't have both sides of the outmost nand be true. So, if Y is true, then X also needs to be true, but if Y is false, then it doesn't matter what X is because the left side of the outermost nand will already satisfy the logic function.
(the above is position 7 in the "appears to be the shortest" handwritten EQU organism)
(X nand Y) nand Y -> X or ~Y
(X nand Y) nand X -> Y or ~X
(Y or ~X) nand (X or ~Y) -> X xor Y
The xor function says that you can have one or the other but not both. So, if X is true (makes the right side of the nand true) then Y can't be true (because it would make the other side of the nande function true as well, making the overall nand false)
(X xor Y) nand (X xor Y) -> X EQ Y
EQU = (((X nand Y) nand Y)) nand ((X nand Y) nand X)) nand (((X nand Y) nand Y)) nand ((X nand Y) nand X))
Symmetry due to pops and pushes (copies)
A better representation might be a tree (if I can get the formatting- ignore the ****'s):
1 *************** X nand Y
2,3 ****** nand Y *** <> *** nand X
4 *************** >nand<
5 *************** <nand>
At level one you perform the first nand. Level 2,3 you get your "or" functions. Level 4 you get two copies of a single "xor" function. Level 5 you join the two "xor" functions by a nand to get your EQ function.
Anyway, if this helps others to understand what's going on, good. If not, sorry for the bad description. If I'm wrong about anything above, sorry about that too.
[ 14. May 2003, 10:40: Message edited by: Micah Sparacio ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. May 2003 10:54
First I should confess that the exigencies of real life are pressing hard on me, so I'll have time only for the occasional "drive by" posting. But as and when I can I'll toss something into the pot.
In one respect, goal posts are clearly being pushed around in this thread and in other discussions of this research elsewhere. Several defenders of IC have observed that there were intermediates available in the conceptual space in which the Avida organisms were evolving, and that somehow vitiates the claim that the research shows that IC structures can evolve. But that's the question - can IC systems evolve in spaces where bits and pieces are available but in which there is not a single-function-preserving pathway from the ancestor to the IC? And the answer in the Avida world is "Yes, they can. Pretty easily, in fact."
Another question involves the appropriate level of analysis. Is it the formal definition of EQU as a logical function, or the sequence of instructions in a particular Avida run that performs EQU? Micah alluded to this question, but it's not clear what his answer is - in fact that's a question he's struggling with, I think. It's also not clear in the case of the flagellum. Dembski estimates probabilities using individual proteins as the unit of analysis, but the specification is in terms of functional elements - propellors, rotors, stators. Does ICness and its superset specified complexity allow one to specify at one level of analysis and calculate probabilities at another?
Micah wrote quote:
I seriously am not trying (flailing my arms) to "save" Behe's theory here. I'm just pointing out that he developed the idea around the notion of mechanical function (mousetrap, flagella) but didn't qualify this fact in his definition. We should be thoughtful enough to see this and to ask, first of all, whether his notion can be applied in the domain of logic functions (which we are finding may not be the case) and then ask whether the core of his argument can be maintained with an adjusted concept to fit in this new domain. That might lead us to a more general notion of IC that incorporates both mechanical and logical function (perhaps others).
I believe you - I do think you're seriously trying to understand what's going on given the Lenski, et al., findings. I think you want to be careful here, though. If one sets out to model a three-dimensional 'mechanical' system in a formalism, be the formalism a mathematical equation or a computer simulation, are there some properties of the mechanical system that cannot in principle by definition be modeled in the formalism? That's what your remarks seem to be tending toward. Adopting that stance has very substantial implications for using any formal modeling technology to learn about the system the formalism models.
Keep track of slippage in levels of analysis here, too, as for example the Avida world as a 'real' system that is the focus of analysis versus the Avida world as an model of (abstract properties of) another system. Does one want to define IC as a concept that cannot be modeled in a formal system, or does one want to define IC as a property that does not characterize certain sorts of systems? ID makes a good deal of fuss about information and its properties, and "information" (in the ID conception, at least) is not a property confined to three-dimensional mechanical structures. Modeling the Caputo draws as a string of 1s and 0s is to represent a mechanical procedure as an abstract model. Surely one wants to preserve the ability to use that kind of abstract model. So there's danger lurking in the effort to constrain ICness to one or another kind of system by definition just because some empirical data suggest a domain in which it is not applicable. It's not necessarily an insuperable danger, but it requires careful attention to avoid being merely an 'IC of the gaps' definition.
RBH
Edited for half a dozen spelling reversals. It ain't a good morning!
[ 14. May 2003, 11:00: Message edited by: RBH ]
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 14. May 2003 11:28
RBH said:
quote:
If one sets out to model a three-dimensional 'mechanical' system in a formalism, be the formalism a mathematical equation or a computer simulation, are there some properties of the mechanical system that cannot in principle by definition be modeled in the formalism? That's what your remarks seem to be tending toward. Adopting that stance has very substantial implications for using any formal modeling technology to learn about the system the formalism models.
I think you're substantially misunderstanding Micah's point. He (and a lot of other people here) is not challenging the ability to do any simulations of IC systems, but rather whether this simulation accurately models what really goes on with IC systems. Surely you would agree with me that computer simulations must capture the essence of the system they're modelling; otherwise, they don't tell you anything. I would argue that this computer model simply doesn't capture the essence of what would be required to show evolution of an IC system. It just doesn't impress me that the system started with all the necessary building blocks, like taking all the parts of a flagellum, scrambling them together in a virtual world, providing the computer with complex "component swapping" functions to move things around, and a very high swapping rate (mutation rate). Given enough time (especailly if you select for subsets of the final flagellum), you're GOING to get a functional flagellum!
Darwinists are always very keen on critiquing arguments that they see as inaccurately describing the evolutionary process (the tornado in a junkyard analogy springs to mind here). However, here's a case which is equally rediculous, has absolutely no relevance to biotic reality (as Charlie d. has implicitly acknowledged) and yet--it's being embraced has having the utmost implications for the evolvability of complex systems!!!
Is there not a double standard at work here? Why don't the Darwinists see the unrealistic scenario and criticize it just as strongly as they criticize the tornado in a junkyard approach? It suggests to me that there is less of an interest in seeking truth and more of an interest in just seeing Darwinian evolution be proven, no matter what. As Micah pointed out, this is cheap cognition, it's not really thinking.
I just want to add one final thing to the debate about Behe's IC definition, etc. I think his original definition had an implicit complexity limit. This is all a probabilities game, and if given all the components of a complex system plus a high shuffling rate, chance alone can shuffle them around into something selectable (like EQU). The probability of that in this program is very high because of the unrealistic scenario the program embodies (and because of its failure to accurately model what biotic reality looks like). The program BEGINS very near where it ENDS--it's just been given a few more NAND functions (and it still needs selection for the intermediate steps!!!). I guess the point I'm making is that it's possible for this program to produce something that's IC by the original definition, yet not tell us anything about the origin of IC systems in the biological world, where real constraints of engineering come into play (again, not because you can't model IC systems, but because the program fails to accurately do so). How are biological proteins co-opted from systems that are entirely different from a flagellum? How do they get co-modified to function together in the new system (surely every part wasn't pre-existing in its final, flagellar form in the cell). Etc. These are the changes which must be addressed by a Darwinian solution, and the computer simulation just glosses over all that. The main difference is one of the implicit (and explicit in Dembski's Specified Complexity) complexity bound.
John
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 14. May 2003 12:02
John: It just doesn't impress me that the system started with all the necessary building blocks, like taking all the parts of a flagellum, scrambling them together in a virtual world, providing the computer with complex "component swapping" functions to move things around, and a very high swapping rate (mutation rate). Given enough time (especailly if you select for subsets of the final flagellum), you're GOING to get a functional flagellum!
But is this not what is proposed as a potential evolutionary pathway for IC systems? And does the program not show that systems which appear to be IC can in fact evolve and evolve through many different pathways.? Thus when John states that 'This is all a probabilities game, and if given all the components of a complex system plus a high shuffling rate, chance alone can shuffle them around into something selectable (like EQU). " he seems to have forgotten that chance alone was shown to not be enough to shuffle them into something selectable like EQU.
If John argues that the program begins close to where it ends then I wonder why it took so many evolutionary steps before the EQU arose? Surely evolutionary distance seems to be larger than what John claims it was. All in all it seems that all the complaints John raises wrt the simulation may very well apply to real life as well. What I find fascinating is how quickly the goal posts for IC seem to be moving now that it seems to run the risk of being falsified.
John raises a lot of good questions wrt the flagellum indicating that the issue wrt ICness and evolution for the flagellum is hardly resolved in fact as Behe stated so well
quote:
The peril of negative arguments is that they may rest on our lack of knowledge, rather than on positive results.
What really surprises me though is that the ID proponents seem to object to all these simulations when in fact they are very helpful in understanding if the argument that IC systems cannot evolve has any real value. It seems that so far our increased knowledge has resulted in closing a few more gaps in our knowledge and has moved the argument of IC further away from relevancy to evolution.
So perhaps given enough time evolution inevitable would have found something like the flagellum.
Surely mutation and selection in Avida is far more relevant than the tornado in the junkyard calculations by Dembski for the flagellum? I would hope that ID proponents would embrace this kind of research which finally would allow them to put Dembski and Behe's arguments to real tests. Real tests mean that also many of the critics' comments and objections can be evaluated.
quote:
It suggests to me that there is less of an interest in seeking truth and more of an interest in just seeing Darwinian evolution be proven, no matter what. As Micah pointed out, this is cheap cognition, it's not really thinking.
Seems to be a potential two-edged sword here.
[ 14. May 2003, 12:08: Message edited by: Pim van Meurs ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 14. May 2003 12:21
Roger:
The parts are "well matched" to the function via ID. The designers of the program picked a limited set of components conducive to the goal of EQU. That isn't anything like a bacterial flagellum in the real world.
Please explain why this is not anything like the ICness of the flagellum? Well matched parts, removing one part destroys function. If you check the program's initial state then other than replication, none of the necessary components were present. Where the components limited (26?), I am sure that they seem to be almost as limited (or even somewhat more) as the number of components in the genome (20 amino acids).
So what part required ID and how was it introduced? Is Roger suggesting that any involvement of intelligence makes the value of an experiment doubtful? Is thus ID present everywhere by virtue of humans doing the experiments and observations? Such a position seems to make science unworkable. Similarly thus one may conclude that science could never determine if IC can arise naturally, and thus making IC unfalsifiable since any attempt to test this would run into the ID objection.
Now back to the work by Lenski et al, perhaps Roger can help us understand why he objects to their experiment and how his objections are relevant? Since most of the tools and data are available to all, this may be time for ID proponents to work out their own scenarios?
For instance what is the relevance of the findings that proteins like RNA form scale free networks which are characterized by a few common forms which extend throughout sequence space and are close to other common forms? A limited set of components, close in sequence space, distributed over sequence space.
[ 14. May 2003, 12:41: Message edited by: Pim van Meurs ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. May 2003 12:33
John wrote quote:
I think you're substantially misunderstanding Micah's point. He (and a lot of other people here) is not challenging the ability to do any simulations of IC systems, but rather whether this simulation accurately models what really goes on with IC systems . Surely you would agree with me that computer simulations must capture the essence of the system they're modelling; otherwise, they don't tell you anything. I would argue that this computer model simply doesn't capture the essence of what would be required to show evolution of an IC system. It just doesn't impress me that the system started with all the necessary building blocks, like taking all the parts of a flagellum, scrambling them together in a virtual world, providing the computer with complex "component swapping" functions to move things around, and a very high swapping rate (mutation rate). Given enough time (especailly if you select for subsets of the final flagellum), you're GOING to get a functional flagellum!
Are you suggesting that the evolution of the flagellum did not have necessary building blocks available via cooption or recruitment, or that those building blocks are not themselves evolvable for other functions and thus become available for cooption? Are the proteins that comprise the flagellum unique to it and not available otherwise? Are there no other known biological systems that use proteins (or similar protein structures) in configurations similar to the functional components of the flagellum - are there no cilia elsewhere doing other things, for example? (I might note that the "component swapping" locution badly misrepresents what happens in Avida's simulation. Since reproduction in Avida is asexual, "swapping" doesn't accurately describe how changes in genomes - mutations - occur. Components can be changed, deleted, or inserted by mutations, but they're not swapped around among organisms.)
John claims that the simulation does not capture what ICness really is. If that's the claim, then what really does go on with IC systems, if not what is abstractly represented in the reported work? What is "the essence of ... an IC system", if not what is represented in the reported simulation? What would be required to show evolution of an IC system? Precisely what kind of simulation, if it failed to evolve IC, would constitute a refutation of the evolutionary claim?
Finally, John wrote quote:
This is all a probabilities game, and if given all the components of a complex system plus a high shuffling rate, chance alone can shuffle them around into something selectable (like EQU).
That is a version of the 'tornado in a junkyard' argument and is an implicitly quantitative argument. If that's what occurred in the simulation reported, it occurred 23 times in 50 tries, with a different sequence of assembly-language instructions (the sequences ranging in length from from 17 to 43 primitive instructions comprising the knock-out defined IC sequence) being shuffled into a non-negligible order (the order of instructions is not irrelevant) in all 23 'successes.' The necessary information for calculating the relevant chance probability of the results that were reported is available: I suggest that John take his mentor's advice and do the calculation. What is the probability of occurrence of those 23 successful-but-different outcomes in 50 attempts by random shuffling of the components? (And while he's doing so, justify his choice of unit of analysis for the calculations.)
RBH
Now will someone PLEASE make me go back to work!
IP: Logged
|
|
|
Printer-friendly view of this topic
