|
Author
|
Topic: Nature Refutes ID?: The Evolutionary Origin of Complex Features
|
RBH
Member
Member # 380
|
posted 14. May 2003 14:32
One more remark on John's notion that 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).
As it happens, the results reported in the paper show that this does not happen, at least not at anything like the frequency with which it happens in the experimental runs. The control runs in which only the performance of EQU had selective advantage failed to produce even one program that performed it. So John's speculation is contradicted by the data. More than ever, a calculation of the probabilities John refers to is necessary to attribute the results to chance. Just as a reminder in case someone has forgotten, evolution invokes chance genetic 'shuffling' (mutations and - in sexual species - recombination) together with selection exercised on phenotypic expressions of those genotypes to account for the evolution of complicated things. If that's John's complaint - that the simulation used chance genetic shuffling and phenotypic selection to produce an IC structure - then he has conceded the point and the case.
RBH
IP: Logged
|
|
|
|
Micah Sparacio
Member
Member # 6
|
posted 14. May 2003 15:27
Throwing ideas out at the risk of being called "sad" or "IDiot" or "laughable" or other such things;-) I didn't know that there was so much at stake in an online discussion forum! I really am sorry for being such a fool (for those who don't understand...don't worry).
Anyway, I'm rocking back and forth on this, but I feel like the key issue is going to rest on complexity issues. It is interesting to note that the authors talk about the evolution of "complex functions." But really, what does it mean for a function to be complex? BTW, Behe implicitly distinguishes between function and those things that are complex. Behe indicates that the complexity in irreducible complexity has to do with the well-matched parts, not the function. The system's function can actually be quite simple.
Anyway, those are my thoughts for the hour. Sorry if they make you sad. Sorry if they are stupid.
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 14. May 2003 15:39
A few participants have been suspicious of the fitness assignment, but those suspicions have not been accompanied by any argument that warrants dismissing the experimental results. The following points summarize the fitness assignment and the analogy with real-world biology:
1. All other things being equal, programs that read input, compute EQU, and output the result replicate faster than those who don't by a factor of 32. (All other things being equal, bacteria with functional flagella are better at reproducing than those without it.)
2. All other things being equal, programs that read input, compute XOR, and output the result replicate faster than those who don't by a factor of 16. (All other things being equal, bacteria with a particular functional system--simpler than a flagellum--are better at reproducing than those without it.)
3. All other things being equal, programs that read input, compute NOR, and output the result replicate faster than those who don't by a factor of 16. (All other things being equal, bacteria with a particular functional system--simpler than a flagellum--are better at reproducing than those without it.)
4. All other things being equal, programs that read input, compute OR, and output the result replicate faster than those who don't by a factor of 8. (All other things being equal, bacteria with a particular functional system--simpler than a flagellum--are better at reproducing than those without it.)
5. etc.
The improvements in replication rate are obviously not quantitively realistic (it would be a serious objection if the reported result was very sensitive to the improvements in replication rate, but it is likely quite robust as long as the more complex functions receive larger improvements in replication rate), but the general assumption that some complicated systems can provide large reproductive advantages should be uncontroversial even among ID advocates. Rather, the ID complaint seems to be that, despite that fully functional flagella provide reproductive advantages, a partially evolved flagellum provides no reproductive advantage per se. The analogous statement is true in the simulation: A partially evolved EQU function provides no increase in replication rate, and the same is true for the other functions. In this, the assumptions of the simulation are in fact biased towards the ID view (because most biologists would likely find it more realistic if a partially evolved flagellum--or EQU function in Avida--provided some part of the total advantage for a complete flagellum). The replication improvements for other functions does not change this, unless ID advocates insist that flagella are literally the only systems that results in reproductive advantages.
It is important to realize in precisely what sense the above increases in replication does and does not bias the Avida world to evolve the EQU function:
(i) No path in genome space between the first ancestor and one/several genome(s) encoding the EQU function has been specified. The decision to reward the complete EQU function means that the "end points" (in quotes because only the researchers, not the evolving population, can decide when to stop the experiment) of the evolutionary journey have been specified. The points which do not correspond to genomes encoding a functional EQU subprogram are unchanged by this.
(ii) The decision to reward other simpler functions means that more points in the genome space are "lifted up" a bit in the fitness landscape. Through direct experiments, we know that a side effect of lifting up the points is that several paths that could be "detected" and followed by the evolving population were created. But this side effect was not controllable--it was indeed a side effect, which is why it is taken to be an indication of a phenomenon that is sufficiently general to be of scientific interest: The mere fact that simpler, but different functions, provide reproductive advantages seems enough to make more complicated functions reachable as well. Exactly which more complicated functions is probably not easy to predict.
(iii) Only fully functional EQU, XOR, NOR, ... programs improved the replication rate. This is what cremates the already dead IC argument (IC in the sense of definitions that don't beg the question by requiring a system to be effectively unevolvable in order to be IC), because a fully functioning EQU program evolved by cooption of fully functioning programs with different functions.
Not everything is lost for the ID advocates, though. For those who are willing to adapt their opinions to the experimental data this is an opportunity for further research. Two interesting questions to investigate could be: The EQU function evolved from simpler functions, but how large can the difference in "complexity" be between the evolved function and the simpler functions that were coopted? How many simpler functions must result in improvements in replication rate before a more complicated function (that also result in improvements in replication rate) can evolve? If the ID advocates are lucky, the answers to these questions might provide a new basis for the long sought disproof of evolution. And the best of all is that such questions can probably be investigated in the Avida world!
Erik
PS. For those who wonder about the meaning of "complex" in this context: In the Lenski et al. article, a function is said to be more "complex" if it requires a longer program to encode it. DS.
IP: Logged
|
|
Kirk Durston
Member
Member # 174
|
posted 14. May 2003 15:51
Regarding IC, moving the goalposts, and the simulation under discussion:
First, Behe's definition, as quoted in this discussion, leaves the goalposts too far apart and I've felt that way for over a year. Furthermore, I see a need to mathematically formalize IC, which can easily be done. Setting aside the second need, let me move the goalposts closer together. Under the definition discussed in this forum, two playing cards leaning against each other, providing shade for a light-phobic bug, would be IC. Remove one card, the other falls down, the shade dissappears, and the bug must move on. Yet such a configuration could undoubtedly be achieved by simply emptying bags of cards off the side of a building on a day with no wind. My point is that the definition provided is too broad. To use Micah's suggestion that we incorporate specified complexity, I would propose that an additional requirement be put forward before a system is IC. The requirement is this: if the distance between the functional system and the next lowest functional configuration of that system exceeds a specified complexity of 70 bits, then the system is IC.
When I look at the simulation under discussion, the specified complexity of the results falls considerably short of 70 bits. That being the case, no ID is required, the simulation shows that no ID is required (ignoring, of course, the carefully designed virtual fitness landscape which is designed to produce the desired results with a probability approaching 1), and so I cannot see how the simulation is relevant to ID, or how it can show that IC systems (requiring more than 70 bits of information) can occur in nature. Notable in this paper is the finding that no EQU function was obtained without the intermediate NAND function even though, according to my proposal, even that would have been possible had they done an additional, and reasonable number of generations. They do express concern about 'stacking the deck', as they put it, by supplying intermediate functions. They justify it by saying that this is how it happens in nature. I have no problem with that, provided the distance between these stepping stones simulates nature as well. In real life, the distance in sequence space between stably folding, functional proteins is many orders of magnitude greater than the distance used in this simulation. If they had used gaps that properly model nature, the simulation would not have worked. In other words, the informational gaps between the intermediate functions they provided and the EQU function are too small to be properly called IC.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. May 2003 16:41
Kirk suggests an additional criterion for determining ICness, namely quote:
The requirement is this: if the distance between the functional system and the next lowest functional configuration of that system exceeds a specified complexity of 70 bits, then the system is IC.
Therefore the Lenski, et al study does not show evolution of an IC structure because quote:
In other words, the informational gaps between the intermediate functions they provided and the EQU function are too small to be properly called IC.
The notion that "informational gaps" in the configuration space of all objects is part of the definition of "irreducible complexity" for a given object (say the mousetrap, or a program that performs EQU) is brand new to ID, as far as I know. This criterion departs in a significant way from the definition of Behe and its later refinement by Dembski, in that it requires additional knowledge beyond the structure of the object being analyzed. In fact, it requires exhaustive knowledge of the configuration space within which the object resides.
Further, with its mention of "function" and in the mention of the "informational gaps" between the precursor (different from EQU) intermediate functions and EQU, conceding that cooption occurred in the simulation in the lineages leading to an EQU-performing genotype, the new criterion requires exhaustive knowledge of all potential precursors regardless of the function of those precursors. Thus while it might be a criterion that seems to save IC, it does so at the price of being impossible to apply to any particular object since it requires omniscience in order to know what the configuration space is. Once again, the discussion moves into an "IC of the gaps" domain. The criterion is merely a backhanded way to introduce consideration of the evolvability of an object into its definition, and not just properties of its structure. But non-evolvability is claimed to be a consequence of ICness, not a defining property.
Finally, this has all the flavour of an ad hoc attempt to save IC. For example, why 70 bits? Why not 68, or 71, or 153? Is there any rationale for any number larger than 1 or 2 or 3 bits other than to save IC?
I will not even mention that an IC system could also potentially evolve by loss of components from a 'higher' more complex system, a situation on which Kirk's new criterion is silent.
RBH
[ 14. May 2003, 16:52: Message edited by: RBH ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 14. May 2003 17:05
Hey Erik,
Brief quibble: I haven't re-checked the article, but wouldn't it be more accurate to say that the EQU function gives a critter 32 more "reproduction points", rather than increasing replication "by a factor of 32".
E.g. a non-EQU organism could start out with 500 points, get EQU, and end up with ~532 (analogous to e.g. food or other resources) points. But a factor of 32 increase would give it 500x32, which is a much bigger number. I'd be surprised if the latter was correct although I don't have time to double-check at the moment.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. May 2003 17:16
yersinia,
It's a little confusing in the paper. It says "If this output matches ... [long snip] ... then the organism's rate of energy acquisition, and hence the execution of its genomic program, is accelerated by the factor shown in Table 1" (p.140, emphasis added). So the rewards are multiplicative, and the factors are rates relative to an organism that performs no logic functions.
RBH
[ 14. May 2003, 17:22: Message edited by: RBH ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 14. May 2003 17:38
Perhaps I can clarify why the "goalpost-moving" charge is a significant one.
It has always been unclear whether things like IC and SC were defined by:
(1) The present traits of a system (e.g., multiple required parts)
or
(2) The evolvability of a system (e.g., those things that can't originate by natural processes are SC)
My subjective opinion based on reading many discussions of the above is that about half of the people (both pro- and anti-ID) assume #1, and the other half assume #2. And even more confusingly, both Behe and Dembski have used their respective terms in senses #1 and #2 (the ambiguity is usually included in the official definitions).
Here is the problem:
If your IC/SC term is defined in sense #1, then one can identify a class of systems meeting the definition, without any consideration of the history of the system, and make an *argument* about the evolvability of systems in that class. This argument can then be tested by examination of the history of various systems in that class. The argument could be right or wrong, but it is a respectable argument.
However, a problem arises when the definition is switched to sense #2. Here, IC/SC is *defined* as something which cannot evolve. Therefore, if the investigation of the previous paragraph turns up a case where IC/SC appears to have evolved, well then, those systems aren't actually IC/SC, because of course they evolved. This is a way of saving the IC/SC "argument" but at the expense of making it unfalsifiable, and reducing the whole thing to meaningless verbiage.
We have seen this switcheroo attempted several times in this very thread, which is why the goalpost-moving charge is raised.
Now, I think that Behe, at least, has realized that for his argument to have legs, IC must be defined in sense #1 (he recommended that the words "by definition" be removed from the definition in a 2001 article IIRC).
However, Dembski has made it quite clear that, despite his regular references to biologists referring to "specificity" and "complexity" of biological systems, what he actually means by SC is "specified event that is wildly improbable under all unintelligent causes", which is a long way of saying "not evolvable". This is sense #2.
What it means for "IC to be a subset of SC" thus becomes very obscure, e.g. in Micah's recent post:
quote:
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.
These definitional difficulties are not trivial, they are fundamental to the whole ID discussion. I am just hoping that people will be clear about what sense they mean the terms in (#1: "present-characteristics" vs. #2: "evolvability) and will stick to them.
yersinia
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 14. May 2003 17:42
Thanks RBH, that makes more sense. Foolish of me to try correcting Erik on a matter of computers... :-)
[ 14. May 2003, 18:06: Message edited by: yersinia ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 14. May 2003 18:05
<delete doublepost>
[ 14. May 2003, 18:05: Message edited by: yersinia ]
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 14. May 2003 19:51
I too only have time for a quick drive-by posting (or two), but I couldn't resist getting into the fray. Of course, I have a different view of IC, as I don't view it as a general principle, but as a very helpful tool in a forensic analysis. Anyway......
Micah writes:
quote:
Look. There is NO question that if you remove a part of the IC core of the standard mousetrap that you buy at the grocery store, it will cease functioning. It will not catch mice. Sure, if your intention was to find an intermediate and to manipulate the remaining parts to get a functional intermediate, you could get something that worked, at least in the imagination. But that is not the point. The point is that if our only action is to remove one part, the system ceases to function. This is an IC system, with an IC core, which is necessary for the function.
Erik replies:
quote:
Fine. Just remember that by your interpretation of Behe's original definition, a flagellum is not IC either. Flagella can perform other functions than propelling bacteria forward in a fluid. They are involved in protein secretion, virulence, and adhesion. Perhaps the motility function ceases if any one of a set of components is removed, but other functions can remain. Another thing that isn't IC, by your interpretation, is a mouse trap, since you can remove components and still have a functional paper clip.
Charlie also makes the same point:
quote:
What matters is that the EQU logical function ceases to work in the absence of a number of those subcomponents (the "core") - which it does. Similarly, the flagellum, functionally defined as an "outboard motor" also is IC, despite that some of its subcomponents may very arguably function as secretion systems, pili, adhesion structures, membrane channels etc.
But is it really as straight-forward as this?
Let's use the mousetrap to paper clip example as an illustration.
For the incomplete mousetrap to function as a paper clip, we also need paper, something that is not needed by the mousetrap. Thus, we can not automatically assume our incomplete mousetrap will function unless we have independent evidence that paper is also available and paper-clipping would be recognized as a function.
Removing any ol' component from the mousetrap will not get us a functioning paper clip. For example, remove the hammer, platform, or spring, and you don't get a functioning paper clip. In fact, I'm not exactly sure what function the holding bar and catch alone are supposed to have. If you think about it, the proposed functions for incomplete mousetraps have been remarkably limited. Of course, one could invent all kinds of arbitrary functions for incomplete mousetraps, where it functions as an art piece, paper weight, something to patch a hole with, something to fill up space on a bench, etc. Yet if we came across a mousetrap missing its hammer, would it really be an arbitrary choice to view it as a functionless mousetrap rather than a paper weight? Not really, as the hypothesis of a functionless mousetrap explains the existence of the spring, catch, and holding bar much better than the hypothesis of a paper weight.
Now, if we return to the flagellum, the alternative functions proposed are rather limited - protein secretion, virulence, adhesion. Yet removal of any ol' flagellar component, such that motility is lost, does not necessarily mean these subfunctions will remain. For example, you can remove one of the motor proteins (such as motA) and still retain something that could fulfill these functions. But if you remove flhA, all of these functions are likewise lost. Recall that there are about 20 structural components that make up the bacterial flagellum. In fact, this probably holds true for most of the flagellar components. In other words, the appeal to alternative functions and cooption only get us so far with regard to the flagellum.
Josh wrote:
quote:
Examination of the paper shows that they attributed higher degrees of fitness to organisms that could perform more complex logic operations, with the "reward" being 2^n where n was the number of logic operations combined. The EQU function required 5 operations, so was rewarded with 32 points; but
intermediate rewards of 2,4,8,& 16 were also allowed for simple functions.
And it's at this point that the appeal to protein secretion, virulence, and adhesion all fail. A glimpse of this can be appreciated from figure 1 in my fifth essay on the flagellum and IC (Here). Remember, the flagellum is not some abstract thing, but something concrete built from the products of fllhA, flhB, fliR, fliQ, fliP, fliF, fliG, fliN, fliE, etc.
BTW, I hope to be adding a new installment (or two) to the flagellum series in the next couple of weeks. Some of the stuff should be very relevant to some claims made in this thread and then I could put more time into the discussion.
[ 14. May 2003, 19:54: Message edited by: Mike Gene ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 14. May 2003 20:14
Mike,
that's just another quantitative difference. Sure, the flagellum has more components than EQU. So what? If IC can evolve up to EQU, it can evolve from EQU up to a slightly more sophisticated and complex function, and from there on (I'd say, to beyond 70 bits, the UPB, or any other insormountable limit du jour). Qualitatively, nothing theoretically precludes IC evolution. Not after this paper.
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 14. May 2003 20:24
Mike Gene writes,
quote:
For the incomplete mousetrap to function as a paper clip, we also need paper, something that is not needed by the mousetrap. Thus, we can not automatically assume our incomplete mousetrap will function unless we have independent evidence that paper is also available and paper-clipping would be recognized as a function.
Um, so we can't safely assume the existence of paper, even though we can safely assume the existence of mice?
Analogies like the mousetrap IMO only function well within very restricted domains -- but if they're going to be extended, you have to extend the same "rules" to both sides or else you'll get inconsistencies in your argument like that above.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 14. May 2003 20:31
charkie d wrote quote:
If IC can evolve up to EQU, it can evolve from EQU up to a slightly more sophisticated and complex function, and from there on (I'd say, to beyond 70 bits, the UPB, or any other insormountable limit du jour). Qualitatively, nothing theoretically precludes IC evolution. Not after this paper.
Just before I read that I got off the phone. It looks like I'll get free access to a multi-processor Beowulf machine in a few weeks. There are lots of interesting extensions of the Lenski, et al stuff, and we (a couple of academic evolutionary biologists and I) are going to start looking at them. I'll let you know how it goes.
RBH
IP: Logged
|
|
|
Printer-friendly view of this topic
