|
Author
|
Topic: Joshua A. Smart: On the Application of Irreducible Complexity
|
Moderator
Administrator
Member # 1
|
posted 05. June 2003 07:44
On the Application of Irreducible Complexity
by Joshua A. Smart
Abstract- The purpose of this paper is to provide a framework and direction for a more rigorous application of irreducible complexity. Critical responses to irreducible complexity have primarily consisted of just-so-stories that substitute vague appeals to chance or other forces for well-thought-out counter-arguments. Intelligent design theorists have noted the lack of substance in critiques of their work, but there has been virtually no attempt to level more specific challenges to evolutionary theory. Intelligent design will not advance within the scientific community as long as it continues to engage in this exchange of generalities. Progress lies in application. To that end, this paper describes a process for applying irreducible complexity, with particular attention to determining the irreducible core of a system.
To read the entire paper, click here.
[ 05. June 2003, 07:44: Message edited by: Moderator ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 05. June 2003 12:54
I am not sure if the author reads these forums but I would like to hear more from Joshua about the application of his ideas and the recent paper published by Lenski et al which seems to suggest that RMNS is sufficient to generate an IC system. In fact the experiment suggests that there are multiple ways to ICness. Does this suggest that ICness through RMNS is contingent and thus RMNS is able to create the complex information similarly to the contingent actions of an intelligent designer?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. June 2003 15:23
Smart wrote quote:
The power of the concept of irreducible complexity is that it invalidates the step-by-step process of evolution, not just the product. (p. 2 of the unpaginated ms.)
I heartily second PvM's remarks. In the Lenski, et al., paper we have data on the complete course of evolution of what I cannot see as anything but irreducibly complex structures - assembly language programs - by random mutation and selection. I would really like to see Smart, or any ID proponent, apply the kind of analysis he outlines in this essay to those data.
Let me briefly recapitulate the relevant findings of the Lenski, et al., paper. They used an ALife platform, Avida, to evolve assembly language programs to perform logic operations starting from an Ancestor program that performed no logic operation. The assembly language programs are analogous to genetic instructions. The logic functions that the programs perform are analogous to phenotypes, that which is exposed to selection. Over the course of a number of evolutionary runs during which simple logic operations were selectively advantageous, 23 different programs evolved to perform one particular complicated function, EQU, that was maximally advantageous. Knockout testing showed that each of those evolved programs had an irreducible core, a set of (genetic) instructions each of which was indispensable to the program's performance of the EQU operation. By both Behe's original definition and Dembski's amended definition of IC, those programs are irreducibly complex.
With the Lenski, et al., data available to us it is possible to apply Smart's "Strict" IC determination method; one need not make assumptions or concessions. And Lenski, et al., did exactly that. Smart wrote quote:
In the Strict approach nothing is assumed, and nothing is taken for granted. Ideally, this is the way every problem would be approached. If meticulously carried out, a Strict approach leaves no one on the short end of the stick. The disadvantage is that taking the Strict approach often involves going through meticulous, time-consuming detail. (p. 4)
Leaving aside the observation that to overturn a highly developed and strongly corroborated central theory in a scientific discipline might well require "going through the meticulous, time-consuming detail" in order to establish a competitor, with the Lenski data we can readily do so. They performed the knockout operation on all of the evolved programs, reporting the number of components in their irreducible cores. All of the evolved programs had an irreducible core, a subset of instructions the loss of any one of which destroyed the program's ability to perform the final function.
The Lenski, et al., data show that several assumptions of the Smart paper are false. Most striking, throughout his essay Smart assumes the conclusion that IDists wish to make about IC structures: that they cannot be produced by evolutionary mechanisms. For example, he wrote quote:
While there are often methods of making a good guess, the selection of components to include would quickly become the target of debate, drawing attention away from the goal of determining whether or not a system could have been produced by evolution. (p. 7)
But that is not the question Smart is addressing, which is how can we determine what is IC? His methods to this point in the essay do not address the question of whether IC structures can evolve, however they are defined in theory or determined in practice. Only by conflating "IC" and "unevolvable" could he phrase that sentence as he did.
That conflation (and the confusion it creates) becomes clear in Section 6, "Evolutionary Method (EV)." Consider the logic of EV as Smart describes it. If three extant structures performing a similar function each have an irreducible core (each IC core somewhat different from the others as determined by knockout), and there is some overlap among them, the IC core of the phylogenetically oldest version is taken to be the 'real' irreducible core. That leads inexorably to the conclusion that both of the other versions in Smart's "generic example" had to have been produced by evolution even though each of them is irreducible by Smart's Strict knockout criterion! How does Smart imagine that to have happened?
In fact, given the knockout-determined IC cores of those two phylogenetically later versions, Version 2 had to have evolved by the loss of one indispensable component (A) of the ancestral Version 1 irreducible core and the gain of two new indispensable components (B and H). Version 3 had to have evolved by adding an indispensable component (H) and the loss of another indispensable component (A, which is present in Version 3 but is not part of its knockout-determined irreducible core). So under Smart's "Evolutionary Method," irreducibly complex systems can be produced by evolution. And, by golly, the additions and deletions Smart invokes in his generic example are among those that evolutionary theorists hypothesize can do it! All Smart is missing is cooption to complete the picture of how IC structures evolve. In order to use Smart's Evolutionary Method, he must grant that at least some IC structures (defined by the knockout procedure) are produced by evolution. That is a delicious irony.
I repeat: the Lenski paper with its associated data form an ideal test bed on which IDists can demonstrate the utility, or lack thereof, of the notion of irreducible complexity. The disadvantage for the IDist who does so, though, is the observed fact that all of those IC programs evolved!
RBH
[ 05. June 2003, 15:38: Message edited by: RBH ]
IP: Logged
|
|
Mike Gene
Member
Member # 149
|
posted 05. June 2003 15:39
RBH: The assembly language programs are analogous to genetic instructions. The logic functions that the programs perform are analogous to phenotypes, that which is exposed to selection. Over the course of a number of evolutionary runs during which simple logic operations were selectively advantageous, 23 different programs evolved to perform one particular complicated function, EQU, that was maximally advantageous.
How many assembly language programs make up the EQU and what are they?
I repeat: the Lenski paper with its associated data form an ideal test bed on which IDists can demonstrate the utility, or lack thereof, of the notion of irreducible complexity.
I demonstrated the utility of IC a long time ago:
http://www.iscid.org/boards/ubb-get_topic-f-6-t-000017.html
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. June 2003 15:53
MG,
The paper and associated supplemental materials are here:
http://myxo.css.msu.edu/papers/nature2003/
RBH
Added in late edit: What Argon said. And the 23 assembly language programs that evolved to perform EQU are listed here.
[ 05. June 2003, 17:32: Message edited by: RBH ]
IP: Logged
|
|
Argon
Member
Member # 276
|
posted 05. June 2003 16:58
Mike Gene wrote:
quote:
I demonstrated the utility of IC a long time ago:
Interesting point. I suspect RBH was thinking of IC's "utility" with regard to determining the evolvability of a certain features.
[ 05. June 2003, 16:58: Message edited by: Argon ]
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 05. June 2003 18:41
Not to take the topic off thread (this will be my only post in reply to RBH) but it seems quite odd to me that the organism (the assembly program) is the IC structure we're talking about in the Lenski system.
So what's the irreducible core? What are the parts of the IC system? Or is all this mute when it comes to logic functions, which are a far cry from mechanical functions?
[Added in for edit -> My advice to Josh, take the Lenski paper with a grain of salt. It might give you reason to refine your conception of IC, but the paper surely doesn't hit at the heart of the concept]
Let's not just take one paper, presented in the last few months, applied within in a static assemply architecture, and use it to completely disregard Josh's work.
[ 05. June 2003, 18:47: Message edited by: Micah Sparacio ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 05. June 2003 19:20
I, for one, would first like to say that this is far and away the best ISCID paper I have seen presented. It is remarkably free of the usual confusions, assumptions, and antievolutionary sniping that plague most of what I've seen.
In fact, I think that this paper makes a significant move towards resolving the confusion over the definition of IC that was apparent among ID-sympathizers in the discussion of Lenski et al's article, and Smart even agrees with RBH's position, that the knock-out criterion should be the fundamental one for determining what constitutes the "IC core."
Smart recognizes in a clear-headed fashion (which I wish was similarly recognized by all IDists) that the issue of defining and identifying IC systems must be one question, and *after this is done* the second question of the evolvablility of this class of systems must be assessed, to see whether the "IC systems can't evolve" argument is, in fact, true.
So in conclusion of my complimentary comments, if ID is to have any future in convincing scientists that they are onto something, they are going to have to make use of clear-headed and low-assumption-content thinking about terms that Josh Smart exhibits.
I have some critical comments which I shall put in another post.
[ 06. June 2003, 02:26: Message edited by: yersinia ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. June 2003 19:25
Micah wrote quote:
Not to take the topic off thread (this will be my only post in reply to RBH) but it seems quite odd to me that the organism (the assembly program) is the IC structure we're talking about in the Lenski system.
I don't think this takes the thread off-topic. After all, one of the defenses against the Lenski data has been to redefine ICness to escape the implications of evolving those programs. As I noted above, in the abstract Avida critters the assembly language programs are analogous to gene strings - the instructions for performing functions - and the functions performed are analogous to phenotypes - that which is exposed to selection.
As I have remarked elsewhere, in the world of ICness one apparently has free choice of level of analysis. It seems to me that the instruction string level is most appropriate when analyzing the Avida programs. I'd have to hear a principled argument for another level, an argument that is absent from discussions of the flagellum, for example. Why are its proteins the appropriate level of analysis for estimating the improbability necessary to attributing specified complexity? Beats me. Smart's discussion of the flagellum is phrased primarily in terms of its genome, after all.
To answer Micah's specific question, the "irreducible core" of the Avida programs consists in the subset of instructions which, when each is replaced by a null instruction one by one in a knockout procedure, results in the program losing the ability to perform EQU. They're marked with red rectangles in the tables on the page I provided above.
Finally, a note on programs that perform logic functions as IC versus "mechanical" functions: is the clotting cascade Behe mentions a "mechanical" function? Is a faithful symbolic representation of, say, the flagellum in a computer simulation a "mechanical" function? Work on those distinctions a bit and you'll find they become fuzzier and fuzzier. I think the attempt to slip around the Lenski, et al., results by somehow claiming they are irrelevant to objects whose mechanical structure (say, a mousetrap) are IC versus those whose abstract structure is IC. Since Micah left the Literature Review thread on this matter I haven't refined the presentation of this, but his briefly suggested distinction there is, I think, either indefensible or (to the extent that it's defensible) trivializes IC to the point that it is pointless to know whether a structure is IC or not.
My advice to Josh would be to first think carefully about the inherent contradiction at the heart of his "Evolutionary Method." Then he might think about the kinds of objects - mechanical, abstract, logical - to which the epithet "IC" applies and does not apply, and why.
RBH
[ 05. June 2003, 19:40: Message edited by: RBH ]
IP: Logged
|
|
John Bracht
Member
Member # 5
|
posted 05. June 2003 20:34
I made this point before (on the literature review board discussion of Lenski et al) but it seems to have gotten lost amidst the critics cries of "evolution of ICness!" and hence not really dealt with.
Here's the problem with the Lenski et al simulation: they don't appropriately model biological evolution and hence their results have no applicability to reality. In particular, the fact that their program's "mutations" consist of swapping out entire functions, or adding/deleting them. There is no analogy in biology for the miracle appearance of fully formed, functional blocks. The closest thing I can think of is a protein that gets swapped out, but in REAL evolution those proteins have to first evolve their novel functionality. Furthermore, since real evolution occurs by mutations at the nucleotide level (point mutations, insertions/deletions etc) the mutations have a high probability of disabling the functional proteins they occur in. In the Lenski simulation, the evolving programs cannot accrue mutations at the function level or below (hence, no non-functional functions can ever evolve). This basically constrains the evolutionary process to the functional areas of sequence space--solving the problem of crossing non-functional canyons, which is the key problem for evolving ICness in the first place!
(RBH, this answers your question about which level ICness needs to be evaluated at: it needs to be generated in a biologically relevant fashion, which means mutations should occur in the source code which forms the functions that then coalesce into more complex entities like EQU.)
So--given this huge artifical setup, and an artifically high "mutation" rate (function swapping), AND selection for intermediate functions, the program produced the "complex" EQU function. Big whoop.
Look. Nobody is arguing that evolution can't take some sort of steps through possibility space. The key question is: How Big? The IC argument suggests that the steps required are too big for the Darwinian process. If your computer simulation (like Lenski et al) can make canyon-leaping strides in one fell swoop, and it knows which canyons to avoid (by virtue of being programmed to operate in the functional areas of possibility space), then sure, it can reach the solution you want. But this is totally artifical and non-Darwinian. Effectively, the program shrinks the space of possibilities to a small enough area that it can efficiently search it by its built-in "mutations". But this cheats--it gets around the very problems which prevent Darwinian evolution from producing ICness in the real world.
I guess my request is for the critics (RBH, PvM) to stop making overinflated claims about how the Lenski et al paper demonstrates the "evolution of ICness and there are multiple pathways to get there". They did nothing of the sort--not, at least, if you define ICness as the biologically relevant entity. And please, no nonsese about this simulation not really making claims about the real world. By claiming the Lenski et al paper as a counterexample, you're implicitly assuming huge implications for real evolution. If the paper really doesn't have implications for the real world, you all need to quit claiming that it has proven the ability of Darwinian processes to produce IC systems.
Sorry for the rant, but I'm feeling like this paper is being accepted uncritically and simply tooted at every opportunity by the ID critics who aren't really serious about carefully evaluating the evidence for and against their claims.
John
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 05. June 2003 20:55
The main "problem" with Smart's paper is basically that, with such a clear grasp of the definition vs. evolvability distinction, and of the various complexities of biology that have to be taken into account when defining multiple-parts-required systems (which is basically what he boils the concept down to), he really should just come on over to the evolutionist side and conceed that IC is not the forbidding barrier to evolution that it has so often been claimed to be!
I shall outline the reasons why:
1) The "varying required parts" problem
Smart explicitly acknowledges (in a way that I don't recall other IDists doing, certainly not emphasizing -- he sort of says that Julie Thomas did this, but as far as I can recall she looked at "parts present in multiple organisms" vs. "parts found to be required in multiple organisms") the common biological fact that homologous multiple-parts required systems with the same function, often have parts that are required in one system, but not in another! As RBH points out, unless one is going to invoke a multitude of mini-ID events for each of these "extra" required proteins in various systems (Hagemann factor is an example, it is at least very important in humans (defects are associated with miscarriages etc.) but in whales it is just a pseudogene), then the "IC barrier" has already been breached, probably by the "incremental indispensability" argument.
2) If you take a "concessionary" viewpoint, then even for the famous ID systems not much is left of the IC systems
For example, for the flagellum, Smart lists the following universally present proteins:
quote:
[note: I am making a few changes based on how I would describe the parts; this is just rearranging, it doesn't lower the parts count]
Cap: FliD
Filament: fliC [/flaA/flaB*]
Hook: flgE [hook cap: flgD?]
Adaptors: flgK, [flgL?]
Drive shaft: flgB, flgC, [/flgF?]
Motor complex: motA,motB
[Rotor "teeth" -- interact with MotA: FliG]
[Rotor/T3SS base ring: FliF]
Export machinery: flhA,flhB, fliR, fliQ, fliP,, fliI, [double comma there BTW]
Unknown fliS - essential, but function unknown
flhF - GTP-binding protein
But, if we follow the concessionary approach, we have in the T3SS homologs of:
quote:
[Rotor "teeth" -- interact with MotA: FliG]
[Rotor/T3SS base ring: FliF]
Export machinery: flhA,flhB, fliR, fliQ, fliP,, fliI, [double comma there BTW]
In the ExbB/D family we have homologs of:
quote:
Motor complex: motA,motB
...leaving:
quote:
Cap: FliD
Filament: fliC [/flaA/flaB*]
Hook: flgE [hook cap: flgD?]
Adaptors: flgK, [flgL?]
Drive shaft: flgB, flgC, [/flgF?]
...and, the interesting thing about this group of proteins includes:
(1) the apparent repetition of the filament-cap system in the hook-hook cap system
(2) sequence and structural similarities between the various rod and hook proteins (and FlgK/HAP1 I think)
(3) similarities between FliC, other filament proteins (some flagella have 6+ filament proteins), and (IIRC) FlgL/Hap3
(4) hook-type III pili extension homologies proposed in recent articles
...which all seems to point towards the potential of all of these proteins being modified copies of each other (many of them are already in the "axial protein family".
The cap proteins are not quite so clearly required parts anyhow:
- it appears that Type III pili might not have them
- flagellar filament protein will spontaneously assemble in certain conditions
- and null mutants for cap proteins are still functional e.g. in bacterial with sheathed flagella (the sheath, an extension of the membrane surrounding the flagellum, apparently concentrates the filament protein enough that it assembles without the cap)
And the cap proteins might even be modified axial family proteins themselves -- e.g., the cap-filament interactions are similar to filament-filament interactions; it takes about 5 filament proteins to complete a turn of filament, and the cap is a pentamer of cap protein.
(some refs for the above are here)
So basically, the mighty flagellum has been boiled down to 3 "parts", corresponding to:
1. A secretion system (BTW, Smart calls it an "excretion" system -- this is probably not the best term IMO :-) ). Secretion systems are ubiquitous functional structures in life.
2. A extracellular extension (secretion system + extension = pili). Pili are ubiquitous functional structures in life.
3. A stator/motor, which just happens to independently associate with the rest of the structure, and just happens to have phylogenetically widespread homologs that function in much simpler contexts.
...each of which are functional in nonflagellar contexts.
Now, we've left out the chemotaxis system and various other details:
(I suspect that FliG is more likely derived from the complement of the MotA/ExbB system rather than having been on a primitive T3SS; chemotaxis has been left out entirely but it is highly variable and not required for functions such as dispersal or even biofilm formation), and Julie Thomas/Josh Smart leave them out anyhow; discussing all of the chaperones etc. would require much more knowledge about what they and their homologs are, plus Smart leaves them out; the origins of a primitive T3SS have not been explained this is a different question, and the detailed functions of the parts are not even well-understood yet; etc.)
...but certainly the "big picture" is that the "unscalable cliffs" of this IC system seem much reduced.
3) Inconsistency in detail requirements
Smart writes, "Critical responses to irreducible complexity have primarily consisted of just-so-stories that substitute vague appeals to chance or other forces for well-thought-out counter-arguments." And actually, I think that there is some truth to this. You can go through the book reviews of Behe and find that very few do more than vague just-so stories. Part of the reason is that detailed discussions are impossible in book reviews, but much of it is that the various philosophers and organismal biologists that replied simply didn't know much about the relevant biochemistry. There were a few exceptions, e.g. Cavalier-Smith (1997, TREE) cited a half-dozen of his own 30-page-plus papers on the origin of life and various systems, work which has continued in a massive fashion to the present day (do a PubMed search on Cavalier-Smith).
However, if Smart is really looking for details, I recommend that Smart instead read the literature on the best-explored of Behe's original IC systems, the immune system. A good introduction is this article:
Evolving Immunity
A Response to Chapter 6 of Darwin's Black Box
http://www.talkdesign.org/faqs/Evolving_Immunity.html
...and particularly the various forms of conceptual flailing that resulted among the IDists on the various ISCID threads discussing it: ISCID immune thread 1, ISCID immune thread 2. (There is also an AE thread with many of the important links etc.)
When presented with detail and peer-reviewed literature on the evolution of this IC system, the IDist response was incoherant. So until someone like Smart delves into this literature and really grapples with it, and explains what exactly the IC system is and what all of these organisms with progressively simpler-yet-functional systems are doing, and how all of these evolutionary hypotheses have been proposed and tested in the field of evolutionary immunology when evolution of this IC system didn't actually happen, all claims about evolutionists lacking in detail will be unimpressive.
Finally, when Smart wrote, "Critical responses to irreducible complexity have primarily consisted of just-so-stories that substitute vague appeals to chance or other forces for well-thought-out counter-arguments" did it not occur to him that this is exactly what ID does? What is "IDdidit" except the vaguest just-so story ever told?
Smart, unlike any other ID sympathizer that I can recall, actually does come up with a reasonable notion of what "detailed" should mean:
quote:
It is not true, as has been claimed, that intelligent design theorists hold any system that has not been given a detailed evolutionary explanation as something that cannot be produced by evolution (see, for example, [5]).
[BTW, does someone recall where Dembski said essentially the opposite of Smart on this point; it was right here on ISCID]
It is true, though, that there has been no detailed evolutionary explanation given for any irreducibly complex structure.
[again, see the immune system and re-evaluate]
A detailed evolutionary explanation does not mean that every base pair and selective pressure involved in the production of the system be accounted for, and every mutation explicitly stated on a timeline. On the other hand it does mean accounting for changes in phenotypes, and the advantages that those changes give.
This final criterion for detail seems quite fair. I think that such things have been acheived for e.g. the immune system and the Krebs cycle. They are clearly constructable for the flagellum, although our evidence is much sparser there. However, what is the comparable standard of detail for an ID hypothesis? There must be some detail, or else nothing is ruled in or out by the hypothesis, and it becomes nonexplanatory and nontestable.
4) Is the Avida experiment which I'm sure RBH will discuss sufficiently. My only point would be that Smart's endorsement of the knock-out criterion would appear to clearly rule "in" the evolved EQU functions as IC.
That's probably enough for the moment. I'm not sure if Smart himself has any interest in the discussion, but if he does, I'd like to say thanks for an interesting article, and that if he follows his logic just a very few steps further he will have one of the more thorough Behe rebuttals out there. ![[Smile]](smile.gif)
[ 05. June 2003, 21:08: Message edited by: yersinia ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. June 2003 22:05
A few more remarks on Micah's posting. With yersinia, I think Smart's paper is a model of lucidity. He very clearly describes the issues and problems surrounding the determination of irreducible complexity for a given object. And I applaud him for that.
My remarks are founded on the assumption that there are really two questions at issue. First, how do we determine whether a given object or process (or (in the Evolutionary Method, a pathway to an object) is irreducibly complex? That is the focus of Smart's essay and it is a clear presentation of the issues surrounding that question.
The other question at issue in the larger debate is whether objects that are determined by one or another means to be irreducibly complex can be constructed by naturalistic evolutionary processes. Smart occasionally slips into the assumption that they cannot, and that is not a settled question by any means.
Smart has done us a service by clearly describing the IC determination issue. One of my suggestions is that he think carefully about the inherent contradiction between the operational method of determining whether an object is irreducibly complex (knockout) and his Evolutionary Method. In the latter, the procedure he suggests for determining which of a set of related objects, each itself IC by knockout, is ancestral or original or exemplary, and which are derived, there is the implicit assumption that the 'derived' IC objects evolved from the original object. The irreducible core of the "earliest known version" of an object is taken to be the irreducible core of ... well, what? Apparently the ancestral form. The Evolutionary Method glides past the question of how the several forms (each of them irreducibly complex by the knockout criterion) arose if one of them is the defining exemplar and the others, though having a different combination of indispensable components by the knockout criterion, are nevertheless somehow derived.
Similarly, Smart moves from this method of defining the ICness of objects to thoughts about irreducibly complex pathways. Following Behe, Smart takes an irreducibly complex pathway to be a sequence of precursors leading to the defining exemplar. Here, in contradiction to Micah's appeal to mechanical functions, Smart is focusing on a sequence of forms, an abstract pathway in a conceptual space. If the 'gap' between two precursor forms is thought to be too wide to be bridged by a plausible evolutionary process, then the pathway is taken to be irreducibly complex. It is noteworthy that an irreducibly complex pathway need not in principle have as its end point an irreducibly complex object. It might be tactically advantageous for research to start with an irreducibly complex object in the effort to determine whether the pathway leading to the object is IC, but I see no principled reason for it to be a necessary property of an irreducible pathway that it end at an IC object. The two - object ICness and pathway ICness - are conceptually independent.
As I have argued elsewhere, a significant obstacle to the progress (and to the acceptance) of ID is that while various writers have suggested methods for detecting design (EF, CSI, SC, IC) that focus on the properties of things, none has been systematically validated on objects and processes of known provenance. No one has done the "detailed, meticulous" work necessary to establish that the various methods in fact do what is claimed for them.
A substantial resource for that project is the information from the Lenski, et al., study that is supplied in the Supplemental Information. Moreover, the complete study can be replicated on a suitable machine - everything necessary has been provided. I can think of no better test bed for developing validation data for the various conceptions that are at the heart of ID. That is why I think it is relevant to Smart's essay, Micah.
RBH
Postscript (that really belongs on the appropriate thread in the Literature Review Forum but is addressed to John Bracht's posting above):
Bracht wrote quote:
(RBH, this answers your question about which level ICness needs to be evaluated at: it needs to be generated in a biologically relevant fashion, which means mutations should occur in the source code which forms the functions that then coalesce into more complex entities like EQU.)
Bracht here displays a fundamental lack of understanding of the Avida simulation's mechanics that vitiates his criticism of the conclusions drawn from it. What he argues mutations should have been is exactly what mutations were in the Avida simulation: they occurred at the level of the source code - the assembly language primitives that formed strings of instructions that coalesced to perform selectively advantageous functions like replication and a subset of all logic operations. There was no block "function swapping" in the sense I understand him to use the term. What has hindered his understanding is the failure to distinguish between the "nand" primitive instruction and the programs' performance of the NAND logic operation. They are different. The "nand" primitive cannot by itself perform the NAND logic operation. In fact, looking through the source code of the 23 evolved EQU-performing programs, there are several in which the "nand" primitive appears in places where it does not appear to perform any function relevant to EQU. The knockout operation reveals "junk" instructions in all of the 23. In one program (Run 120) the primitive "nand" instruction appears nine times, four of which are "junk" occurrences by the knockout criterion - no logic function is lost when any of those four is knocked out! Hence Bracht misunderstands the mechanics of the simulation.
Second, the core question is whether the complex objects - assembly language programs - that evolved in the Avida simulation are irreducibly complex on any reasonable reading of the several definitions of that phrase. On the original Behe definition and on the modified Dembski definition they are IC: both definitions preserve the knockout criterion for operationally determining ICness. What Bracht must argue is that the evolutionary operators in the Avida simulation are so implausible that the emergence of 23 different programs that performed the EQU logic operation are somehow accounted for by other than the kind of evolutionary operations that characterize biological evolution. The question is not whether the simulation models biology, it is whether it models evolutionary processes and operators veridically enough to warrant the conclusion that ICness can evolve. So far Bracht has not addressed that question beyond complaining that the simulation achieved its results by using what he deems to be a too-high mutation rate so that somehow, "function swapping" achieved the outcome. But that complaint depends on his misunderstanding of what mutated, not at how high a rate mutations occurred. Lenski, et al, ran appropriate control conditions to rule out precisely the explanation Bracht seems to be offering, that finding programs that perform EQU could have occurred by the chance assemblage of the functional blocks he thinks were being "swapped" around in the mutation process. They weren't.
Later this summer, when I get the simulation running on the local Beowulf cluster, I intend to test his mutation rate claim. I'll post the findings here.
RBH
[ 05. June 2003, 22:09: Message edited by: RBH ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 05. June 2003 23:10
It occurs to me suddenly that everything would be much clearer if the term "irreducibly complex pathway" were just dropped. The words "irreducibly complex" appear to mean entirely different things when in the phrases
"irreducibly complex system"
vs.
"irreducibly complex evolutionary pathway"
In the first case, if we take the present-characteristics definition, then the meaning of "irreducibly complex" is quite clear: complexity that can't be reduced (while retaining function X).
In the second case, referring to pathways...the words "IC" bear no relation to the pathway's complexity or irreducibility. An "IC pathway" is not, for example, a pathway that can't be reduced (whatever that would mean). What is meant is an unworkable/unlikely pathway, with varying degrees of unlikeliness depending on e.g. the number of neutral steps required (according to Behe's original usage; there are problems as neutral mutations do spread to fixation all the time but this is a detour). But if you mean "highly improbable pathway" then you should just say that, and not confuse the issue by adding "irreducible complex" just because it is your favorite term or sounds spiffy.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 06. June 2003 00:59
I found Joshua Smart's paper to be very clear and well-reasoned. I especially liked the description (and examples) of how to determine whether a given structure was irreducibly complex, under various different criteria, and the discussion of the strengths and weaknesses of each criterion.
However, like others, I would note that the connection between having an irreducibly complex core and being unevolvably complex is not clear-cut. The two can be conflated only if the IC core is also the most probable evolutionary ancestor. For example, structural complexes may most easily be built by incorporating a huge number of components that interact weakly (creating a huge, inefficient glob that performs whatever strucutral function is advantageous, but which is easy to create since it doesn't much matter what is in the glob, only that there is one), and then throwing away components while refining the structure of the others, until a minimal set remains. This would produce a system that contained only the IC core and was highly efficient, but the chance of directly producing that core from scratch would be irrelevant.
Of course, the complex systems as a refined glob scenario may not occur often or at all in real biological systems. What is really needed is an extra test that could tell us when an IC system must have been created from simpler systems upwards to the core, instead of refined downwards from large messes. Unfortunately, I can't yet think of a good way to do this yet, but I would encourage people to think about how it might be done.
In the meantime, we may get some of our most instructive results from programs like Avida, where we know that any answer that is constructed was a result of pseudorandom mutation and selection. If Avida produces IC systems, it is important to understand why and how, and to what extent the same factors could be playing a role in biology. (Simply noting that Avida is not biology is not enough; rather, the question would be whether there are attributes of Avida that are essential for development of IC systems but which are absent in biological systems (or vice versa).)
IP: Logged
|
|
|
|
|
Printer-friendly view of this topic