|
Author
|
Topic: William A. Dembski: Evolution's Logic of Credulity: An Unfettered Response to ...
|
William A. Dembski
Member
Member # 7
|
posted 04. December 2002 15:04
Evolution's Logic of Credulity: An Unfettered Response to Allen Orr
by William A. Dembski
William_Dembski@baylor.edu
ABSTRACT—Allen Orr wrote an extended critical review (ca. 6000 words) of my book No Free Lunch for the Boston Review summer 2002 (http://bostonreview.mit.edu/BR27.3/orr.html). The Boston Review subsequently contacted me and asked for a 1000 word response. I wrote a response of that length focusing on what I took to be the fundamental flaw in Orr's review (and indeed in Darwinian thinking generally, namely, conflating the realistically possible with the merely conceivable). Of course Orr had the last word, with the Boston Review giving him 1000 words to reply to my response (for the exchange see http://bostonreview.mit.edu/BR27.5/exchange.html).
In his reply Orr takes me to task for not responding to the many particular objections he raised against my work in his original review, suggesting that this was the result of bewilderment on my part and intelligent design running out of steam and not, as was the case, for lack of space. This sort of rule-rigging by Orr and the Boston Review -- give the respondent a little space, and then let the original author crow about winning -- is to be expected. I actually find it encouraging, taking it as an indication of intelligent design's progress. Orr's review and follow-up hardly spell the death-knell for intelligent design or for my work in this area. Sooner or later (and probably sooner) Orr will find himself in a forum on intelligent design where the rules of engagement are not rigged in his favor. I look forward to his performance then.
To read the entire paper, please click here
[ 08. January 2003, 07:25: Message edited by: Moderator ]
IP: Logged
|
|
yersinia
Member
Member # 324
|
posted 04. December 2002 18:07
Dr. Dembski writes,
quote:
But what if we weren't sure that there even were any car keys? The situation in evolutionary biology is even more extreme than that. One might not be sure our hypothetical set of car keys exist, but at least one has the reassurance that car keys exist generally. Indirect Darwinian pathways are more like the supposed leprechauns a child is certain are hiding in his room. Imagine the child were so ardent and convincing that he set all of Scotland Yard, indeed some of the best minds of the age, onto the task of searching meticulously, tirelessly, decade after decade, for these supposed leprechauns, for any solid evidence at all of their prior habitation of the bedroom. And then imagine that in all those decades, the detectives, driven by gold fever for the leprechaun's treasure let's say, never flagged in searching out and postulating new ways of catching a glimpse of a leprechaun, a leprechaun hair, a leprechaun fingerprint, any solid clue at all. After these many decades, with not a single solid clue to show for all that work, what should one say to the aging parents of the now aging boy if these parents decided there were no leprechauns in the boy's room? Would it be logical to shake your finger at the parents and tell them, "Absence of evidence is not evidence of absence. Step aside and let the experts get back to work." That would be absurd. And yet that, essentially, is what Orr and his fellow evolutionary biologists are telling us concerning that utterly fruitless search for credible indirect Darwinian pathways to account for irreducible complexity.
...and...
quote:
And apparently not by indirect Darwinian pathways either -- the absence of scientific evidence here is complete (critics who claim otherwise are bluffing). What's more, appealing to unknown material mechanisms is even more tenuous.
Unfortunately, it is Dembski who is bluffing. His bluff was called over on the ISCID immune system thread:
http://www.iscid.org/boards/ubb-get_topic-f-6-t-000152
...where even many IDists or ID-friendlies (such as Mike Gene and Martin Peonie) appeared to concede that Darwinists had found a substantial amount of evidence that the (supposedly IC) immune system(s) had evolved, just like Behe said they couldn.html't. Certainly they would agree that "solid clues" and the equivalent of "fingerprints" are readily available in the case of the evolution of the immune system.
Assertions that Dembski makes, like "utterly fruitless", are therefore without basis and his continuing to repeat them after his error has been demonstrated, in rather a lot of detail, with rather a lot of references to peer-reviewed literature, on his own discussion board, indicates that he has little knowledge of the field he criticizes, namely biology, and his arguments will only appeal to those who are similarly underinformed.
yersinia
IP: Logged
|
|
Noel Rude
Member
Member # 516
|
posted 04. December 2002 21:54
Materialists endlessly lecture the opposition on the nature of science -- so it's good to see this rebuttal from Bill Dembski. It's important we not forget philosophy, that we not just focus on ID as a scientific procedure or methodology, in which case we will lose ourselves in the thicket of technical detail and constantly be accused of not being a "science".
But as for methodology -- I've posted on this elsewhere on ISCID (forget where in my senility), but I think it worth mentioning again: the methodology of historical linguistics is similar to ID methodology and provides a good analogy.
Refutability and Evidence
We cannot prove that any two languages are not related, we can only garner evidence that they are, and when we do there are, as in ID, no false positives. Chinese and English, for example, may not be related historically but we cannot prove this. To demonstrate that languages are related we look for potential cognate vocabulary, and should we amass a sufficient amount (complexity) we then check for regular sound and grammar correspondences (specification). We might quibble over the amount of complexity and specification necessary to establish a relationship, and the relationship may only pertain to a portion of the language (e.g. a sizable chunk of English vocabulary is Latinate, but the more basic vocabulary and the grammar are not) -- but everyone agrees that there does come a point when a relationship has been established.
Now it could be, as I said, that English and Chinese are related but at such a distance that we simply have not been able to catch on to it as yet. English and Hindi, for example, are related -- but this is not readily apparent. Sound change -- regular sound change -- has obscured many cognates. A classic example is English wheel and Sanskrit/Hindi chakra -- which share not one single sound but do mean the same thing and are relatable by regular sound change to the parent Proto-Indo-European *kwekwlos 'wheel'. Whereas it has been established by such means that English and Hindi are cognate languages, it cannot be established to the same degree of certainty that English and Chinese are not also related.
Scientific evidence, as Demski reminds us, and the evidence of linguistic relationship, is never absolute -- it could be that all the remarkable similarities between English and Hindi are coincidental -- but the odds of this are astronomical -- probably disallowed within all the time provided by the cosmologists.
But suppose I want to demonstrate a relationship between English and Chinese -- promisory notes and pleas that the evidence will be found mean nothing. I've got to provide the evidence which, as I said, must be both complex and specified. Dembski's Causal Adequacy? Language change is a well known phenomenon, we observe it over even a decade or two and we can trace it in written languages over millennia. We know how dialect change over time in, say, Vulgar Latin has given us our modern Romance languages. We even have good theories as to how change is instigated and spread, which brings me to the following:
Structural and Functional Change
Biological evolution involves the transmission of genes, historical change in language does not alter the gene pool (though the Darwinists on another computer network -- FUNKNET -- are arguing this right now). But this does not mean that language change is completely random -- structural changes are unidirectional, as, for example,
k becomes ch becomes sh becomes s becomes h becomes zero
a word becomes auxiliary/adposition becomes affix
and function also changes along predictable, one-way paths, such as
'go/come' becomes FUTURE TENSE
'will/control' become FUTURE TENSE
'give' becomes 'to' becomes DATIVE becomes ACCUSATIVE
Even though language is as much design as anything, it provides an example of change that is evolutionary but only in the sense of a continuous recycling of structure and function along one-way paths. This equating of structure and function is analogous to the biological situation. And when Bill Dembski speaks of direct and indirect Darwinian pathways, I wonder about possible analogues in linguistics -- but draw a blank at the moment.
[ 05. December 2002, 08:49: Message edited by: Noel Rude ]
IP: Logged
|
|
Mark Szlazak
Member
Member # 391
|
posted 05. December 2002 11:13
Boy those leprechaun examples sure get around! My former collegue Dale Nansel used a leprechaun exterminator analogy in a letter to the editor back in the late 1980's. I've used leprechauns in characterizing materialism in another thread on this board.
Also, a previous post mentions a supposedly confirmed Darwanian pathway for an aspect of immune system function. If this is the case, then how does it relate to the direct and indirect issue mentioned in Dr. Dembski's response?
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 05. December 2002 12:27
I am a bit confused about the leprechaun argument. Would that not be an example of a false positive for intelligent design?
Science on the contrary is learning more and more about the possible pathways and the evolutionary relationships of for instance the flagellum and other 'IC' systems
So perhaps the leprechaun makes for a better analogy for intelligent design?
[ 05. December 2002, 12:32: Message edited by: Frances ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 05. December 2002 14:50
Dembski's "unfettered" response to Allen Orr contains material that has appeared elsewhere recently. In particular, his remarks on evolutionary algorithms, fitness functions and the "displacement" issue were tried out a few weeks ago on ARN in the What Genetic Algorithms Can Do thread. It is not clear that the several responses to that tryout had any effect on the finished "unfettered" essay. In particular, Gedanken provided strong points in criticism on that thread and I'll draw on his posts as well as my own in this post since the "unfettered" essay contains verbatim copies of parts of Dembski's ARN posting.
Constraints and Models
In Section 4, "Displacement - The No Free Lunch Principle" (beginning on p. 10-all page numbers from my version of Word), Dembski wrote quote:
Obviously, the way around NFL is to start constraining a given class of fitness functions. So you don't like time-dependent fitness functions that vary independently of the progress of the evolutionary algorithm in reaching a solution; therefore constrain this class of fitness functions so that they depend on progress to a solution. So you don't like fitness functions closed under permutation (see Igel and Toussaint); therefore focus on classes that are not closed under permutation. All such focusing and constraining imparts information. Provided that information is both complex and specified, I show you never get more specified complexity out of such evolutionary processes than was programmed into them through such constraining. (p. 11)
"Constrain" and "like" and "the way around NFL" are strangely pejorative locutions to use in this context. The object of the game is to model and test the natural processes hypothesized by evolutionary biology to account for biological phenomena. The "constraining" condition on the model is fidelity to nature. When one focuses on classes that are not closed under permutation one is focusing on the class of fitness functions that are biologically relevant. The question at issue is about the sort of fitness functions that describe biological nature, and the answer is fitness functions that are not closed under permutation. Hence the NFL theorems are not applicable to the fitness functions relevant to modeling biological evolution and Dembski's NFL-based remarks are merely irrelevant. To observe that the NFL theorems don't apply to the relevant class of fitness functions is not "getting around" them; it's pointing out the misapplication of the NFL theorems in "No Free Lunch."
And Dembski's assertion quote:
So you don't like time-dependent fitness functions that vary independently of the progress of the evolutionary algorithm in reaching a solution; therefore constrain this class of fitness functions so that they depend on progress to a solution.
is simply incoherent. Once again, it's not a question of "don't like," it's a question of what is an appropriate model of biological evolution, where "appropriate" means one can map the relevant properties of natural fitness functions in a co-evolving biological system onto the properties of the fitness functions enabled in a model. Biological fitness functions and their induced fitness landscapes change with changing co-evolving populations, and thus one is directed to that class of fitness functions by the phenomena, not by some arbitrary liking for them. (Parenthetically, while I won't develop the point here, Dembski's "progress to a solution" locution betrays a fundamental misunderstanding of evolution in biology. As I have remarked elsewhere, biological evolution is not a goal-directed solution-optimization task. To model it as such, as MESA does, is to both misrepresent evolutionary biology and beg the question of teleology.)
Gedanken said it well in the "What Genetic Algorithms Can Do" thread on ARN: quote:
But isn't the reason they are not "liked" because they don't properly model the real world? The real world process is most certainly one in which the fitness landscape is sculpted by the existence of other organisms in the current environment. Hardly a smuggling of "design" information to properly model the real environment. "So you don't think the fitness functions are closed under permutation; therefore focus on classes that are not closed under permutation." But fitness functions that do not build on the shoulders of the current environment are not accurate representations. If only trivial problems are "closed under permutation" then it is no "smuggling" act to model with more realistic real-world models.
Clearly the existence of multiple species/populations and the fact of competition among and within them, with the attendant selective pressures and the resulting differential reproductive success, are in some sense "specified." If I write an evolutionary algorithm as a model to study co-evolution, I must make provision in the program to allow (but not require) it to occur. But with nothing more than differential reproduction consequent on variations in the ability to utilize environmental resources, the several populations shape and change each other's fitness landscapes. Not only do I not determine specifically how those fitness functions vary with time when I write the program, I don't - I can't! - know specifically how they did so without keeping careful historical records of the evolutionary runs and doing laborious post-run analyses of those records to ascertain the interactions among the several species that arise. And it is not, as Dembski claims, that "complex specified information" is smuggled in by the model parameters. The 'information' and 'complexity' and 'specification' in the co-evolutionary biological world are generated by interactions within that world.
"Smuggling" versus Veridical Modeling
Dembski makes the same claim (verbatim) here as he did in his response to my OP in the "What Genetic Algorithms Can Do" thread on ARN: quote:
Orr's shifting fitness landscapes that magically generate specified complexity float in from nowhere. As soon as you specify them and the process that varies them, you'll find that you've introduced specified complexity.
But "you" don't specify fitness functions so they magically float in from nowhere, either in nature or in an evolutionary algorithm. The algorithm itself, via interactions among co-evolving populations, "specifies" them in a process akin to bootstrapping. The "process that varies them" - co-evolution - is just the process hypothesized by evolutionary biology: Dembski's criticism is that the model represents the phenomena it models!
Parasitism, symbiosis, pollinator/pollinated, and predator-prey interactions are among the more obvious biological phenomena in which the fitness landscapes of one species are defined and deformed by evolutionary changes in another species. The several fitness landscapes on which a parasite evolves are in part controlled by the evolutionary responses of the host species to the (changing) parasitic load. Hence "specifying" general conditions in which parasitic co-evolution is allowed to occur in an evolutionary algorithm is modeling the biological world to test hypotheses. One doesn't have to explicitly provide "complex specified information." Given the opportunity, the evolutionary process modeled in the algorithm generates it on its own, so while it may be "complex" in Dembski's idiosyncratic improbability-against-a-uniform-PDF sense, it is a lawful outcome of known evolutionary processes and thus not "designed" in the sense of the Explanatory Filter.
Dembski says quote:
The interesting thing is that a co-evolutionary process doesn't buy you anything that you didn't already have with a standard evolutionary process once the processes in question are specified. In the latter case you only have to specify a given fitness function. In the former you have to specify how the fitness functions vary with time.
That's an accurate description of what one does in writing an evolutionary model, but, ironically, it is not a criticism, it is an affirmation. It says the model is a veridical representation of the biological phenomena. In writing an EA one does define the conditions in which changes in fitness functions (and in the several fitness landscapes induced by evolutionary operators) are allowed to occur. One does not "specify" the precise form and time course of changes, just the general conditions (the choice guided by theory) that allow them to occur if the algorithm's behavior induces it. But since evolution is determined in part by the variability made available by random mutations, one cannot determine in advance exactly how the fitness functions and the associated landscapes will change. That is historically contingent and depends on the variability thrown up by random mutations. However, the processes hypothesized by evolutionary biology, once they are embodied in the model, behave much as they are hypothesized to occur in nature, constituting a passed test of the theory that generated the hypotheses. Once again, Dembski's complaint seems to be that the model represents the biological processes being modeled.
Co-evolution
Dembski claims that Section 4.10 of NFL covers the problem of co-evolving fitness landscapes. He writes quote:
In that section I show that co-evolving fitness landscapes are mathematically equivalent to evolution with respect to a fixed fitness landscape. The argument requires reconceptualizing the configuration space so that co-evolving fitness, as it were, gets built into it. (p. 13)
Again I quote Gedanken at length from the ARN thread: quote:
Now the 5 pages of section 4.10 actually do appear to recognize the problem that the previous 45 pages ignore: that neither "blind search" nor the NFL theorems are based on a representation of how evolution is really considered to work.
There are two dimensions of aspects to consider:
1) The evolution really is "co-evolution" as is recognized in section 4.10. Humans could not live alone without a large fraction of the present non-human life on our planet in support. We could not evolve or change because we could not even continue to live without all the supporting food and other cycles of our environment. The section 4.10 correctly points out that the NFL theorems don't apply because they are based on a fixed fitness landscape. (Humans could not evolve to be better hunters without something to hunt!)
2) Then the basis of each next "step" in evolution is to build on the shoulders of the last step. This aspect is both of the "search" and of the "fitness landscape". (See aspect "1" for the basing of the present step on modified "fitness landscape" built on the shoulders of the previous time step.) Dembski's search example of starting with a fully randomized Rubik's cube and performing a "blind search" to attempt to reach a fully solved puzzle is misleading. The evolutionary step that is performed is much more like a Rubik's cube that is only one twist from complete. Then most random trials fail, but they should then be reset before trying the next. Eventually the "goal" twist will be found which has greater 'value' of some sort.
The problem is of course where the 'value' comes from in the fitness landscape. Dembski claims that it must be smuggled in, because (like the "Methinks?" problem) the simulations are all designed with human written goals encoded into the program. But of course all programs are written by humans, and the goals must by definition be "specified" because they are written computer statements. The real question is how well the fitness descriptions and the simulation aspects actually model the real world.
If they accurately model the real world, then we must observe a couple of things. The "displacement problem" must mean that aspects of the real world provide source of the CSI. The real world's characteristics must be providing the specification. And if this is evidence that the world as a whole is "designed" then I have no argument. For this is not an argument against the correctness of evolutionary algorithms as an assessment of the operation of evolution in the real world. That the CSI is built into the basic relationships of physics would be no conflict for evolutionary theory. Any claims that the GA model does not correctly model the physical world would be completely reasonable consideration. (And if some "goal" were smuggled in by such failure to accurately model the real world, that is a valid criticism of the GA model at hand. But remember that such models are usually attempts at Western reductionist abstraction, and do not attempt to completely model the whole physical world but rather a small and abstracted part of it. Claims that a model is an abstraction and not complete is not a valid criticism of theory that claims to be approximated by such simulations. (emphasis added)
Appropriate ID Research Assuming "Displacement"
Finally, if Dembski is right about "displacement," and if biological evolution is in the end merely an elaborate copying system transcribing "specified complexity" from the non-biological physical environment, thereby displacing the source of SC to the non-biological environment, then the IDist research focus should obviously shift to the physical environment as the locus of designer interventions and quit fooling around with biology. See Section 5 of Wein's critique of NFL for the inconsistency in Dembski's position in this respect.
As is obvious, I don't believe that Dembski is right. It is clear to me that the kind of "specified complexity" (or "specified information" or whatever neologism is appropriate) of interest can in fact be generated by (among other things) regular co-evolutionary interactions both in nature and in appropriately built EA models of natural evolution. But ID is welcome to take on physics. Perhaps the infinite-wavelength zero-energy signals by means of which unembodied designers are said by Dembski to be able ("in the limit") to transfer designs to flesh and blood will generate some interest there.
RBH
IP: Logged
|
|
Noel Rude
Member
Member # 516
|
posted 06. December 2002 11:04
Back in the late 50s certain linguists, skeptical of Skinnerian behaviorism, wanted to show that the output of Natural Language was compatible with the creativity we intuitively sense is there. And so Bill Dembski mentions "Chomsky's demonstration that ... finite state automata are incapable of generating certain languages". This is because Chomsky and his followers thought that embedding in complex sentences is what allows for the open ended output of language. But language is hierarchical from the smallest features of sound and meaning up to the longest conversations and most ponderous books and essays, and so if perchance we found a language without complement clauses (sentence level embedding), it would be gotten around. Semantic embedding ('I want her to write me') need not be coded by structural embedding -- you could say, "I want it, she will write me" -- where reference and coherence principles operating across main clauses do the same trick. The process is already there in all languages, and this is how some languages do say 'I want her to write me'. Thus open-endedness has been demonstrated in linguistics on both logical and empirical grounds.
[ 06. December 2002, 11:07: Message edited by: Noel Rude ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 06. December 2002 12:22
Noel wrote quote:
Semantic embedding ('I want her to write me') need not be coded by structural embedding -- you could say, "I want it, she will write me" -- where reference and coherence principles operating across main clauses do the same trick. The process is already there in all languages, and this is how some languages do say 'I want her to write me'. Thus open-endedness has been demonstrated in linguistics on both logical and empirical grounds.
Two brief remarks. First, that's why I was always more comfortable with generative semantics (and the general notion that meaning rather than form is core) than with transformational generative grammars. The Chomskyan notion of the centrality of syntax bothered me when I first read Syntactic Structures more than 30 years ago, even though his 1959 review of Skinner's Verbal Behavior legitimized work that cognitive psychologists had been doing out of sight of (and ignoring the strictures of) radical behaviorists since at least as early as Bartlett in the 1930s. The Sapir/Whorf hypothesis that emerged from anthropological linguistics was a strong stimulus to look at something other than stimulus-response pairings, too, before Chomsky.
Second, though I've been away from these particular issues for three decades, I always liked "generative" and "open-endedness" better as characterizations of human natural languages rather than "creative." The latter seems to me to carry a heavy connotational load that may or may not be justified.
RBH
[ 06. December 2002, 12:24: Message edited by: RBH ]
IP: Logged
|
|
Noel Rude
Member
Member # 516
|
posted 06. December 2002 15:32
RBH says quote:
I always liked "generative" and "open-endedness" better as characterizations of human natural languages rather han "creative." The latter seems to me to carry a heavy connotational load that may or may not be justified.
That's interesting because Chomsky, whatever his formalist/structuralist position (not to mention his politics), has always been accused of being an anti-Darwinist, especially among we functionalists (who tend to be ardent Darwinists). I remember how in his Language and Mind Chomsky went out of his way to distinguish between human creativity (for which he said he had no explanation) and the syntactic mechanism through which it operated.
IP: Logged
|
|
Daniel Edington
Member
Member # 421
|
posted 06. December 2002 22:51
Quite frankly I agree with Orr, in that Dembski's response is indicative of the fact that his response was the result of bewilderment and the fact that intelligent design is running out of steam.
Too be sure, Dembski didn't get anywhere near as much space to respond and thus could not have responded to all the claims in Orr's review. However, Orr was not complaining about Dembski not responding to all of his claims. He was upset that Dembski not only picked out of his entire review a point he felt was trivial, but Dembski also misrepresented this point as the central idea of his review.
As far as his comments on Irreducible complexity..
quote:
Last, I can't help but wonder why Dembski's so worked up about irreducible complexity in the first place. Irreducibly complex systems do show specified complexity, but so do non-irreducibly complex ones. METHINKS IT IS LIKE A WEASEL is specifically complex (at least if it were longer) but it's not irreducibly so. So why the special treatment? Dembski seems to imply that irreducible complexity is special because it shows some structures can't be reached by smooth fitness functions. But this is refuted by scaffolding and incremental indispensability. The fact is that irreducible complexity plays no definable role in Dembski's view specifically and poses no challenge to Darwinism generally. The idea is dead and it's time the ID community gave it a proper burial.
Orr is quite correct in his appraisal of IC. It has been refuted, in addition to this it has also been show that the idea of irreducible complexity has not been experimentally demonstrated in biological systems.
The question is: If we take a biological system (say the bacterial flagellum) and destroy or remove a componant. Say then the system stops functioning (at least as a flagellum.) What does this result show? Does it show that the system is irreducibly complex and that no darwinian type process could have produced it? Or does it show that the darwinian type process that produced it is effectivily irreversible?
To be sure this experiment is inconclusive, at best. Without experimental confirmation of ireducible complexity, the idea is dead. I hope you all can attend the services.
[ 06. December 2002, 22:55: Message edited by: Daniel Edington ]
IP: Logged
|
|
Noel Rude
Member
Member # 516
|
posted 07. December 2002 21:57
Have been following some of these ISCID threads and find it a bit discouraging. Many of the critics of ID seem discouraged as well -- what's wrong with those ID folks that they can't see the light? All of which reinforces the impression that we differ at a profound philosophical level and that we simply are not going to come to a meeting of minds by arguing "scientific" detail. In response to the sharp criticism leveled at Dr. Dembski in this thread I'd like to brainstorm on the following three points where some of us might be able to reach agreement.
The Purpose of Science is to Find God
Here I think theists and animists and atheists can find some agreement -- if science (which, after all, is the serious quest for knowledge) -- if science has a purpose then surely it cannot exclude searching for the highest levels of purpose -- why and how did we get here? Etc. The philosophical materialists will argue that science has come a long way in authenticating their stance, ID will argue the opposite, and the more open minded will say that the intellectual ferment is healthy. Only the sharp demarkationists and theistic evolutionists will argue that the Purpose of Science is NOT to find/discredit Purpose -- and hopefully with them we can simply agree to disagree. Thus I think that Phillip Johnson in another thread made his point well -- we need to come to a meeting of minds at the philosophical level before we can begin to understand each other -- science issues from our philosophy -- cultures lacking the proper philosophy will not have science -- and perhaps if our culture drifts too far from its roots it will lose (or misuse) the science bequeathed by those roots.
Evolution is Evidence
I like to shock my colleagues by asserting that Evolution is evidence for Creation. But of course my Darwinist friends think it is evidence for Evolution. Both parties (the Darwinists and the Creationists) are guilty for the misunderstanding (I think others in ID have noted this). Because Darwinism is a theory of evolution, Darwinists tend to assume that any evidence that evolution has occurred is evidence for their theory. The Creationists who argue from a certain interpretation of Genesis also tend to assume that evolution and Darwinism are one and the same thing. But they are not.
Darwinism was an attempt to explain evolution, and therefore any evidence that evolution has occurred cannot be taken as ipso facto evidence for the theory. Because ID is neutral to the thinking of theologians we can build our theory on the design constantly emanating from ourselves: We can observe ID in action from every possible angle, and when we do we will see that it too is evolutionary -- the only evolution we can actually observe is teleological -- man's technology (I believe Bill Dembski has been talking about this too). Therefore evolution is evidence, not just for Darwinism because Darwinism demands it, but also for Design because ID predicts it. Thus the proper question is: Which model best predicts the kind of evolution that paleontologists and microbiologists are able to describe.
Now what if it turns out that, say, for the bacterial flagellum, or, say, the origin of life itself, we are able to imagine a highly improbable materialist solution -- say so improbable that it could only have happened once in cosmic history -- which would we opt for, the Darwinian solution or ID?
Isn't your answer to this question dependent on your philosophical predisposition?
Remember -- once you let design in the door -- it's there! -- be it in anthropic phenomena, the origin of life, evolution ... allowing it in at one juncture makes it available elsewhere. Materialists and IDists alike should be able to agree on that.
Anyway ... I think it is important for ID that we constantly remind ourselves and others of the difference between evolution and the materialistic doctrine that says chance and necessity alone can account for it. We need to remember that the first preference of the materialists was an eternal, steady-state cosmos. Evolution requires explanation of a different order than the mere description of what IS.
Research Program
Here I think ID must be very careful. If we are scientifically inclined -- and also in view of the socio-political esteem "Science" enjoys in the culture -- we could let our biology envy get us lost in the forest of research programs with the promise of great discovery.
What if it turned out that after years of such programs little of anything was discovered specifically because of ID? What if describing what IS depends less on theories of origin than on simple investigative prosedures? In my field of linguistics some give lip service to Darwin, and though some will disagree with me here many will agree: Virtually nothing in linguistics has been discovered due to a Darwinian approach. My impression is that the same applies in biology -- 150 years has given us a grand Materialist Myth and lots of story telling -- but I'll wager that the dramatic progress in biology has little to do with anyone's Darwinism.
But don't get me wrong. We should milk our theoretical models for all they're worth -- both Darwinism and ID -- we need all the inspiration we can get in our research programs. And I'll bet that ID will be quite fruitful. But if not, so what?
[ 08. December 2002, 00:37: Message edited by: Noel Rude ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 07. December 2002 23:21
Thanks to RBH for the kind words. (And we’ll see later about thanking RBH for drawing me into ISCID this way. I’ve thought about it from time to time but avoided so far.)
I thought that a little extension or clarification of might be in order.
Dr. Dembski’s Unfettered Response states:
quote:
Actually, I'm quite aware of all the points Orr raises in this paragraph. What's more, I never argued that Darwinism requires or entails that evolution proceed toward simplicity. My point was simply that Darwinism, in itself, does not mandate increasing complexity and inherently favors simplicity. Thus, if we see increasing complexity, something besides Darwinism must be at work. Now Orr offers several rationales for why we should expect increasing complexity strictly on Darwinian grounds (e.g., the irreversibility of certain changes and the lower wall of complexity below which things are dead). But all of these rationales are post hoc -- in each case the opposite might well have happened and Darwinism would still be true. Thus we can imagine (and even program on computer) Darwinian evolutionary scenarios in which reversibility has a selective advantage, in which arms races are won by simplifying, and in which the lower wall of complexity is an absorbing barrier where maximal fitness is conferred by being maximally simple.
(My emphasis)
I don’t think that Orr disagrees on scenarios in which evolution tends toward simplifying. In fact this would be inherent in Dr. Dembski’s comments on “scaffolding” that Orr remarks on in his review, with regard to Darwinian explanations for Behe’s irreducible complexity. Although Orr comments that other pathways could amount to continual increase of complexity toward IC, his comments indicate that both would be found in Darwinian explanations, and that in fact Dembski also accepts the reasonableness of both.
The question at hand in a significant part of Dr. Dembski’s entire book NFL, as well as the main point of Unfettered Response is the claim that Darwinism doesn’t predict large aggregate periods of building of complexity that Orr identifies. Orr shows us that for life to come from zero complexity to today’s level of complexity, that over the entire period of life that evolution must have had a net or average trend toward greater complexity. (This would in fact be the case if at present life’s complexity had stabilized, and that on average all life is no longer becoming complex -- precisely because we are averaging over the entire time period that includes the startup time period from zero complexity.)
However Dr. Dembski has significantly failed to support the point that “…Darwinism, in itself, does not mandate increasing complexity and inherently favors simplicity. Thus, if we see increasing complexity, something besides Darwinism must be at work.” Even though he accuses Orr of not reading section 4.10 of NFL:
quote:
The footnote to which Orr refers occurred in section 4.8 of No Free Lunch. Perhaps Orr stopped reading after that section because if he had continued to section 4.10, he would have found an entire section (and thus considerably more than a "brief moment") devoted to fleshing out that footnote, a section titled "Coevolving Fitness Landscapes." In that section I show that coevolving fitness landscapes are mathematically equivalent to evolution with respect to a fixed fitness landscape. The argument requires reconceptualizing the configuration space so that coevolving fitness, as it were, gets built into it. The upshot is that coevolution introduces no new mathematics and therefore no way out of the displacement problem. If displacement is a problem for evolution with respect to a single fitness landscape, then it remains a problem for coevolving fitness landscapes. That's why in the footnote that Orr cites I state -- and continue to maintain -- that my argument remains intact. Orr should have read further.
These statements would lead one to think that Dr. Dembski is claiming that by representing the overall evolutionary process as a single search algorithm over a fixed (or strictly time sequenced) fitness function, combined with the NFL theorems, is therefore evidence against Darwinian evolution producing complexity!
Dr. Dembski has made no such demonstration in section 4.10. As I understand it, most applications of the NFL theorems would be to individual search models of individual organism’s properties -- not aggregate models of the entire life structure of the planet Earth. (We might model all such populations of the Earth in an aggregate model, applying to individual populations respectively, but then we have the circumstance to be avoided of coevolution, which invalidates the NFL theorem’s applicability. The “fitness functions” and “evolutionary algorithms” being discussed are normally referring to individual populations respectively, rather than to aggregate of all lifeform.) But let’s concede that such aggregate modeling was realistic, just to see where it takes us. The model must then search the cross product space of all life development, in order to avoid the coevolution problem. Dr. Dembski has then simply shown that they “are mathematically equivalent to evolution with respect to a fixed fitness landscape,” with emphasis on the word “a” meaning a singular aggregate (possibly fixed time series) fitness landscape.
Now for Darwinism to have a difficulty with producing increasing overall complexity (as a computational model abstraction as a search algorithm) we have to have some reason or evidence that the specific fitness functions and “search algorithm” will have difficulty producing increasing complexity. Dr. Dembski is relying on the NFL theorems in this regard as well, which imply that taken over all possible fitness landscapes (which may be taken as information contexts j in J, as described in NFL section 4.6), that on average they will do no better than “blind search”.
But of course in a model of the entire world’s progress of biological evolution, as envisioned in section 4.10, there is only a single overall information context, a single overall fitness function applicable to the cross-product of the entire life system in the overall model presented. There is no ensemble of different fitness functions over which to average, as only one multi-trillion year trial is represented. There is no basis presented for a claim that the particular fitness function of the aggregate biosphere and possible states of the biosphere in this particular Earth will not produce a relatively efficient search process over the eons.
Dr. Dembski in essence admits that the issue of displacement is at the core of chapter 4 -- that the NFL theorems do not in themselves provide any argument against Darwinian evolution.
“Thus, if we see increasing complexity, something besides Darwinism must be at work” is simply not implied by anything in chapter 4 on NFL theorems. Displacement of the information context to the environment and patterns of interaction does not provide evidence that Darwinian mechanism is incorrect.
The only point of chapter 4 is that the information for successful evolution of greater complexity, if it is modeled accurately by an aggregate evolutionary algorithm, is the following: The information of what works, what can provide greater complexity in many circumstances, must come from the environment and the interactions of the present state of life on the planet for each step in evolution. This is a singularly unremarkable point, as it is precisely what the biologists who model Darwinian algorithms were saying all along.
I will close by repeating my own remarks, which RBH emphasized above:
quote:
If [evolutionary algorithms and Dembski’s analyses] accurately model the real world, then we must observe a couple of things. The "displacement problem" must mean that aspects of the real world provide source of the CSI. The real world's characteristics must be providing the specification. And if this is evidence that the world as a whole is "designed" then I have no argument. For this is not an argument against the correctness of evolutionary algorithms as an assessment of the operation of evolution in the real world. That the CSI is built into the basic relationships of physics would be no conflict for evolutionary theory.
I emphasize this (yet again) because I do not believe (as many appear to believe) that science is creating a philosophical argument against design of the universe and its relationships, when it learns about the relationships in evolutionary science. And casting the argument as a demonstration that either the philosophical point has some reason to be questioned -- or Darwinian evolution is wrong -- does not provide evidence specific to one of those points.
[ 07. December 2002, 23:23: Message edited by: gedanken ]
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 08. December 2002 10:21
Gedanken,
Welcome to Brainstorms. Very nice post. Keep up the good work.
IP: Logged
|
|
|
Printer-friendly view of this topic