|
Author
|
Topic: Responses to Criticisms of Specified Complexity
|
Nel
Member
Member # 614
|
posted 03. May 2003 14:46
DNAUnion would be ideal for this thread ![[Frown]](frown.gif)
But I'd like to give my take on these criticisms, to be fleshed out more as time goes on as this is a quick and dirty critique of the criticisms (and I'm still learning all this stuff myself). I'll try to get to as much of the criticisms as I can going in order, so I'll be able to respond to at least one per day. I hope the moderator doesn't mind this.
C1: I don't regard the UPB as too stringent a number to be scientifically useful. There are sciences, such as cryptography, that make use of such probability bounds, for example, they work with a probability bound of 10^90. However, I agree that a more useful probability bound can be given but this is likely to be worse then the UPB. For example, it is estimated that instead of an entire universe of possibility is available, realistically, for a flagellum for example, it is estimated that 100 million years elapsed between the time the earth was at a state where life was habitable, and the first primitive organisms. How much time is really available for, say, flagella to evolve via RM&NS? It would be a probability bound much worse than 10^-150. Note this would directly connect an ID inference with a certain taxonomy or history of life, or it may propose one itself based on current evidence.
C2) As far as definition 1 being complexity as defined as improbability and complexity as defined as the number and interrelationship of each of the parts, I regard this as equivalent. For if the number of parts and interrelationship between them is given as a high probability for their formation then the number of parts and the relationship between them is not that complex (consider for example, metabolons).
C3) I think that the current calculation in No Free Lunch (concerning the flagellum) gives us a way to measure the probabilities for a given biological structure.
C4) I don't regard the comparison of biological machines and non-biological machines as an analogy, they are literal machines. Especially when the comparison will be most clear with the invention of self-reproducing nanomachines. But even if it was just an analogy, the argument still has enough teeth.
This will have to do for now, I will get to the rest later.
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 03. May 2003 14:55
Nelson: C3) I think that the current calculation in No Free Lunch (concerning the flagellum) gives us a way to measure the probabilities for a given biological structure.
The calculations in NFL provide one to calculate the probabilities for a strawman scenario rather than to measure the probabilities of a given biological structure based on evolutionary pathways. While it is interesting to show that the flagellum could not have arisen purely by chance, the calculations obviously ignore the relevant pathways.
Or as Miller remarks "I have no doubt that to the casual reader, a quick glance over the pages of numbers and symbols in Dembski's books is impressive, if not downright intimidating. Nonetheless, the way in which he calculates the probability of an evolutionary origin for the flagellum shows how little biology actually stands behind those numbers"
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 03. May 2003 15:41
I already spoken of the relevance here. First if it was only chance that Dembski was calculating, it would still be relevant. Because both Darwinism and ID would be irrelevant if it could likely form by spontaneous chance. Secondly, because the calculation focuses on function and perturbation, and localization, applying the Darwinian mechanism to the flagellum, as can be seen by the calculation, is still nothing but pure chance assembly. Only by plugging in functional intermediates (for which we have one) can the probability still be made lower, but by definition, IC systems have no functional intermediates unless you invoke alternative functions all the way up, but none have been given except for the type III system which seems to post-date flagella, but even if it didn't it would still be mutlitple pure chance events which is still random assembly. So the calculation is highly relevant contrary to Miller's claim. Pim by the way, you sound suspicously like Frances.
[ 03. May 2003, 15:41: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 03. May 2003 16:54
quote:
Only by plugging in functional intermediates (for which we have one) can the probability still be made lower, but by definition, IC systems have no functional intermediates unless you invoke alternative functions all the way up, but none have been given except for the type III system which seems to post-date flagella, but even if it didn't it would still be mutlitple pure chance events which is still random assembly.
The event's can't be "pure chance" if they depend on previous steps. Since each possible mutation depends on previous generation, such probabilities depend on the natural selection steps at that generation. And the very existence of the Type III secretory system shows that evolution is likely to produce steps that could be part of a cooption like might be necessary for the flagellum. And note in statement "which seems to post-date flagella". So really "but even if it didn't it would still be mutlitple constrained chance events". This is of course consistent with "random mutation natural selection".
Was there a listed criticism that relates to this issue (in terms of "specified complexity") that discusses the problem with the IC issue of sequence?
Remember that even if Dembski's EF were reliable, one would have to show "complexity". And at least one issue deals with fact that complexity as defined by Dembski really means low probability, so what we are looking at is specified low probability case. Contrary to the assertion, there has been no presentation that demonstrates that the flagella is acutally low probability.
And of course "but by definition, IC systems have no functional intermediates unless...". So how would one show that the flagellum is actually IC by such a definition? It requires showing that there are no functional intermediates. Since that has not been shown, it does not apply, so saying that something occurs here "by definition" is irrelevant.
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 03. May 2003 17:47
Gedanken, I find that it helps to use the traditional notation with different symbols for the unknown quantities and the constants representing their possible values. So let me do this:
DATA is a stochastic variable which can take values data-1, data-2, data-3, ...
ID is a stochastic variable taking the value true if the DATA was intelligently designed and the value false otherwise. (Note that this differs slightly in form from my previous usage.)
Suppose we observe that DATA = data-17. Then Dembski would search for a certain type of superset of data-17 and reject the hypothesis ID = false if
Pr(DATA inside specification of data-17 | ID = false)
is sufficiently low. In contrast, a Bayesian would base decisions and inferences on the values of
Pr(ID = true | DATA = data-17) = Pr(DATA = data-17 | ID = true) Pr(ID = true) / Pr(DATA = data-17),
Pr(ID = false | DATA = data-17) = Pr(DATA = data-17 | ID = false) Pr(ID = false) / Pr(DATA = data-17).
My original (and only) point in my response to you was simply that in the particularly interesting cases (i.e. when strict use of Dembski's decision criterion leads us to accept ID = true) the likelihoods will massively dominate over the prior probabilities.
Do you agree that the natural Bayesian approach is to break down the analysis like this?
Erik
[ 03. May 2003, 19:42: Message edited by: Erik ]
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 03. May 2003 18:27
Rex Kerr, I forgot about your comments on my C15. Yes, specificational resources are supposed to take care of this, but it can be successful only if intelligence is not as magical as Dembski (and ID advocates in general) seems to think. Dembski seem to think that something magical happened when, say, Shakespeare wrote the play "Hamlet" or when Beethoven composed his music. Shakespeare and Beethoven, we are supposed to believe, did something far beyond what any "natural" processes could have done. But if intelligent agents can magically create instances of "specified complexity", then what prevents intelligent agents from magically inventing specifications for any events they take interest in? When Dembski discusses specificational resources, he assumes that intelligent agents are subject to the same limitations as non-intelligent processes, but everywhere else he assumes that intelligence is magical.
Regarding C13, you wrote that this objection need only come into play when evaluating intermediate levels of intelligence. But that is true for at most half of the objection. C13 contains two objections: The intelligence-might-be-like-a-heap objection and the objection for the a priori choice of the EF over, say, the DEF. Why should we use the EF rather than the DEF? Furthermore, in Dembski's work "intelligence" it seems that intelligence is not something that is determined by whatever convention is the most convenient in the present case, but rather something of ontological significance. How do we know that the lower end of the intelligence spectrum is not sufficiently magical to create "specified complexity"?
Regarding C14, note that the definition of "specification" includes a subjective, psychological relation. Thus, even if all the details were supplied, we would still have to trust (or not) the user of Dembski's method that the subjective relation is satisfied. For instance, if I tell you that my knowledge of parabolas explicitly and univocally identifies the function f : R --> R, f(x) = x^2, how would you know? Is there any information I can supply to make you confident that my statement is (in)correct?
Regarding C22, this objection, unlike complaints about missing details in applications, is an in principle argument saying that there is no consistent interpretation of the independence criterion that will work. It may well be related to the missing details, but I like to separate criticism of applications from criticism of the underlying theoretical framework.
Erik
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 03. May 2003 18:40
Nelson: So the calculation is highly relevant contrary to Miller's claim.
It's "highly" relevant only as much as it allows us to reject some strawmen pathways. As you said, despite being IC there may be pathways yet to be discovered which make ICness itself not a very useful indicator of ID after all as I suspected. As Gedanken has shown, in addition to several others, Dembski's calculations address a strawman pathway. Relevant only to the extent that such a pathway was proposed but since evolutionary pathways have 'yet' to be proposed, it may seem to be too soon to infer design. The mere existence of pathways with alternate function as proposed by you makes any ID inference based on elimination too soon unless one wants to accept that ID inference can suffer from false positives?
I would like to hear more about your ideas about ID proposing hypotheses though because I believe that only through such approach ID can address the many criticisms it has received so far based on their eliminative approach to detecting design.
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 03. May 2003 21:54
[Replaces earlier post]
Yes, Eric, thanks.
You are correct. I made couple of simple changes (the slight changes noted in my introduction) and rederived my result using the standard Bayesean form as you presented -- with the same results.
The change was to define ID=true as that the ID put the DATA in the rejection region, rather than ID just implying that the ID affected observation. (This is not necessary, and then one has to carry a likelihood associated with ID which would probably as you say dominate that value.) Then Bayes rule applied to falling in rejection region rather than specifically at the data value, which is the more important change. This made one of the likelihoods drop out, just as before. Results were the same, as both were mathematically derived from same conditional probability equations, so baring any algebraic mistakes I should have gotten the same result whether using Bayes rule or the premises of Bayes rule.
The result is that the likelihood (from the EF) in ratio to the prior probability of the "designer" acting dominates as long as it is correct, and the prior probability of the designer acting to put the DATA in the rejection region (or combined prior and likelihood) is greater than the "chance" likelihood.
But make a mistake that puts the "chance" likelihood greater than the probability of the designer acting, and that is flipped on its head.
Another instructive notation is rather than ID=true, one could use CAUSE=id or CAUSE=not-id. The not-id can be expanded into an array of non-id causes such as c(i) for i=1..N, and variable is thus CAUSE=id or CAUSE=c(i) i=1..N. (Even list 'id' as c(0).) This expands the list of "probabilistic resources" explicitly, making its place in the calculation more obvious.
I've got to write up a paper before presenting any more here. Thanks.
[ 04. May 2003, 15:23: Message edited by: gedanken ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 04. May 2003 14:44
Ged,
It is precisely the fact that quite a few steps from a type III secretory system to a flagellum contains unselectable steps that points directly to random assembly. In other words, the very structure-function (as an IC system) of the bacterial flagellum tells us that there in no step leading up to the flagellum that might depend on the last one. The calculation directly shows this.
The very existence of the Type III secretory system does not show that evolution is likely to produce it, especially when it post-dates the bacterial flagellum. For example, as I showed previously, a summary of some of the unselectable steps, you have the problem of the 6 part export machine. There is no evidence that any subset of this export machine carries out alternative function. You would want to add FliE since it's a great proto-filament, but FliE has no obvious role outside the flagellum either. Not even the Type III secretory system requires it. Then you have the rod, which is three components that are functionally indivisible, so adding each additional component of the rod adds more unselectable steps.
The calculation that was done in No Free Lunch gives the flagellum a low probability of 10^-2000 (approx.). And here is why the post-dating of the T3SS comes in to play. As I discussed at ARN, this is just an illustrative example, not quanlitative,
We go with an organism with a "protein supermarket" each with a
protein of probability 10^-39 (A)
|
|
|
A type III secretory system with probability: (B)
10^-663 * 10^-600 *10^-122 = 10^-1385
|
|
|
|
|
And a bacterial flagellum, taking the Type III secretory system into account, with probability:
10^(-780) (C)
It may be a lot more than 10^-780, I'm not sure if Dembski did the Pperturb for this.
Again, this is just illustrative, whats important is how relevant this calculation becomes when we start applying it to pathways. As one can obviously see, simply bringing up the T3SS is not sufficient to bring down the probabilities that natural selection would obtain the flagellum. Look at the large amounts of random assembly between a protein and the T3SS, between the T3SS to the bacterial flagellum. One can get more specific here, bringing in penalties that come with co-option and gene duplication, etc, evolutionary dead-ends, things like that. Which is why Dembski says there is a lot of work to be done. But to say that the calculation is completely irrelevant be easily be shown to be false, as one can see from the quick and dirty calculation of an evolutionary pathway I show above, which still is quite below the UPB. What if one can show that between each system, there are functional steps? But even a sequence of functionally different intermediates would be a lot of random assembly and would most likely still yield a low probability. More so then, what if it's partial function all the way up to theh fully functional flagellum? What if the type III secretory system was a type of flagellum that caused movement like the flagellum? Then like the protein with 1 amino acid, the probability gets smaller and smaller at each step, leading to a smaller probability at (C).
Now, with an IC system, you're gonna run into this problem by it's very definition. However, with a non-IC system, such as hemoglobin, you will have functional intermedates all the way through, so selection might handle such a system, or, at least there are functional intermediates.
[ 04. May 2003, 14:49: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 04. May 2003 15:34
C5) Dembski doesn't say that "nature can't create new information". Dembski states that
quote:
"Natural causes are … incapable of generating CSI. This broad conclusion I call the Law
of Conservation of Information … ."
CSI is "complex specified information" not just information. Dembski states that chance and necessity can generate CI but not CSI, CSI is information that is independently identifiable by
an external pattern (i.e. irreducible complexity). He does define this rather rigorously in the form of the UPB or 500 bits of information.
C6 Is also based on an error. The specification of the flagellum is not based on an analogy from outboard motors, although the fact that we invented outboard motors before we found one attached to a bacteria doesn't hurt.
Dembski states:
quote:
This is not to say that for the biological function of a system to constitute a specification humans must have independently invented a system that performs the same function. Nevertheless, independent invention makes the detachability of a pattern from an event or object all the more stark.
NFL pg 289
Specification refers to biological function.
C7: The very determination of CSI as was originally formulated (IC was not used to "patch" any hole) showed that you cannot get a specification from something that you can get from gradual accumulation. Specification always referred to a pattern that is not reducible to it's consituent parts. You will never get it from a gradual accumulation of CI, because then the S in CSI would be ad hoc.
So C7 is: SC is not necessarily irreducible complexity. For example, on pages 16 to 17 one can see an example of a specification where the professor was attempting to determine which pattern of coin flips was random and which was not. A student can attempt to "trick" the professor by doing some binary arithmetic, in doing so however the pattern became a specification that was not read off the event.
C8: IC is falsifiable in that if the function can be reduced to a particular component it is not IC and therefore not SC. C8 also is quite mistaken about what Dembski actually said about several pathways, for example, in the spinning just fine thread, Dembski states:
quote:
Even if the trajectories are continuous, it still remains to be shown that they are Darwinian. I'm entirely comfortable with the evolution of the immune system, for instance. In fact, I'm comfortable with full common descent. My argument, always, is that evolutionary pathways to irreducibly complex systems, even if they exist, are non-Darwinian.
This is quite different from what C8 presented. An evolutionary pathway may be given but if it's not Darwinian, the the trajectory itself (as I show in the Immunity thread) may be the result of intelligent design.
C9: The inductive argument of "no false positives"
is not unlike the scientific method's use of a null hypothesis, see for example, the null hypothesis "there are no ets" and how it is used in SETI:
http://www.setileague.org/editor/null.htm
[ 04. May 2003, 15:35: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 04. May 2003 15:39
quote:
Now, with an IC system, you're gonna run into this problem by it's very definition.
If it's by "definition" then you separately need to demonstrate those very points to know that it matches the definition. So then calling it "IC" does not imply that as a consequence and calling it "IC" says nothing because one has not done the work.
However if one did the work (and the definition of IC was that it met those conditions) then the reason for what you claim would not be because it was "by definition", rather the reason would be bewcause it met the conditions (and being "IC" would be the consequence, not the premise).
However I will leave the issue of whether one can demonstrate the conditions you claim (or if that is even something meaningful) to the biologists, and when I get some more time I'll also bring it up my definitions thread again.
My point is that you can't claim what you say is true "by definition" and then start with a claim that it is IC as the premise for meeting those consequences -- that is circular reasoning. This is why the slippery definitions of IC are so important -- one definition claims the consequences as coming from another observation, another definition claims those consequences as the conditions for declaring IC. Then circular reasoning is used by using the former definition, then slipping to the latter and claiming those consequences "by definition" but never completely supporting them along the way. That should relate to this thread, but I don't have the list in front of me. It definitely relates to my "sequence of tests" thread, even you seem to claim otherwise.
--
But now are you also claiming that the T3SS is itself "IC"? I think that by your argument above it must have an "irreducible core" according to the way you present such things. So why don't ID promoters also champion the ICness of the T3SS? I think that your claim that all possible pathways must be "unselectable" is unsupported.
[ 04. May 2003, 15:49: Message edited by: gedanken ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 04. May 2003 15:47
Ged,
My post demonstrates the unselectable steps inherent in the IC flagellum. I didn't simply assert ICness, I demonstrated it. Also, by showing that the 6 part export machine requires unselectable steps as well, I was pointing to the ICness of the T3SS. So as an IDer, I do "champion" it's ICness.
IP: Logged
|
|
gedanken
Member
Member # 594
|
posted 04. May 2003 15:56
Thankyou Alonso, I think that aspecs of what you said (about T3SS being IC per se) will be important in my Sequence of tests for IC thread. I'll have to note this and recall it when I get back to that. I'm not refering to the specifics of T3SS, rather to the structural form of the argument you made, as "sequence of tests.." is supposed to be about the generic argument and not specific cases.
I think we share issue of the fuzziness of definitions between this thread on criticisms of SC and that thread on definiitons of IC and tests for IC.
[ 04. May 2003, 15:57: Message edited by: gedanken ]
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 05. May 2003 17:18
C10: Above, I argue that Dembski indeed take into account the hypothesis that the type III secretory system evolved to the flagellum. However, as of right now the sample space of testable evolutionary pathways seems empty.
C11 is based on a gross error. This person defined specified complexity as the elimination of non-ID hypothesis. However, this is a consequence of specified complexity and not part of the definition. Specified complexity itself is defined as low probability associated with an external pattern, as I say in a previous post. This adds everything to the discussion since "specified" used with complexity always triggers a design inference in many areas of both science and common intuition. Paul Davies's discussion of specified complexity is actually equivalent to Dembski's notion of specified complexity,
quote:
To bring out this point clearly, consider the way in which the four bases A, C, G and T are arranged in DNA. As explained, these sequences are like letters in an alphabet, and the letters may spell out, in code, the instructions for making proteins. A different sequence of letters would almost certainly be biologically useless. Only a very tiny fraction of all possible sequences spells out a biologically meaningful message, in the same way that only certain very special sequences of letters and words constitute a meaningful book. Another way of expressing this is to say that genes and proteins require exceedingly high degrees of specificity in their structure. As I stated in my list of properties in Chapter 1, living organisms are mysterious not for their complexity _per_se_, but for their tightly specified complexity."
Notice here, how Davies distinguishes the external pattern from complexity, as Dembski does. The meaning exhibited by the code he likens to the pattern we see in human languages (more relevantly to computer languages and machines -- IC). In each case, the meaning of the message is not reducible to the individual bases. Thus, it is external. This is identical to Dembski's notion of specificity.
C12: That no one has been applying the calculation is a testament, in fact, to the lack of will exhibited by Darwinists to test their pathways. However, absence of evidence, in this case, is not necessarily evidence of absence. There has been other calculations being done (i.e. for the T3SS on ARN), and some that I am aware are currently being worked on (i.e. to assess co-option). This does not necessarily testify to the utility of the EF, to say so would clearly eliminate many models that have yet to be replicated even by the author of the model, such as A Pessimistic Estimate of the Time Required for an Eye to Evolve," Proceedings of the Royal Society, London B (1994) 256, 53-58, and yet has been cited (albeit incorrectly) by the likes of Dawkins, Mark Perkah, and talk.origins.
C13: There is an emphasis on what we would expect to find from the acts of intelligence. Does this necessarily require a definition of intelligence? Maybe not. Does it require that we can correctly distinguish an act of intelligence from an act of purely natural process? I think that Dembski and Behe have already progressed in this area by the notions of specified complexity. Nonetheless, it wouldn't hurt to make our inferences to intelligent acts more refined by shooting for rigor. If I'm not mistaken, a paper by Micah and some writings by Chris Langan is addressing C13.
C14: What C14 is describing is a fabrication, not a specification. Note, Dembski has shown how one can rigorously distinguish between a specification and a fabrication in NFL (in his discussion of hypothesis testing), one of these ways is through the use of a "rejection region". Background information noted as K "explicitly
and univocally identifies the rejection function f". The background knowledge can be seen in the external pattern proposed, such as irreducible complexity. It cannot be ad hoc, and must be testable (you can't just say it!)
There are about 8 left, but I don't have the book with me, so I leave the rest for tomorrow.
[ 05. May 2003, 17:26: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 05. May 2003 23:45
I would like to voice my discomfort and concern with the comment made by Nelson
quote:
C12: That no one has been applying the calculation is a testament, in fact, to the lack of will exhibited by Darwinists to test their pathways
I can accept that Nelson may be frustrated with the lack of progress made by ID proponents to provide for relevant probability estimates for evolutionary pathways and I am not upset by his obvious logical fallacy here, in fact I have grown quite accepting of them but in this case he accuses, without much evidence Darwinists of a lack of will to test their pathways.
I'd hope that this board will not tolerate Nelson's comments.
Let me also show why his comments are erroneous for several reasons.
First of all let me point out that it is the eliminative nature and the use of probability that requires the design inference filter to eliminate Darwinian and other chance or regularity pathways in order to be able to infer design. So it is obvious from the start that it is not the Darwinist's responsibility to provide for such probability estimates but rather the responsibility and task of the ID proponent.
Secondly let me point out that Darwinist are not unwilling to provide evidence for their proposed pathways but probability estimates are but one way of achieving such and in fact given the nature of the beast, probability estimates are intrinsically difficult if not impossible, as Dembski and others have found out.
Thirdly Darwinists are not unwilling to consider the evidence and falsification of their ideas and in fact in many of instances they use whatever available data to formulate ways to support or eliminate their hypotheses. Probability estimates for their pathways, due to their nature of being impossible to accurately calculate are not required nor even necessary. Unless of course one argues from a purely probabilistic eliminative nature like the design inference.
That Nelson somehow holds Darwinists responsible for the failure of design proponents to provide for probability estimates and elimination of pathways I could understand but that he actually accuses them of being unwilling to test their pathways is not only insulting but also at odds with reality.
Such manner of discussion is neither productive nor effective and I hope that Nelson will (come to) realize this.
[ 06. May 2003, 00:24: Message edited by: Pim van Meurs ]
IP: Logged
|
|
|
Printer-friendly view of this topic