|
Author
|
Topic: Responses to Criticisms of Specified Complexity
|
Rex Kerr
Member
Member # 632
|
posted 08. May 2003 00:10
Nelson, C16 is a very specific mathematical defect in the formulation of calculating the number of specifications that was introduced in NFL but not in TDI, and I don't see at all how your response is relevant. In fact, I have already formulated a solution to the problem, so perhaps C16 it is already addressed, if my solution is adopted. For details, see this thread.
The response to C17 is not particularly illuminating, given that C17 complains about a lack of details. Can you point me at a reference for or text of the "Logical Underpinnings"? If that doesn't demonstrate how to compute the number of specifications of a given level, then I would like to know that, too.
Unfortunately, I must disagree with the response to C18. The most difficult parts of the calculation for the cake and Gettysburg address were omitted--namely, the careful delimiting of the space from which random and chance hypotheses were drawn; the demonstration that all relevant hypotheses were tested; and a demonstration that the specification identified the object in a statistically independent and sufficiently precise way.
In particular, the cake is an example only of multiplying three numbers together--p_orig, the probability that the cake ingredients exist; p_local, the probability that they'll be colocalized; and p_config, the probability that they'll be mixed in the proper order and manner to result in a cake. (Incidentally the choice of those three distinct factors is largely arbitrary.) Nowhere does he demonstrate that a cake is specified, or even that this is the right way to evaluate the probability of a cake occurring by chance. Furthermore, since he's using it as an example, he puts in ludicrous values such as 50 ingredients for a cake (?!), and fails to take into account extremely important factors such as that not all ingredients exist (or are used) in the same proportions. Although this could be remedied somewhat by extensive use of upper bounds on probabilities, this is not done, rendering the cake calculation not only incomplete as a design inference, but equally incomplete as a probability calculation. However, I'm not criticizing Dembski here--he was only using it to illustrate a conceptual point. The Gettysburg example is similar--here he's illustrating the use of perturbation probabilities, with the numbers invented on the fly for illustrative purposes. I don't want to see numbers invented--for a real calculation, we need to know how to come up with the real numbers, and use it on an example that we are certain was designed.
Also, a nuclease is not a reverse topoisomerase; it's more like a reverse ligase; hence my surprise at Nae-I.
Finally, even if there were only 100M years available (quite debatable), we don't know how complex the molecular systems were in bacteria that long ago. Even if we had a guess, we couldn't necessarily replicate it in the lab, as right now there are about 10^30 bacteria on the planet; even if the planet was, say, a billion times less favorable to bacteria (10^21 bacteria), over a 100M years we're still something like 16 orders of magnitude short in lab experiments.
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 08. May 2003 14:52
I don't have time for a full response at the moment (i'll be back later today or tomorrow) however, Rex's post contains things I already addressed except for this little bit:
Rex writes:
quote:
Also, a nuclease is not a reverse topoisomerase; it's more like a reverse ligase; hence my surprise at Nae-I.
Actually a nuclease is a reverse topoisomerase. However, since I don't want to make a "yes it is" "no it isn't" tit for tat, I'll simply quote the literature:
quote:
A DNA topoisomerase can be viewed as a reversible nuclease
http://www.garlandscience.com/pdfs/ch05final.pdf
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 08. May 2003 15:57
Nelson:
a reversible nuclease is most certainly not a reverse nuclease. I am sure you can see the difference between the two terms.
A topoisomerase's "core" reaction is simmetrical (a nucleic acid phosphodiester bond is cut, its energy "stored" onto an aa residue, and then transfered back to the nucleic acid as the phosphodiester bond is formed again). Thus, it is "reversible".
An endonuclease reaction, like Nae I's, is asimmetrical, and so is a ligase reaction, and in fact one is not the reverse of the other either, although of course the nucleic acid substrates and products of certain endonucleases and ligases are reversed.
[ 08. May 2003, 15:58: Message edited by: charlie d. ]
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 08. May 2003 15:57
While noting the occasional irritation and impatience, I can’t help but see the humor in much of this. Are we expecting or requiring Dr. Dembski to do what we are saying is “impossible” to do? LOL We might consider more closely what appear to me to be some assumptions shared by both Dr. Dembski, his critics, and otherwise uninvolved biologists working with probability models.
Positively (or maybe not), I would suggest that this implicit and otherwise undeveloped argument, that some (mysterious?) system of causal dependencies is being ignored by Dembski needs to be examined more closely. Biological evolution is trivially a matter of causal dependencies. It is significantly a problem in the systematic relaxation and even elimination of such dependencies. This is exactly what permits notable evolutionary advances (rather than the usual dithering, tinkering, and twiddling) to be made, and is directly related to our definitions of “evolvability.”
Sorry, but Darwin didn’t have a clue about this. (And I will support the argument from the literature.) We might update our arguments, otherwise critics like Dembski may conclude that we don’t even know evolution at all, and therefore he can safely ignore the criticisms as its not his responsibility to keep us informed about developments in evolutionary biology. Must this debate occur in the 19th century, or could we transpose it into the 21st century? (Hmmm… Gettin’ a little irritated and impatient myself.)
I find Dr. Dembski’s (probability) thesis entirely unaffecting. I don’t consider it “wrong.” I consider it irrelevant! (Sorry, Doc.) I can anticipate that theory will continue along the lineaments of the past forty years, and that we will eventual cease to treat the improbabilities and uncertainties inherent in any biologically realistic evolutionary process as theoretical “invariants” (Or is that “embarrassments?”), but as the very factors that life forms adapt to!
Did I manage to offend just about everyone? I hope so! LOL
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 09. May 2003 03:53
As charlie said, reverse and reversible mean different things. I grant that after the fact, it's not too surprising that an endonuclease can be mutated to be reversible (esp. for someone who knows about microscopic reversibility from physics/chemistry). My surprise arose from not having anticipated this--and it makes me wonder what other things (such as functional intermediates) would be completely unanticipated yet would seem sensible enough once explained.
However, on its own, this is pretty inconsequential. I'm more interested in answers to my other questions, or precise references to places where they have been answered.
Also, Janitor, posts are easier to comprehend and understand in detail if a scholarly tone is adopted instead of an internet discussion board style.
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 09. May 2003 16:05
Rex,
C16: Rex, thanks for more detail on this. Your first mistake in that thread is that specifications are not all simpler than the specification you realized (R). The only restriction on the complexity of all the specifications in the set is that it cannot be more complex. Also , (P(R|H_i)>=P(T|H_i)) doesn't mean that all of them are rarer. It is the same mistake. This (P(R|H_i)>=P(T|H_i)) shows that the probability of the specification cannot exceed R, but it can be equal to it. I noted this when I mentioned that the complexity of specifications in the set varies and that a specification can possibly match the event.
Your example is a specification but it is not complex. It is not a coincidence that this error is the same as the coin toss/911 error. For example, suppose that you set the specification that I will correctly guess the outcome every time I roll a die, you roll 6 times, the probability is f(6) = 1/46656 = .002% which is improbable (using f(x) = [n!/(y!(n-y)!)](1/6)^y(5/6)^(n-y)). To see this in plain language, if you tell me you can get a lot of results before getting a 1 by rolling the dice, and you show me how many times you got a 1 on the first shot, I come back the next morning and after hundreds of tries, you show me all the number of ways you were able to roll the dice without getting a 1, I'm gonna say "so?" Tell me you are going to get a 1 after f(6) tries and I will be impressed.
Moreover, your "fix" is kind of disingenuous. All you want to do is relax 3 and 4 so that you can make the set of specifications infinite. And therefore, say anything is specified (or nothing is). There is no reason from the example for a "fix" nor is there any reason to relax 3 or 4, since the set of specificational resources can be equal to R. There doesn't need to be any simpler specifications. There doesn't need to be a certain amount of specifications either.
I don't understand what your problem with my response to C17 is. My response shows how one can describe the complexity of an event (a bacterial flagellum) without doing it an injustice. Indeed, in science, the more we look at the details, the more complex it usually turns out to be.
I don't think that you adequately responded to my reply to C18. Porig is actually calculating getting all the ingredients of the cake by chance in a place where a blind selective process (Plocal) can select the correct parts, and that such a process will build it given those parts. It is quite obvious that a cake is specified as the pattern "cake" is external to it's particular ingredients and must be "assembled" by an intelligent agent. The particular number of ingredients is not important as it was an illustrative calculation, I agree with that. But the example was not so illustrative as to be useless. One can extend the calculation, same for the bacterial flagellum, and use it to infer design.
We might both be wrong with respect to topoisomerase (I havn't really looked into it because it doesn't matter much) but I don't think it has any relevance to whether I should drop my ID inference because a future explanation may be found that will do away with all those unselectable steps. It is the clinging on to those promises for future explanations that makes Darwinian evolution unfalsifiable.
With respect to the number of bacteria that can be used in a lab experiement, I still have no idea why you are using the number 10^30 bacteria as a prerequisite to evolve an IC system. From the time that life was habitable to the time that we find bacteria in the fossil record (which look almost identical to modern bacteria) was a mere 100 million years, if it is true that there are 10^30 bacteria now, thats the result of 3.5 billion years. 100 million years, 3.5 billion years, thats a big difference.
From this we can definitely use a sample space that does not require hording large amounts of bacteria (???). Your stringinet argument against evolutionary experiments in the lab pretty much destroys the relevance of all lab experiments that have evolutionary significance.
There is nothing about large amounts of bacteria that will get all those unselectable steps towards an IC system anymore selectable.
There is nothing magical about 10^30 bacteria that would get us evolved bacterial flagella. Since evolution is a stochastic process, theoretically, from a Darwinian point of view, you could get it with 10^2 bacteria. From a Darwinian point of view, then, it is merely contingent that evolution of phyla took place in less than 10 million which is about the time it took to vary finch beaks.
Also, statistical tests do not require hording bacteria at all, so Dembski's mathematics would be ideal even if you would need such large amounts of bacteria.
[ 09. May 2003, 16:32: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
Erik
Member
Member # 160
|
posted 09. May 2003 19:47
Some final remarks on Nelson Alonso's latest reply to me:
C10: First of all, the term sample space does not mean the set of hypotheses we consider (at least not in conventional statistics nor in Dembski's unconventional version), as you seem to think. Perhaps the term is never used in "No Free Lunch" (maybe Dembski called it "reference class of possibilities" instead--I don't remember), but the sample space is the set of possible outcomes/data sets/experimental results. The set of hypotheses is something different (at least in non-Bayesian statistics). Second, you replied to my observation that Dembski fails to supply details with an unreferenced quote from Dembski's writings, where he makes a general comment on how hypotheses are classified and which hypotheses are considered (the "relevant" ones). That is not even a clear statement of which hypotheses he considered in his flagellum calculation, and it certainly doesn't clearly specify all the details necessary to verify that the calculation is correct w.r.t. Dembski's stated standards. The details needed are:
1. The sample space.
2. The event under study.
3. The set of hypotheses that are considered.
4. The rejection function(s) used.
5. The background knowledge used to explicitly and univocally identify the rejection function(s).
1, 2 & 4 must be given in mathematical detail (not just vague prose) and 3 should be detailed enough to make clear exactly which probability distributions are implied. It is unclear to me how detailed the account of 5 is supposed to be. The reason that it is important to be formal with this is that the amount of "specificational resources" explodes when we allow specifications that are not mathematically precise (e.g. when we allow rejection functions whose precise values have not been fixed everywhere on the sample space). Despite your claims that these details exist, you have not managed to provide any of them.
C11: You seem to be arguing that it is part of the definition of "external pattern" that it could not have arisen according to a non-ID hypothesis and that "external pattern" is part of the definition of "specified complexity", which, if true, merely proves my point. Your unreferenced (why won't the Brainstorming ID advocates write out the full reference to their quoted material?) Paul Davies quote simply does not show that his notion of "specified complexity" is equivalent to Dembski's. When we look into the details, the resemblance quickly disappears. For instance, nowhere in Davies's notion is the idea that "specified complexity" is unlikely to arise by chemical or biological evolution.
C12: The test you likely have in mind likely don't exist (because they would not be reliable). There have been many (other) tests of evolutionary biology, some of them statistical and some of them more informal. If you read scientific journals like Evolution and Journal of Molecular Evolution you will see that evolutionary biologists go far beyond stating logical possibilities. However, even if evolutionary biologists had not properly tested their models, that still wouldn't change Dembski's failures. Alleged flaws in evolutionary biology can never compensate for the flaws in Dembski's ideas -- only if ID advocates have learnt to answer criticism of their concepts and ideas without making reference to alleged failures of others will they have a good candidate for a theory.
C13: It is simply not true that intelligent agents can accomplish everything that "chance" and "regularity" can accomplish. For instance, at no time have we observed an intelligent agent create a full scale tornado or earthquake. The intelligent agents (humans and animals) we have observed are quite limited and everything they do consists of coopting and interacting with non-intelligent processes.
It is true that Dembski's EF has a similarity to null hypothesis testing, but so does the DEF, so the criticism stands.
C14: In reponse to my criticism, you claimed that a rigorous calculation exists. The problem is that C14 is not about any calculations. It is about the problem of how to verify that a rejection function is "explicitly and univocally" identified by background knowledge. Your reply is irrelevant.
Erik
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 10. May 2003 00:29
Nelson: We might both be wrong with respect to topoisomerase (I havn't really looked into it because it doesn't matter much) but I don't think it has any relevance to whether I should drop my ID inference because a future explanation may be found that will do away with all those unselectable steps. It is the clinging on to those promises for future explanations that makes Darwinian evolution unfalsifiable.
I find Nelson's statements remarkable for a variety of reasons. First of all a design inference seems to ultimately depend on exhausting the set theoretic complements of the disjunction chance or design. It seems that under such a definition any design inference for the topoisomerase seems to be too soon. Secondly, Nelson suggests that both may be wrong which means that contrary to the statement that the design inference does not suffer from false positives, Nelson believes otherwise. I agree with Nelson. Thirdly Nelson suggests that evolution is not falsifiable because "it is clinging on to future explanations". But that is a fallacy. Evolutionists are quite willing to admit that the evolutionary pathways for toposomerase at this moment remain unresolved. However admitting ignorance does not mean that either evolution has been falsified or that evolution cannot be falsified. In fact if it can be shown that no regularity or chance pathways to topoisomerase exist then evolution could be falsified. Given the problems however with formulating such pathways in a probabilistic manner I do not foresee that the design inference will be able to resolve this issue anytime soon.
So we may at most conclude that "we don't know" in the absence of any testable hypotheses and due to the lacking evidence.
Nelson still seems to be under the false impression that it is up to the Darwinist to show detailed probabilistic pathways that show how evolution happened. While such pathways would surely strengthen the Darwinian argument, there is no requirement for the Darwinian scientist to be held to such limited rules of scientific inquiry. Remember it is not the scientist who proposes that we have enough evidence to calculate the probabilities for such pathways in sufficient detail. Requiring such accuracy would be like requiring the ID proponent to give a detailed step by step explanation of how the designer did it. Surely the ID proponent would object to such requirements. But an ID proponent who relies on the design inference to infer design has the unfortunate task to show how the ""the set theoretic complement of the disjunction regularity and chance" can be infered through a careful and complete formulation of probabilistic pathways for chance and regularity.
Thus while the ID proponent seems to have the enviable task to obtain relevant probabilistic estimates before one can reject particular hypotheses, the Darwinain proponent is not constrained by the approach chosen by Dembski in his design inference.
So when Nelson asks of examples of such pathways as evidence of their lack, he merely seems to be pointing out that the task set for the ID proponent may be far more complicated than expected.
Finally Nelson suggests that Dembski did take into account Darwinian pathways, selection etc in his perturbation calculations but as Nic and others have shown Dembski did nothing of the kind. In fact work by Adami and Schneider have shown how complexity can increase in the genome through the simple processes of chance and regularity (mutation and natural selection). In fact the open system allows information from the environment to increase the information in the genome and thus the specification remains external to the genome imposed by the environment. It seems that as such the environment seems to meet most if not all the requirements of being a causal agent of intelligent design, suggesting that the definition of intelligent design as formulated above may not be sufficient to rule out the environment acting on the genome thru regularity on the variation generated by the replication of the genome.
In the most recent paper in Nature, Adami, Lwensk, Pennock et al seem to show how evolutionary pathways can be quite complicated in that they may include co-option, the reuse of existing simpler systems, the hitchhiking of detrimental mutations which become essential in a later step. I would like to add to this the effect of neutral mutations, the observation that protein space seems to be characterized by a few common structures which can be found throughout sequence space and also 'close' to eachother meaning that with a few mutations different structures can be reached. We may also understand better the concept of irreducible complexity in terms of neighborhood of protein structures. Forward and reverse pathways may be very different and the observation that many reverse pathways can lead to the destruction of the function, this does not mean that there are no forward pathways to the structure. In fact their experiments suggest that irreducible complexity may be nothing more than an artifact of us looking back in time without having access to the past.
It would be interesting to see Dembski's calculations applied to these experiments to see if one were to infer design. It should not be too hard to apply Dembski's perturbation approach to the examples?
All in all it seems to me that the calculations performed by Dembski seem to ignore the shape of protein structure space versus its sequence space, the possibilities of essential hitch hiking of detrimental mutations, the effect of neutral mutations, the closeness of protein structure space, the effects of selection and co-option.
I would argue that if so many relevant evolutionary pathways have been ignored how one can reach a design inference ? It seems far too early to jump to such a conclusion. Only through hard work can the ID proponent work through the task of providing evidence that the probability estimates for likely pathways can be eliminated succesfully.
I would like to know if anyone is aware of ID proponents who are working through such calculations? On this group Nelson seems to be most interested in evaluating these pathways and while I believe that some of his interpretations of what Dembski has or has not done are erroneous, I do consider Nelson one of the better proponents of the ID inference. Certainly his enthusiasm in exploring the implications of the ideas proposed by Dembski et al should be applauded and shouls serve as an example to other ID proponents.
In an effort to propose a positive contribution to ID research I hereby propose that the scenarios in the Adami, Lenski Pennock et al paper in Nature are examine through the ID inference filter and Dembski's calculations for perturbation, localization etc to determine what the ID inference would have been if we had not known the full details of the pathways that led to the observed structures.
While their work may not be directly relevant to ID I do believe that ID could use this work to investigate the value of the ID inference.
[ 10. May 2003, 00:38: Message edited by: Pim van Meurs ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 10. May 2003 00:40
Nelson: We might both be wrong with respect to topoisomerase (I havn't really looked into it because it doesn't matter much) but I don't think it has any relevance to whether I should drop my ID inference because a future explanation may be found that will do away with all those unselectable steps. It is the clinging on to those promises for future explanations that makes Darwinian evolution unfalsifiable.
I find Nelson's statements remarkable for a variety of reasons. First of all a design inference seems to ultimately depend on exhausting the set theoretic complements of the disjunction chance or design. It seems that under such a definition any design inference for the topoisomerase seems to be too soon. Secondly, Nelson suggests that both may be wrong which means that contrary to the statement that the design inference does not suffer from false positives, Nelson believes otherwise. I agree with Nelson. Thirdly Nelson suggests that evolution is not falsifiable because "it is clinging on to future explanations". But that is a fallacy. Evolutionists are quite willing to admit that the evolutionary pathways for toposomerase at this moment remain unresolved. However admitting ignorance does not mean that either evolution has been falsified or that evolution cannot be falsified. In fact if it can be shown that no regularity or chance pathways to topoisomerase exist then evolution could be falsified. Given the problems however with formulating such pathways in a probabilistic manner I do not foresee that the design inference will be able to resolve this issue anytime soon.
So we may at most conclude that "we don't know" in the absence of any testable hypotheses and due to the lacking evidence.
Nelson still seems to be under the false impression that it is up to the Darwinist to show detailed probabilistic pathways that show how evolution happened. While such pathways would surely strengthen the Darwinian argument, there is no requirement for the Darwinian scientist to be held to such limited rules of scientific inquiry. Remember it is not the scientist who proposes that we have enough evidence to calculate the probabilities for such pathways in sufficient detail. Requiring such accuracy would be like requiring the ID proponent to give a detailed step by step explanation of how the designer did it. Surely the ID proponent would object to such requirements. But an ID proponent who relies on the design inference to infer design has the unfortunate task to show how the ""the set theoretic complement of the disjunction regularity and chance" can be infered through a careful and complete formulation of probabilistic pathways for chance and regularity.
Thus while the ID proponent seems to have the enviable task to obtain relevant probabilistic estimates before one can reject particular hypotheses, the Darwinain proponent is not constrained by the approach chosen by Dembski in his design inference.
So when Nelson asks of examples of such pathways as evidence of their lack, he merely seems to be pointing out that the task set for the ID proponent may be far more complicated than expected.
Finally Nelson suggests that Dembski did take into account Darwinian pathways, selection etc in his perturbation calculations but as Nic and others have shown Dembski did nothing of the kind. In fact work by Adami and Schneider have shown how complexity can increase in the genome through the simple processes of chance and regularity (mutation and natural selection). In fact the open system allows information from the environment to increase the information in the genome and thus the specification remains external to the genome imposed by the environment. It seems that as such the environment seems to meet most if not all the requirements of being a causal agent of intelligent design, suggesting that the definition of intelligent design as formulated above may not be sufficient to rule out the environment acting on the genome thru regularity on the variation generated by the replication of the genome.
In the most recent paper in Nature, Adami, Lwensk, Pennock et al seem to show how evolutionary pathways can be quite complicated in that they may include co-option, the reuse of existing simpler systems, the hitchhiking of detrimental mutations which become essential in a later step. I would like to add to this the effect of neutral mutations, the observation that protein space seems to be characterized by a few common structures which can be found throughout sequence space and also 'close' to eachother meaning that with a few mutations different structures can be reached. We may also understand better the concept of irreducible complexity in terms of neighborhood of protein structures. Forward and reverse pathways may be very different and the observation that many reverse pathways can lead to the destruction of the function, this does not mean that there are no forward pathways to the structure. In fact their experiments suggest that irreducible complexity may be nothing more than an artifact of us looking back in time without having access to the past.
It would be interesting to see Dembski's calculations applied to these experiments to see if one were to infer design. It should not be too hard to apply Dembski's perturbation approach to the examples?
All in all it seems to me that the calculations performed by Dembski seem to ignore the shape of protein structure space versus its sequence space, the possibilities of essential hitch hiking of detrimental mutations, the effect of neutral mutations, the closeness of protein structure space, the effects of selection and co-option.
I would argue that if so many relevant evolutionary pathways have been ignored how one can reach a design inference ? It seems far too early to jump to such a conclusion. Only through hard work can the ID proponent work through the task of providing evidence that the probability estimates for likely pathways can be eliminated succesfully.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 10. May 2003 01:42
Nelson, I don't think you understand my counterexample in C16. I set it up so there are no equalities, and the correct accounting for specifications does not depend on whether the probabilities are complex or not. Furthermore, TDI works regardless of whether an event is complex or not--it's just a lot more convenient to use the UPB than to set up a new rejection region each time. Also, my fix does not allow infinite specifications--you just need to count all simpler specifications in your specificational resources, whether or not those specifications specify the event you've observed.
I also asked what "Logical Underpinnings" is. Perhaps this contains the clue that links what you're saying to my objection C17, but right now I can't see the connection at all. We mostly describe things that exist, and thus our concepts are limited by reality. How can we then verify that our language is independent of a phenomenon we're studying, as is required for our specifications? I really don't see how your answers address this at all--but then I don't understand where the 10^20 comes from, what "Logical Underpinnings" is, or why you think I'm worried about a descriptional injustice. I think that our description will appear to be a simple specification (of a process of great complexity) and fool us into thinking we have design in nature, when in fact we have design in defining our words.
Also, you seem to have completely missed my point for C18. I want to see a positive example where all details are worked out in full! The cake clearly does not have all details of a design inference worked out in full; instead, it has the probability calculation worked out in part. Without a positive example worked out in full, it is very difficult to know whether it is even possible to apply the algorithm that Dembski has produced.
Finally, I don't understand your conclusions with respect to bacteria. I am simply observing how many bacteria that there are, and speculating that evolution of bacteria on the entire planet has hugely more sampling power than a lab experiment does. Thus, lab experiments may not be able to reproduce some of the things that evolution has, particularly when numerous successive rare events need to be selected for (since we can't get the rare events in the lab in the first place, because we have too few bacteria and don't keep them long enough).
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 10. May 2003 16:51
Rex,
Only have time for a few comments to Rex.
Rex, I perfectly understand your counterexample in C16. If your specification is that you will roll the dice once and not get a 1, then the specification is f(1). There is no need to relax 3 or 4, the specification in f(n) , is equal to R and this is consistent with his discussion of SpecRes. This is a specification but it's not complex. Also, any outcome is not specified complexity. A complex example is if your specification was f(10) and you roll the dice 10 times and you don't get a one, my jaw will drop. Despite the fact that it's not below the UPB.
Also, if you relax 3 then you allow both more complex, equally complex, or less complex specification into the set of specificational resources, in other words, pretty much everything.
Also, you missed my point with C18, I don't really care if you want to see a calculation worked out in full for a cake (or any designed object), if you did what I discussed in my response to you and Dembski's example is more than enough to do it yourself.
[ 10. May 2003, 16:53: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 10. May 2003 17:02
Nelson: Also, you missed my point with C18, I don't really care if you want to see a calculation worked out in full for a cake (or any designed object), if you did what I discussed in my response to you and Dembski's example is more than enough to do it yourself.
I think you miss Rex's point here namely that "Without a positive example worked out in full, it is very difficult to know whether it is even possible to apply the algorithm that Dembski has produced." If your argument is that sufficient data exists then the ID inference could benefit from such a worked out example. Since you seem to claim that there is enough information to do it yourself, I would suggest that you provide us with the details of the calculation and prove use critics wrong. Not only would this be immensely useful in furthering the arguments by Dembski but it may help to establish some examples of fully worked out examples of ID inference calculations. That by itself would be considered quite a positive contribution to the discussion on ID.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 11. May 2003 08:33
Nelson, if I roll a 10^150-sided die in my example, rather than a 6-sided die, then all the specifications are complex as measured by the Universal Probability Bound. (Granted, it's hard to actually perform 10^150ish rolls until a one is generated, but one can generate exactly the same distribution by recursive bisection, which only requires on the order of log(n) computations.)
And, anyway, the point of my example is that everything is specified, and yet the specificational resources are always 1. Doesn't that strike you as rather undesirable? Our specification makes it look like we've always hit the bullseye, except we're paying attention to a different bullseye each time.
Also with respect to C16, Dembski suggests that we should favor the non-chance hypothesis whenever p * SpecRes * ReplRes is less than 1/2. See, for example, NFL p. 79. Therefore, by Dembski's published method, my counterexample generates a result of specified complexity on every die roll, unless you arbitrarily pick a huge number for ReplRes.
Finally, relaxing (3) only admits more probable specifications, not every specification: condition (4) explicitly rejects more complex specifications.
With respect to C18, I'm not really in the mood to post a bunch of verbose floundering to demonstrate that I really can't generate a calculation worked out in full. Maybe at some point I will so you can see just how hopeless I am at performing such calculations. (I suspect that the problem is partly with the presentation, given that I'm decent at performing other calculations, as you may have noticed from some of my posts on here.) However, right now I am going to appeal to the burden of proof being on the one who presents a new idea. Positive examples are widely recognized as quite valuable in establishing the validity of an idea. I've done a fair few of them in math classes (e.g. demonstrating the GCD algorithm). It really should not be too much to ask, given that one is asking to overturn 150 years of heavily researched and remarkably successful science on the basis of the mathematical accuracy and applicability of the General Chance Elimination Argument!
Also, I should point out that I'm not just trying to be fussy here. Actually, I strongly suspect that there are no cases of specified complexity. Not a flagellum, not a cake, not Midsummer Night's Dream, not this posting, and not a watch in the desert. I could well be wrong--it's a hunch, not a proof. But right now, that there is even one case of specified complexity is also at the level of a hunch, given the steps that have been skipped in all examples we've seen so far.
IP: Logged
|
|
Nel
Member
Member # 614
|
posted 11. May 2003 13:21
Rex,
I'll respond to your post more fully tomorrow, but for now , can you fully discuss what steps have been skipped in Dembski's calculation that I havn't already shown that they exist?
How can SpecRes = 1? If the specification is f(10) then I have F(1) through f(10) as specifications. If you are saying that the specification is whatever happens, and only what happens, then it's not detachable from the event, it's a fabrication, your counterexample fails from the get-go. Either way your counterexample is stuck in a rut.
Can you show me a bacterial flagellum that works without it's stator? If you can't, how can you say that there are no examples of specified complexity?
Can you show me how you remove just one of the letters in this sentence and still retain it's meaning? If not, how can you say there are no examples of specified complexity? The meaning of these sentences have nothing to do with the computer processes that have brought them to your screen, it's not even reducible to any of these single letters. And if I were to do a probability for getting these words onto your computer screen, how likely do you think that would be?
If I do the Pdco calculation for your post, would that be a positive, fully worked out example of specified complexity?
Anyway, I'll apply all of this, especially to C16, tomorrow. Happy Mother's Day to all your moms ![[Big Grin]](biggrin.gif)
[ 11. May 2003, 14:51: Message edited by: Nelson_Alonso ]
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 12. May 2003 10:11
Sorry, Rex Kerr, but I was under the impression that this was an Internet discussion board. Nor will I make any pretense to “scholarship.” But Mr. Moderator may disabuse me.
In The Design Inference, Dr. Dembski states he will attempt to formalize the intuition that we all share, i.e., the inference or discrimination that we all make with such apparent alacrity, ease, and reliability. That inference is, plainly, not that life forms are designed, but that anything is designed.
The critical demolition of Dembski’s formalism doesn’t change the fact we (including natural scientists) do this. Ironically, all it does is deprive Dr. Dembski’s critics of any formal basis for making arguments such as, “Life was not designed. It evolved.” (In case anyone thinks I’m being one-sided it also deprives Dembski-proponents from making the opposite argument, “Life was designed. It did not evolve.”) Of course, his critics may simply concede that their argument, such as it is, is not made on any formal, rigorously logical and scientific, basis. I’m sure Dr. Dembski would appreciate it and indeed may be quite gratified to hear it.
The common assumption of Dembski and his critics is that evolution and design are two different things and we are therefore capable of making a scientifically meaningful distinction here. But reviewing these discussions I think I might reasonably conclude that there is no scientific or logical basis for making this distinction in our minds. (I believe Dr. Dembski has recognized that possibility in his writings.)
I have attributed the false assumption to the conflation of Darwin’s arguments, peremptory as they are, against “special creation” as in any way addressing design. In my own “unscholarly” way I have simply consulted the source and lo and behold what do I discover there? Darwin himself makes an inference to and argument from design! To support his particular vision of evolution! I.e., Darwin plainly does not do what it is (carelessly) assumed or interpreted that he did. Where does the word “design” even appear in Origin? (I guess I’ll have to look that up, because I can’t recall that it ever appears there.)
Even though I am hardly a devotee of Darwin (or Dembski), I feel that whenever I make such “design” inferences, analogies, and arguments I stand well within the tradition of Darwin and that Darwin’s own even most passionate and committed defenders (Who may not have read Darwin?) don’t.
LOL Dr, Dembski appears to me to be more “Darwinian,” or at least more consistently so, then many of Darwin’s defenders! (I have noticed that Dr. Dembski is often expected to respond to some distinctly non-Darwinian arguments, such as the intuitively appealing argument that selection positively conditions variation. Darwin said the opposite.)
The problem remains with this common assumption. And I have been quite insistent that it be solved. If it can be solved. And if it can’t?…
IP: Logged
|
|
|
Printer-friendly view of this topic