Home » Evolution, Informatics, Intelligent Design » “Life’s Conservation Law: Why Darwinian Evolution Cannot Create Biological Information”

“Life’s Conservation Law: Why Darwinian Evolution Cannot Create Biological Information”

Here’s our newest paper: “Life’s Conservation Law: Why Darwinian Evolution Cannot Create Biological Information,” by William A. Dembski and Robert J. Marks II, forthcoming chapter in Bruce L. Gordon and William A. Dembski, eds., The Nature of Nature: Examining the Role of Naturalism in Science (Wilmington, Del.: ISI Books, 2009).

Click here for pdf of paper.

1 The Creation of Information
2 Biology’s Information Problem
3 The Darwinian Solution
4 Computational vs. Biological Evolution
5 Active Information
6 Three Conservation of Information Theorems
7 The Law of Conservation of Information
8 Applying LCI to Biology
9 Conclusion: “A Plan for Experimental Verification”

ABSTRACT: Laws of nature are universal in scope, hold with unfailing regularity, and receive support from a wide array of facts and observations. The Law of Conservation of Information (LCI) is such a law. LCI characterizes the information costs that searches incur in outperforming blind search. Searches that operate by Darwinian selection, for instance, often significantly outperform blind search. But when they do, it is because they exploit information supplied by a fitness function—information that is unavailable to blind search. Searches that have a greater probability of success than blind search do not just magically materialize. They form by some process. According to LCI, any such search-forming process must build into the search at least as much information as the search displays in raising the probability of success. More formally, LCI states that raising the probability of success of a search by a factor of q/p (> 1) incurs an information cost of at least log(q/p). LCI shows that information is a commodity that, like money, obeys strict accounting principles. This paper proves three conservation of information theorems: a function-theoretic, a measure-theoretic, and a fitness-theoretic version. These are representative of conservation of information theorems in general. Such theorems provide the theoretical underpinnings for the Law of Conservation of Information. Though not denying Darwinian evolution or even limiting its role in the history of life, the Law of Conservation of Information shows that Darwinian evolution is inherently teleological. Moreover, it shows that this teleology can be measured in precise information-theoretic terms.

  • Delicious
  • Facebook
  • Reddit
  • StumbleUpon
  • Twitter
  • RSS Feed

197 Responses to “Life’s Conservation Law: Why Darwinian Evolution Cannot Create Biological Information”

  1. Was this published in a peer reviewed journal?

  2. Looks like a book chapter. There are, however, two articles mentioned in the end notes. The prospective journals are blacked out, so it is unknown as to whether they will be peer reviewed.

  3. Thank you Dr. Dembski and Dr. Marks for formalizing what we understand intuitively.

    Unfortunately, I don’t think the fundamentalists across the aisle will concede so easily. It will be interesting to see how they respond (no Weaseling, please).

    Please do keep ‘racheting’ up the rigorous reseach you are doing.

    ID is humbly :) ‘letting the dogs out on scientific dogmatism’.

  4. ‘..information is a commodity,that,like money obeys strict accounting principles.’
    Aaah, like Bernie Maddoff, or the wise guru, Allen Greenspan, or more egregiously Reagan’s “supply side” fiasco.
    If LCI follows these “strict accounting principles” I’d more likely come down on the side of random mutation, and selection for fitness, if you don’t mind.
    Rob.

  5. Aaah, like Bernie Maddoff

    Not unless you are saying Bernie Maddoff is synonymous with money. There certainly was accounting with regard to the money along with tragic consequences for many.

    or more egregiously Reagan’s “supply side” fiasco.

    You have to have been born after 1988. Hint: don’t trust your teachers, and ponder this.

  6. Critter @1:

    Was this published in a peer reviewed journal?

    I personally hope not. Who needs the approval of a bunch of politically correct groupies? We don’t need groups with hidden agendas to tell us that our stuff is good. Peer review, to me, is synonymous with ‘@rse review’.

    Sorry. I just don’t like mainstream science too much. The entire world should be our peers. If your stuff is good, it will survive and win out in the end. If not, it will die. How’s that for a simple theory of evolution?

  7. trib,
    I was born in 1966, lived through the nightmare of Reagonomics, and Thatcherism, and have lived to see the obvious (to me at least) consequences. I witnessed Reagan’s disgrace in dealing with terrorist Iran, and his disgrace in supporting white South Africa.
    Again, if LCI follows the strict accounting principles of moneterist ideals, I suggest it is more akin to ideology than to science.
    I couldn’t follow your link, sorry.

  8. What I’m shocked at is, no one seems to be reacting to a key quote in this post.

    Though not denying Darwinian evolution or even limiting its role in the history of life, the Law of Conservation of Information shows that Darwinian evolution is inherently teleological.

    I suppose it wouldn’t be news to regulars here, but here we have Prof. Dembski arguing in a paper that, even if “Darwinian Evolution” is true (in fact, it seems to be assumed it is true), the facts on the ground indicate it’s teleological.

    I’m sure many people would question such a view. But considering how often ID is billed as evolution-denying, isn’t this an eye-opener?

  9. It seems to me that what this paper is saying is that evolution was designed, and I agree. Please correct my interpretation if it’s wrong.

  10. rvb8 -
    “Again, if LCI follows the strict accounting principles of moneterist ideals, I suggest it is more akin to ideology than to science.”

    You are confusing economic accounting principles with policital economy. Regardless of what you think of Reagan’s economic policies, or economic performance under the Reagan administration, or what you label monetarist ideals, it is still true, and must be true, that the total flow of money must sum to zero (outflows = inflows).

  11. 11
    Timothy V Reeves

    Thanks very much; this looks very interesting, I’ll give it a good look.

    Mapou says at 11

    It seems to me that what this paper is saying is that evolution was designed, and I agree.

    …that might be the opinion I’m beginning to form as well.

  12. I urge people to read the paper before commenting. What in the world are references to Reaganomics and Thatcherism doing in this thread? We (Bob Marks and I) argue that active information obeys an accounting principle. That claim needs to be taken on its own terms. If there’s a tight analogy with money, so be it. But there’s also a tight analogy, so we argue, with energy — does anyone on this thread deny that energy obeys strict accounting principles? Anyone here prepared to challenge conservation of energy? If so, you need to contact Al Gore and offer some solutions for global warming.

  13. nullasalus: You quote “Though not denying Darwinian evolution or even limiting its role in the history of life, the Law of Conservation of Information shows that Darwinian evolution is inherently teleological.” But you didn’t quote the following sentence, which concludes the thought: “Moreover, it shows that this teleology can be measured in precise information-theoretic terms.” That’s the kicker, and it’s why this paper is properly an ID paper rather than a theistic evolution paper.

  14. Dr. D you are completely right with regard to politics and the discussion of your paper.

    Money, btw, is just a container of information. If the information it contains should be found to be inaccurate or unreliable, money will be worthless. It’s why it is pointless to try to make everyone a millionaire by printing and distributing $1,000 bills.

    I will now be a good boy and read the paper.

  15. Thank you, Dr. Dembski, for offering us a substantial topic to discuss!
    Where, then, does the LCI Regress end? In fact, it may not end, implying that the information that enables an alternative search to succeed in locating T was present always. Alternatively, it may end because an external information source added the information needed to locate T. One option suggests front-loading of information, the other direct input. Both options evoke intelligent design.

    However, the authors criticize both possibilities earlier.

    Theistic evolutionists attempt to make room for God within this Darwinian scheme by claiming that God created the universe so that Darwinian processes would produce living forms. Accordingly, God brings about the natural forces that produce living things but makes sure not to interfere with those forces once they are in operation. Though logically possible, theistic evolution offers no compelling reason for thinking that nature is a divine creation.
    And earlier

    There’s no sense in which human beings or any other multi-celled organisms are latent in single-celled organisms, much less in nonliving chemicals.

    I found the citation of MESA somewhat amusing. Why should the existence of a program be more important than the results obtained with the program? If you are going to put MESA up against AVIDA you should be able to point to some results. The ISCID.org page does not list any results. In any case, MESA was meant to counter WEASEL, not AVIDA, so the authors are vastly ovestating its relative importance.

    As a mathematical result, I look forward to how the authors will apply it in the controlled environment of computational evolutionary algorithms. I especially look forward to their reconciliation of their LCI to Holland’s Schema Theorem. While Barricelli may be of historical interest, a rigorous discussion properly begins with Holland.

    I will also be interested to see how the authors complete the research program of creating a true conservation Law. Having shown information cannot be created is it also true that it cannot be destroyed? Here, the authors will have to grapple with results from research into black holes.

    Ultimately, computational evolution uses resources not yet accounted for by the authors, such as the contingency of one query on previous results (the existence of time), and a source of (pseudo)random numbers. So we are still left with the question of whether the laws of nature (fitness function) plus time, plus heat (randomness) can create local structures of arbitrary complexity at the cost of entropy increases in other parts of the system.

  16. 16
    Timothy V Reeves

    William at 12

    I’ve never quite grasped this parallel between information and matter-energy conservation.

    A disordered sequence of bits, perhaps generated by a coin, is packed with “surprisal” value and therefore packed with information for the uninitiated observer.

    But then let’s say the observer suddenly learns that the sequence is in actual fact generated by a simple chaotic algorithm that can be expressed in a few bits; the burden of information initially apparent to the observer then seems to have disappeared into nothingness. Conversely if the observer forgets all about the algorithm the information jumps back out of nothing.

    The problem here seems to lie in the fact that information as a quantity binds together both observer and observed into one system – the observer becomes a variable in the system and this produces results that confound attempts to find strict analogies between physical quantities like mass-energy and quasi-objective quantities like information.

    However I think this is really only a technical quibble as I’m sure the real moral behind the paper is that the improbability of life has to be matched either by a logical hiatus to be found either in evolution itself or directly in living structures in the form of IC. But I’ll leave final judgment until I’ve read the paper.

  17. Dembski and Marks write,

    Theistic evolutionists attempt to make room for God within this Darwinian scheme by claiming that God created the universe so that Darwinian processes would produce living forms. Accordingly, God brings about the natural forces that produce living things but makes sure not to interfere with those forces once they are in operation. Though logically possible, theistic evolution offers no compelling reason for thinking that nature is a divine creation.

    It seems that Dembski has confused theistic evolution with deism, which strikes me as odd since Dembski has a theology PhD (I think). Also, the last line above doesn’t seem correct – even deism, and even more so theistic evolution, would consider nature a divine creation.

    Theistic evolution posits that God is present in all events – not in an “interfering” way, but rather as an active participant. Christians don’t doubt that God is subtly guiding their lives towards the ends that God desires, so I don’t see why they would doubt that God could likewise guide evolution.

    I’m pretty sure this is not the thread on which to discuss this, and it may not relate to Dembski’s thesis because this participation by God could include a constant imparting, or changing, of information. But the paragraph quoted above seems to me to both mischaracterize and unreasonably dismiss theistic evolution as a religious perspective that can accept evolutionary science.

  18. Hazel makes several claims about deism and theistic evolution, including:

    It seems that Dembski has confused theistic evolution with deism, which strikes me as odd since Dembski has a theology PhD (I think). Also, the last line above doesn’t seem correct – even deism, and even more so theistic evolution, would consider nature a divine creation.

    Saying so immediately after quoting Dembski:

    Though logically possible, theistic evolution offers no compelling reason for thinking that nature is a divine creation.

    This problem is recurrent. Assertions of belief do not provides grounds for belief.

  19. Does not all Christian theology consider nature a divine creation? I’m puzzled by what point I am missing.

  20. 20

    The paper includes this interesting aside:

    The wedding of teleology with the natural sciences is itself a well established science—it’s called engineering. Intelligent design, properly conceived, belongs to the engineering sciences.

    Are you alluding to some new conception of intelligent design that places it within the engineering sciences? If so, I’m sure that engineers would like to hear more about it.

    If not, perhaps you and Dr. Marks could illustrate your point by providing some “war stories” of times when engineers, as an important part of an engineering project, studied patterns in nature to determine whether or not they were the result of intelligence.

  21. Excellent essay Dr. Dembski and Dr. Marks.

    Others,

    Please don’t begin to discuss TE vs. Deism, etc, as this does a disservice to Dr. Dembski and Dr. Marks’ work. The paper isn’t about that and only references it tangentially.

    The key issue is that replication, mutation and selection by themselves do not create information as long as the wrong fitness function is used. And to find a “correct” fitness function (relative to your target) out of the space of all possible fitness functions requires an input of information not less than the active information, which is the information gained over blind (null) search. (Formally, log(q/p), where p is the probability of finding your original target using null search and q is the probability of finding the target using your new assisted search.)

    So adding a fitness function and selection scheme doesn’t resolve the information problem, it only keeps it the same or increases it, since the space of possible fitness functions is exponentially larger than the original search space and now you must perform a search for a good fitness function.

    This is what should be the topic of discussion, in my opinion.

    BTW, Proximity Reward Search vs. Proximity Neutral Search in Weasel 2.0 shows the same principle in an empirical manner. Proximity Reward Search uses an information rich target specific fitness function out of the space of all possible fitness functions and so can easily find the target. Proximity Neutral Search uses different fitness functions out of the space of possible fitness functions (and users can use their own), which do not necessarily encode target specific information. The result? Proximity Neutral Search usually fails to find the target, even though the replication, mutation and selection are present as they were before. Check it out for yourselves.

    Atom

  22. Please consider my posts at 17 and 19 unnecessary and tangential, and carry on. Sorry.

  23. Sorry hazel, didn’t mean that as a jab towards you. I’d just hate to see yet another UD thread get side-tracked and us lot are easily distracted. :)

  24. I’ve only read the paper once through, and that is never enough to really gain a comprehensive grasp of any scientific paper (if such a thing is even possible), and so I will refrain from commenting on most of the content until I have had more time to analyze its contents.

    However, this phrase (already noted by nullasalus in comment #8) definitely caught my attention:

    “Though not denying Darwinian evolution or even limiting its role in the history of life, the Law of Conservation of Information shows that Darwinian evolution is inherently teleological.”

    My first reading of Dr. Dembski and Dr. Marks’ paper is that they do indeed accept that evolution has occurred and that natural selection is one of its principle mechanisms. What their analysis seems to indicate is that the common interpretation of natural selection – that it consists of the demographic outcome of variety, heredity, and fecundity – is insufficient to explain the observed changes in phenotypic complexity indicated in the record of evolution (presumably the fossil and genomic records). Their analysis (which I stress I have not reviewed in detail as yet) indicates that some other process (dynamics and source unspecified) provides a teleological “guidance” for the evolutionary process.

    Dembski and Marks cite N. Barracelli’s (1962) conclusion that the commonly asserted formulation of evolution by natural selection of mutation + inheritance + reproduction + selection isn’t sufficient to produce the observed diversity (and presumably also functionality/”adaptedness”) of life:

    “The selection principle of Darwin’s theory is not sufficient to explain the evolution of living organisms if one starts with entities having only the property to reproduce and mutate. At least one more theoretical principle is needed, a principle which would explain how self-reproducing entities could give rise to organisms with the variability and evolutionary possibilities which characterize living organisms.”

    Barricelli, N. (1962) “Numerical Testing of Evolution Theories, Part I: Theoretical Introduction and Basic
    Tests,” Acta Biotheoretica 16(1–2) (1962): 69–98. Reprinted in David B. Fogel, ed., Evolutionary Computation: The Fossil Record (Piscataway, N.J.: IEEE Press, 1998), pp. 170-171

    Furthermore, they apparently base this conclusion on a mathematical analysis of the dynamics of informational change, based primarily on “no free lunch” theories of informational change over time. I say “apparently” as I have not yet fully dissected the content of their paper, and so cannot yet state definitively if this is the case.

    Their suggestion is intriguing, as the problem of teleology (quite apart from its connection with theology) is a perennial one in both philosophy and science. Perhaps the most intriguing aspect of their analysis is that Dembski and Marks conclude their paper with a section entitled “A Plan for Experimental Verification”. In it, they assert (in the basis of their analysis of a theoretical analysis of the dynamics of informational change) that:

    “Evolving systems require active information. How much? Where do they get it? And what does this information enable them to accomplish? Tracking and measuring active information in line with the Law of Conservation of Information is the plan we propose for experimentally verifying intelligent design…” Dembski, W. & Marks, R. (2009) Life’s Conservation Law: Why Darwinian Evolution Cannot Create Biological Information, (in press), pg. 34

    They go on to suggest that:

    “Just as information needs to be imparted to a golf ball to land it in a hole, so information needs to be imparted to chemicals to render them useful in origin-of-life research. This information can be tracked and measured. Insofar as it obeys the Law of Conservation of Information, it confirms intelligent design, showing that the information problem either intensifies as we track material causes back in time or terminates in an intelligent information source. Insofar as this information seems to be created for free, LCI calls for closer scrutiny of just where the information that was given out was in fact put in.” ibid

    My only quibble at this point is the phrase highlighted above. In my view, it would seem that a successful outcome to the empirical research program suggested by Dembski and Marks would indeed confirm that evolution is teleological, but whether “teleological” equals “intelligently designed” seems to me to be a completely separate question.

    I will try to explain why I believe this is the case in a later post.

  25. 25
    AmerikanInKananaskis

    Hmm. One of my comments critical of this “article” appears to have been deleted.

  26. Let’s be clear where our argument is headed. We are not here challenging common descent,
    the claim that all organisms trace their lineage to a universal common ancestor. Nor are we
    challenging evolutionary gradualism, that organisms have evolved gradually over time. Nor are
    we even challenging that natural selection may be the principal mechanism by which organisms
    have evolved. Rather, we are challenging the claim that evolution can create information from
    scratch where previously it did not exist. The conclusion we are after is that natural selection,
    even if it is the mechanism by which organisms evolved, achieves its successes by incorporating
    and using existing information.

    So we all agree on common descent from a universal common ancestor, gradual evolution of morphology, and that natural selection sifts the “fittest” from the available variety. The only area of contention is whether the source of variety is produced by mutations (using that in its broadest form, Allen) or whether the new information is provided by some teleological force or entity (not sure ab out best choice of word here). That certainly sets ID in clear distinction from YE creationism.

  27. Mr Atom,

    What is the fitness function of biological evolution, and how was it searched for? From what space of alternatives was it chosen?

    How do we go from a definition of active information to an effective procedure for tracing it, as the authors suggest is a proper research program?

    The input to a blind search is a list of random numbers, which on average has to be half the size of the search space.

    The input to an evolutionary search is the fitness function, a similar list of random numbers, a clock, and a lattice of tuples (sets of independent and dependent variable values) representing knowledge of previous queries.

    Taken as given the definition of active information by Dembski and Marks, we still have to trace the active information to its source, which could be any of the multiple inputs. The active information could even be hidden in the list of random numbers (think of Cavuto’s biased coin).

    To me, being able to partition active information among these multiple sources will be a significant result of the LCI research program. I am looking forward eagerly to further progress.

  28. Mr AmerikanInKananasKis,

    I agree your comment is gone. Most unfortunate.

    If I recall, you commented on the section:

    But the fact that things can be alive and functional in only certain ways and not in others indicates that nature sets her own targets. The targets of biology, we might say, are “natural kinds” (to borrow a term from philosophy). There are only so many ways that matter can be configured to be alive and, once alive, only so many ways it can be configured to serve different biological functions.

    You simply stated a disagreement with this terminological borrowing. Can you be more expansive in your disagreement? Thank you.

  29. Nakashima wrote:

    What is the fitness function of biological evolution, and how was it searched for? From what space of alternatives was it chosen?

    If we think of a fitness function that takes in all the current environmental factors as well as the current state of organisms within that environment (think co-evolution and density dependency), we will never be able to enumerate that fitness function, since the number of possible permutations that we’d have to assign fitness values to would greatly exceed the number of particles in the universe. So we have to start small.

    Take the phrase, “METHINKS IT IS LIKE A WEASEL” to borrow a well-discussed phrase. If we limit ourselves to the 26 capital latin letters plus one space character, we have 27 ^ 28 possible permutations of that length (28 letters long). Now lets say we want to assign a “fitness value” to each of those permutations, ranking them in some order. If we limit ourselves to integer values between 0 and 9, we would have 10 ^ (27 ^ 28) unique ways to assign those values. In other words, we have that many unique fitness functions we could choose from.

    There are estimated roughly 10 ^ 80 atoms in the universe, so it would be physically impossible to enumerate all of the fitness functions even for 28 letter strings, even if we “wrote” one function on each atom in the cosmos. We can forget about doing so for millions of base pairs of DNA and organismal interactions.

    Luckily, we don’t need to.

    Continuing with my Weasel example, let’s say we now limit ourselves to the Proximity Reward fitness function which ranks permutations based on distance from our target. How many Proximity Reward functions are available to choose from within our space of possible fitness functions? Roughly 10 ^ 40. Not coincidentally, this is roughly the same number of permutations in our original search space. (In other words, we can have a single-peaked smooth fitness function like Prox Reward that has its peak at any one of the 27 ^ 28 elements in our original search space.) So selecting the single-peak function that has its peak at our target, from all the other single-peaked functions that have their peaks somewhere else other than the target is just as hard as our original search.

    Furthermore, since there are many other types of fitness functions other than single-peaked proximity reward functions, the search for a fitness functions is actually exponentially harder than our original search.

    Now, viewing this from an information perspective as Dembski/Marks do, we can see that the reduction of the fitness functions to a suitable one requires an information input not less than the active information we gained in using our fitness function. If “nature” limited the fitness function to one that finds biological targets readily, then the information needed to reduce the set of fitness functions to that particular one (or set of them) cannot be less than the information we gain by using that fitness function. In other words, reducing the space of possible fitness functions to the small space of “good” fitness functions is as hard or harder than reducing the space of all configurations to the small space of our target in the original search.

    We still have to account for the information.

    Atom

  30. Alan — So we all agree on common descent from a universal common ancestor, gradual evolution of morphology, and that natural selection sifts the “fittest” from the available variety.

    No. We all agree that the paper doesn’t challenge those things.

  31. Can we all agree that the more information a searcher has, the more likely the search will be successful?

    Can we all agree that there is a cost in acquiring the information to increase the probability of finding new information?

    Can we all agree that evolution is purpose driven which means there is a target (survival) which requires pre-existing information?

  32. Mr Atom,

    I agree that a large calculation is unrealistic for biological evolution. We have no evidence that bacteria (for example) have the computational resources for such a thing. So I agree with you that that the fitness function across broad classes of biological evolution is extremely simple, such as “if resources > threshold, reproduce = true”. This is a completely local function, not even testing the environment.

    Of course, the problem with such a simple fitness function is finding a way to assign very much active information to it. So it may be that the active information has to be located in the contingent history, not in the fitness function. I am looking forward to results in this area of research!

  33. Can we all agree that the more information a searcher has, the more likely the search will be successful?

    If this is information about the location of the object that the searcher is looking for, I would have thought so, yes.

    Can we all agree that there is a cost in acquiring the information to increase the probability of finding new information?

    No, because I can’t parse the statement, sorry.

    Can we all agree that evolution is purpose driven which means there is a target (survival) which requires pre-existing information?

    Definitely not!

  34. 34

    William Dembski,

    I agree that this is an ID paper rather than a TE paper, and didn’t mean to imply otherwise. I simply meant that so often the accusation against ID is that it (among other falsehoods) entails the denial of evolution, or common descent, etc.

  35. Can we all agree that there is a cost in acquiring the information to increase the probability of finding new information? . . .No, because I can’t parse the statement, sorry.

    I guess for this to make sense you have to start with the acceptance (or at least consideration) Manfred Eigen equating the problem of life’s origin with uncovering the origin of information as per the paper, and the claim that biology is information.

    Once this is done, however, even if you don’t accept the premise that information is perfectly conserved it should be apparent that information is required to narrow (and increase the chance of success) of a search greater than by doing it blindly.

    And that time, effort, searching etc. would be required to acquire the information needed to make the search for the other information successful.

  36. I guess for this to make sense you have to start with the acceptance (or at least consideration) Manfred Eigen equating the problem of life’s origin with uncovering the origin of information as per the paper, and the claim that biology is information.

    I need to start with being sure what is meant by “information” in this context. I think I know what is generally understood by it, qualitatively, but I get the feeling it has a rather more loaded meaning and is, somehow, quantifiable.

  37. …it should be apparent that information is required to narrow (and increase the chance of success) of a search greater than by doing it blindly.

    Yes,seems reasonable. A street map is a great asset in an unfamiliar city.

  38. And that time, effort, searching etc. would be required to acquire the information needed to make the search for the other information successful.

    I need to find the bookstore to buy the street map. OK.

  39. Wesley Elsberry remarks, at AtBC, on the paper

    …in his 1997 talk at the NTSE conference in Austin…Dembski provided a quantification of the amount of information that natural selection could fix per generation as -log_2(1/n) bits, where n was the number of offspring an organism had. Bill Jeffereys asked Dembski if a mutation for, say, a coat color in dogs had a different amount of information simply because of a difference in the number of pups a bitch might have in her litter? “Active information” shares exactly the same defect.

    He seems to share my problem with the attempt to present “information” as a scalar quantity.

  40. Alan– but I get the feeling it has a rather more loaded meaning and is, somehow, quantifiable.

    I didn’t see it defined in the paper. I’m guessing he was using this one which would be scalar.

  41. …the difference between two forms of organization or between two states of uncertainty before and after a message has been received.>/blockquote>Now this illustrates my problem. No problem to estimate the comparative amount of information in two stretches of DNA of different length (though maybe not if one sample is of tandem repeats). No problem to compare the information in two street maps of different quality or scale. But can you sum the information in DNA with information in the street map?

    The same problem arises with intelligence. Assessments of intelligence in people are made comparatively against an average.

  42. Oops.

    Previous post should have said:

    From your link, I get:

    …the difference between two forms of organization or between two states of uncertainty before and after a message has been received.

    Now this illustrates my problem. No problem to estimate the comparative amount of information in two stretches of DNA of different length (though maybe not if one sample is of tandem repeats). No problem to compare the information in two street maps of different quality or scale. But can you sum the information in DNA with information in the street map?

  43. Aaah, like Bernie Maddoff

    Actually, a Ponzi scheme is an excellent case of conservation of money, since the whole problem is that in the long run no more money can come out than goes in. That said, there’s a meaningful and widely-used sense in which fractional-reserve banking (for example) creates money, and in general there’s a murky and shifting relationship between money and wealth, the latter of which is certainly not conserved. So I’d be leery of the money analogy as it’s likely somewhat inaccurate or at least liable to breed misunderstandings and tangential gotcha arguments (“I am shocked by Dembski’s demonstrated ignorance of even basic economics” etc. etc.).

  44. But can you sum the information in DNA with information in the street map?

    Alan, that’s one of the foundational points of the paper. Look at page 12 with regard to Active Information.

    What is the probability of finding a particular building in a city with a certain number of streets and buildings in a blind search? What is the probability of finding it with a map?

    What is the probability of a tendon forming in the appropriate spot without DNA? With DNA?

    You can make a comparison between them with regard to information content.

  45. One of the things I have found missing in all discussions of “information” that I have read (including Dembski, Marks, Kolmogorov, and Shannon) is a consistent and testable definition of meaningful information. By that, I mean the quality (or qualities) that distinguish between a random series of bits and a “meaningful” message/signal.

    For example, the following six strings of letters contain the same three letters, yet only two of them are “meaningful” while the rest of them are “meaningless” (to people who can read English, of course):

    dgo dog gdo god odg ogd

    We can note the following about the empirical qualities of the six three-letter strings:

    1) they contain exactly the same letters

    2) they also contain exactly the same number of letters

    3) the only thing distinguishing between them is the sequence in which the letters is presented in each string (reading from left to right or from right to left, it makes no difference which)

    4) “something” about the two “meaningful” strings – “dog” and “god” – distinguishes between them and the other four, yet this is clearly not an intrinsic property of the strings themselves

    5) the “meaning” of the two “meaningful” strings – “dog” and “god” – inheres not in the strings themselves, but somewhere else

    6) The “meaning” in the two “meaningful” strings is analogous (and possibly more than just analogous) to the “meaning” that inheres in the codons in mRNA, the anticodons in tRNA, the corresponding three-base strings in DNA, and the sequences of amino acids in proteins which they specify, all of which taken together form the fundamental basis of all biological information.

    7) Neither Kolmogorov nor Shannon definitions of “information” can be used to determine which of the strings in the set of six I provided is “meaningful” and which ones are “meaningless”.

    Ergo, it seems to me that none of the current theories of information, including the one presented in the post that heads this thread, address the question of “meaningful information” directly. It also seams to me that the question of where “meaningful” information comes from, and what makes it “meaningful” (and what makes other strings, such as those in non-coding DNA “meaningless”) may be tied in some way to Dr. Dembski and Dr. Mark’s concept of “active information”.

    Finally, as Dr. Dembski and Dr. Marks point out in their paper, the question of information is clearly tied to the question of teleology. I agree, but would extend this realization by asserting that it is my intuition that this is primarily because teleological processes require “meaningful” information. Unfortunately, this is only an intuition; I have neither evidence nor logic with which to explain it, nor (to my knowledge) does anyone else.

    Ergo, until we can fully understand what makes information “meaningful”, we will not fully understand either information (in all its forms) nor teleology (in all its forms).

    I am neither sufficiently conversant with the mathematics of current information theory nor the current status of semantics to know if my questions are addressed in them. I suspect they are not, but if someone reading this thread has information to the contrary, I would greatly appreciate it if they would enlighten me.

    One more thing: it may be (as Wittgenstein suggested) that the questions I have asked here are unanswerable. I would like to know if this is the case as well. Does anyone know?

  46. Just for fun, copy the following line and paste it into a document with a spell-checker. The result is enlightening, and possibly intriguing as well:

    dgo dog gdo god odg ogd

    If your spell-checker is like my spell-checker, something interesting happens when the spell-checker hits the string. The same “interesting” event also happens in your mind when you “read” the string. Are these “interesting” events related to the phenomenon about which I was asking in comment #45, and if so, how and why?

  47. 47

    Allen,

    Two noteworthy takes on “meaningful information”:

    P. Vitanyi, 2006, Meaningful Information

    D. H. Wolpert, 2008, Physical Limits of Inference

  48. Allen,

    Here are two references, one of which was discussed recently here but I think superficially

    The Capabilities of Chaos and Complexity by Abel

    http://www.mdpi.com/1422-0067/10/1/247

    Information Theory in Molecular Biology by Adami

    http://arxiv.org/abs/q-bio.BM/0405004

  49. Allen,

    It seems clear that in biology, “meaningful” information corresponds to functional proteins, nucleotides, etc. While the meaning of words in a human language is quite arbitrary, there is no such arbitrariness in biology; either the amino acid sequence is functional or not.

  50. Mr jlid,

    There is a position that the choice of chirality was arbitrary, and we use L-amino acids and D-sugars by chance. In our biology, an L-sugar is non-functional, not metabolized, which makes it attractive as a kind of sweetener. It would still “function” in that way! :)
    So meaning is still contextual, even in biology.

  51. 51

    Allen,

    —-”Ergo, it seems to me that none of the current theories of information, including the one presented in the post that heads this thread, address the question of “meaningful information” directly. It also seams to me that the question of where “meaningful” information comes from, and what makes it “meaningful” (and what makes other strings, such as those in non-coding DNA “meaningless”) may be tied in some way to Dr. Dembski’s and Dr. Mark’s concept of “active information”.

    Replace “Dr. Dembski and Dr. Marks” with “Dr. Dawkins’ and the Weasel simulation” and you have a point.

  52. It seems clear that in biology, “meaningful” information corresponds to functional proteins, nucleotides, etc. While the meaning of words in a human language is quite arbitrary, there is no such arbitrariness in biology; either the amino acid sequence is functional or not.

    So I have a genetically engineered bacterium that contains the appropriate sequence for producing human insulin. I can use a broth of these bacteria to produce insulin commercially. I experiment by substituting a different sequence of the same length. I produce my new protein and it does not have any biological function that I can find. Which bacterium contains more information?

    I can see which bacterium is more useful; which bacterium contains more meaningful information, to take on Allen’s word, but how can I know without testing the new protein.

    What if that protein turned out to have a completely novel and beneficial function in humans, such as dissolving the plaques that cause Alzheimer’s? Would it then have more information?

    More generally, how can I claim that information is conserved if I can’t quantify it?

  53. The four papers mentioned above are all available on free access. Wolpert is hard going, Tom!

  54. PS Perhaps Paul Vitanyi should be credited with coining the phrase “meaningful information”.

  55. Alan Fox asks:

    More generally, how can I claim that information is conserved if I can’t quantify it?

    And even more fundamentally, why is it called the “Law of Conservation of Information” when it doesn’t rule out the loss of information?

    Conservation laws are about quantities that remain constant.

  56. In #51 Clive Hayden wrote:

    “Replace “Dr. Dembski and Dr. Marks” with “Dr. Dawkins’ and the Weasel simulation” and you have a point.”

    OK:

    ”Ergo, it seems to me that none of the current theories of information, including the one presented in the post that heads this thread, address the question of “meaningful information” directly. It also seams to me that the question of where “meaningful” information comes from, and what makes it “meaningful” (and what makes other strings, such as those in non-coding DNA “meaningless”) may be tied in some way to Dr. Dembski’s and Dr. Mark’s concept of “active information” “Dr. Dawkins’ and the Weasel simulation”.

    To me, that seems to have rendered this particular quotation meaningless, at least in the context of this discussion. Was that your point?

    Indeed, what exactly was your point, Clive? You gave no clue in your comment, and I must admit I’m too obtuse to get it now…

  57. In #52 Alan Fox wrote:

    “So I have a genetically engineered bacterium that contains the appropriate sequence for producing human insulin. I can use a broth of these bacteria to produce insulin commercially.”

    A very interesting point, because the human insulin made by the bacteria, which is OC exactly the same insulin as we find in humans, has no biological “meaning” for the bacteria. This is one of the reasons we we produce the human insulin in bacteria: they don’t use it, and therefore don’t modify it, rendering it of questionable use in humans.

    Note that only a specific sequence of nucleotides in humans produces a “meaningful” version of human insulin, whereas the exact same sequence of nucleotides in bacteria produces a completely “meaningless” version of exactly the same polypeptide.

    If I am reading Dembski and Marks’ paper correctly (and it is OC quite possible I am not), then the nucleotide sequence that specifies the amino acid sequence for human insulin is “active information” in humans, but not so in bacteria, even though it is the exact same sequence. As Nakashima-san points out, context seems to be everything when it comes to “meaningful” information.

    In the same comment Alan Fox also wrote:

    “I experiment by substituting a different sequence of the same length. I produce my new protein and it does not have any biological function that I can find. Which bacterium contains more information?”

    I believe that if one is speaking of either Kolmogorov or Shannon information, the answer is that both the functional and the non-functional nucleotide sequences have exactly the same amount of information, but this is manifestly not the case if one is speaking of “meaningful” information.

    Ergo, we have the very peculiar (and to me, very interesting) situation in which the exact same sequence of nucleotides (i.e. theoretically the exact same string of bits/information) contains absolutely crucial “meaningful” information when produced in its full biological context (i.e. if a human can’t “read” the information in this bit string, s/he is dead), but in a different biological context, the exact same string has no “meaningful” information at all. Indeed, if the accumulation of human insulin inside the bacterium has any negative effect on the bacterium (say, by reducing the ability of the bacterium to make its own “meaningful” proteins), then the exact same string of information can even have negatively “meaningful” information in the bacterium.

    Let me emphasize once again that the “meaningfulness” of the bit string is clearly not an intrinsic characteristic of the bit string itself. Rather, all of the “meaningfulness” of the bit string can only become manifest when it is “expressed” in a particular biological context.

    Once again I ask the question, does the foregoing interpretation of “meaningful” information match Dembski and Marks’ definition of “active information”? I think perhaps it does (at least in some ways), but I’m not certain.

  58. If, indeed, “meaningful” information (as demonstrated earlier in this thread) is synonymous with “active” information, as defined by Dembski and Marks, then we have the very interesting situation described by Alan Fox in comment #52:

    “What if that protein turned out to have a completely novel and beneficial function in humans, such as dissolving the plaques that cause Alzheimer’s? Would it then have more information?”

    If we assume that such a thing is possible (and I can see no biological reason that it cannot be), then indeed a bit string that is “meaningless” in bacteria (i.e. one that has been completely artificially constructed by the experimenter, as described in Alan’s comment) has come to be imbued with biologically “meaningful” information, simply by having it expressed in a different context (i.e. not by changing anything intrinsic in the bit string itself, and therefore not by changing its Kolmogorov or Shannon information).

    Let me return to Dr. Dembski and Dr. Marks’ definition of “active information”. If “active information” is indeed synonymous with “meaningful information” as demonstrated by the gedankenexperimenten performed above, then the “activeness” of “active information” is entirely a function of context.

    To me, this implies that genetic information (i.e. the information contained in the nucleotide sequences of DNA and RNA and the information contained in the amino acid sequences of the proteins for which they code) is rendered “meaningful” (and therefore “active”) completely by its context within a living cell/multicellular organism. And, by the same argument, the “meaningful” information that makes up the organism is rendered “meaningful” via the interactions between that organism and its ecological environment. Which, in turn, seems to imply (at least to me) that for any kind of information to be “meaningful” in an evolutionary sense (i.e. for it to have evolutionary consequences), it must be expressed in an ecological context (in the same way that, to paraphrase G. Evelyn Hutchinson, an evolutionary “play” only becomes “meaningful” when performed on an ecological “stage”.

    Ergo, it ultimately the ecological context of any form of biological information that renders that information “meaningful” (and therefore “active”) in a biological sense. Which strongly implies (at least to me) that the “missing information” that Dembski and Marks refer to in their paper (i.e. the information that is “smuggled in” and that makes “active information” active) is just what evolutionary biologists say it is: the totality of the information contained in any given ecological environment which renders the biological information contained in the phenotypes (i.e. “bodies”) of the organisms that live in it “meaningful”.

    If this be error and upon me prov’d, I never writ, nor no information
    Were ever made “meaningful”
    By being expressed in context…

    …maybe? (sorry, Bill)

  59. Sorry, I somehow dropped the phrase “appears to me that” following the word “it” in the first sentence of the next-to-last paragraph above, rendering its meaning less “meaningful”.

    Perhaps my subconscious is playing little tricks on me, eh? Or maybe it’s because I had a mug of cocoa/coffee with my lover’s chocolate cake last night, and have been awake intermittently ever since.

    Interesting: the caffeine in the coffee and the phenylethylamine in the chocolate have a different “meaning” in the plants that make them. In the coffee and cacao plants, they function as herbivore repellents, but in me, they function as sleep repellents…

  60. To all:

    Does the foregoing imply that “meaningful” information can, indeed, be created out of “airy nothing” (which seems to give a local habitation and a name), and therefore is there perhaps something not quite right about the concept of “active information”, at least insofar as it is asserted that it must be conserved?

    Just curious…

  61. Damn, damn, DAMN: left out an “it” in the mangling of Shakespeare in the previous quote…oh, the Bard (and my old director) would be so PISSED (indeed, he often was when I did this during performances, thereby altering the “meaning” of a line in ways the Bard of Avon never intended).

    To the trail of meaningful “words” there is no end…

  62. …but the trail of “meaningless” bit strings is infinitely longer.

  63. The birds are finally singing again; time to do something (anything!) else…

  64. 64

    “Damn, damn, DAMN”

    Relax Allen. From the standpoint of reason, the past few days here have not been so kind to you. Maybe its time to take a break. :-)

  65. This thread has a very strange “feature”. If I come into it first (that is, by entering UD from somewhere else) and then try to register to post comments to it, I get a message that says “Comments are closed”. However, if I first go to some other thread and register, the comment box opens up there. If I then return here, the comment box is open here, too.

    Another weird side effect of “meaningful” information, perhaps? My attempt register to comment on this thread “means” one thing here, but something completely different at some other thread, even thought the information I attempt to enter is exactly the same.

    Context is everything…

  66. then the nucleotide sequence that specifies the amino acid sequence for human insulin is “active information” in humans, but not so in bacteria, even though it is the exact same sequence.

    Allen, as I understand the paper the active information in the sequence would be the difference between the probability of it occurring in a blind search vs. the probability of it occurring in the (successful) “alternative search”.

    So, whether the sequence is used or not the information content would be the same. I comparison might be having a “men at work” sign at a work site or having it in a storage room.

  67. And even more fundamentally, why is it called the “Law of Conservation of Information” when it doesn’t rule out the loss of information?

    As I understand it, the paper says there is no information loss.

  68. Dembski uses the phrase “meaningful information” once in the context of language. We can certainly have a thread of 300+ comments by everybody under the sun on the metaphysical significance of “meaningful information” but it may not have meaning with this paper.

    Maybe if we are confused we should concentrate on “active information” and try to get that into layman’s English so we can try to understand the paper.

    Certainly we can also try to understand the Law of Conservation of Information and what that means in terms of layman’s English. Does it mean that if a book is burned that we lost information, a potentially natural act, or does it mean something else. I am not sure any of us know so it may be fruitful to try and ascertain what it implies as opposed to sneer from our ignorance. If it is a vapid concept it will come out eventually.

    I believe that the idea behind the paper is to show that using an alternative search criteria to find something desired, for example, buried treasure, requires as much information to produce the search criteria as the blind search does. So we have a more efficient search but the information to enable this more efficient search has a price.

    To use Alan Fox’s example of buying a map in a bookstore would require information to prepare the map, build the bookstore, find a bookstore that has the map, and then find the right map in the store.

  69. Mr MacNeill,

    While you are pursuing interesting questions on meaningful information, I think you should not be so quick to identify that term with active information as used in the book chapter presented here.
    For one thing, the actual definition of information and how to measure it is a “princess in another castle” – note 48 refers to forthcoming articles.

    The authors do attempt a simpler exposition of what the calculation of active information is at that point in the paper. I must say that I find it confusing.

    In the WEASEL, for instance, Dawkins starts with a blind search whose probability of success in one query is roughly 1 in 1040. This is p. He then implements an alternative search (his evolutionary algorithm) whose probability of success in a few dozen queries is close to 1. This is q.

    And then proceed to calculate log(q/p). But this is comparing apples and oranges. In the above, p is expected success of a single query, while q is the expected success of a set of queries.

    I can think of two ways to fix this situation. In the first, we compare expected number of trials. ‘q’ is several dozen, p is half the size of the search space. But in this formulation log(q/p) will be the inverse of the previous formula.
    The second option is to compare the probability of success of a single query. Not a problem for p – all the queries have the same probability of success. But for q, what do we do? The queries of the first generation of the evolutionary algorithm are also random, they have the same probability of success as the first algorithm. The queries of the last generation have a far higher expected probability of success. You can’t just take the average, that has other problems.

    I am really at a loss to understand why the authors perform a calculation using p and q when the definitions of each are so different. Perhaps the forthcoming articles will clarify this.

  70. I think that q represents the probability of finding the specific entity given the new search methods. For example, q is much higher than p once we know that the treasure is on Bora Bora.

    By the way Bora Bora is in my opinion the prettiest island in the world. I have certainly not searched for them all but a more efficient search was driven by articles on it, pictures of it, stories about it, actually visiting it, photographs I took of it and then comparing it with information on other islands etc. All of which have an information cost.

    And also my personal experiences as to what makes something pretty which may be analogous to whether the protein is functional or not. Or maybe it isn’t analogous. But that hasn’t nothing to do with the current paper.

  71. 71

    Alan:

    Far from stopping science, ID would then be a search for a context that makes information meaningful. Thus we could see someplace where there is information, defined as the exclusion of real possibilities, and then search for a context within which that information is meaningful. For instance, it is puzzling why the vertebrate eye is backwards when it could have been forwards. That would then inspire a search for a context which would make that information meaningful.

  72. Nakashima #69: You’re right about the apples and oranges, although the problem is easily fixed by normalizing both searches to a single-query search on an m-fold Cartesian product space where m is the maximal query count. In any case, the active information associated associated with single-hill fitness functions remains unchanged, so given that p increases by going to the Cartesian product means that the information problem only gets worse.

  73. Is the ONLY counter-”argument” to act obtuse over the words “active information”?

    You guys must be really proud of yourselves.

  74. 74

    Sorry I meant “Allen” not “Alan”.

    The law of conservation of energy states that energy can neither be created nor destroyed, however it can be changed into an unusable form, which is detailed by the second law. These laws state that no work can be done without increasing entropy, that is losing energy to an unusable form. Perhaps the law of conservation of information will turn out to work the same way. For instance, Alan Fox’s example of transferring the human insulin gene to a bacterium would cause no loss in information, but it would cause a loss of meaningful information.

  75. joseph:

    Your comment #73 added no content whatsoever to this discussion, but rather was the equivalent of s**ting in the punchbowl. Every comment here, including both the positive and negative ones, has included substantive content except yours, which simply functioned as a rock tossed through the window.

    Whatever the moderators decide to do about your seemingly cumpulsive tendency to make ad hominem comments, it is clear to me that further interaction with you is pointless and counterproductive. From now on, regardless of whether they actually appear here, your comments are “invisible”.

    Talk about “meaningless” information…

  76. tragicmishap in #74:

    You make an interesting point. However, I’m not certain that there is necessarily a loss of meaningful information when the gene for human insulin is transplanted into a bacterium. If one defines the “meaning” of the transferred sequence as the protein for which it codes, this “meaning” has not been lost as the result of the transfer, as the protein is still made. However, if one defines the “meaning” of the transferred sequence as the biological function which the protein mediates, then indeed that “meaning” has been lost, because that function does not occur in the bacterium.

  77. Once again, the “meaning” of the transferred sequence depends entirely upon the context within which it is expressed.

  78. Dr. Dembski,

    Thank you for such a quick response! I’m afraid I have to ask you to unpack it a little for me. Can you spell out the new definitions of p and q? Especially given that the queries resulting from the evolutionary algorithm are contingent on previous queries, I’m not sure how to normalize them into a single query. Perhaps I have misunderstood.

  79. Nakashima: the new p for the Cartesian product becomes 1-(1-p)^m (for m = 1 this is just p), the latter p being the old p.

  80. Dr Dembski,

    Hmmm.

    m = 10^40 (maximal number of queries)
    old p = 10^-40
    1- old p = close to 1, but still less than 1
    (1-old p)^m = close to 0
    1-(1-old p)^m = close to 1 = new p

    Then q/new p is going to close to 1 also. Very different result than the previous calculation. Unless there is a new q also?

    Getting on a plane now, will look at the thread again in about 8 hours. Thanks again for your reply.

  81. 81

    Ha! I just now got to the section of the article titled “Entropy” (pg. 26).

    Dr. Dembski, are you suggesting that constant influx of information provided by intelligence can counteract the effects of the second law? Or perhaps not directly counteract but blunt those effects?

    “It seems,
    then, that information as characterized by the Law of Conservation of Information may be
    regarded as inverse to entropy: increased information indicates an increased capacity for
    conducting successful search whereas increased entropy indicates a decreased capacity for doing
    the work necessary to conduct a search.”

    Analogy would be a computer that is slowly losing flops with an intelligent programmer constantly increasing the information content of the search algorithms to compensate.

  82. 82

    Allen, I think maybe we could come up with a general context for biological meaning that is independent of specific organisms. It would involve protein shape-space and would be theoretical binding rather than context dependent in the manner you suggest. Hopefully once we can accurately model structure and binding for proteins and other biomolecules then we could also predict any function it might have from its structure. This would include protein-protein binding. We can already do this to a limited extent.

    I get these references from Behe’s Edge of Evolution and refer to research on the binding profiles of antibodies generated by the immune system.

    Perelson, A. S., and Oster, G. F. 1979. J.Theor.Biol. 81:645-70

    Segel, L. A., and Perelson, A. S. 1989. Immunol. Lett. 22:91-99

    De Boer, R. J., and Perelson, A. S. 1993. Proc.Biol.Sci. 252:171-75

    Smith, D. J., Forrest, S., Hightower, R. R., and Perelson, A. S. 1997. J. Theor.Biol. 189:141-50.

  83. Allen,

    You are correct- my apologies.

    Now to the point-

    Reducibility- as in can biological information be reduced to matter, energy, chance and necessity?

    IOW all YOU have to do to refute Dembski/ Marks, and ALL of ID is to demonstrate such reducibility.

    However given the paper by Lincoln and Joyce on sustained RNA replication, the reducibility argument is in serious trouble.

    It would also help your position if you actually knew what it is that makes organisms what they are.

    Where is that information? And can it be altered in such a way to account for the diversity of living organisms?

    For example with eyes PAX6 can be transferred from mice to fruit-flies but the fruitflies develop fruitfly eyes.

    And even though we know a great deal more about eyes/ vison systems that Darwin did, the “evidence” for their evolution is still the same.

    Doesn’t that make you wonder, even just a little, that your position isn’t up to the task?

  84. Nakashima: The relevant m is the number of steps it takes Dawkins’s algorithm to converge with high probability on the target (i.e., METHINKS*IT*IS*LIKE*A*WEASEL). That m is less than 100.

  85. Dr. Dembski,

    As far as I can see, nowhere in the paper do you and Dr. Marks express skepticism regarding the ability of Darwinian evolution to account for the diversity of life. Rather, you seem to grant that Darwin was right that random mutation and natural selection are sufficiently powerful, provided that fitness has a teleological origin.

    That seems like a huge departure from your former indictment of Darwinian theory as flawed and unsupported by the evidence.

    Have you in fact shifted your position?

    How has the ID camp reacted to your paper?

  86. Beelzebub: This paper was written under the supposition that common descent holds and that natural selection is the principal mechanism behind it. Writing under a supposition does not mean accepting it. My own views of the truth of the matter are clearly spelled out in THE DESIGN OF LIFE (http://www.thedesignoflife.com). In particular, I think that irreducible complexity at the molecular level (especially in the origin of DNA and protein synthesis) provides compelling evidence for discontinuity in the history of life.

  87. Dr. Dembski,

    I understand what it is to assume something for the sake of argument, and I do it myself quite often. I’m just surprised that you didn’t state that the assumption was contrary to your own position, when a short disclaimer would have made this clear, e.g.:

    The authors remain skeptical that Darwinian evolution explains the full diversity of life on earth. However, we show in this paper that if it does, it is necessarily teleological.

  88. On another topic, commenter DiPietro at AtBC has pointed out an apparent circularity in your paper. I elaborate on this below.

    You write:

    Active information is to informational accounting what the balance sheet is to financial accounting. Just as the balance sheet keeps track of credits and debits, so active information keeps track of inputs and outputs of information, making sure that they receive their proper due.

    You define active information as log(q/p), where p and q are the probabilities of success of the null search and the alternate search, respectively.

    But if the active information of the “output” is defined as log(q/p) and the active information of the “input” is defined as log(q/p), where p and q refer to the probabilities of success of the same null search and alternate search, respectively, then p is the same for input and output, and so is q.

    The LCI then reduces to this tautology:

    log(q/p) ? log(q/p)

    This seems like a fatal problem.

    Later in the paper you seem to offer a way out in your discussion of plans for experimental verification of the LCI:

    Tracking and measuring active information to verify intelligent design is readily achieved experimentally. Consider, for instance, that whenever origin-of-life researchers use chemicals from a chemical supply house, they take for granted information-intensive processes that isolate and purify chemicals. These processes typically have no analogue in realistic prebiotic conditions. Moreover, the amount of information these processes (implemented by smart chemists) impart to the chemicals can be calculated. This is especially true for polymers, whose sequential arrangement of certain molecular bases parallels the coded information that is the focus of Shannon’s theory of communication.

    The problem is, you seem to be equivocating on the word “information”. The LCI applies to active information, but here you are referring to the Shannon information of polymers. Nowhere in the paper do you show that active information and Shannon information are equivalent or commensurable.

    If you try to fix this by calculating the active information of the polymers, rather than the Shannon information, then you run into the fatal tautology problem elucidated above.

    Could you comment?

  89. In the comment above, the “less than or equal sign” got eaten by WordPress despite showing up correctly in the preview window.

    The tautologous statement should read as follows:

    log(q/p) <= log(q/p)

  90. Beelzebub: The three conservation of information theorems proven in the paper are elementary but they are not trivial, as you suggest — that should be evident from the fact that the third of these theorems proves and then extends the standard no free lunch theorem. As it is, the active information input occurs in a higher-order search space, the active information output occurs in the original search space.

  91. Writing under a supposition does not mean accepting it.

    Fair enough to take an opposing point of view but it didn’t become clear.
    Arguments at UD appeared often arbitrary and sometimes contradicting during recent months (e.g. the Hitler-Darwin discussion, dispending and reinstating the EF). I doubt that it is currently possible for readers who come here the first time to understand what ID is about. I am afraid that in its current state UD isn’t

    Serving the Intelligent Design Community

  92. 92

    Bill you wrote,

    “the Law of Conservation of Information shows that Darwinian evolution is inherently teleological.”

    To nit pick though- this is actually a contradictory statement. What is meant here is that Evolutionary theory must be inherently teleological and hence a neo-Darwinian viewpoint is fundamentally flawed in respect to the law of conservation of information.

    In other words you are posting and pitting information against chance as the mechanical explanation for the origin of novelty. No?

  93. 93

    And Bill if I could invoke here a little Kant with his synthetic vs analytic modes of judgment and reasoning and how it can be used to shed light on the relevance of the two mechanisms.

    First I note that the arugment for chance mutation is inherently weak because it synthetically applies arrangement to already existent systems. In other words it applies a synthetic explanation of chance to the variables within a model. Now take the position of information as the primary explnaaiton for novelty- information merely takes what is already there then critcally analyzes it – and puts the data into a comprehendible model.

    So the invocation of chance as a mechanism is inherently synthetic- the God of their dreams – hence my favorite slang for the new atheists, “chance worshipers.”

    Now albeit the law of conservation of information is also synthetic- and hence the inference to necessary influx of novelty from “elsewhere” is also synthetic- but the fundamental back bone of the theory rests on pure empirical experience. That is the synthetic argument of the conservation of information is rooted in mathematical rationalization- hence it is valid and sound as 2+2= 4. Not perfect but for all practical purposes strongly cogent.

    So the bottom line is that the Darwinian Evo model is an obviously inherently atheistically (or to be scientific and philosophical opposed to theological) “anti-teleology” driven “synthetic” explanation for life’s origin.

    So my point is that the DE model is not about the data but the interpretation of it. We are weighing here informational based construction vs random chance based construction.

    My argument is that information is much closer to an analytic judgment than chance is. I feel like appealing to Locke here- as if to say there is something “more real” and “measurable” about the conception of chance than there is of information.

    I conclude that the ID model of informatics is apparently much more scientifically sound than that of the DE model.

  94. 94

    {Correction above} ^

    …something more real and measurable about “information” than there is about “chance.”

  95. Would Dr Demski have time to comment on whether his concept of “active information” has any equivalence to “meaningful information” as discussed upthread? Does he agree about the difficulty of assigning information quantitatively to, for example, DNA sequences without a priori knowledge of their potential functionality?

  96. Previous comment error. S/B

    …assigning information content to…

  97. Dr. Dembski,

    Since you didn’t respond to it, are you ceding my point about the incommensurability of active information and Shannon information?

    Moving on, you argue that the LCI is not tautologous because information input occurs in a higher-order search space than information output, so that p and q are different for the input and output when calculating the active information.

    I thought I would try to apply this idea to the two scenarios you mention in your paper: an OOL experiment, and evolution itself. In doing so, I ran into trouble right away.

    Fitness landscapes are what distinguish Darwinian evolution from a blind search. To apply the concept of active information to evolution, then, we must compare the probability of finding a target using blind search to the probability of finding it under a specified fitness regime.

    One problem with this is that Darwinian evolution does not specify targets in advance. In fact, the target space is determined by the fitness regime. If you change the fitness regime, the target space changes.

    Second, the fitness regime is determined by the physical environment(s) in which evolution takes place. For a complete accounting of the active information in the environment, we need to account for the active information in the search that “found” the environment. How can this be done without knowing the search space of all possible universes, plus the search algorithm that was used to “find” ours, plus the size of the target space within the space of all possible universes?

    Third, the active information of a Darwinian “search” is not independent of the target. A particular fitness regime therefore contains very little active information with respect to some targets, and a huge amount with respect to others.

    Turning to the OOL example, you and Dr. Marks write that “Tracking and measuring active information to verify intelligent design is readily achieved experimentally,” and you propose the idea of measuring the active information of the chemicals used in OOL experiments and comparing it to the active information of the target molecule(s).

    In trying to apply the idea of active information here, I ran into more problems: what does a “blind search” for purified chemicals look like? What is its probability of success? Do I also need to calculate the active information of the glass beakers used to carry out the experiment? What, again, about the active information of the universe in which the experiment takes place? What does a blind search through the space of all possible universes look like?

    I’m skeptical that all of this is “readily achieved experimentally.”

  98. A final point for the night:
    You and Dr. Marks state that “searches, in successfully locating a target, cannot give out more information than they take in.” Yet blind searches can successfully locate a target without any information input at all. They can even do this rather quickly when the search space is small.

    If so, is the LCI really a law?

  99. I have loads of questions about this paper but here is a basic one. The paper is based on the idea of a target and a function which tries to find that target. But in the case of biology the “target” and the function are the same thing. Put simply: the function is “will survive” and the target is “will survive”.

    I will try to demonstrate this with an analogy. Imagine a different “search space”. A steep but uneven slope. A small boulder is dropped at a random location on the slope (all positions are equally likely). But of the course the boulder will bounce and tend to fall downwards. You could treat this as the boulder exploring a search space for a target. The (continuous) search space is the entire area of the slope. The target is the bottom of the slope. This area is tiny compared to the area of the slope. The proportion is p. However, there is a rather high probability that the boulder will end up at the bottom of the slope (not quite 1, it might become lodged again) because of the laws of gravity. Call this probability that the boulder will reach the bottom q.

    Does it make sense to talk of a separate fitness function – gravitational attraction – which is used to search for lower positions? Does the LCI imply that the law of gravity has an information content of log(q/p)?

    Compare this to the situation where you are looking for a particular buried treasure somewhere in the slope. You are then given the information that with probability q the treasure is at the bottom of the slope. Now it makes sense to differentiate the search function and the target and to talk about the additional information that you have been given.

    I submit that the case of evolution is much closer to the first than the second case.

  100. Pardon a footnote:

    On accounting vs wealth creation — hopefully relevant.

    In an accounting system, sums of money move around between accounts in such a way that ASSETS = LIABILITIES (where the latter incorporates owner equity).

    That is always true, whether we are dealing with a wealth making or a wealth destroying enterprise; even a Ponzi scheme. (The difference between the two is a matter of creative function in a given environment — as a rule, a major issue of highly intelligent design. [Translating: it is possible to make a lot of money by blind luck, but that is highly improbable.]

    As a rule, wealth creating enterprises inject very intelligent organisation, which gives a context in which there is a build up of new customer accounts such that on a sustained basis, overall sales less costs of sales and expenses gives rise to a healthy profit, which when accumulated is money-denominated wealth. So, the accounting equation is not violated, but associated with the algebra [and T- accounts and balance sheets and income statements etc are all in effect applications of algebra, with various conventions and generally accepted principles of praxis] is a real world process that is the root of growing wealth, a process that is as a rule highly designed. (On Reaganomics, Thatchernomics etc, I will simply say that from the time when the major Western nations decided to take the brunt of the recession at the turn of the 1980′s to break the stagflation spiral of the 1970′s [complete with the infamous Phillips curve gone mad as workers built in more and more inflation expectations into wage demands], the world has moved to a much lower inflation, sustained economic growth regime. Just as, until those two worthies came along, it was thought that the USSR was a more or less permanent destabilising factor in the world, one armed with a few dozens of thousands of nukes, north of 40 – 50,000 tanks (many aimed at the Fulda Gap) and global ambitions backed up by decades of geostrategic power plays, which at the time were plainly winning the global balance of power. In my native land, that global contest came down to an unofficial civil war And, once the oil crises of the 1970′s hit, economic trend lines became a lot less predictable; indeed trend line based “forecasting” lost its lustre, not only in economics but in practical management. [So, I think a fairer, more balanced reading of the 1980's than has been suggested above is indicated.])

    Similarly [but not equivalently], the Dembski-Marks paper is pointing out that while in principle functional information can materialise out of lucky noise, the odds are very much against it. And, taking the 43 generation search now notorious Weasel 86 program example as a case in point [whether partitioned search form, which is a valid interpretation, or implicitly latched form makes no practical difference . . . ] we see that odds of 1 in 10^40 or so of hitting the Weasel sentence on one guess fall to near certainty of hitting it in say 100 generations on a suitably latched ratcheting, cumulative search. That is, someone has hit on a way to make what is credibly practically infeasible to something that is a practical proposition.

    How?

    ANS: By injecting a well-tuned search.

    But, surprise, that search has in it a lot of information on the target and how to get there.

    So, we have now opened a second information account: the search account. And, it turns out that while we can in principle get to a good search by chance, the odds, in general, are even longer than that of hitting the original target in one guess.

    So, we now have a search for a search. Which of course can go off on an infinite regress . . . or else truncates somewhere.

    Where?

    ANS 2: In general, by observation, we see that the successful search for a well-tuned search (at whatever level of observable regress ultimately applies) that involves functionally specific and complex information [recall 500 - 1,000 bits is a reasonable universal threshold for complexity . . . the cosmos as a search engine would most likely be stumped to find islands of function if they require at least that much information . . . ] is carried out by intelligent designers. (So, the new information has come from a fresh account, not out of the magic of lucky noise. There is no free lunch here.)

    But, what are (a) information, (b) functionally specific complex information [FSCI], (c) function, & (d) intelligence?

    a –> Information, per the UD glossary [courtesy materialism-leaning Wikipedia cited as admission against presumed interest]: “ . . that which would be communicated by a message if it were sent from a sender to a receiver capable of understanding the message . . . . In terms of data, it can be defined as a collection of facts [i.e. as represented or sensed in some format] from which conclusions may be drawn [and on which decisions and actions may be taken].” (I think we need to insist that people reckon with the WACs and glossary, instead of recycling long since cogently answered objections and infinite regresses of demands for definitions etc. Also, observe, the just cited definition implicitly terminates on iconic exemplars, and builds in the implication that if something looks sufficiently like that, it is information. [Which brings us to the value and legitimacy of reasoning by key examples and sufficiently similar cases, i.e of analogy. To blanket reject analogous reasoning is to fall into self-referentially inconsistent, self-refuting, selectively hyperskeptical absurdity. For, we form many key concepts -- learn -- by analogous reasoning.])

    b –> FSCI, per same glossary: “complex functional entities that are based on specific target-zone configurations and operations of multiple parts with large configuration spaces equivalent to at least 500 – 1,000 bits; i.e. well beyond the Dembski-type universal probability bound.” Onward, TBO in the 1984 TMLO — the first technical level ID book — summarise the conclusions of OOL researchers by the early 1980′s: “. . . “order” is a statistical concept referring to regularity such as could might characterize a series of digits in a number, or the ions of an inorganic crystal. On the other hand, “organization” refers to physical systems and the specific set of spatio-temporal and functional relationships among their parts. Yockey and Wickens note that informational macromolecules have a low degree of order but a high degree of specified complexity.” [TMLO (FTE, 1984), Ch 8, p. 130.]

    c –> Function: Here, as TBO summarise, we discuss systems that transform inputs to yield outputs and/or outcomes. In the relevant context, we therefore have parts that are integrated in accord with a pattern across space and time [spatio-temporal . . . relationships"] — an architecture, which must fulfill a certain criterion: it must “work” in some context ["functional relationships"]. So, it must be organised in such a way as to foster the transformation of inputs into outputs, yielding advantageous outcomes. Many examples are observed in the world of technology, and also in biology. In cases of multipart irreducible complexity (for a core, the removal of any one part destroys function) and of complex specification (the arrangements of parts store large quantities of information) of known origin, such entities are designed, and indeed, it is hard to see how such function can reasonably — per search space challenges — come about apart from intelligent input. [Notice the termination of the chain of definition on a study of empirical examples and an inductive generalisation therefrom.)

    d --> intelligence per the same glossary, and courtesy Wiki again: “capacities to reason, to plan, to solve problems, to think abstractly, to comprehend ideas, to use language, and to learn.” (Note, tis is in the end defined relative to observed cases and generalisations therefrom. if a new case is sufficiently close on family resemblance criteria, it will be accepted as a new case of intelligence.]

    So, we see that as a practical matter, and per probability, a first level search space challenge can be solved by using intelligence to cut down the effective search space. To do so requires creation of a well-tuned search algorithm, which is a higher order instance of functionally specific complex information. So, the search space problem has not gone away, just moved up a level. We either face an infinite regress, or an appeal to increasingly improbable lucky noise, or else we can simply revert to what we observe: such chains of higher order searches tend to terminate in the work of an observed intelligence, with high reliability on empirical investigation. [E.g. None of eh evolutionary computing algorithms are credibly listed as originating in lucky noise.]

    And, we may define the concept of active information as the information gap between what is achievable on a random one-step search algorithm, and what is achievable on a well-tuned search. The increment in information that makes the search practicable in the real world is the injected active information, which can be quantified, per accounting principles [noting that we are dismissing lucky noise as empirically incredible].

    So, it turns out that such active information is as a general rule traceable to intelligence; lucky noise not being a credible source once we are past a threshold of relevant complexity. [Evolutionary materialists: you can knock down the above by simply providing a good counter-example . . . ]

    All of which brings us back tot he significance of identified signs of intelligence,and that active information joins the list of such signs — indeed, it brings out the force of CSI/FSCI and IC, as well as linguistic textual information and algorithmically functional information.

    GEM of TKI

    PS: Odds of 1 in 10^40 come down to 1 of 2^133 choices. So, if we have a functionally anchored, uniquely functional state requiring 133 bits to store, it is not a feasible search on the span of the earth’s resources. Using the same squaring trick as was used with the UPB, we see that 266 bits sets up a second order search of 10^80 cells, so it is very unlikely that we would find something as “common” as 10^40 functional configs in that space. (That is, since there probably have not been that many proteins formed on earth since its origin, we have a credibly insurmountable search for proteins of length sufficient to absorb that many bits. That looks like about 62 AAs. OOL on earth is in trouble, and the origin of novel body plans — which require a lot of functionally integrated proteins to work — is in trouble.)

  101. PPS: Re MF: the key idea is that many, many configs of organic molecules are possible, but very few of these will perform as a viable biofunctioning cell [multicellular organisms being just that]. To do that, they have to be very specifically organised, and the parts have to be just as specifically composed [esp. proteins]. To be self-replicating, there has to be a blueprint that stores information on how to build the machines step by step, including how to replicate the blueprint and preserve it from rapid deterioration. And, such has to be both encapsulated and joined to physical implementation machines — in effect we have highlighted that a living cell has to have in it a computer guiding an automaton. All of this, so constructed as to work and compete in a given ecosystem or pre-biotic environment.

  102. Mark Frank,

    Good point- in biological evolution per the MET there isn’t any target beyond “survival”.

    There aren’t any bacterial flagella to search for.

    There aren’t any vision systems to search for.

    There is no way to know if flagella or vision systems are even obtainable given a starting population or populations that never had either.

    So there isn’t any search- nevermind a search for structures that don’t exist.

    Don’t you think this poses a bigger question mark for your position?

  103. beelzebub,

    Can a blind search find a target that doesn’t exist, even given a small space to search?

  104. Mark Frank and Joseph: I think this is where we fundamentally differ — on whether evolution has targets. You say no. Mark thinks that a river finding its lowest place or a rock tumbling down a hill and finding its lowest place is more representative of evolution. Bob and I, on the other hand, see evolutionary processes traversing configuration spaces and locating things like bacterial flagella. These satisfy independent functional specifications. These specifications are not least action principles like those in physics and of the sort that you are putting forward. They are specifications of mechanical systems that require engineering principles to characterize and understand. So how is it that evolutionary processes managed to locate them? Our answer in a nutshell: active information. Our paper develops that answer.

    Addendum: I need to add this to the previous remark, namely, when a rock finds a particular valley because it rolls down some hills, we might say the hills provide the information for it to do so. Fair enough. The hills, in our terminology, supply an alternative search. But an alternative search for what and in comparison to what probabilistic baseline for blind search? Presumably, the blind search that forms the contrast class to the alternative search is a random search over a flat landscape, any point of which is equiprobable for the rock to land. But what was it about that point that made it salient? In the case of life, independent functional requirements make the points of biological configuration space salient regardless whether the search is blind or directed. In the case of the rock, where it lands by null (blind) or alternative search is a matter of indifference — unless one adds an independent functional requirement. If, for instance, the rock landed precisely where treasure was buried, then the information from the hills landscape would be relevant to design (most likely, in this case, the treasure burier would have chosen the place to bury treasure on the basis of a least action principle). So, even though the concept of specification is not explicit in this paper, it is there implicitly

  105. Mark Frank — A small boulder is dropped at a random location on the slope . . .

    The active information content in that event would be an interesting thing to ponder.

    The probabilities of it reaching its rest shouldn’t be much different between chance and design.

    If you knew the mass of the boulder, the resistance of the slope, the force at which it was set in motion etc. you should be able to calculate exactly where it stops without regard for chance or design.

    So the probability as to where that boulder ends is 1, just as it were designed, so I guess that means there would be little active information in the event.

  106. 106

    Mr Joseph,

    There aren’t any vision systems to search for.

    Indeed, which has allowed many different vision systems to be discovered. There is no one, perfect vision system planned out for all creatures, nor even one system per niche. The distribution of vision systems makes sense given common descent and nested hierarchies, however.

  107. kairosfocus,

    Your comments nearly all the time contain very useful information. However, they are also nearly all the time extremely long and to use an old expression, “contain everything but the kitchen sink” and because of that become less useful because many will not read through them or get lost on dealing with peripheral issues.

    I would use part of your name here, namely the focus part, and make your comments shorter and to the point. Over time all your points will get made several times but it is difficult to pick out the most relevant at the moment. For example, somewhere deep in your post you discuss active information and that seems to be a matter of misunderstanding here. The post would be much more effective if limited to this one concept. The other items could come out piece by piece if necessary when something get challenged.

    You are a valuable resource for us here but the tendency to quote the entire encyclopedia each time makes you less effective. I personally would be interested in a thorough discussion of active information and LCI and just what these means in lay terms. Otherwise the thread will go off on any merry way one tries to steer it.

    Your discussion of active information immediately did away with any confusion of it with meaningful information which was a potential hijack of the thread.

  108. Nested hierarchies and common descent are not topics of this thread.

  109. @ Jerry,

    I hope Mr. M. takes your (sound) advice. It may be better received coming from an ID proponent. It could save us all developing scroll finger syndrome. :)

    You say:

    Your discussion of active information immediately did away with any confusion of it with meaningful information which was a potential hijack of the thread.

    I have to say this is not true, for me, at least. The introduction of “meaningful information” was an attempt to clarify what is meant by “active information” and how it might be quantified. Perhaps Dr. Dembski will weigh in with a clarification of what he and Dr. Marks mean by the term.

  110. Dembski and Marks did define “active information” but in terms of symbols in an equation. “Meaningful information” was used by Allen MacNeill several times in one comment but only once in the paper so I don’t think this is what active information is. Active information was used about 50 times in the paper.

    I am sure active information is meaningful but is all meaningful information active information. Probably in some context but not in others so it may not apply in Dembski and Marks formulas. The active information seems to be the additional information that is added to the search to make it more efficient. For example, the map where the treasure is. And my guess is that the amount of this information plus the new search information equals or exceeds the information required for the blind search.

    If that is not it, then I would appreciate someone providing input on it.

  111. Here is a very brief layman’s explanation of the paper in one page but it does not answer all the questions and probably opens a lot more.

    http://www.salvomag.com/new/ar.....embski.php

  112. Mr Jerry,

    Thanks for helping keep us focused!

  113. Dr. Dembski,

    I hope you’ll find time to respond to my previous two comments, as they raise serious questions about the applicability of the LCI.

    Link 1
    Link 2

    Also, I hope you’ll answer my earlier question about why you chose to name it the “Law of Conservation of Information” if information is not conserved:

    And even more fundamentally, why is it called the “Law of Conservation of Information” when it doesn’t rule out the loss of information?

    Conservation laws are about quantities that remain constant.

  114. Nakashima,

    “Indeed, which has allowed many different vision systems to be discovered. There is no one, perfect vision system planned out for all creatures, nor even one system per niche. The distribution of vision systems makes sense given common descent and nested hierarchies, however.”

    I want to understand your position at it is related to my response in the Sternberg post to Allen.

    Does the statement above render moot the neo-Darwinian evolutionist explanation for advantageous selection of survival?

    I ask this because in many evolutionist papers I’ve read, the reason given for “evolved” components is for more “optimum” survival.

    From Wiki, on Survival of the Fittest(history and evolving meanings):

    “What Darwin meant was “better adapted for immediate, local environment”, tracking changing environments by differential preservation of organisms better adapted to live in them. The theory is not tautological as it contains an independent criterion of fitness.[6]”

    “In the introduction he(Darwin’s 5th publication of On Origins of Species) gave full credit to Spencer, writing “I have called this principle, by which each slight variation, if useful, is preserved, by the term Natural Selection, in order to mark its relation to man’s power of selection. But the expression often used by Mr. Herbert Spencer of the Survival of the Fittest is more accurate, and is sometimes equally convenient.”[11] By the word “fittest” Darwin meant “better adapted for immediate, local environment”, not the common modern meaning of “in the best physical shape”.[6]”

    What is the definition of “local environment?” Is it not “niche?”

    Can this lead to a statement that evolutionist do believe a vision system can be “better adapted for immediate local environment” or “niche.” Thus allowing for “preservation of organisms.” It does not insure success of the organism outside its “local environment,” but it is qualified as “better adapted” for survival.

    In that case a vision system can be argued to be “better adapted” for survial in local niche.

    Does this then contradict inverted fish eyes? Or maybe I should ask, do inverted fish eyes contradict Darwin’s principle? If so, was Darwin wrong?

    I ask this because I see this principle utilized in evolutionist papers all the time.

    Is there criteria or definitions that I misunderstand?

    thanks

  115. DATCG asks:

    Does this then contradict inverted fish eyes? Or maybe I should ask, do inverted fish eyes contradict Darwin’s principle? If so, was Darwin wrong?

    DATCG,

    Evolutionary theory says that if a better variation arises, it will tend to be preserved by natural selection. It does not say that all possible better variations will arise. An inverted eye is unlikely to “uninvert” itself, because it would require too many simultaneous changes to the genome to achieve an uninverted eye that performs better than its predecessor.

  116. 116

    Mr DATCG,

    At the risk of trying Mr Jerry’s patience –

    My statement did not render moot any neo-Darwinian hypothesis. It was predicated on such a hypothesis.

    Success in an ecological niche relates to function. In this example, many niches reward entities with better light gathering and organizing systems. The critical word is “better”. Comparative, not superlative. It doesn’t have to be the best, just good enough to fulfill the function. Spending any more time and energy than good enough is a waste. Example: Hawks fly around in the open air, are often thousands of feet away from their prey. Their niche requires excellent vision. Jungle dwelling apes rarely have to see further than ten feet away, because the jungle is incredibly dense. Their vision is tuned to objects 10 feet away or less. Building a hawk’s vision into an ape is a waste of energy during development. That’s why blind cave fish lose their eyes, to spend the energy somewhere else.

    But each function, such as sight, can have multiple implementations. A niche such as the ocean (big niche, I know) hosts cephalopods, vertebrates, and arthropods, each with a different vision system. Each is good enough, none is perfect.

    To bring this back around to the thread topic, Drs. Dembski and Marks are not saying in their paper that each kind of eye is telicly pre-ordained. I think that by claiming that by claiming some teleology is necessary, they are saying that if life is inevitable and/or diverse in certain circumstances, it is because of those circumstances, not because of the blind process of reproduction, variation and selection.

  117. Beezebub,

    I’m aware of what theory states today. But, doesn’t your response assume there is a verted eye that is better than inverted? How do we know? And isn’t each situation an environmentally dependent niche for survival by better adaptations.

    Some argue scientist can design better eyes, but until an eye is actually engineered, we may not discover why inverted or verted works better for adaptation in each niche.

    But I’m taking this post off topic as I now see in above comments.

    Unless maybe active information serves as to limit the blind search guess of inverted versus verted eyes for better adaption in each environmental niche. But I’m not sure if I understand that is the application of Dembski/Mark’s paper.

  118. Nakashima,

    Thanks, trying to get back on topic. In my response to Beelzebub, maybe it relates to our discussion as well.

    Beelzebub,

    “Evolutionary theory says that if a better variation arises, it will tend to be preserved by natural selection.”

    A process that tends to preserve better variable information is one that conserves information, no?

    In that case, if I’m understanding correct, though vision systems are not all optimal, each is preserved as optimal for the niche at that specific time. And time can be relative – long or short – dependent upon the amount of active information and fundamentals of external stimuli at that specific moment in time. Otherwise, the species is superceeded by another, yet with core information that allows it to build upon previous information.

    I guess the question then is do we find core information present in eyes that are common among all vision systems? Whether verted or inverted? Whether environmental changes or not?

    In the example of cave fish that Nakashima and I discuss, there is not actual loss of information, only a switch that turns off other cascading events. Yet can immediately be turned back on within one generation.

    Now if loss of information is true for the fish(i.e. all genetic information is lost) then that is a defintion of entropy, correct? But that also represents a potential loss of energy and cost as well, correct?

    Thus conservatoin of information like conservation of energy can be/are linked together?

    What am I not understanding?

  119. More questions related to Conservation of Information.

    Is Conservation of Information proportional to amount of work done or cost to maintain the information?

    So less work = less energy cost, information deteriorates as entropy takes over.

    So the Conservation of Information violates the 2nd law of thermodynamics? At least for specific time? In our case, 13.7 billion years since the theory of Big Bang?

    So, active information was funneled in from the beginning that guides the process of increased complexity of specific functional information that violates 2nd law?

    Is that was it postulated in the paper?

  120. 120

    DATCG,

    You’re veering way off course here. The LCI is not about the conservation of genetic information. Also, the LCI isn’t even a conservation law, as I pointed out in an earlier comment.

    I would urge you to read the paper first and then come back to this thread if you still have questions.

  121. I believe that there is a serious misunderstanding here between the medium in which information is encoded and the meaning of that information. From my own analysis of Dembski and Marks’ paper, it seems clear to me that their treatment applies only to the former, whereas the real issue in biological evolution (indeed, all forms of evolution) is the latter. Even if it is the case that the individual components of a “concatenation” of information may be “conserved”, despite alterations in the order of those components (e.g. the nucleotides in a strand of nucleic acid or the primary structure of the amino acids in a polypeptide), the “meaning” of the information so encoded is not an intrinsic characteristic of either the “bits” in the message or their sequence. Rather, the “meaning” of such a “string” is a function of its relationship to the objects and processes “for which it stands”, and this relationship depends completely on the context within which the information has been encoded, transmitted, decoded, and its meaning made manifest in the structures and functions of the organism.

    I am currently working on a more formal treatment of the idea of “meaningful information” and its relationship to teleology (inspired at least in part by Dr. Dembski and Dr. Marks’ paper) and will be posting comments on my progress on this treatment at my blog:

    http://evolutionlist.blogspot.com

  122. Another thing to consider vis-a-vis Dembski and Marks’ analysis:

    On page 34, paragraph 5 of their paper, the authors present the following analogy:

    “Just as information needs to be imparted to a golf ball to land it in a hole, so information needs to be imparted to chemicals to render them useful in origin-of-life research.”

    A similar analogy was made by Mark Frank in comment #99, in which he described what happens when a ball rolls down a “rugged” hill. This analogy is very similar to the one made by C. H. Waddington in his analysis of “epigenetic trajectories” in biological development. Waddington’s model is directly applicable to Dembski and Marks’ analysis because it demonstrates how genetic (i.e. “internal”) and epigenetic (i.e.”external”) information can dramatically alter the developmental and evolutionary trajectories of developing and evolving entities.

    Consider the ball rolling down a hill. If the hill is perfectly smooth, then the only prediction one can make with any confidence is that the ball will wind up at the bottom of the hill. However, if the hill has grooves in it (and especially if those grooves are shaped like “watersheds”; that is, they branch and get deeper the further down the slope the ball rolls), then very slight changes in the initial trajectory, along with very slight deviations in trajectory as the balls rolls downhill, can have extremely significant effects on the ball’s final location.

    In Dembski and Marks’ model, it appears to me that the grooves in the hillside are what they are referring to as “active information”. That is, information that is external to the developing/evolving (i.e. changing) organism and/or population that guides it to a specific end point. Such a process is indeed teleological, but it is important to realize that in Waddington’s model, the movement of the ball itself has the effect of defining and altering the grooves in the hillside. That is, as organisms develop and evolve, they alter their environment in ways that then alter their further development and evolution. This continuous feedback process means that the totality information that produces the actual trajectory of the ball is located both in the ball itself and in its environment.

    Therefore, for Dembski and Marks’ model of “active information” to be truly representative of biological reality, it must somehow incorporate not just the “onboard” information in the organism (or DNA molecule, or whatever), but also all of the information in every component of that entity’s environment.

    Furthermore, one of the most salient characteristics of biological information is that it is, indeed, highly “meaningful”. That is, the actual function of a DNA sequence is not an intrinsic characteristic of that sequence alone. On the contrary, it depends completely on what that sequence “means” from a biological standpoint.

    Consider a non-coding sequence of nucleotides in DNA. If a retrovirus transposes an active promoter sequence “upstream” from that non-coding sequence, it seizes being a non-coding sequence. If the mRNA for which it now codes is translated into a protein, it is unlikely that this protein will have a function.

    Unlikely, but definitely not impossible. This is because many non-coding sequences in eukaryotic DNA are either pseudogenes or degenerate transposons, retroviral cDNA sequences, or repressed regulatory genes (such as those coding for cellular growth factors). If the latter are activated, the result is cancer, a very “meaningful” biological consequence of a seemingly “meaningless” retrotransposition event.

    The point I am trying to make here is that there may, indeed, be something to the idea that information defined as discrete “bits” may be conserved (i.e. rearranging their sequence will have no effect on the proportion of each bit present in the aggregate). However, this has no necessary connection with the biological “meaning” of the various arrangements of the “bits. But, as any molecular geneticist can attest, it is the sequences of “bits” and their “meaning” (i.e. their relationship to the biological objects and processes for which they code) that makes all the difference.

  123. Something else to ponder with ball on the hill exercise:

    Let’s say you standing there waiting for the ball to come rolling to a rest behind you so you can use it for a seat.

    If the landing area is small enough and the ball is funneled to it, there is a reasonable expectation you could do this.

    Now let’s say the landing area becomes much larger and the balls can come down from many different angles. The likelihood of you getting a seat becomes less.

    Now, let’s say you pick a spot to wait where the force of gravity can’t carry the ball to you but you have the expectation some unknown power will kick in to get you your seat.

    In that case, my friend, you must be an evolutionist :-)

  124. Re #104

    “So, even though the concept of specification is not explicit in this paper, it is there implicitly”

    Yes. Your definition of information is crucially dependent on the concept of specification. And I think you need to be more explicit about how this relates to targets.

    You define the information content of an outcome as -log(p) where p is the probability of an outcome meeting a target. But any event can meet an infinite number of different targets. For example, a bacterial flagellum might meet the target of:

    * not being fatal to the bacterium

    * enhancing the bacterium’s fitness

    * adding to the bacterium’s mobility

    But also

    * being less than 1 micron long

    Each of these targets has different probabilities. Therefore the information content of an outcome is relative to the chosen target. If LCI is to be a law then there must be an objective way of deciding what it is the salient target to use.

    I thought you addressed this through your concept of specification. As far as I am aware your latest definition of specification is this one. This defines specification in terms of – fits a pattern that can be described simply. But in that case the boulder example should also count as a meeting a target. “Bottom of the valley” is a pattern among all possible locations that can be described very simply.

    But then a very simple law – things move downwards – causes boulders to meet a target which would otherwise have a low probability. There are very few locations at the bottom of valleys compared to all locations. So if I have correctly described the relationship between target and specification then it appears the LCI is wrong.

  125. 125

    Bill Dembski,

    I’ve read your chapter closely. (Clever segue from Marx’s lead soldiers to Dawkins’ characters, you tireless cultural warrior. But I’ll get to the essential pun in my next comment.)

    You give a lucid, honest, and interesting presentation of your philosophical perspective in the first two sections. I’ve long felt that you should say outright that intelligence creates information.

    [Go here for more excerpts of T. M. English and G. W. Greenwood, "Intelligent Design and Evolutionary Computation," Chapter 1 of Design by Evolution: Advances in Evolutionary Design, edited by P. F. Hingston, L. C. Barone, and Z. Michalewicz (Springer 2008).

    That was brazen promotion of a new book, but this is to point you to the 1996 paper in which I actually proved the main "no free lunch" theorem in terms of "conservation of information": Evaluation of Evolutionary and Genetic Optimizers: No Free Lunch (footnote 1 gives full bibliographic data). I don't understand why you've cited later stuff.]

    On next to section 3.

  126. 126

    Bill Dembski,

    It hurts to see you jump from interesting philosophy to “Pop Goes the WEASEL” in section 3. You misunderstood the WEASEL long ago, and fixated somewhere along the way on the notion that Dawkins “smuggles in information.” The active information of the program is a direct consequence of Dawkins’ (vague, pop-sci) model of evolution. Your “smuggling” characterization would make sense if Dawkins were modeling evolutionary search for a single genetic or amino-acid spelling of a protein — and this seems to be how you regard the WEASEL sentence. In that case, he would indeed be using a fitness function to direct a stochastic process to a “target.” But the program makes absolutely no sense as a model of biological evolution from Dawkins’ perspective if you regard fitness that way. There is no biological plausibility in cumulative selection — the process Dawkins said he was illustrating — of bases or amino acids in a target sequence, and we all know that Dawkins would never say otherwise.

    What is, in fact, subject to cumulative selection in biological evolution? Phenotypic traits, also known as, appropriately enough, characters. A biotic lineage can accumulate fit characters gradually. And what determines whether a character is fit or not? The environment. This is Darwinism 101 — just what Dawkins was trying to teach, whether or not he succeeded.

    Think of a critter in the WEASEL program as expressing traits in 28 dimensions. The WEASEL sentence is then an abstract model of the environment. It specifies, for each of the 28 dimensions, which of 27 possible traits (characters) is fit. In Dawkins’ simplistic model, fit traits make identical, additive contributions to overall fitness of the virtual critter, and unfit traits simply make no contribution — there are no deleterious traits. This choice in modeling fitness is the source of active information. And I don’t see a bit of “smuggling” in it. Dawkins was trying to show how his model of evolution differed from searching for the combination that opens a lock. The active information reflects the difference he said he was illustrating. You cannot accumulate correct numbers in a combination by trial and error, but you can accumulate fit traits, provided the rate of mutation of characters is not “too high.”

    A great deal of confusion over the WEASEL program arises from folks’ assumption that it implements a genetic algorithm in the traditional sense of mimicking evolution of genotypes. There are also programs that mimic evolution of phenotypes (physical traits). Read a brief survey of evolutionary computation, and then ask yourself, “Are the WEASEL critters like genotypes, or like phenotypes?” Answering with the latter is a no-brainer for me.

    My objective here is not to embarrass you. Scrutinizing your arguments over the years has taken me into valuable regions of “concept space” I would not have entered otherwise. But I have to say that this whole WEASEL business is just a waste of time and energy. I want to see you challenge us with your strongest arguments. How is it the WEASEL shows up alongside a proposal for a new law of nature?

    Onlookers: Another WEASEL war is precisely what I do not want. Please let this be an open letter to Bill. He’s welcome to contact me by email if he does not want to respond here.

  127. There aren’t any vision systems to search for.

    Nakashima

    Indeed, which has allowed many different vision systems to be discovered.

    That doesn’t even make any sense.

    How can something that doesn’t exist be discovered?

    There is no one, perfect vision system planned out for all creatures, nor even one system per niche. The distribution of vision systems makes sense given common descent and nested hierarchies, however.

    Common descent can at best provide a LINEAGE and a lineage is not a nested hierarchy.

  128. beezebub:

    Evolutionary theory says that if a better variation arises, it will tend to be preserved by natural selection.

    “Better” is relative. Also random effects can and do take away even the best adapted organisms.

    But anyway trhere is a HUGE difference betwen the survival of the fittest and the ARRIVAL of the fittest.

  129. Wm Dembski:

    Bob and I, on the other hand, see evolutionary processes traversing configuration spaces and locating things like bacterial flagella.

    In a design scenario that makes sense.

    In a non-design scenario, however, that is a non-starter.

    And that is my point- their position doesn’t have anything to search for.

    ID has those specifications that can be searched.

  130. 130

    Bill Dembski,

    In your third theorem, you make the fitness function something that induces a search, and that is more or less consistent with your exposition indicating that a search uses a fitness function. You claimed in this thread that the theorem generalizes the conventional (folkloric) NFL theorem. But in the original NFL analytic framework, search algorithms operate upon given fitness functions, and there is no notion of a target in the sample space (domain of fitness functions). Furthermore, performance is measured on the sequence of observed values of the function, not the sequence of visited points in the sample space. Your target is in the sample space, and “hitting the target” amounts to measuring performance on the sequence of visited points, rather than the sequence of observed fitness values of those points.

    I think your failure to see this crucial distinction is, again, that you have fixated on the notion that people use fitness functions to “smuggle in” information.

    The only way I’ve ever reconciled your notion of NFL with the original is to regard the target as a dichotomous fitness function — the characteristic function of the target, to be precise. This is shot to pieces when you make the fitness function something independent of the target. The most natural way to obtain a target in the original framework — oddly enough, you mentioned this yourself, in some communication with me years ago — is to threshold the fitness function. That is, the target is the set of all points in the sample space with satisfactory fitness.

  131. Tom English,

    Talk about smuggling in information.

    Curious, do you cover the NCSE, ACLU, and other Orwellian political police mongors and organizations in your book?

    You know, the kind where they actually tell scientist what they may or may not say in their thesis and research papers to be published?

    Nothing like scientific fascism.

  132. Beelzebub,

    I read your earlier comments prior to my response to you. I think the Conservation of Information as a Law is debatable.

    As to your authoritarian order to return and read the paper before posting again, please point out specifics in the paper to make your point. Otherwise, I consider the remark simply a hostile, sneering attitude and condescending in nature. Not a serious request, but a intellectual blow-off.

    thanks

  133. 133

    Mr Joseph,
    “Better” is relative. Also random effects can and do take away even the best adapted organisms.

    Yes. Evolution does not produce the best, only the better, a comparative measure local in time and space. And, as you point out, with no guarantees.

    But anyway trhere is a HUGE difference betwen the survival of the fittest and the ARRIVAL of the fittest.

    Which is why OOL is usually considered a separate topic of investigation from the rest of evolution.

    To a point in a previous comment of yours – the sum of all known lineages is a nested hierarchy.

  134. Evolution does not produce the best, only the better, a comparative measure local in time and space.

    And there could be many variations that are “better”. And each variation could be competing against the others.

    But anyway trhere is a HUGE difference betwen the survival of the fittest and the ARRIVAL of the fittest.

    Which is why OOL is usually considered a separate topic of investigation from the rest of evolution.

    The OoL doesn’t have anything to do with what I said.

    To a point in a previous comment of yours – the sum of all known lineages is a nested hierarchy.

    Only if one doesn’t know anything about nested hierarchies.

    Ya see nested hierarchies are built on CHARACTERISTICS not descent.

    With common descent characteristics can be lost as well as gained.

    Nested hierarchies demand additive characteristics.

  135. In #134 joseph wrote:

    “With common descent characteristics can be lost as well as gained. Nested hierarchies demand additive characteristics.”

    On the contrary, the loss of a characteristic is just as significant as the gain of a characteristic, and just as important to the construction of a nested hierarchy. All nested hierarchies are based on the comparison between “ancestral” characteristics (called “plesiomorphies”) and “derived” characteristics (called “apomorphies”). These are used to construct cladograms (i.e. nested hierarchies) in which shared derived characteristics (called “synapomorphies”) are used to infer the branch points in the cladograms. Such synapomorphies need only be derived characteristics, and the loss of a previously existing characteristic counts as an apomorphy just as easily as the gain of a previously non-existent characteristic.

    The protocols for constructing nested hierarchies were first systematically set out by Willi Hennig, a German entomologist and systematic biologist. Perhaps his most important publication in this regard is Grundzüge einer Theorie der phylogenetischen Systematik. Deutscher Zentralverlag, Berlin, (1950), in which he clearly laid out the algorithms that are now used by virtually all systematists, both inside biology and in many other fields.

    The only people who think that nested hierarchies are purely additive are people who think that evolution must be relentlessly progressive (which apparently includes virtually all ID supporters). Unfortunately for them, there is abundant empirical evidence that this simply is not the case.

  136. Tom English,

    I find your comments extraordinary and by that not in any positive sense. They seem to be of someone who is angry and attempting to call out Dr. Dembski and in some places on minutiae. And then have the gall to suggest Dr. Dembski contact you as if you are the sole judge on this paper.

    If you were truly interested in a dialogue between colleagues there is no way you would have used this forum as a means of discussing problems with the paper. You would have gone directly to Dr. Dembski with your points and if he did not respond, approach him here in a friendly manner asking questions and not berating him.

    The rest of the know nothings here who in reality know nothing about the paper behave as typical for them. We expect this childish behavior from them but we would not expect it from some one who actually understood the content of the paper, has published on the subject and can offer constructive criticisms.

    Interesting game you are playing.

  137. Why is it wrong to respond to the paper here, as the paper was the subject of a post here and thus made the subject of comment. Dembski would not have posted if he didn’t want comments.

    Also, if I’m not mistaken, Tom English is a former colleague of Dembski’s and has been a part of the history of Dembski developing his ideas.

  138. 138

    There was a post about the two Tom Englishes a couple of years ago.

  139. Hazel,

    Nobody is objecting to comments. I cannot imagine what Tom M. English is up to but it does not seem like anything positive or else he would have taken a completely different tact.

    And by the way I know who Tom M. English is from his past comments on this site. He was banned about a year ago for what I believe were derogatory comments.

  140. For those interested, Tom M. English has a website

    http://www.boundedtheoretics.com/

  141. Dr. Dembski, it’s encouraging that you’re actively participating in this thread. I have several issues to bring up, but I’m sure you have a lot of demands on your time, so I’ll make my points a few at a time.

    - In this paper, as well as previous active info work, it’s pointed out that the reduction of a search space constitutes active information. But for every search space, there are ways to define a proper superset by relaxing constraints, e.g. expanding an alphabet. By viewing these constraints as contingent, we can view any search space as having an unbounded amount of built-in active information. Can you comment on this? How do we non-arbitrarily decide which aspects of the problem should be considered contingent and which should not?

    - I see a problem that runs through all of your CoI theorems: Proving that the probability of selecting an alternate-search-whose-chance-of-success-is-at-least-q is less than or equal to p/q does not prove that selecting an alternate search doesn’t improve performance on average.

    That’s an awkward statement, so I’ll provide a counterexample: Consider a scenario in which the selection of an alternate search has a 1/3 chance of doubling the odds of success (q=2*p), and a 2/3 chance of having no effect (q=p). On average, selecting an alternate search improves performance (active info), yet the scenario accords with your theorems, i.e. the probability of choosing a search with a chance-of-success of at least q is always less than p/q.

  142. 142

    jerry,

    I am an equal-opportunity offender. Mark Perakh was right in saying that active information would end up playing a role analogous to that complex specified information had played in Dembski’s earlier work. I was right in insisting that active information was in fact quite different from complex specified information.

    This leads to an important point. Dembski has suggested that he and Marks are filling in the mathematical details he left out of No Free Lunch: Why Specified Complexity Cannot Be Purchased without Intelligence. Active information and complex specified information are radically different measures. Active information is measured on a search, relative to a null search and a target. Complex specified information is measured on an event, relative to a probabilistic model and a semiotic agent.

    Only by using the term information equivocally can Dembski claim to have made good on his promise to provide a Law of Conservation. The claim in his book was that complex specified information is conserved (actually, with a “leak” of up to 500 bits). The claim in the chapter is that active information, quite different from CSI, is conserved.

  143. 143

    I meant to write, Only by using the term information equivocally can Dembski claim to have made good on his promise to provide a Law of Conservation of Information.

  144. Allen MacNeill:

    On the contrary, the loss of a characteristic is just as significant as the gain of a characteristic, and just as important to the construction of a nested hierarchy.

    With a loss of characteristics you lose containment:

    A SUMMARY OF THE PRINCIPLES OF HIERARCHY THEORY

    nested hierarchies involve levels which consist of, and contain, lower levels.

    And a loss of containment equals a loss of NH.

    Dr Micael Denton wrote a thorough refutation of the premise that evolution leads to/ expects a nested hierarchy in “Evolution: A Theory in Crisis”. Perhaps you should read it and respond to the refutations it contains.

    But hey perhaps you can provide a nested hierarchy based on a loss of characteristics.

    That would help your case but I know you won’t post one because it doesn’t exist.

    BTW I have already posted that evolution does NOT have a direction.

    So I don’t know what your issue with that is.

  145. Tom English,

    I find the whole discussion interesting in ways different than most. First, no one seems to understand how the term “information” is being used let alone “active information” though we seem to have bit of a grasp on what that generally means. And we understand it is very different from CSI which I claim no one here can define. A few people will get upset at that characterization but I have been reading about it at various places for over 3 years and still cannot find a good definition.

    Without much of an understanding of the paper some of the anti ID people are willing to attack the article on things they do not comprehend. What these amateurs think they are doing is a mystery. I guess they seem to feel they must defend Darwinian evolution at all times even if they do not understand the threat.

    You come along who obviously know what the article is about and attack it on some peripheral stuff such as Dawkin’s Weasel. Why waste your time on something so trivial when the essence of an aspect ID is at stake. About a 1000 comments have been written here on Weasel in the last 4-6 weeks and all to naught. Maybe you don’t know that but it has been a colossal waste of cyber time and disc space.

    Who are you trying to communicate to? I do not use “with” because I don’t get the idea you want a conversation or are anxious to teach us what this is about. To Dembski? It cannot be to the peons here who do not understand more than the bare outline of what the paper is about. If it is to Dembski, you picked a strange way and it reflects badly on yourself.

    None of us are willing to take the paper at face value as the beginning of the establishment of ID as a legitimate discipline at least in terms of information. I doubt any of us understand the mathematics with all its implications which would seem to be a requisite for an evaluation. We hope it will validate ID on an information basis because we all believe intuitively what the paper is saying is valid and it would be nice to be able to point to something that it rigorous and compelling from a mathematical point of view.

    But I do not know what your objective is. If you think the paper is flawed, then lay it out. We may not be able to understand it but I bet Bill Dembski will and he may comment or he may not. But one thing for sure he is not going to respond to what you have already written.

  146. Jerry #145

    You write:

    “no one seems to understand how the term “information” is being used let alone “active information” though we seem to have bit of a grasp on what that generally means.”

    ” I doubt any of us understand the mathematics with all its implications which would seem to be a requisite for an evaluation. ”

    Surely the mathematics and the definition of information are straightforward? I agree the implications are more subtle but we are all as well qualified as each other to discuss them. As Dembski says in #90 above the theorems are “elementary but not trivial”.

    For example, the information content of an outcome is defined as log base 2(p) where is the probability of that outcome meeting a target. That’s pretty elementary.

    But what is the value of redefining a probability as “information”. In fact it may confuse more than help. As I tried to emphasise above this definition entails that the “information” content of an outcome is relative to a target. This fundamental point is rather easily hidden when you talk about bits of information rather than probabilities.

  147. 147

    Mark Frank writes:

    As I tried to emphasise above this definition entails that the “information” content of an outcome is relative to a target. This fundamental point is rather easily hidden when you talk about bits of information rather than probabilities.

    Mark,

    Don’t hold your breath for a response from Dr. Dembski. I made a similar point two days ago:

    Third, the active information of a Darwinian “search” is not independent of the target. A particular fitness regime therefore contains very little active information with respect to some targets, and a huge amount with respect to others.

    Dr. Dembski has not responded to that point or to the others I raised.

  148. “As Dembski says in #90 above the theorems are “elementary but not trivial”.”

    If I had about 6 months with nothing else to do, I could probably master the mathematics here but I abandoned formal mathematics long ago. I had a fellowship in a PhD program for math at Duke University and in the first year one of the courses was titled Analysis. A simple sounding name but Analysis was the hardest course of all the initial graduate courses. Half the students in the class were second year PhD students and nearly everyone struggled with it. Calculus is a small offshoot of Analysis. The book was about 800 pages and every page was mainly a long progression to some very complicated theorems in the discipline.

    One of the last theorems we proved was the “Central Limit Theorem.” This is a basic or elementary theorem of Statistics. However, the mathematics behind the proof took a year of esoteric mathematics to prove it. So yes, as Dembski say, somethings are elementary but not trivial.

    Now I have no idea how non trivial his theorems are but by saying they are elementary does not make them easy to understand or prove.

    Also the term “information” seems to be used in different ways.

  149. Re #147

    Beelzebub

    I see you made the same point. It is reasonable for Dr. Dembski not to respond. We are all busy and no one is obliged to take part in an internet discussion. Perhaps someone else will respond?

  150. In #144 joseph wrote:

    “…perhaps you can provide a nested hierarchy based on a loss of characteristics. That would help your case but I know you won’t post one because it doesn’t exist.”

    It’s difficult to do this without providing a diagram (i.e. a cladogram), but here goes (it helps if you draw your own cladogram using the definitions listed below; just draw a big “check mark” – that is, a large letter “V” with the right-hand line extended to the right, and then follow the directions given):

    Consider a very simple set of metallic hardware fasteners, consisting of:

    1) a ten-penny nail

    2) a deck nail

    3) a wood screw

    4) a stove bolt

    One can construct a nested hierarchy (i.e. a cladogram) of these four objects, using the following nodes and internode definitions:

    1) the outgroup is the ten-penny nail, as it lacks threads (put the name “ten-penny nail” at the end of the left-hand line of the big “V”);

    2) the first internode definition (i.e. the first synapomorphy) leading to the first derived clade is “threaded” (put a hash mark and this label up and to the right of the point of the big, lopsided letter “V” you started with);

    3) the first derived clade is “deck nails” (put a line branching off to the left from the extended right-hand line of the letter “V”, and label the end of this left-hand-sloping line “deck nails”);

    4) the second internode definition (i.e. the second synapomorphy) is “slotted heads” and “finer threads” (put a hash mark and these labels up and to the right of the branch point between the leftward sloping line that leads to “deck nails” and the right-hand line of the letter “V” you started with);

    5) the second derived clade is “wood screws” (put a second line branching off to the left from the extended right-hand line of the letter “V” above the second hash mark, and label the end of this left-hand-sloping line “wood screws”);

    6) the third internode definition (i.e. the third synapomorphy) is “loss of sharpened point” (put a hash mark and this label up and to the right of the branch point between the leftward sloping line that leads to “wood screws” and the rest of the right-hand line of the letter “V”);

    7) complete your cladogram (i.e. your “nested hierarchy”) by labeling the end of the far-right-hand line “stove bolts”.

    You should have a large letter “V”, slanting to the right and with four branches off of it, reading in order from left to right: “ten-penny nails”, “deck nails”, “wood screws”, and “stove bolts”. This is a simple cladogram for metal hardware fasteners.

    Now, in the context of your request, notice that the the third internode definition (i.e. the third synapomorphy) is “loss of sharpened point”. I have provided you with a “nested hierarchy” that does, indeed, have a loss of a character as one of its defining characteristics. Ergo, your assertion, “I know you won’t post one because it doesn’t exist” has been conclusively and demonstrably falsified.

    One might object that the cladogram/nested hierarchy I provided in the above example is for “designed” objects (i.e. objects that are the product of “intelligent design”, in this case the “designers” of metallic hardware fasteners). Therefore, I will now provide an analogous cladogram/nested hierarchy using vertebrates (for which, I assume, you have no problem with the idea that they have evolved in the pattern indicated):

    Consider a very simple set of vertebrates, consisting of:

    1) a lobe-finned fish (i.e. a member of the Rhipidistia)

    2) an ancestral amphibian (such as a Labyrinthodont

    3) an ancestral mammal (such as Cynognathus)

    4) a derived protocetacean (such as Indocetus ramani)

    5) a modern whale (such as Physeter macrocephalus)

    One can construct a nested hierarchy (i.e. a cladogram) of these five taxa, using the following nodes and internode definitions:

    1) the outgroup is the lobe-finned fish, based on its anatomical characteristics, especially the bone structure of its fins (put the name “lobe-finned fish” at the end of the left-hand line of the big “V”);

    2) the first internode definition (i.e. the first synapomorphy) leading to the first derived clade is “tetrapod anatomy” (put a hash mark and this label up and to the right of the point of the big, lopsided letter “V” you started with);

    3) the first derived clade is “Labyrinthodonts” (put a line branching off to the left from the extended right-hand line of the letter “V”, and label the end of this left-hand-sloping line “Labyrinthodonts”);

    4) the second internode definition (i.e. the second synapomorphy) is “mammalian skeletal structure”, especially the structure of the bones at the hinge of the jaw (put a hash mark and this label up and to the right of the branch point between the leftward sloping line that leads to “Labyrinthodonts” and the right-hand line of the letter “V” you started with);

    5) the second derived clade is “Cynognathus” (put a second line branching off to the left from the extended right-hand line of the letter “V” above the second hash mark, and label the end of this left-hand-sloping line “Cynognathus”);

    6) the third internode definition (i.e. the third synapomorphy) is “skeletal modifications for locomotion in water” (put a hash mark and this label up and to the right of the branch point between the leftward sloping line that leads to “Cynognathus” and the rest of the right-hand line of the letter “V”);

    7) the third derived clade is “Indocetus ramani” (put a second line branching off to the left from the extended right-hand line of the letter “V” above the second hash mark, and label the end of this left-hand-sloping line “Indocetus ramani“);

    8) the fourth internode definition (i.e. the fourth synapomorphy) is “loss of hind legs” (put a hash mark and this label up and to the right of the branch point between the leftward sloping line that leads to “Indocetus ramani” and the rest of the right-hand line of the letter “V”);

    7) complete your cladogram (i.e. your “nested hierarchy”) by labeling the end of the far-right-hand line “Physeter macrocephalus” (i.e. sperm whales)”.

    You should have a large letter “V”, slanting to the right and with five branches off of it, reading in order from left to right: “lobe-finned fish”, “Labyrinthodont”, “Cynognathus”, Indocetus ramani, and “Physeter macrocephalus“. This is a simple cladogram for the phylogeny of modern whales.

    Once again, in the context of your request, notice that the the fourth internode definition (i.e. the fourth synapomorphy) is “loss of hind legs”. I have provided you with another “nested hierarchy” that does, indeed, have a loss of a character as one of its defining characteristics. Ergo, your assertion, “I know you won’t post one because it doesn’t exist” has again been conclusively and demonstrably falsified.

  151. Allen,

    I am not sure that “cladogram = nested hierarchy”

    Nested hierarchy follows the rules I linked to.

    Did you read it?

    It is from the ISSS.

    Cladograms do not have to follow such rules.

    Cladograms are based on shared characteristics only.

    So the problem is you conflating a cladogram with a nested hierarchy and then trying to use that to refute my claim.

    IOW Allen once again you use dubious tactics to try to make a point.

    I told you to read “Evolution: A Theory in Crisis”- Denton goes over this.

    But anyway Allen, if you take a cladogram based on living organisms- do you see the lines and Vs? Each point and each node is a TRANSITIONAL form. A form with a MIX of characteristics.

    If we included ALL (alleged) transitional forms in a cladogram, what do you think it would look like?

    Would it look nice and neat with distinct tips? No. It would be a mess beacuse characteristics can be lost or gained, mixed and matched.

    So all you have done is refute a strawman.

    You should be very proud of yourself.

  152. Clade:

    The term “clade” did not exist in the older Linnaean taxonomy, which was by necessity based only on morphological similarities between organisms. The concept embodied by the word “clade” does not fit well into the rigid hierarchy that the Linnaean system of taxonomy uses; indeed, cladistics and Linnean taxonomy are not really compatible.

    Linnaean taxonomy demands that all organisms be placed neatly into a rigid, ranked, hierarchy of taxa, such that one individual kind of organism must belong in one of each of the categories: species, genus, family, order, class, phylum and kingdom. Because of this necessity to “file things away neatly”, the Linnaean system is often very convenient indeed in organizing such things as large museum reference collections, however it does not represent well the process of change that actually happens over evolutionary time.

    Because clades can be nested at any level, they do not have to be neatly slotted into a rank in an overall hierarchy. In contrast, the Linnaean taxa of “order,” “class” etc. must all be used when naming a new taxon. They cannot be avoided, and each one implies a certain (admittedly very poorly defined) level of diversity, which is supposed to be equivalent throughout the system.

  153. beelzebub and Mark Frank,

    You don’t need Dembski to respond.

    All you need to do is take something that is alleged to have active information and show it can arise via nature, operating freely.

  154. I’m a little bummed that I joined the party after Dembski left. I’m guessing that the issues brought up by several commenters here, including myself, will remain unaddressed.

    But just for good measure, I’ll throw some more out for anyone who cares:

    - The LCI, as stated in the paper, is false for the simple reason that no constraints are specified for defining the higher-order search space. Given the freedom to define it as we please, we can always define it such that the probability of finding a good search is higher than the performance gain of a good search over the null search.

    To remedy this would require the addition of a condition such as “… with the higher-order search space reasonably defined …”. This is, of course, vague language, but no more vague than leaving it out.

    - I join Tom English in congratulating Marks and Dembski for coming right out and saying, “Intelligence creates information.” Unfortunately, no evidence for this assertion is offered. I submit that, according to their definitions, active information is created by luck, not by intelligent agents. Here’s my reasoning:

    The amount of active info associated with something is the log-transformed, baseline-relativized probability of that something finding a solution (which may be a low-order target or a good search ). If intelligent agents have a knack for finding solutions, that means that there is a high probability of them doing so, hence a large amount of active info associated with intelligent agents. So when intelligent agents find solutions, that success is accounted for not by the creation of active info, but by the pre-existing active info associated with the agents.

    On the other hand, when a good search is found by something that is unlikely to find it, this is a case where something with high active info came about by something with low or no active info. Thus, luck creates active info.

    - Quotes regarding information seem misapplied to the active info framework, since the authors of the quotes were certainly not talking about active info. For example, Marks and Dembski have quoted Brillouin several times in their active info papers. But Brillouin was talking specifically about deterministic computation, which he viewed as non-information-producing precisely because of its determinacy. What does his point have to do with conservation of performance for typically non-deterministic searches?

    - In a comment above, although not in any active info papers that I’m aware of, Dembski restricts the notion of intrinsic targets to points in the search space that are salient. As Mark Frank pointed out, this is a concept that would need to be fleshed out objectively if the LCI is to be applied to situations in which a target isn’t given.

    Dembski also mentioned “independent functional specifications” as an example of salience. My question is: Independent of what? If he means “independent of the probability of the outcome matching the specification”, as he did in his specified complexity work, then targethood depends on the choice of search. It seems that having targets appear and disappear depending on our choice of search violates Marks and Dembski’s assumptions.

    - In Dembski’s previous work, he stated, “It follows, therefore, that how we measure information needs to be independent of whatever procedure is used to individuate the possibilities under consideration.” Agreed, but the active info measure is not independent of how the search space elements are individuated. So different but equally accurate models of a problem could yield different information measurements. Why has Dembski stopped seeing this as a problem?

  155. R0b,

    I suggest you go to Tom English’s site and ask him. He knows about this stuff. You may not get the same answer that Bill Dembski would give or even get an answer. They do not seem to agree with each other on some areas of this topic.

  156. Beez,

    I cannot presume to speak for Dembski, but I don’t understand your objection and this may be why you haven’t gotten an answer from anyone here, including him. If you could clarify your concern more clearly perhaps someone here can address your question or give another take on it.

    As it stands, I doubt if your short post articulated your concern well enough for us to understand why you see an issue. (Or maybe I’m just obtuse, which is always a live option.)

    Atom

  157. R0b wrote:

    The LCI, as stated in the paper, is false for the simple reason that no constraints are specified for defining the higher-order search space. Given the freedom to define it as we please, we can always define it such that the probability of finding a good search is higher than the performance gain of a good search over the null search.

    Since no one is responding to your point I’ll attempt to address it, as discussions are no fun when people ignore you.

    The beauty of Dembski’s COI theorems, if the formal mathematics underlying them holds, is that they show the relationship of information transfer from higher level to lower level searches. Namely, they bound it by the relation that the amount of information used at the higher level to select a “good” fitness function from the space of possible fitness functions (that reduction of possibilities is information in the sense of reducing uncertainty and constraining possibilities) will always be greater than or equal to the amount of information gain in the lower level search, measured in terms of improved search performance. (There are other ways to measure the information gain in the lower level search rather than by a posteriori measurements of search performance, but Dembski’s measure is fine and works.)

    In other words, the gain in search performance at the lower level is proportional to the information input at the higher level, via the reduction of “fitness function space” to a single (or set of) fitness function(s) that gives us at least this increase in performance.

    This is all very abstract, so let’s try to make it slightly more concrete.

    Let’s begin with a search for the string “I AM AN ENGLISH SENTENCE” from the search space of all 24 letter permutations of the 27 letter alphabet (26 uppercase plus one space character). Our original search space is 27 ^ 24 elements big, roughly 10 ^ 34. So we have a base probability (intrinsic difficulty) of 1 in 10 ^ 34. (This much should not be controversial so far, if I did my arithmetic right)

    Now let’s say we want to use an evolutionary search to improve our chances of finding the target using an appropriate fitness function. (As Weasel Ware 2.0 shows empirically, not just any fitness function will work.) We have to reduce the space of possible fitness functions to the space of ones that “work”, or allow us to find the target in a reasonable number or queries.

    But how many possible fitness functions are there for that search space?

    If we limit the number of possible “fitness values” to n, we get n ^ (10 ^ 34) possible fitness functions. So the search space for fitness functions is exponentially larger than our original search space, except for the trivial case where we only have one possible fitness value (we’d only have one function in that case.)

    Depending on how many “good” fitness functions there are, this search may be harder than the original search. (It is Dembski’s point that it is.)

    So to get to your concern, the higher order “search space” is simply all the ways to assign values between 0 and n to each of the permutations of the original space. So it is already implicitly defined. Now if we want to reduce this higher order space to only a subset of the possibilities, this reduction incurs an informational cost. That cost is greater than or equal to the active information gained (by the original search) by such a reduction.

    So choosing a good fitness function costs information, and allows the original search to perform better than blind search by that same (or less) amount of information, using Dembski’s metric of active information (q being the improved probability of success, -log_2(q) being the information associated with that new probability.) It is either the same or less, but never more.

    If we want to assume some fitness functions are more “likely” than others, this change of probability distributions also incurs an informational cost. Demsbki shows this using the probability distribution version of his theorem, which again shows that the higher order reduction incurs a cost at least equal to the active information.

    So unless you can put your objection another way, I don’t think the freedom to reduce the space of possible fitness functions violates COI in any way; it would seem to underscore it. So your claim that Dembski’s COI is “false” is probably presumptuous and incorrect.

    I’m open to further clarification of your concern, as I may have missed a vital part of your argument.

    Atom

  158. 158

    jerry,

    There’s a backstory I’m not going into here. You might consider that Dembski has posted at DesignInference.com,

    Jeffrey Shallit I and Jeffrey Shallit II. [8Jul05] My end of a sharp exchange with Jeff Shallit. Jeff was a teacher of mine at the University of Chicago in the 1980s. I took away some useful insights from his course on computational number theory. I’ve valued him as a critic even though my public denunications of him have been a bit over the top. Perhaps some day we will be able to put our differences on the table dispassionately.

  159. 159

    Folks,

    While I’ve said that the Law of Conservation of Information is not what Dembski promised in his book No Free Lunch, I will say also that the chapter is not “written in jello.” There are some clear and important points to be debated.

    Perhaps we could get a fresh thread from the management, now that we’ve read and contemplated the chapter.

    Ironically, Dembski and Marks get a free lunch from discussion like this. They don’t have to mingle with the riff-raff, but they can exploit anything useful that comes along in revising the chapter. It doesn’t look like a final draft to me.

  160. I was going to say this earlier but Tom English just said it. Bill Dembski uses this site for ideas and he has been open about this for as long as I have been contributing. I have no idea if anyone has found a flaw let alone a fatal flaw in the current paper. But my guess is that as Tom English has said this was not written in jello and Dr. Dembski was looking for some fine tuning.

    So all you anti ID people out there, take pride in your contributions to ID. They have been many. So maybe as Tom English asks, Bill will put up LCI Two. And maybe Tom will explain to us what it is all about in a kinder and gentler way than his earlier comments. I know I have lots of questions but they are of a basic level and a clarification level and maybe the topic is just too beyond me.

    By the way if I were Bill Dembski I would not answer too many questions in a forum like this. Bill tends to get quoted often by people looking for something negative to say about him and anything said here would not be used in any positive way. It is just how the anti ID people work. They are such a nice considerate lot.

  161. jerry wrote:

    Bill tends to get quoted often by people looking for something negative to say about him and anything said here would not be used in any positive way

    Ain’t that the truth.

    Atom

  162. 162

    jerry,

    What you said about your course in analysis reminded me of the concept of logical depth, developed by Charles Bennett. The Central Limit Theorem can be derived by a fairly small program containing the requisite axioms, but the program evidently has to run a long time to obtain the result. Loosely speaking, the program length is the algorithmic information of the theorem, and the running time (computational cost) is the logical depth of the theorem.

    I just had a look at a classic paper by Bennett, Logical Depth and Physical Complexity (1988). Most of what he says in the first 5+ pages is not only easy to follow, but also utterly brilliant. I found the intro highly relevant to the present discussion, and I recommend it highly to everyone.

    For Bennett, the value of an object is the time cost of obtaining it from a compact program. That is, an object low in algorithmic information may be high in value because it takes a long time to compute. Dembski and Marks focus on information cost, ignoring time. I believe there’s a lot to be learned in considering the difference.

  163. Atom, thanks for your comment. If there’s any ID proponent here who can hash out these issues, it’s you. Here’s my ridiculously long-winded response.

    From the paper:

    The Law of Conservation of Information (LCI). Any search that proportionately raises the probability of locating a target by q/p with respect to blind search requires in its formation an amount of information not less than the active information I+ = log(q/p).

    To assess the information cost of an efficient search, we essentially invent a story of how the search came to be. The LCI doesn’t tell us what this story should be, other than that it should involve selecting something from some space, and we calculate the information cost by dividing the number of good somethings by the size of the space.

    Given an evolutionary search that efficiently finds “I AM AN ENGLISH SENTENCE”, you say that it was the fitness function that was selected. It may just as well have been the algorithm, or any other aspect of the search, or the search itself, as in Marks and Dembski’s measure-theoretic CoI theorem.

    And what space was it selected from? You say it was from the space of all fitness functions that map all 24-letter sequences (from an alphabet of capital letters and spaces) to n numerical values. But why can’t we say, like Marks and Dembski did in their WEASEL analysis, that it came from the space of all fitness functions that indicate proximity to a target?

    Interestingly, Marks and Dembski say in endnote 49 that they should have defined the space to be more expansive, as you did. They say, “In limiting ourselves to fitness functions based on the Hamming distance from a target sequence, we’ve already incurred a heavy informational cost.” But — and this is my key point — we have to limit ourselves somehow or we have a completely undefined space.

    For instance, how does your higher-order search know the domain and codomain of the fitness function that it’s searching for? Shouldn’t the space include all fitness functions, not just those with that particular domain and codomain?

    And how does it know to search for a fitness function, as opposed to an algorithm or something else? Shouldn’t the space include everything?

    As I said, we have to limit ourselves somehow or we have an undefined space. But the LCI doesn’t tell us what limitations to impose on the higher-order space. So it seems that we’re free to limit it as we please. We can limit it to only “good” fitness functions if we want, in which case finding a good fitness function incurs no information cost, and the LCI is rendered false.

    Marks and Dembski could specify in the LCI how a higher-order space of fitness functions should be defined, but higher-order spaces can also be defined in an infinite number of ways that don’t involve fitness functions. As the paper says, “the specific forms by which null and alternative searches can be instantiated is so endlessly varied that no single mathematical theorem can cover all contingencies.” Thus my admittedly poor remedy that the LCI include a condition that the higher-order space must be “reasonably” defined.

    Interestingly, this problem was introduced by this paper. Previously, the only CoI theorem was a more detailed version of their measure-theoretic CoI theorem. This theorem specifies a definition for the higher-order search space that works for any search. Under that definition, the LCI certainly holds. If Marks and Dembski had stuck with that single definition, this problem wouldn’t exist.

    But perhaps I’m taking the LCI too literally. Here’s one of its characterizations from the paper: “Thus, instead of LCI constituting a theorem, it characterizes situations in which we may legitimately expect to prove a conservation of information theorem. LCI might therefore be viewed as a family of theorems sharing certain common features.”

    If the LCI is merely a family of theorems, and there are situations in which we would not legitimately expect CoI to hold, is it right to call it a law?

  164. 164

    R0b mentions the Brillouin quotation Dembski and Marks often use:

    The [computing] machine does not create any new information, but it performs a very valuable transformation of known information.

    I have highlighted the part that dovetails nicely with my post on logical depth (162). Bennett addresses the value of information, and Dembski and Marks do not.

    In fact, most computations of classical computers are irreversible. You can’t tell your laptop computer to run in reverse to recover its earlier state. Information is lost (entropy increases, and your lap gets hot) as your laptop machine works to make irreversible transitions from one state to another.

    Biological reproduction generally increases, and natural selection generally reduces, the entropy of a population — this is a common and uncontroversial observation. Novel information arises when biota work to reproduce, and something extrinsic (the “environment”) impinges on the process. Parents do not create the novel information in their offspring. To tie this to Brillouin, we may interpret natural selection as a value-increasing transformation in which information is reduced. What constitutes value in biotic information is more subtle than most people make it out to be, and I’m going to leave the concept fuzzy for now.

  165. “Biological reproduction generally increases, and natural selection generally reduces, the entropy of a population — this is a common and uncontroversial observation.”

    No problem here. At least I think not but it will depend upon what is meant by entropy in this situation. Natural selection can lead to survival in a single step process (introduction of a new environment) or to extinction in a multi-step process (several changes in environment.) In the multi-step process key genetic elements may have been weeded out by adapting to a previous environment and thus make the population less likely to survive a new environment.

    “Novel information arises when biota work to reproduce, and something extrinsic (the “environment”) impinges on the process.”

    Here we are getting into a sort of “No Man’s Land” and by that I mean it depends on exactly what you mean whether it is true or not. Novel information has to be defined. In the concept of a gene pool there are multiple versions of many genetic elements, e.g. alleles. There may not be any member of the pool that has every possible combination and sexual reproduction generates a unique individual but no unique genetic element has arisen.

    A second scenario is that there is recombination during sexual reproduction to produce a unique genetic element that was not in the original gene pool and the gene pool is now larger and there is truly a unique new element. But usually these genetic elements are of no major consequence in life’s journey from microbe to man but just provide some new variant.

    A third possibility is that some of the gametes were mutated by one of the many known processes and then during sexual reproduction this new genetic element was incorporated into the gene pool.

    Thus, two and three add new genetic elements. Natural selection or environmental pressures may determine if a new genetic element survives or what percentage of the population gets the new genetic element possibly taking thousands of generations or more to reach a stable level.

    From what I have read the supposed cause for major evolutionary change is this third process. Now I have no idea how this plays out in the information scenario that is being analyzed in the subject paper.

    “Parents do not create the novel information in their offspring. To tie this to Brillouin, we may interpret natural selection as a value-increasing transformation in which information is reduced. What constitutes value in biotic information is more subtle than most people make it out to be, and I’m going to leave the concept fuzzy for now.”

    Yes it is fuzzy with or without any reference to the subject paper. The term value increasing is fuzzy. What is value increasing in one environment could be detrimental in another but I am not sure if that is what you mean. As you said and I agree, fuzzy.

    Natural selection tends to cull the gene pool over time and this process leads to a gene pool that may not be able to survive some environment and thus, go extinct. This is certainly a loss of information and does not seem to be the same information that is referred to in LCI.

    Yes, fuzzy!!!

  166. 166

    R0b,

    I mentioned that I enjoyed reading the opening philosophical sections of the chapter. But there is a hugely important philosophical issue begging to be addressed, and Dembski and Marks have ignored it: What is a probability? It seems that you are touching on this omission.

    There are people far better qualified than I to debate Dembski on probability interpretation. But I am certain that it is a crucial issue in the present context.

    To treat the negative logarithms of probabilities as physical information, Dembski and Marks must stick consistently to physical probabilities. The consensus on physical probability is that it must be related to a repeatable physical experiment. You mentioned the absence of “constraint” on probabilities, and I think I have just identified what is required.

    Dembski and Marks never discuss how their probability measures model physical searches. They go from pure math regarding a regress of probability measures to a claim that they have a characterized physical reality, doing nothing to establish the probabilities as physical. In short, they have reified (hypostasized) abstract mathematical entities. At first blush, I would say that virtually none of their searches corresponds to an experiments repeatable in the observed universe (i.e., the relative volume of the subset of realizable searches is vanishly small).

  167. R0b,

    Thank you for your comment. I think your point is now clearer.

    I think this is where Dembski’s point about LCI being applied to individual situations is important. Once a person defines their search set-up and how it has improved performance over blind search, the applicability of LCI can be shown, as Dembski did to Dawkins’ Weasel.

    In the footnote you mentioned, Dembski showed that even limiting himself to just the proximity reward functions, you still have as many different proximity reward functions as you have elements in your original search.

    If we take into account the other possible fitness functions for our evolutionary search, the informational cost is even greater.

    So you question becomes, in essence, “What if we define our higher order search as just the elements that are ‘good’ fitness functions?” In this case we are still looking at a subset of the possible functions, so the cost is there.

    But why doesn’t this apply to choosing an evolutionary search (with all fitness functions) out of all the other search algorithms? Shouldn’t we also take into account the informational cost in the algorithm reduction?

    The answer is no, because simply choosing one search strategy over another does not improve search performance (as shown by the NFL theorems.) So it becomes irrelevant to us, since we’re only interested in explaining the improvement in our original search and showing that the improvement gain in the original search comes at a cost equal to or greater than the active information.

    If we take into account all the other informational costs that do not contribute to search improvement, then the informational cost can get much larger than the active information. But if we limit ourselves to only the reductions that lead directly to search performance improvement, this cost cannot be less than the active information, per Demsbki and Marks’ paper.

    Atom

  168. Atom:

    So you question becomes, in essence, “What if we define our higher order search as just the elements that are ‘good’ fitness functions?” In this case we are still looking at a subset of the possible functions, so the cost is there.

    Yes, but my point is that the cost is always there, because every set is a proper subset of another set.

    According to Marks and Dembski, the appropriate way to define the higher-level space of fitness functions is to fix the domain and codomain, and vary the mapping. But that space is part of a larger space in which the domain and codomain also vary. And that’s part of a larger space that includes things other than functions. Marks and Dembski ignore the information costs associated with selecting a space from these superspaces, so why can’t we ignore the cost of selecting the “only good functions” set from a superset?

    Your point that our choice of algorithm has no effect on performance is a good one, assuming that the fitness function is not fixed and that all available algorithms sample without replacement. In such a case, it makes no difference whether the algorithm is fixed or variable. But we could, if we wanted, fix the fitness function and vary the algorithm.

    Different definitions of the higher-level space will yield different information costs. If you apply the three different CoI theorems to the WEASEL example, you’ll get three different information costs. So information cost is not just a function of the problem at hand, it’s also a function of how we model it. Given the lack of constraints in the LCI, we can model it such that the information cost is less than the performance gain, thus falsifying the LCI.

  169. Tom:

    But there is a hugely important philosophical issue begging to be addressed, and Dembski and Marks have ignored it: What is a probability?

    I think that this question begs to be addressed in all of Dembski’s ID work. It’s a rock that has broken many a shovel while digging through the mathematical, empirical, and metaphysical layers of Dembski’s arguments.

  170. R0b,

    Thanks for your response.

    You wrote:

    Given the lack of constraints in the LCI, we can model it such that the information cost is less than the performance gain, thus falsifying the LCI.

    In my post I emphasized the minimal information cost for reductions that led to lower level search improvement. As I said explicitly, not all reductions will aid in your lower level search, and thus are irrelevant for LCI purposes. But those that do aid in search improvement will incur an informational cost not less than the q/p improvement measure, or following the paper, the active information.

    Just because we can arbirarily expand/inflate the information cost as much as we’d like by including irrelevant (non-performance improving) reductions does not mean we can reduce it below the active information when we include the relevant reductions.

    BTW, we can always arbitrarily inflate any probability calculation (“what are the chances of getting heads on a coin flip…with that particular coin out of all the coins of history?”), so if your objection were viable, we could never meaningfully calculate the probability of anything.

    I believe Dembski’s paper makes clear the relevant subset, since it is always in reference to an improved original search, therefore, only search improving reductions need to be included, but they all must be included.

    Again, the importance becomes on choices/reductions that improve the search performance. Even if those are the only reductions we will include in our total (and we could always include more), the LCI still holds.

    So I don’t believe you can “model [the higher-level search] such that the information cost is less than the performance gain” when you include all the costs associated with performance improving reductions; if you can, I bet it would be pretty easy to show where you’re ignoring relevant (performance improving) reductions. Go ahead and give it a shot; see if you can select a subset in a higher order space such that this subset improves search performance by ratio q/p and yet has an informational cost less than the active information. I don’t think you can. Either that choice resulted in a search improvement, and so that reduction would need to be added to the cost (by definition) and will be greater than or equal to the active information (per the theorems), or it wouldn’t result in an improved search, which contradicts our premise that it does.

    Atom

  171. 171

    Mark Chu-Carroll of Good Math, Bad Math has reviewed the Dembski/Marks paper.

    He is not impressed, to put it mildly.

    Link

  172. beelzebub,

    Can either you or Mark Chu-Carroll demonstrate that nature, operating freely giving rise to active information?

    THAT is the only way to refute the paper.

    Nobody cares who is impressed or not. People care what can be demonstrated.

    IDists can demonstrate agencies giving rise to active information….

  173. 173

    Mr Joseph,

    As I read Mr Chu-Carroll, he is agreeing with Dr Dembski that evolution works by active information. He has situated that active information in the structure of the search space – evolution is searching physico-chemical systems in 3 spatial and one temporal dimension at a certain range of temperatures and pressures, but still far from equilibrium. That is a subset of search spaces that is far easier for evolution than other arbitrary searches, most of which resemble white noise.

    Dr Dembski’s differences with Mr Chu-Carroll only begin with questioning the source of this active information. Dr Dembski has said the source in teleogical, Mr Chu-Carroll is silent on the issue.

  174. Nakashima-San:

    1] the complexity of organization of the cell and its genetic component defies blind search, starting with the need to get to the shores of islands of function.

    2] posing hill-climbing algorithms that use randomness to get variations that fitness functions reward or punish begs this prior question.

    3] Active info relates to the info needed to get TO islands of function in combinatorially explosive config spaces. (Notice that ratio of odds on blind random search landing in the target vs a search that gets you there in a much more feasible span of effort.)

    4] We observe 2 sources of high contingency: chance/stochastic and purposeful. Known cases of complex organised function uniformly trace to design — for the obvious reason.

    5] The relevant large increments in biofunctional info to explain first life or body plans — 100′sK to 10′s or 100′s M bits — thus on inference to best explanation credibly come from design, i.e teleology.

    6] Can you — or any other supporter of the power of chance to get us to the shores of function, supply an observed case of chance doing the lucky noise trick?

    7] If not, you have not got a good empirical case, I am afraid.

    GEM of TKI

  175. Here we go again. Just what is a search function, a fitness landscape, active information?

    Evolution supposedly works by creating a modification in the genome through let’s say Allen MacNeill’s 50+ engines of variation. Each such change is a search. It is a blind search for nothing. It has no objective except that it may be stimulated by some environmental pressure but in most cases it will be the result of a random event to the genome with no particular environmental pressure.

    The search which seems to be a misnomer because there is no actual search going on is really a wandering into various combinations of DNA. And these various combinations of DNA can affect the organism. This is fairly simple to understand.

    The easiest change to understand is that a protein will be modified and this modification can have some effect on the organism. The fur may now be white and not brown. If the effect results in more offspring then we say that natural selection has “selected” this change for keeping. Most likely it will be damaging or neutral and be eliminated except some neutral ones may be kept because of a random process or for other reasons.

    Now this is a little like looking for buried treasure and while the big treasure is buried on Bora Bora and this island is quite out of the way even if you somehow made it to Tahiti. But the analogy is that all over the world there is buried treasure and while most of the digging will come up empty occasionally a dig will find something valuable and near these valuable finds will probably be some useful items that the treasure buriers included that make the original treasure more desirable and for which the diggers can easily get to.

    But the analogy is not quite the same for the genome because while the nearby treasure is usually considered additional loot, a nearby genomic element is just a possible replacement for what has already been found. The treasure analogy would have to be modified so that any additional treasure would somehow just replaces one you already found.

    What are the odds of a treasure seeker getting to Bora Bora when it is digging in Omaha or Barcelona or Manila. Not very likely in a trillion years let alone the 3.5 billion years since life appeared on earth. Especially when the treasure seekers are constrained by moving on foot and will essentially continue to dig around the original find and might feel obsessive and dig in places where they already dug. It becomes obvious to treasure diggers that after a while there is nothing more to dig for because each new hole produces less and less and almost all are producing nothing. And the search pattern when that happens is to go back to the original site and dig some more around it. They do not widen the search because they find no value in digging empty holes when the original site may provide some additional trinkets.

    But if someone gave the treasure diggers a map, and instructions on how to build a boat, then they may get to Bora Bora and many other places where other large treasures are buried.

    Modern evolutionary biology says that a few loners will defy the masses and go out alone digging holes all over the place in a random fashion and by doing so, a few of these loners have found some new treasure. But there is no evidence that any of these wandering treasure hunters ever found anything of much values let alone a jackpot.

    The point of this little analogy is that there isn’t any specific treasure but that the search finds many minor small treasures but is really unlikely to find the really big ones which are buried quite far away from where they are digging, maybe on another continent or even under the sea.

    The active information is the map and boat in the treasure hunt. In evolutionary biology it is ??? My guess it is intelligence. Because intelligence can certainly do it just as it provides the map and boat for the treasure hunters.

    Maybe others can define the active information differently and just how does a fitness landscape fit into this scenario.

  176. Atom, I’m not sure what you mean by performance-improving reductions, so I’ll go back to your “I AM AN ENGLISH SENTENCE” example to concretize the discussion.

    You propose a higher-level space consisting of all functions f:A->B where A is the set of all 24-letter sequences and B consists of n fitness values. As you point out, there are n^(27^24) such functions.

    I submit that, in proposing that space, you are already incurring an infinite information cost. I’ll call your space X, and note that it is a subset of space X+Y (“+” meaning union). I’m going to define Y in a thoroughly contrived way:

    Y is the set of functions f:C->B for which condition0 is true. C is the set of all sequences of any length, B consists of n fitness values, and condition0 is the condition that all 24-letter sequences in f map to a fitness of 0.

    Assuming that your algorithm searches the space of 24-letter sequences, the functions in Y provide no benefit to the search since they always return 0 for 24-letter sequences. Since Y is infinitely large, it overwhelms the finite number of “good” functions in X, so the percentage of good functions in X+Y is 0.

    Reducing the space from X+Y to X improves performance in the sense that it increases the percentage of good functions in the higher-order space. (I don’t know any other sense in which a higher-order space reduction can improve performance.) But this is an infinite reduction, incurring an infinite information cost.

    I submit that if you can ignore the infinite information cost incurred by reducing X+Y to X, then I can ignore the infinite information cost incurred by reducing X+Y to the set of only good searches. In doing so, the LCI is falsified.

  177. Mr Kairosfocus,

    1] the complexity of organization of the cell and its genetic component defies blind search, starting with the need to get to the shores of islands of function.

    Agreed, blind search is a complete mischaracterization of OOL and evolution.

    2] posing hill-climbing algorithms that use randomness to get variations that fitness functions reward or punish begs this prior question.

    What was prior was not a question, it was an assertion. However, it should be noted that hill-climbing algorithms rely much more on history than randomness to acheive results. So your combination of “blind search” and “randomness” are getting you off on the wrong foot in understanding evolutionary processes.

    3] Active info relates to the info needed to get TO islands of function in combinatorially explosive config spaces. (Notice that ratio of odds on blind random search landing in the target vs a search that gets you there in a much more feasible span of effort.)

    Agreed. My reading and questions to Dr Dembski still leave me fuzzy on some of the details of active information, but I think I have basic understanding of the concept. You can see some of my questions above, and some of Dr Dembski’s responses.

    4] We observe 2 sources of high contingency: chance/stochastic and purposeful. Known cases of complex organised function uniformly trace to design — for the obvious reason.

    If you look at the Human Competitive Awards, you might rethink that.

    5] The relevant large increments in biofunctional info to explain first life or body plans — 100’sK to 10’s or 100’s M bits — thus on inference to best explanation credibly come from design, i.e teleology.

    Sir, these are numbers you repeat over and over again. This is handwaving, not fact. But even if I took it seriously, let’s assume that the very large number of cells populating the early oceans of Earth, relicating perhaps as fast as every half an hour, could in a year only fix one bit of information. Then it would take them only 100 million years to reach your number of bits necessary to develop diverse body plans. Since history is much more important than randomness, Deep Time has to be taken very seriously as the source of active information.

    6] Can you — or any other supporter of the power of chance to get us to the shores of function, supply an observed case of chance doing the lucky noise trick?

    I have, in universes with simpler laws of nature, such as this Evoloops CA example shown here. Doing the same in our universe is what scientists have been doing ever since Miller-Urey.

    7] If not, you have not got a good empirical case, I am afraid

    I am always ready to listen to a good empirical case. If there is a suite of show-stopping experiments demonstrating how OOl is doomed to failure, by all means point them out to me. I am quite willing to listen to incremental results, because all that the conventional scientific community has at this point is incremental results.

  178. R0b,
    I understood your point but I don’t know if you understood mine.

    By reductions that improve performance, I meant relative to the performance of blind search.

    You wrote:

    I submit that if you can ignore the infinite information cost incurred by reducing X+Y to X, then I can ignore the infinite information cost incurred by reducing X+Y to the set of only good searches. In doing so, the LCI is falsified.

    In going from X+Y to X, you didn’t improve the performance of the search over blind search, for this simple reason: the set of all possible fitness functions / strategies, when averaged over all problems, is no better than blind search and the performance of the set of all n^(27^24) fitness functions, which is a proper subset of X+Y, still has a performance equal to blind search when averaged on our problem (meaning you average the performance of all the fitness functions in our set on our problem.) You can see this intuitively as well, since for every fitness function in our reduced set that encodes correct information about the target, we have a negated fitness function that would encode incorrect information about the target.

    The reduction from all strategies to just the evolutionary strategy with all fitness functions of n^(27^24) did not improve search performance. Both sets average to blind search performance.

    So your reduction actually didn’t improve search performance, so is irrelevant in terms of informational cost. We’re only concerned with reductions that improve the lower-level search performance over blind search.

    Further, you propose that we can create an artificially inflated space which isn’t implicitly defined, but is created to cause a problem. This may be true (I’ll have to over your reasoning again), but you ignored my point that we could do this with any probability calculation, and arbitrarily inflate the search space (see my coin example in the previous post.) if you point were valid, we’d never be able to calculate infomration from any probability, since we can always say “You’re ignoring the reduction of uncertainty from the contrived/inflated space to the ‘real’ space, which is infinite.” Remember, Shannon information relies on the probability of receiving a set of characters to calculate information content…so if your point is valid, you’ve “falsified” not just Active Information, but information theory itself.

    So your point probably isn’t valid.

    Atom

  179. jerry:

    The active information is the map and boat in the treasure hunt. In evolutionary biology it is ??? My guess it is intelligence. Because intelligence can certainly do it just as it provides the map and boat for the treasure hunters.

    But if an intelligent entity makes a map based on pre-existing information, she is simply “shuffling around pre-existing information” as Marks and Dembski say, which unintelligent entities are also capable of doing. And I know of no evidence that an intelligent entity can make a map to a treasure without having pre-existing information regarding the location of the treasure.

    If anyone thinks that intelligence does have that capability, I’ll reissue a challenge: Can someone use their intelligence to find a 32-character (capital letters and spaces) string with an MD5 hash of cb6ba5a8daf75b7d50fef95cecae78d7?

    If intelligent agents create active info, as Marks and Dembski claim, then they can only do so by luck, not by any inherent ability to create active info. Active info is defined as better-than-random probability of something occurring successfully. If there is a better-than-random probability that active info will be created, then that active info already exists, and its so-called creation is actually a mere shuffling of it. Therefore, it’s self-contradictory to speak of something that has a better-than-random chance of successfully creating active info.

  180. Atom, we’ve definitely talked past each other, and I’m not sure how to get back on the same page. In [157] you described a higher-level space of just fitness functions, not fitness functions and strategies. I had assumed that an evolutionary strategy was a given, and the higher-level search consisted of finding a good fitness function. My X+Y example describes a higher-order space of functions only, not strategies, so there is no reduction of strategies involved. And since the strategy is a given, there isn’t any NFL-like comparison of different strategies’ performance over all fitness functions.

    I’ll have to wait until my next comment to address the question of whether this issue applies to all probability calculations and to Shannon info.

  181. Atom:

    BTW, we can always arbitrarily inflate any probability calculation (”what are the chances of getting heads on a coin flip…with that particular coin out of all the coins of history?”), so if your objection were viable, we could never meaningfully calculate the probability of anything.

    But that’s not inflating a probability calculation, it’s two different probabilities, i.e. the probability of flipping a heads given that I have a fair coin and I flip it, and the probability of flipping a heads with that particular coin, where me having that coin is not a given.

    But elaborating on your point, and hopefully not distorting it too much, let’s imagine two different conversations.

    Conversation #1.

    Atom: 1982 pennies are slightly biased. If I flip this 1982 penny, there is a 51% chance that it will come up heads.

    R0b: But what are the odds of this being a 1982 penny? Assuming that you got it from my coin jar, the odds are 1 out of 367. (I fastidiously keep track of the dates on all of the coins my jar.) Or assuming that you got it from a penny collection that has 1 penny from every year from 1970 to 2009, the odds are 1 in 40. Or assuming–

    Atom: Uh, I gotta go.

    Conversation #2.

    Dawkins: This WEASEL search has a much better probability of success than random sampling. There’s a 50% chance that it will find the target in 40 generations.

    Marks and Dembski: But what are the odds of it being that efficient? Assuming that you chose the fitness function from all functions that map 28-letter sequences to an 8-bit fitness value, then the odds are 10^-3742. Or assuming that you chose the search from all possible searches uniformly distributed according to the distribution they confer on the search space, the odds are 10^-8930. Or assuming–

    Dawkins: Uh, I gotta go.

    The point being that the information cost depends on what we assume about the origin of the search. If the LCI doesn’t hold for all possible assumptions, which it doesn’t, then it should place explicit restrictions on those assumptions.

  182. Atom, with regards to Shannon information:

    Shannon info is a relative measure based on epistemic probability. It measures the reduction of uncertainty in the receiver, so it’s explicitly relative to the receiver’s prior knowledge.

    The active info framework, on the other hand, attempts to provide absolute measures by regressing probabilities to a point of no prior conditions. But it can’t do so, because an ultimate, unconditional search space is an undefined search space, and there’s no way to derive probabilities from it.

  183. Joseph (#153):

    You don’t need Dembski to respond.

    All you need to do is take something that is alleged to have active information and show it can arise via nature, operating freely.

    I could be wrong here, but IF I read Dembski and Marks correctly, then if active information was found to arise “via nature, operating freely”, then this would, in fact, have been caused by intelligence smuggling the information in somehow.

    IF this is the case, then I guess that the oft repeated argument that ID would be falsified if natural processes could produce CSI is wrong.

  184. R0b,

    I agree that there may have been some talking past points. I’ll try to get this back on track.

    I thought you had expanded the higher order space to include different search strategies (this wasn’t clear to me), so if I mischaracterized your argument, I apologize.

    Let us assume an evolutionary strategy is a given.

    We will further assume a base search space, which will be the permutation space of all 24 letter long base 27 (per our alphabet) strings. We agreed this space had 27^24 elements.

    Now our evolutionary strategy will have to use a fitness function (following the standard implementation of an evolutionary search). What is the search space of possible fitness functions?

    First, we’d want to use deterministic functions that assign only one fitness value to each element in our original space. Furthermore, we want to limit the number of fitness functions, which we will do in two ways. First, we limit the function to only take as inputs the elements of our original search space (in other words, the domain is all x such that x is a permutation in our original set.) Secondly, we will limit the possible output values of the function to integers between 0 and n, so that our search space becomes well defined. This I will label Reduction A.

    Given this set-up, we can now calculate the informational cost of choosing one fitness function from that new set in a straight forward manner. (Call this Reduction B.)

    But the question becomes, if I understand you correctly, why do we include the informational costs of Reduction B and not of Reduction A (which is infinite)? More importantly, if we can ignore the informational cost of Reduction A, why can’t we also ignore the cost of Reduction B?

    If this is not you position, then please clarify, because I have misunderstood you.

    If so, I will reiterate my earlier response. Reduction A is the reduction from the set of all possible fitness functions (setting n to ∞, effectively) which as you correctly point out is a reduction of an infinite set of possibilities to a finite set, which would incur an infinite informational cost.

    But as I correctly pointed out, Reduction A does not improve search performance over blind search. Showing this is easy. Imagine you perform your search using fitness function 1 of your reduced set (assuming that you can order the fitness functions in our reduced set, which you can), then use fitness function 2, then fitness function 3, etc, until you’ve performed the same search using all of the fitness functions. You then average the performance of all the functions and will find that your evolutionary strategy performed only as well as blind search.

    So if Reduction A results in a subset that still only performs as well as blind search, either a) the reduction didn’t improve search performance, and so incurs no informational cost, or b) it did improve search performance, which means that the original set of all possible functions (the set prior to Reduction A, which is infinite) somehow performs worse than blind search. But since that set includes all possible fitness functions, it will perform as well as blind search, per the NFL theorems. (I could be wrong on this point, since my understanding of the NFL isn’t as strong as some of the other commenters, but I think the NFL would apply in this case as well.)

    So if Reduction A didn’t improve search performance, it is irrelevant to our calculation. We can go further than Dembski (I believe) and define our higher order search baseline as the smallest set that 1. assigns a value to each and every permutation in the original space and 2. still, when averaged, performs only as well as blind search. That would be an objective baseline to measure our subsequent reductions from.

    If you find this disagreeable, then please show a reduction from a set that performs as well as blind search to one that performs better, and show how this reduction does not incur an informational cost of at least the active information.

    Atom

    PS You are correct in your point on Shannon info about reduction in receiver uncertainty. However, I don’t think you understood my larger point, being that we could inflate any uncertainty/probability/information measure, even that of an observer, by including irrelevant reductions. (What about the reduction for the receiver to limit him to his current state, from all possible messages he could have expected, to just a few? We only consider the issue from a baseline, being defined for us in the Shannon case as the receiver’s current state of uncertainty, but implicitly defined for us in the Dembski case, using the criterion I outlined.) Regardless, it was a side issue which isn’t necessary to understanding my argument and I brought it up only as a way of hopefully getting you to see my original point. I will drop it.

  185. Addendum,

    I should make explicit that I’m not considering the trivial set of only one fitness function that assigns the same value to all permutations as the baseline set, or for that manner, any set that is smaller than our reduced set. For a reduction to make sense, the reduced set needs to be a subset of the baseline set. Sorry I didn’t spell that out explicitly.

  186. 186

    Atom (and R0b),

    It seems that the key claim is that a regress of mechanism gets materialistic explanation nowhere in accounting for active information because there are at least as many alternatives in a higher-order space of material configurations as in the lower-order space. Dembski and Marks seem not to object to treating the known universe as a finite computing machine, and I’m going to proceed more or less along those lines. A huge fraction of alternatives we can allude to in mathematics have no physical realization simply because they require excessive resources to “fit in the universe.”

    In the third theorem, there are

    (M + 1) ^ K

    fitness functions, where K is the size of the base-level search space Omega. For any binary representation (e.g., the machine language of the computer you’re using), almost all fitness functions have no description of length much less than K log (M + 1). Even though Dembski and Marks indicate that M is large, I set M = 1 for simplicity. Now the typical fitness function requires K bits to describe.

    As Dembski and Marks observe, if Omega is the set of all length-100 sentences over a 20-amino-acid alphabet, then K is about 10^130. But Seth Lloyd estimates that the observed universe registers at most 10^120 bits of information. The upshot is that if the entire known universe were searching a space of descriptions of fitness functions, only a minuscule fraction of the descriptions would be sufficiently compact to arise:

    1 / 2^10000000000 [10 zeros].

    I have to note the absurdity of this scenario. We are within the universe, and to posit the existence of an entity that can observe a succession of states of the universe begs the question of the existence of a supernatural entity. Similarly, we cannot regard the evolution of the universe as a search process. We cannot frame the universe that has in its unfolding included us as an alternative to a null universe. There is no way to assign a physical probability to the universe. Thus we cannot associate active information with the universe. Some fitness functions are physically possible, and others are not — and you cannot attribute the mere existence of physical constraints to intelligence.

  187. Dr. English,

    Thank you for your contribution. I feel that your exasperation, however, has led you to make a couple leaps towards the end of your comment.

    You wrote:

    We are within the universe, and to posit the existence of an entity that can observe a succession of states of the universe begs the question of the existence of a supernatural entity.

    Whoa whoa whoa. No one I’m aware of was discussing a “supernatural entity” nor assuming one. Demsbki and Marks paper is about the mathematics underlying conservation of information; to begin discussing metaphysical interpretations is beyond this thread, as the contents of the paper itself have barely begun to be discussed.

    The universe can only instantiate at most a fraction of the total number of possible fitness functions, which is correct. So a reduction has already taken place due to the physical constraints. But this reduction, in as much at it improves the performance of our original search, would incur an information cost of at least the active information, if the math in the paper holds. You have not criticized the math, only its application, so I’ll assume it does hold.

    Now, you may argue “You cannot calculate this informational cost, since we don’t know the ‘probability’ of the universe.” It is true that we don’t know the probability of the universe. But the paper also provides a measure theoretic version of the theorem, which would apply if the probability distribution differed substantially from the uniform in a way that eventually assisted our lowest level search. (Even if the constraints are necessary, that is probability of 1 for that one state and zero for the others.)

    In short, you’d have a search-for-a-search-for-a-search. The universe would be assigned a (non-?)uniform probability (reduction 1, measure theoretic version) for assigning the set of possible fitness functions (reduction 2, measure theoretic version), from which we choose our actual fitness function (reduction 3, fitness theoretic version.)

    Unless I’m missing something (which is always a possibility) the LCI would seem to also hold vertically, for your tri-layered search.

    Atom

  188. Atom, I think we’re getting pretty close to the same page.

    I think a major discrepancy in our thinking is your association of information cost with performance averaged over all of the functions in the higher-order space. For instance:

    So if Reduction A results in a subset that still only performs as well as blind search, either a) the reduction didn’t improve search performance, and so incurs no informational cost

    [Emphasis mine]

    One counterintuitive aspect of Marks and Dembski’s framework is that information cost is not based on the average performance of elements in the higher-order search space. Rather, it’s based on the fraction of those elements that perform at a level of at least q. Information cost does not tell us whether the average performance of the higher-order space is better or worse than the null search. It only tells us what the odds are of randomly selecting a search that performs at least as well as the given alternate search.

    Consider that the set of functions that indicate proximity to a target performs no better on average than the larger set mentioned in endnote 49, i.e. they both perform on average the same as the null search. Yet Marks and Dembski say that the reduction from the latter to the former entails a heavy information cost.

    More later, probably after Mothers’ Day.

  189. Dr. English, an addendum,

    I have been thinking about my response and wanted to make a distinction. When I say we could possibly deal with the physical constraints (which I referred to as a form of “necessity”), what I meant was physical necessity, given the number of particles in the universe. I don’t want this confused with logical necessity, which wouldn’t make sense to treat as contingent (obviously, by definition).

    I just wanted to make sure I was clear on that point. Given that there is no logically reason we’re aware of that the universe has this number of particles, which causes a reduction to take place, then measuring a cost on that reduction could be meaningful (via the tri-level search outlined above.) If however there is a logical necessity to that number of particles, the reduction requires no explanation, as necessary entities are their own explanation.

    Atom

  190. “no logically reason”* => “no logical reason”

  191. 191

    Atom,

    You dropped my qualifier in “physical probability.” See more on this in the new thread Bill started.

    I’m guessing that you, like me, are more engineer than philosopher. I accuse myself of a serious error in neglecting computational complexity in my investigation of NFL. Dembski and Marks are making the same error in focusing entirely on information costs. There are huge distinctions in search programs when time and memory are limited. I don’t have to go with Seth Lloyd in saying that the universe literally is a computer to say that there are analogous distinctions in nature.

    This discussion has turned interesting at just the wrong time. I really need to put on the blinders and deal with the end-of-semester drudge work.

  192. Atom,

    Hopefully we’ve gotten past the confusion about average performance vs. information cost. I can’t remember my train of thought from a few days ago, so I’ll just reiterate the point that you’re disputing:

    1. Information cost depends on the definition of the higher-order search space.

    2. We can define the higher-order search space to contain only good searches, thus making the information cost zero and falsifying the LCI.

    3. In response to the objection that this higher-order search space must incur an information cost from an even higher-order search space, we can point out that this is true for all search spaces that have a non-zero probability of yielding a good search. If the LCI requires us to regress probabilities all the way up, then we’re stuck with an infinite information cost in every case.

  193. R0b wrote:

    One counterintuitive aspect of Marks and Dembski’s framework is that information cost is not based on the average performance of elements in the higher-order search space. Rather, it’s based on the fraction of those elements that perform at a level of at least q. Information cost does not tell us whether the average performance of the higher-order space is better or worse than the null search. It only tells us what the odds are of randomly selecting a search that performs at least as well as the given alternate search.

    R0b,

    You’ve almost got it. I didn’t say that the average performance of the higher level search was used to calculate the incurred cost, only that it can be used as an objective basis for deciding which informational costs are relevant, and hence, must be accounted for. It also provides a handy method for setting an objective baseline for for the higher level informational cost measure.

    My reply has been consistent and I fail to see any issue with using the method I outlined to define the higher order space in a non-ad hoc way.

    Atom

  194. R0b wrote in the next post:

    1. Information cost [of the higher order reduction] depends on the definition of the higher-order search space.

    Correct and agreed.

    2. We can define the higher-order search space to contain only good searches, thus making the information cost zero and falsifying the LCI.

    No we can’t, since doing so would result in a search performance on the lower level search. If a reduction leads to search performance on the lower level search, then we cannot ignore that cost. If it leads to no search improvement (and no hinderance, since we can contribute negative active information), then we can ignore it.

    3. In response to the objection that this higher-order search space must incur an information cost from an even higher-order search space, we can point out that this is true for all search spaces that have a non-zero probability of yielding a good search. If the LCI requires us to regress probabilities all the way up, then we’re stuck with an infinite information cost in every case.

    Either that, or a source that can generate information without relying on search spaces. But this is a side issue.

    Atom

  195. ” in a search performance on the lower level search ” = ” in improved search performance on the lower level search”

    Sorry, I type too fast sometimes.

    Atom

  196. Atom,

    Okay, I think I’ve finally got it. Sorry it took so long to sink in. I think your idea for defining a higher-order baseline is a good one, but I don’t believe it works with Marks and Dembski’s framework.

    First of all, back in [168] where I agreed with your point about all algorithms performing equally over the whole set of fitness functions, I was wrong. Marks and Dembski’s model is not, in general, NFL-compatible. The problem is that Wolpert and Macready define the goodness of a search in terms of the codomain of the fitness function, but Marks and Dembski define the target independent of the fitness function, as Tom English pointed out above.

    Consider an algorithm that finds the WEASEL target with the following logic: It randomly selects points in the search space until it finds a point whose fitness plus the number of the query is even. In other words, if it’s the 3rd query and the fitness is 127, then the condition is satisfied. After finding such a point, it immediately goes to “METHINKS IT IS LIKE A WEASEL”.

    No matter what fitness function we use, this algorithm will likely find the target within a few queries. So how do we apply your condition that the higher-order space of fitness functions must have the same average performance as the null search?

    I think that coming up with generally applicable constraints on the higher-order space definition is harder than meets the eye. As it says in the paper, the ways to search and to metasearch are endlessly varied, and the higher-order space definition can include or exclude any aspect of any conceivable search.

  197. R0b,

    Thank you for the reply. You wrote:

    Consider an algorithm that finds the WEASEL target with the following logic: It randomly selects points in the search space until it finds a point whose fitness plus the number of the query is even. In other words, if it’s the 3rd query and the fitness is 127, then the condition is satisfied. After finding such a point, it immediately goes to “METHINKS IT IS LIKE A WEASEL”.

    This strategy would no longer be using a standard evolutionary strategy, which could find different targets simply by using different fitness functions, but would constitute a new search strategy/algorithm. We could also say “What about an algorithm that simply tries one query, no matter what the fitness function, then goes to the target?” or any other variation of that. But these are different search strategies, so the fitness function method I outlined isn’t directly applicable, since they aren’t really evolutionary strategies in the normal sense of the word.

    However, your example wouldn’t escape the LCI.

    Going with your new set-up, we can see that there exists a similar set-up for every target in your lower level search space: for example, it could go to “Meblinks it is like a weasel” after satisfying the condition. So why did we choose the one algorithm that goes to our target rather than to “Meblinks…”, “Rethinks…”, “hstjdins…” or any other of the 10^40 permutation choices we have?

    More importantly, what is the minimum informational cost incurred by going from the set of all such algorithms (bounded by our original search space) to the set that chooses “Methinks…” with the same efficiency as the algorithm you constructed?

    As I mentioned, the “goto” target of your algorithm could have been any of the roughly 10^40 permutations in the original search space, so we have at least 10^40 algorithms to choose from. The search for your particular algorithm (or one that performs equivalently well) is as hard as, and likely much harder, than our original search.

    The LCI still holds.

    Atom

Leave a Reply