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

_{William Dembski

May 1, 2009

Evolution, Informatics, Intelligent Design}

Share: Facebook; Twitter; LinkedIn; Flipboard; Print; Email

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.

Comments

Allen, Two noteworthy takes on "meaningful information": P. Vitanyi, 2006, Meaningful Information D. H. Wolpert, 2008, Physical Limits of InferenceT M English_{May 2, 2009
May
05
May
2
02
2009
07:48 PM
7
07
48
PM
PDT}

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?Allen_MacNeill_{May 2, 2009
May
05
May
2
02
2009
07:21 PM
7
07
21
PM
PDT}

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?Allen_MacNeill_{May 2, 2009
May
05
May
2
02
2009
06:56 PM
6
06
56
PM
PDT}

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.tribune7_{May 2, 2009
May
05
May
2
02
2009
05:15 PM
5
05
15
PM
PDT}

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.).anonym_{May 2, 2009
May
05
May
2
02
2009
04:45 PM
4
04
45
PM
PDT}

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?Alan Fox_{May 2, 2009
May
05
May
2
02
2009
04:19 PM
4
04
19
PM
PDT}

...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.
Alan Fox_{May 2, 2009
May
05
May
2
02
2009
04:17 PM
4
04
17
PM
PDT}

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.tribune7_{May 2, 2009
May
05
May
2
02
2009
03:56 PM
3
03
56
PM
PDT}

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.Alan Fox_{May 2, 2009
May
05
May
2
02
2009
03:14 PM
3
03
14
PM
PDT}

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.Alan Fox_{May 2, 2009
May
05
May
2
02
2009
03:06 PM
3
03
06
PM
PDT}

...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.Alan Fox_{May 2, 2009
May
05
May
2
02
2009
03:04 PM
3
03
04
PM
PDT}

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.Alan Fox_{May 2, 2009
May
05
May
2
02
2009
03:01 PM
3
03
01
PM
PDT}

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.tribune7_{May 2, 2009
May
05
May
2
02
2009
02:37 PM
2
02
37
PM
PDT}

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.nullasalus_{May 2, 2009
May
05
May
2
02
2009
02:12 PM
2
02
12
PM
PDT}

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!Alan Fox_{May 2, 2009
May
05
May
2
02
2009
02:03 PM
2
02
03
PM
PDT}

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!Nakashima_{May 2, 2009
May
05
May
2
02
2009
01:48 PM
1
01
48
PM
PDT}

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?tribune7_{May 2, 2009
May
05
May
2
02
2009
01:21 PM
1
01
21
PM
PDT}

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.tribune7_{May 2, 2009
May
05
May
2
02
2009
01:14 PM
1
01
14
PM
PDT}

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. AtomAtom_{May 2, 2009
May
05
May
2
02
2009
12:51 PM
12
12
51
PM
PDT}

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.Nakashima_{May 2, 2009
May
05
May
2
02
2009
12:28 PM
12
12
28
PM
PDT}

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.Nakashima_{May 2, 2009
May
05
May
2
02
2009
12:12 PM
12
12
12
PM
PDT}

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.Alan Fox_{May 2, 2009
May
05
May
2
02
2009
12:02 PM
12
12
02
PM
PDT}

Hmm. One of my comments critical of this "article" appears to have been deleted.AmerikanInKananaskis_{May 2, 2009
May
05
May
2
02
2009
11:43 AM
11
11
43
AM
PDT}

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.Allen_MacNeill_{May 2, 2009
May
05
May
2
02
2009
11:13 AM
11
11
13
AM
PDT}

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. :)Atom_{May 2, 2009
May
05
May
2
02
2009
11:05 AM
11
11
05
AM
PDT}

Please consider my posts at 17 and 19 unnecessary and tangential, and carry on. Sorry.hazel_{May 2, 2009
May
05
May
2
02
2009
10:57 AM
10
10
57
AM
PDT}

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. AtomAtom_{May 2, 2009
May
05
May
2
02
2009
10:54 AM
10
10
54
AM
PDT}

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.Freelurker__{May 2, 2009
May
05
May
2
02
2009
08:41 AM
8
08
41
AM
PDT}

Does not all Christian theology consider nature a divine creation? I'm puzzled by what point I am missing.hazel_{May 2, 2009
May
05
May
2
02
2009
08:07 AM
8
08
07
AM
PDT}

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.Charlie_{May 2, 2009
May
05
May
2
02
2009
07:56 AM
7
07
56
AM
PDT}

Prev 1 … 4 5 6 7 Next

You must be logged in to post a comment.

Leave a Reply