How do evolutionary processes create information?

| 69 Comments

It seems that ID has chosen to rekindle the ‘how does evolution create information’ question. See for instance “Richard Dawkins on the Origin of Genetic Information” at EvolutionNews.org where spokesperson Luskin presents this question. And yet, the question has been answered many times, so why are ID activist ignoring these explanations or pretending that it has not been answered succinctly and successfully?

One of the basic claims of ID is that processes of regularity and chance cannot create complex specified information. ID relies here on an equivocation of the term ‘information’ since ID’s definition of information is merely a measure of our inability to explain it. In other words, unlike the complexity and information that science can explain, ID relies on that which science cannot explain (yet?) and calls it complexity or information.

Confused? I bet… Many ID proponents have similarly fallen victim to the bait and switch approach here.

So whenever ID states that science cannot explain complex specified information, all one has to do is point out the tautological nature of the claim. When ID then switches to the more common definition of information and complexity, it is trivial to show how evolutionary processes can indeed generate in principle information and complexity.

The real question then becomes: Where these processes indeed involved in the evolution of life on earth? While science provides a rich framework to study these questions, ID is left at the sidelines, unable to contribute anything relevant since it refuses to constrain its designer, it refuses to provide pathways and processes.

And remember, whenever science proposes a pathway, all ID can do is reject a strawman version of it, namely a pathways based on pure chance. Of course, any non trivial scientific pathway is inaccessible to the calculations needed by ID to make its case.

Back to the question of information and complexity. How does science explain it? Not surprisingly via very simple processes of regularity and chance: namely selection and variation. As many have shown, these simple processes are sufficient to explain the information in the genome. So now the question is not “how does science explain information in the genome” but “how well do science’s explanations perform”? For that we have to take existing genetic data and determine actual pathways. This historic reconstruction is not simple, although there now exist a handful of examples where science has indeed reconstructed the pathways, consistent with evolutionary theory. ID may of course argue that science still has not provided all the answers, but the mere fact that contrary to ID’s predictions of an Edge, science finds why evolution succeeded.

A good example comes from the work on evolvability and RNA. Contrary to ID’s predictions, RNA shows scale free networks, which themselves can be explained by simple processes of gene duplication and preferential attachment. These scale free networks provide a rich environment for evolution to succeed since it both contributes to the robustness as well as the evolvability of RNA. The reason is that most RNA structures are close to most other RNA structures in sequence space. In other words, most any RNA structure can, via mutations in its sequence, reach any other RNA structure where most of the mutations are in fact neutral. Such findings help understand why evolution appears to proceed in stasis followed by rapid changes. This is exactly what the evidence suggests and the work on RNA has explained this evidence.

So perhaps ID proponents can help us understand how ID explains the origin of information in the genome? But it is unlikely that we will here any further details on this matter. ID has chosen to remain scientifically vacuous

Dembski Wrote:

As for your example, I’m not going to take the bait. You’re asking me to play a game: “Provide as much detail in terms of possible causal mechanisms for your ID position as I do for my Darwinian position.” ID is not a mechanistic theory, and it’s not ID’s task to match your pathetic level of detail in telling mechanistic stories. If ID is correct and an intelligence is responsible and indispensable for certain structures, then it makes no sense to try to ape your method of connecting the dots. True, there may be dots to be connected. But there may also be fundamental discontinuities, and with IC systems that is what ID is discovering.”

Finally, I would like to remind the reader that even if ID were correct that evolutionary algorithms cannot do better than random search, random search is an almost trivially effective search See for instance this link

Tom English Wrote:

The obvious interpretation of “no free lunch” is that no optimizer is faster, in general, than any other. This misses some very important aspects of the result, however. One might conclude that all of the optimizers are slow, because none is faster than enumeration. And one might also conclude that the unavoidable slowness derives from the perverse difficulty of the uniform distribution of test functions. Both of these conclusions would be wrong. If the distribution of functions is uniform, the optimizer’s best-so-far value is the maximum of n realizations of a uniform random variable. The probability that all n values are in the lower q fraction of the codomain is p = qn. Exploring n = log2 p points makes the probability p that all values are in the lower q fraction. Table 1 shows n for several values of q and p. It is astonishing that in 99.99% of trials a value better than 99.999% of those in the codomain is obtained with fewer than one million evaluations. This is an average over all functions, of course. It bears mention that one of them has only the worst codomain value in its range, and another has only the best codomain value in its range.

69 Comments

Dembski is of course correct, in that we have no clue how Dembski’s god magicks reality into being, and no clue how to GET a clue. You either accept or reject that he’s on the right track; you can’t check it out. Those who reject Dembski’s view of how his god does it can’t in principle be answered through scientific means, nor is there any serious effort to do so. They must be answered through political and administrative means, as has traditionally been the practice in theocratic systems.

The DI’s claim that magic is science probably isn’t taken at face value even by Johnson or Luskin. Explicitly, the goal is to position magic as science with the minimum level of verisimilitude to provide creationist school boards, judges, and ideally legislators and Presidents with a nominally plausible rationale for using civil authority to impose and enforce religious doctrine.

And thus the wedge: If only we can get government’s stamp of scientific approval on creationism, we can use the pervasive authority of the State to indoctrinate children young enough so that verisimilitude is no longer necessary. The critical mass of creationism-supporting voters is already there across much of the US, we have a born-again President who’s been stuffing the Supreme Court with religious lackwits; the wedge is working.

The effort to protect the word “information” from the DI’s carefully orchestrated semantic void is certainly worthy. As Orwell taught us so well, we can’t think straight when the words we think with have been hijacked. Yet I don’t think the average Kansas-style voter really much cares what “information” means in any rigorous or technical sense. It’s just WORDS they can use to justify convictions that the educational system, in order to minimize hassles and complaints, sidestepped around correcting.

I’m not qualified to evaluate the technical meanings of “information” or “scale free networks” or “the lower q fraction of the codomain” (huh?), nor am I inclined to make what I recognize would be the significant effort of reaching that point. I either accept that evolution is creative as I understand creativity, or that it is not. If I start with the unquestionable conviction that goddidit, then evolution didn’t do it. Now, do I want my child to have a solid technical understanding of science, or do I want my child to get into heaven? The DI’s overriding goal is to convince me that these choices are mutually exclusive, at least until we can get Jesus back into science classes where he belongs (and get anti-Jesus science OUT).

I note that Tom English (or I assume it to be him) is a member of the Evolutionary Informatics lab:

http://www.evolutionaryinformatics.org/

I was wondering if he’d turned IDist or whether or not the lab is more than simply a bit of ID propaganda.

These people wouldn’t know the technical definition of information from a hieroglyph. If they knew the first thing about Shannon’s Information Theory they’d know that it is precisely random processes that are the sources of information - the more random the process, the more information it generates.

So, in short, the information comes from mutations and the close match of genome to environment, IDers so called “specification,” comes from natural selection.

There, I explained it in one sentence with two paragraphs of background - the second paragraph being the unwritten one about natural selection.

The adaptive immune system is a straightforward example of biology creating information. Vertebrates can make antibodies to a vast array of substances, including novel chemical compounds. Not only does the immune system create immense variation, but it does so an a time scale of weeks. This is a time scale that even creationists can comprehend.

Interesting, Tom English is an affiliate indeed, a recent addition

Thomas M. English, Ph.D. Bounded Theoretics Lubbock, Texas 79410 USA Thom.English(at)gmail(dot).com

English is no friend of ID though.

Surely there’s more than one Tom English.

Not one which has authored the same papers…

Although there are two Tom English’s According the UcD Thomas D and Thomas M English

Thomas M English is the one I was referring to

Tom M English Wrote:

My vita is available, as it has been, at my web site. I am not going to counter with a biosketch — I hate the things. But I will hint at why my work is more relevant to ID than is that of a creationist interested in biodiversity.

In 1996, six years prior to the release of Bill Dembski’s No Free Lunch, I argued that “no free lunch” in search is a consequence of conservation of (Shannon) information. This should have a familiar ring for many of you here. I also established that, contrary to intuition, optimization is easy under the assumptions of the “no free lunch” theorems. It took some time for IDists to catch on to that — there are still some who have not. (By the way, this was the first theoretical paper I ever wrote, and it is far from my best work on “no free lunch.”)

I gave well-publicized tutorials at major conferences when Bill was completing NFL:

• “Introduction to ‘No Free Lunch’ in Optimization and Learning,” 2001 Congress on Evolutionary Computation (CEC2001), Seoul, 2001 • “No Free Lunch: A Gentle Introduction to Conservation in Optimization andLearning,” Sixth International Conference on Parallel Problem Solving from Nature (PPSN VI), Paris, 2000

In 2000, I showed that almost all fitness functions are algorithmically random. This is definitely something that IDists need to comprehend, but that most have not. Bill seems to deal with it now by stipulating fitness functions “of interest” in his work. Presumably a function is of interest only if it is compressible.

I hope IDists will understand that what drew me into the debate over ID was an overlap in Bill Dembski’s interests and my own. I am more comfortable with the topics of search, information, and complexity than are most IDists. In fact, I have spent considerable time studying Bill Dembski’s arguments, and I would be highly surprised if Thomas D. English can relate them more accurately than I can.

As quoted from the same UcD thread above

I chatted with a Tom English last year regarding Dembski’s website. I don’t know what his middle initial is, though. I just emailed him about the situation.

The problem, as I see it, as that the term “information” is applied as an analogy (like “genetic blueprint,” “genetic code,” etc.), yet DNA is NOT information. Yes, we can glean “information,” as humans define it, from the “code,” but it is not information in and of itself. It is a set of chemical reactions that have been honed by natural selection into recurrent processes, just as the principles of physics shape crystalization of water elements into snowflakes. What sets life apart from other physical phenomena is self-reproduction, and we are getting close to understanding possible mechanisms by which that may have happened. Once it is started, there is no need for an informant of the “information” process, because it is not technically “information.”

Steve S suggested that I share what I wrote to him and PvM. I originally planned to silently take the heat for affiliating myself with the EvoInfo lab. I believe that scholars should associate freely, and without explanation. The reason I’m deviating from my plan is that I am appalled to see the mileage the ID movement is getting out of the affair at Baylor. The lab is not doing ID, and I am not an ID convert.

Shall I recite my catechism? A 1987 Supreme Court decision torpedoed instruction in creation science in public-school science classes. An attorney by the name of Johnson salvaged everything in creation science but the “making” and christened the tub intelligent design. Thus from the outset there has been no intellectual legitimacy in the concept of ID. It is simply a lawyer’s strategy for passing legal muster while giving up as little as possible. Of course, there has followed a substantial effort to make it appear that the dogma of a sociopolitical movement actually has intellectual roots. The majority of “scholarly” writing on ID is propaganda.

Tom Wrote:

I intended for all in the know to know I was posting under an alias at PT. I once told Bob Marks I didn’t care to make public statements on the EvoInfo (virtual) lab, and I’m using the alias to stick to what I said, if only legalistically. Please do not make my identity widely known.

The reason I’m posting is that I’m getting pissed, watching the [ID movement] use the affair at Baylor to get ID cast in a sympathetic light in the media. I joined the lab to protest Baylor’s infringement of Bob Marks’ academic freedom. In my opinion, it’s impossible to discriminate against ID, because it’s a sociopolitical instrument that has no intellectual legitimacy.

If you look at the discussion page of the Wikipedia article on “no free lunch theorems,” which I maintain, you’ll see that I decided in June, prior to the controversy, that there was no ID at the EvoInfo lab’s site. My concern at the time was only to avoid bias in the article. It happens that Marks’ definition of evolutionary informatics covers most of my research of the past twelve years, no matter its motivation. It also happens that I was expelled from a Baptist institution 30 years ago for opposing discrimination against women, and I have strongly supported freedom of expression ever since – especially for those I don’t agree with. It seemed to me I was, like it or not, the perfect person to step forward and support Marks. I resisted for a while, worrying about what sort of games I might be sucked into. Now I’m in the difficult position of not only backing a guy everyone I know in the IEEE Computational Intelligence Society regards well, but aiding and abetting the ID movement. I believe that individuals are higher than causes, and academic freedom is a higher cause than opposition to ID, so I’m really just whining about being in a tight spot.

I don’t know the pub date for the book. […] I think I parsed Dembksi’s latest definition of specified complexity more closely than anyone else has. I rewrote his expressions with more explicit notation to expose the details of what he was saying. I also identified some severe computability problems it seems no one else has. I’m not saying I did more than Wes Elsberry and Jeffrey Shallitt did with an earlier version. But I think I’ve been lucky enough to address the final morph of specified complexity. It appears Dembski has given up on the concept.

I have contacted my editor to see if I can share my chapter with you. He’s Down Under, so it may take a while to get in synch.

Thanks Tom. I have been following you over the last few years and I was surprised to hear that you had joined Marks’s lab. Although, once you explained your reasoning, I am accepting your reasoning even though, as you may have noticed by now, ID seems to be getting quite some mileage about how ID is being censored, even though, as you point out, the papers and the work by the lab have little relevance on ID.

The three papers on the site may be seen as arguing against evolutionary processes as a source of information, even though on closer scrutiny, they have far less to argue than one may expect from Dembski’s bragging

WD: It’s too early to tell what the impact of my ideas is on science. To be sure, there has been much talk about my work and many scientists are intrigued (though more are upset and want to destroy it), but so far only a few scientists see how to take these ideas and run with them. There’s a reason for this slow start. My work in The Design Inference was essentially a work on the philosophical foundations of probability theory, trying to understand how to interpret probabilities in certain contexts. This led naturally to some ideas about information and the type of information used in drawing design inferences. My book No Free Lunch was a semi-popular overview of where I saw the ID movement headed on the topic of information. The hard work of developing these ideas into a rigorous information-theoretic formalism for doing science really began only in 2005 with some unpublished papers on the mathematical foundations of intelligent design that appeared on my website (www.designinference.com). With the formation of Baylor’s Evolutionary Informatics Lab just this month and work by me and my colleague Robert Marks on the conservation of information (several papers of which are available at www.evolutionaryinformatics.org), I think ID is finally in a position to challenge certain fundamental assumptions in the natural sciences about the nature and origin of information. This, I believe, will have a large impact on science.

Source

Your work and the work by Marks seem to overlap quite nicely, it’s too bad that circumstances have caused the work to become so controversial.

Btw, should Dembski’s statement be seen as an admission that his work so far has been far from stellar in making the ID argument?

That would be quite some news…

before engaging a creationist in a discussion on information, you need to very clearly ask “by information, do you mean an intended message by an intelligent being”? then beat with a stick until they indicate they are prepared to argue about something sensible

PvM,

Evolutionary processes search for biosystems just as tornadic processes search for trailer parks. In other words, I think it’s a huge mistake to regard evolution as a search process. Search implies a target, and what Marks and Dembski may well show is that it is infeasible for “evolutionary search” to hit certain “biological targets.” That would be fine with me. They would count it as evidence for a supernatural source of information, but others would come to the fore with search-free models of evolution.

As I see it, complex specified information is dead, and active information is its replacement. But I hasten to add that while CSI had serious shortcomings, active information is reasonable and interesting.

I don’t see evolutionary processes as creating information, but as latching it in. I’m going with haploid organisms here. In the stage of evolution known as reproduction-with-variation, the variation is not reasonably attributed to evolutionary processes. The organism “tries” to make a perfect copy of itself, and random errors are caused by, for instance, thermodynamic effects. Errors can either increase or decrease the Kolmogorov complexity (algorithmic information) of the genome. Any change in the information of the genome indicates that there was information in the random errors. I think it’s very important to place the source of information (randomness) outside the evolutionary processes, partially to avoid “creation,” and partially to reflect the fact that evolutionary processes get information “for free.” Incidentally, Paul Davies, commenting on the informational physics of evolution, suggested that evolution is analogous to a Brownian ratchet. I like that idea, though I’m not sure how close the analogy is.

Well, isn’t this thread of to a kick ass start. PvM’s post is thorough, and the comments are excellent.

PvM Wrote:

ID relies here on an equivocation of the term ‘information’ since ID’s definition of information is merely a measure of our inability to explain it.

Acute.

PvM Wrote:

So now the question is not “how does science explain information in the genome” but “how well do science’s explanations perform”? For that we have to take existing genetic data and determine actual pathways.

This addresses so many problems with ID. Information in the genome isn’t interesting for biology in the light of evolutionary mechanisms. (Which unfortunately for ID doesn’t require teleology.)

And when we use a specific information measure to characterize some structure, the question remains if it is descriptive and above all predictive. Is it useful, and how useful is it?

This is also a fundamental difficulty with ID’s idea of criticizing models such as EA’s or biologically inspired EA’s for failures because of “experimenters input of information”. These models are describing and testing predictions. Any artificial source of information (if needed) is a natural part of the experiment, and doesn’t mean they fail as specific tests. If that would be the case, all experiments testing scientific predictions fails.

snaxalotl wrote:

“before engaging a creationist in a discussion on information, you need to very clearly ask “by information, do you mean an intended message by an intelligent being”? then beat with a stick until they indicate they are prepared to argue about something sensible.”

I agree. I once gave a lecture and said something about the information in the genome. Someone raised their hand and asked’ “does that imply intelligence”, meaning of course does that imply an intelligent cause for the information. I replied that there is informatioin in the periodicity of a pulsar but that does not mean that the pulsar is intelligent or that an intelligence was required in order to create it. However, intelligence is required to interpret the information. I don’t know if that is the best answer I could have given, but as you point out, that was certainly the assumption implicit in the question.

Tom,

Thanks for sharing your thoughts with us. I commend you on your courage in defending academic freedom. If you are still reading posts here, perhaps you could give us the benefit of your expertise. What do you think is the best definition of information and what do you think is the best way to measure it?

By the way, I completely agree that evolution is not a “search” in the ordinary sense and that it definately does not involve a “target” in the ordinary sense. Creationists and ID proponents always seem to inject anthropomorphisms into their arguments, I wonder why that is?

Tom English:

I would like to thank you for taking a swift and decisive action, that should quell speculation and rumors. It was also interesting to see your reasoning and concerns.

That doesn’t mean that there doesn’t remain concerns to discuss. (When did it ever? :-P)

Tom English Wrote:

I decided in June, prior to the controversy, that there was no ID at the EvoInfo lab’s site.

I note that Marks and Dembski’s papers criticize evolutionary biological models especially, and that they do so based on ID ideas that the modeled structure “smuggles in” information. As this isn’t a neutral view on algorithms but a specifically chosen example my own decision is different.

That difference in opinion shouldn’t be a problem. After all, we should not want to block Marks or Dembski from presenting their ideas.

Tom English Wrote:

I think it’s very important to place the source of information (randomness) outside the evolutionary processes,

Certainly information (measured as randomness) is created by the environment, but it is also created by biological mechanisms such as sexual recombination or biological processes such as selection or drift (as in both cases traits are fixed in a probabilistic process).

Since information is a property of a system it seems difficult to me to tease out how much is created by engaging with the environment and how much is created by the evolutionary process itself.

Incidentally, discussing analogies, some population models are analogous to bayesian inference models used in machine learning. The populations genomes (the distribution of alleles over the population) is analogous to a learning machine that learns of the environment by trial and error.

[Btw, which knowledge becomes obsolete when the environment changes, so has to learn anew. The natural specification of the object, how the resulting traits work out in fitness terms, to measure (as well as the object itself) changes with the environment. So this isn’t exactly analogous to the minimal description length algorithmic information measure, at least as far as this layman understands.]

Tom English Wrote:

Paul Davies, commenting on the informational physics of evolution, suggested that evolution is analogous to a Brownian ratchet.

I’m not sure what you, or Davies, means here? AFAIK a Brownian ratchet is an unphysical perpetual motion suggestion. How would such an analogy bear on the real world?

I believe there is a typo above.

The probability that all n values are in the lower q fraction of the codomain is p = qn

Unless I am seriously misreading, that should be q to the power of n, not q multiplied by n.

For example, if I choose 3 random variables from a uniform distribution with range of zero to one, the probability that the first one is 1/2 or less is “1/2”.

When I have chosen two, the probability that both are 1/2 or less is, of course, (1/2) * (1/2), or 1/4.

The probability that I will randomly choose three that are each 1/2 or less is 1/2^3, or 1/8.

Multiplying q by n gives us the expected value of the sum of all the random variables, not the probability that all the variables fall within some fraction of the distribution.

infeasible

see, Tom, herein lies the problem.

Neither marks, nor Dembski, nor yourself, nor I know how to define the real world probabilities of hitting a specific “target” wrt to a specific trait.

even knowing ALL the competing selective pressures, even in simple systems like John Endler’s Poecilliids (guppies) in Trinidad, it is still not easy to predict exact probabilities with any confidence. general directions to expect, yes, but exact probabilities? no.

hence the entire argument from “improbability” is completely flawed to begin with.

it really is that simple.

mental masturbation and playing with models is one thing, but what we see in the real world has to fit the models, too.

If the IDists are going to talk about a technical subject WHY DON’T THEY MAKE TECHNICAL DISTINCTIONS?

Information information information, I’m sick to death of information. Is it really that hard to say SEMANTIC INFORMATION versus SHANNON INFORMATION versus.…?

Tom English, I’m glad to hear from you and have some things cleared up. I am surprised to read, though, that imperfect replication of DNA is not part of evolution, or an ‘evolutionary process’ as you put it. I think what you are saying, sans surprising use of terms, is that variation is the source of information rather than selection plus drift. Some would say it is the combination of these, but it is all part of evolution.

Torbjorn the unprintable, you must be thinking of some other Brownian ratchet. The biological ones are not ‘unphysical’. Search molecular motor brownian.

I’m currently reading The Touchstone of Life by Werner R. Loewenstein. From this, I get the impression that evolution doesn’t so much “create” information as reflect self-organization within flows of information. One of his starting points is that information is originally carried in photons and they of course are abundant. If his is the better concept, then the question isn’t so much how information is created but rather how it is captured and how and why the molecular forms which carry it develop and interact the way they do (a question which evolutionary biology addresses and ID doesn’t). I guess I wonder if anybody is familiar with Loewenstein enough to say a) if I’m understanding him correctly and b) if his treatment of information and biological evolution is mainstream and productive of useful lines of research.

Robert Marks’ good past research: some people are perfectly reasonable as long as the subject is not creationism.…

Dembski and Marks at Baylor: it may just be that Baylor and Dembski have a history and Baylor does not want him back, period. He and Marks can do all the research they wish with Dembski affiliated with his Seminary, provided they can figure out anything to do that suits the creationist agenda. All that has happened is that Dembski’s attempt to claim to be from Baylor again, and to have a grant from a regular foundation, did not work. He is a seminarian, and the grant was actually part of his support from Disco. It’s not like Dembski now has no affiliation; he just doesn’t have one to match his ego. Sometimes in life, one just can’t have one’s heart’s desire.

Perhaps what I am going to say here has some relevance to both the question of the connection between ID and Marks’s virtual lab, and to Tom English’s remark about Dembski’s model of evolution as a search algorithm.

In the past I posted at least two entries right here on PT and also on Talk Reason where I argued against Dembski’s model - a search for a small target in a large search space. (See for example here). As usual, Dembski ignored my comments, which is OK - my intention was far from a desire to convince Dembski.

Now about Tom English’s comment (I never knew there are two Tom Englishs: the one I have some knowledge about is imho a highly qualified expert on NFL). Now Tom asserts that Marks is a great scientist and has nothing to do with ID. While Marks indeed may be an excellent scientist, I take the liberty to doubt assertion of his being not in cahoots with Dembski. Here are some facts (besides having Dembski affiliated with Marks’s lab). Some time ago a Swedish mathematician Olle Haggstrom published an article critical of Dembski’s concepts (see here). After a while, a reply to Olle was posted authored by Marks and Dembski (see here). This reply maintains that evolution necessarily has certain intrinsic targets and therefore Dembski’s model is valid. This article is imho unsubstantiated; it brazenly asserts that all Olle’s arguments not only do not disprove Dembski’s thesis but in fact support it. Such a chalenging view of Marks/Dembski is not supported by any substantial arguments but just declared as self-evident. Maybe Marks is great in his field, but his (with Dembski) anti-Haggstrom paper seems to show, first that he shares Dembski’s pro-ID views, and, second, that perhaps he is not really great beyond his field. Just IMHO of course.

All of this has no bearing on Tom’s decision to join Marks’s lab and I wish him success in that endeavor.

I’m glad that discussions on the actual content of Marks and Dembski’s work are starting to pop up. Here are my own beliefs on the EIL work:

1) Casting the concepts in information terms serves only to create confusion. The issues are much clearer when expressed in terms of probabilities. According to Shannon surprisal, which is the information measure that D&M use, D&M’s exogenous information measures the amount of information in the success-or-failure outcome of the baseline search, not the amount of information in the search parameters. Likewise with endogenous information, and active information doesn’t measure the amount of information in anything.

2) The “active information” measure yields a number that has no useful significance. It doesn’t tell us anything of use that we didn’t already have to know in order to calculate it.

3) The “active information” measure is always relative to a somewhat arbitrary baseline. Apparently, the baseline should be a blind search, but over what search space? And using what search structure? For instance, M&D’s response to Schneider discusses two related blind searches, one far more efficient than the other (so they say) because of its search structure. Which of these should we choose as a baseline for Schneider’s ev search?

4) Applying the EIL concepts to evolutionary processes in the real world requires that a target be defined. How do we do this? D&M’s notion of “intrinsic targets” has very obvious logical problems.

As an interesting sidenote, a very obvious discrepancy can be found by comparing Schneider’s Evolution of biological information to M&D’s response. It turns out that the problem is on M&D’s side, and it stems from a bug in one of their MATLAB scripts. The problem is so huge that M&D are going to have to rewrite that paper. 50 panda points for whoever can find the bug.

Perhaps what I am going to say here has some relevance to both the question of the connection between ID and Marks’s virtual lab, and to Tom English’s remark about Dembski’s model of evolution as a search algorithm.

In the past I posted at least two entries right here on PT and also on Talk Reason where I argued against Dembski’s model - a search for a small target in a large search space. (See for example here). As usual, Dembski ignored my comments, which is OK - my intention was far from a desire to convince Dembski.

Now about Tom English’s comment (I never knew there are two Tom Englishs: the one I have some knowledge about is imho a highly qualified expert on NFL). Now Tom asserts that Marks is a great scientist and has nothing to do with ID. While Marks indeed may be an excellent scientist, I take the liberty to doubt assertion of his being not in cahoots with Dembski. Here are some facts (besides having Dembski affiliated with Marks’s lab). Some time ago a Swedish mathematician Olle Haggstrom published an article critical of Dembski’s concepts (see here). After a while, a reply to Olle was posted authored by Marks and Dembski (see here). This reply maintains that evolution necessarily has certain intrinsic targets and therefore Dembski’s model is valid. This article is imho unsubstantiated; it brazenly asserts that all Olle’s arguments not only do not disprove Dembski’s thesis but in fact support it. Such a chalenging view of Marks/Dembski is not supported by any substantial arguments but just declared as self-evident. Maybe Marks is great in his field, but his (with Dembski) anti-Haggstrom paper seems to show, first that he shares Dembski’s pro-ID views, and, second, that perhaps he is not really great beyond his field. Just IMHO of course.

Tom also asserts that, unlike Dembski’s CSI, “active information” (discussed in particular in Marks/Dembski’s anti-Haggstrom article) is a useful and reasonable concept. I agree that this concept as such may be construed as reasonable. However, the question is not whether AI as a concept has contents, but rather whether or not evolutionary algorithms can only succeed if the AI is ether front-loaded or supplied from outside sources. This question is related to both Dembski’s “displacement problem” and the essence of the NFL theorems. Neither Dembski nor Marks offer any evidence that AI necessarily must be added to what they call “endogeneous information.” They simply claim that this is so (which is just another representation of the “displacement problem.”) In fact, as long as we stay within the framework of the NFL theorem, they are only valid for “black box” algorithms which by definition have no access to any information besides that accumulated by the search algorithm in the course of exploration of the fitness landscape and gleaned exclusively from that landscape. They neither possess a front-loaded AI nor receive it from outside during the search. This however does not prevent certain specific algorithms to immencely outperform blind search, which is just a routine occurrence. Therefore all the talk about AI is as irrelevant to biological evolution as the talk about CSI of the NFL theorems.

All of this has no bearing on Tom’s decision to join Marks’s lab and I wish him success in that endeavor.

Neither marks, nor Dembski, nor yourself, nor I know how to define the real world probabilities of hitting a specific “target” wrt to a specific trait.

even knowing ALL the competing selective pressures, even in simple systems like John Endler’s Poecilliids (guppies) in Trinidad, it is still not easy to predict exact probabilities with any confidence. general directions to expect, yes, but exact probabilities? no.

hence the entire argument from “improbability” is completely flawed to begin with.

it really is that simple.

mental masturbation and playing with models is one thing, but what we see in the real world has to fit the models, too.

I have made similar remarks in arguing that specified complexity is generally not computable, and mostly agree with you. But something you might consider is that exact probabilities are not necessarily required. In search, some quantities grow very rapidly and others shrink very rapidly. It is conceivable that someone could establish an inequality on very imprecise quantities that would make a persuasive argument that “evolutionary search” could not have “hit targets” without an extrinsic source of information.

To expand a bit on what I said above, it seems to me that teleology is inherent in the search metaphor. Most of us believe, at least when we stop and think, that it’s just a metaphor, or perhaps that it’s a model that must be taken with a grain of salt, and that evolutionary processes do not have the end of hitting targets. In my opinion, Marks and Dembski take the metaphor literally, and it is important to keep that in mind.

And when we use a specific information measure to characterize some structure, the question remains if it is descriptive and above all predictive. Is it useful, and how useful is it?

You might have a look at Paul Vitanyi’s recent work on the Kolmogorov structure function. It is not only up-to-date, but gives the most accessible treatment of Kolmogorov complexity and the structure function I have ever seen.

Quite a number of researchers have looked at the Kolmogorov complexity of genomes. Perhaps this measure of information is not relevant to all investigation, but it certainly is to some.

Apology for multiple posts - it was some glitch in PT’s software, not due to my intention.

Tom English:

Thanks for the kind reply!

Tom English Wrote:

It seems to me that ontogeny is much like a chain of behaviors extended through trial and error.

That is certainly an idea along the lines I was discussing. The model is discussed here.

As Dawkins has noted somewhere, the genome still bears on memories from distant times (Hox genes et cetera). Ontogeny is such a slow changing expression of the genome.

Of course, as always evolution goes both ways. Parasites often evolve simplified traits because their hosts provides a stable environment.

Henry J:

Henry J Wrote:

I.e., that line of “argument” supports common ancestry

Indeed. Actually, the whole a priori probability vs a posteriori likelihood confusion that creationists indulge in is really an example of where scientists instead find support.

If any one outcome of evolution is unlikely, we also know that this improves the precision in the likelihood methods that are used to find the phylogenetic trees. (The subset of likely trees will decrease.) And which incidentally confirms the prediction of nested hierarchies in evolutionary theory.

Returning to the post and the problems with ID, this further illustrates the depth of ignorance and stupidity in creationism. They suspect and deride the very methods that all scientists, not only biologists, use to arrive at knowledge. All the while pretending that they are interested in scientific support.

Vile creatures. Who ordered them? :-P

fnxtr Wrote:

That does put an interesting perspective on it, doesn’t it. ID says, “Look how unlikely life is!”. The fossil record says, “No kidding: almost everything goes extinct.” Where’s the teleological ‘design’ in that? Unlike the human world, biological history is not written by the winners.

Indeed! And the relative proportion of complex organisms that have existed and gone extinct may give some indication about probabilities of what kinds of life are most likely to exist (e.g., bacteria, or beetles). The more complicated and delicate the life form, the larger the flow of energy required in sustaining it (big entropy producers). These may be the ones that may be ultimately the hardest to sustain because they exist in states that are the least probable. Doesn’t bode well for humans if true.

Torbjörn Larsson Wrote:

Hmm. IMHO it doesn’t work out directly, since the mechanisms including fixation under selection is varying, in effect the same problem that the original ratchet had.

No, you were right the first time. The more realistic “ratchet” has other degrees of freedom, and may begin walking in a different direction if selection conditions change.

Torbjörn Larsson Wrote:

Hmm. IMHO it doesn’t work out directly, since the mechanisms including fixation under selection is varying, in effect the same problem that the original ratchet had.

No, you were right the first time. The more realistic “ratchet” has other degrees of freedom, and may begin walking in a different direction if selection conditions change.

Mike Elzinga:

Perhaps, but I think that depends on the time scales and other details of the involved processes such as fixation. It isn’t obvious (at least to me) which is why I said “[not] directly”.

Torbjörn Larsson Wrote:

but I think that depends on the time scales and other details of the involved processes such as fixation.

I am not clear in my mind about the meaning of “fixation”, but certainly at the molecular level, things can happen very quickly (dt = dp/F), because the forces are large relative to the changes in momentum.

But more to the point, it is the flow of energy due to temperature differences (kinetic energy per degree of freedom) that is important at any scale. Damping (as in the pawl of the ratchet and pawl motor) is simply another way of saying that energy flows out of the pawl allowing it to latch and not bounce around randomly due to hits from every direction. If energy can flow out of the pawl, there is a temperature difference (net transfer of momentum in a given direction), and the ratchet and pawl can walk uphill.

Mike Elzinga said: I am not clear in my mind about the meaning of “fixation”, but certainly at the molecular level, things can happen very quickly (dt = dp/F), because the forces are large relative to the changes in momentum.

IIRC, Fixation is one allele for a gene replacing all the other alleles that the species used to have for that gene.

Henry

PvM said: Interesting, Tom English is an affiliate indeed, a recent addition .

Thomas M. English, Ph.D. Bounded Theoretics Lubbock, Texas 79410 USA Thom.English(at)gmail(dot).com

English is no friend of ID though.

As an ID’er I am very happy to see that Tom English has joined the Marks Informatics group. I would certainly consider him a friend. I have very much appreciated Tom’s critical contributions thus far and I remember him from the ARN board. Being a “friend” in science does not necessarily mean not being a critic. In science, critics and proponents work together.

Hi Tom,

Thanks for the link to the very good Grunwald Vityani paper.

About this Entry

This page contains a single entry by PvM published on September 16, 2007 12:41 PM.

Changes are a-commin’ was the previous entry in this blog.

Today’s Bob Jones “Biology for Christian Schools” Howlers is the next entry in this blog.

Find recent content on the main index or look in the archives to find all content.

Categories

Archives

Author Archives

Powered by Movable Type 4.381

Site Meter