Criticisms of Dembski’s No Free Lunch go unnoticed … again and again

| 139 Comments

Over at Uncommon Descent, Eric Holloway has declared that the critics of William Dembski’s 2002 book No Free Lunch actually accept that the No Free Lunch Theorem applies to evolution. He uses as his evidence the replies to Dembski’s use of the NFLT by Allen Orr and by David Wolpert (who co-wrote the original NFL paper). They had argued that evolution was a more complicated process than the simple model used in the NFLT, a model that for evolution would associate fitnesses with genotypes in a simple search for the genotype of highest fitness. So aren’t computer scientists (Wolpert) and biologists (Orr) implicitly acknowledging that the NFLT theorem applies to any such simple model, and prevents it from searching effectively?

But there have been other criticisms of Dembski’s use of the NFLT, and Holloway does not cite them. I summarized them in a 2007 article I wrote in Reports of the National Center for Science Education. And in the matter of the use of the NFLT my criticisms were actually not new — as I noted there, the fundamental point had been made many times since 2002, originally in a 2002 article by Richard Wein, and also in articles by Jason Rosenhouse (2002), Mark Perakh (2003), Jeffrey Shallit and Wesley Elsberry (2004), Erik Tellgren (2005), and Olle Häggström (2007). I will immodestly claim that my article is the clearest of these many clear articles.

So what is this oft-repeated criticism? When we have a simple model of evolution with genotypes and phenotypes, the NFLT argues that if we average over all the ways that set of fitnesses could be associated with the genotypes, that a simple model of search that climbs uphill on the fitness surface cannot do any better than a random search by pure mutation (one which is unaided by natural selection). That is disastrously bad. It sounds like it says that natural selection in such a model cannot work at all.

But notice the averaging part. It is critical to Wolpert and Macready’s theorem. In effect, it says that we are dealing with an infinitely rough fitness surface. If we change a genotype by making one mutation — changing a single position in its DNA — we arrive at a genotype whose fitness is randomly chosen from the whole set of possible fitnesses. In effect, a single mutation has the same effect as mutating every site in the genome simultaneously. (I apologize for shouting, but the point is not being noticed over at UD).

Of course real biology doesn’t work like this. Mutations are on average worse, but they mostly don’t instantly reduce the organism to rubble. In the real world, nearby genotypes are usually similar in fitness — often a bit worse but sometimes a bit better. In the NFLT world essentially all mutations are disastrous, and evolution would not work. So the No Free Lunch Theorem does not model real biology, not even in a simple model of evolution searching for genotypes of higher fitness on a fitness surface.

So far Holloway has not cited any of these criticisms, and when asked by a polite commenter whether there are any such criticisms, he has simply declared that

I spent some time reading the critics, and this bore [sic] my frustration. I could not find one author who treated Dembski’s work fairly! If someone could fairly refute Dembski’s work I’d be all over it, but I haven’t found anyone! Instead its all passive aggressive ad homineum [sic] and brow beating, with ample burning of strawmen, very tiring to read.

So the discussion at UD continues, hermetically sealed in a self-reinforcing bubble (though I notice now that in that discussion Elizabeth Liddle has tried to raise the relevant point).

Note added 8/29/2011: Eric Holloway has now replied to this post in a post he made recently at Uncommon Descent. For my response to this reply, see the two comments I have made below dated 8/29/2011 at 1:17am and the one following that.

139 Comments

I’m not entirely following all of the NFLT argument for ID because it screams “red herring” on the face of it, and it is uninteresting to work to learn something just to confirm that yes, it is as bogus as it appears …

… but it seems the root fallacy is like that of a supply officer for a military unit who determines the average size of the troops in the unit, and then obtains uniforms for them all in that one size.

I think it’s the other way around. It’s like a supply officer saying that it is impossible to order uniforms for a unit because the sizes of troops are random.

Of course, one can go out and measure a specific unit and order the uniforms that fit them.

Reed A. Cartwright said:

Of course, one can go out and measure a specific unit and order the uniforms that fit them.

Or even, for that matter, obtain a set of uniforms with quantities declining on both sides of the mean – which is how logistics works in the real world.

“Random does not mean incoherent.”

Well in my army, everyone get hand tailored uniforms because the military-sartorial complex is too powerful in Washington. Gotta spend the tax dollars somewhere.

Reed A. Cartwright said:

Well in my army, everyone get hand tailored uniforms because the military-sartorial complex is too powerful in Washington. Gotta spend the tax dollars somewhere.

Huh? Tailored fatigues? Was this the same Army I was in? Reminds me of the graffiti I saw in a latrine in Basic at Ford Ord: “If my mother wanted me to be a soldier, I would have been born with green baggy skin.”

But that may have been the era before starch went away.

So what has Dembski said in reply to these criticisms?

He says: “I don’t have to match your pathetic level of detail.”

I think the whole point of that article on UD was to make that ONE quotemine of Wolpert.

Uncommon Descent used to be amusing because of the unstable nature of Dembski and DaveScot (maybe I’m crazy, but I always wondered if the latter was Berlinski - but of course, extreme arrogance, foul temper, and claims of mathematical genius are a common trait cluster).

Now it’s just a rubber room for a tiny cult of not-very-bright deniers. I doubt if there’s much of a recruitment rate.

Possibly relevant - there has been some evidence of increased acceptance of evolution in polls since 2005.

I’ve often complained about polls that bias the question with implications that accepting evolution contradicts religion. This latest Gallup poll simply used a “yes, no, don’t know” structure. The response of science supporters has been, understandably, to complain about the mere 39% “yes” rate, but look at the rest. Only 25% outright deny evolution. I personally see it as a sign of honesty that 36% admit they “don’t know”. Also, when broken down by education, the “yes” to “don’t know” ratio increases, which suggests honest answers. I’d certainly like to see more “yes”, but a 39:25 “yes/no” ratio is probably an improvement over past levels. http://www.gallup.com/poll/114544/d[…]olution.aspx

Also, FASEB did a poll, the raw results of which I can’t find, in 2008. They claim a 61% rate of acceptance of evolution among Americans. That poll was apparently structured in the biasing “did God create humans of did they evolve” format, but nevertheless, “only evolution” and “evolution guide by God” apparently combined to 61%. I need to find the raw questions and how the sampling was done, but that seems like an improvement over similar polls of the past. Caveat - FASEB is a “pro-science” source, of course.

Obviously, we can’t be sure whether any of this is related to ID. But one possible tentative conclusion is that the whole ID scheme backfired. Bringing evolution into the media again and provoking scientists to make strong rebuttals mildly to moderately increased public understanding and acceptance of evolution. As I said, this is just a tentative, hypothetical idea, but it’s intuitively credible.

In particular, it fits with my personal experience from the early heydays of ID - when I tried to actually explain ID and counter it with an explanation of evolution to actual neutral people from non-science backgrounds, they saw the logical flaws in ID much more quickly than I expected (which is evidence of how dishonest and/or conflicted creationist trolls are).

harold said: Uncommon Descent used to be amusing because of the unstable nature of Dembski and DaveScot (maybe I’m crazy, but I always wondered if the latter was Berlinski - but of course, extreme arrogance, foul temper, and claims of mathematical genius are a common trait cluster).

“Mad yes, scientist no.” There was something flamboyant and outrageous about Dembski. ID now seems to be dominated by the Denyse O’Luskin crowd, who are simply dull and whiny.

Eric Holloway once said this:

Interestingly, Kolmogrov [sic] complexity is uncomputable [sic] in the general case due to the halting problem. This means that in general no algorithm can generate orderliness more often than is statistically expected to show up by chance. Hence, if some entity is capable of generating orderliness more often than statistically predicted, it must be capabable [sic], at least to some extent, of solving the halting problem.

It’s like gibberish for weasels. And GodEntity can make immovable rocks He can move. At least to some extent. And “Kolmogrov” is spelled that way for many paragraphs.

https://me.yahoo.com/a/Qjp5enRppdSW[…]SLhOc-#9be9d said:

Eric Holloway once said this:

Interestingly, Kolmogrov [sic] complexity is uncomputable [sic] in the general case due to the halting problem.

AIUI, “in the general case” is the same flaw (or directly analogous to) the one Joe cited above.

Kolmogorov - in reference to a single, actual programming language (rather than the infinite variety of possible programming languages), this problem disappears.

NFLT - in reference to a single, actual real world (rather than some average over “all the ways” fitnesses could be associated with genotypes), this problem disappears.

In effect, a single mutation has the same effect as mutating every site in the genome simultaneously. (I apologize for shouting, but the point is not being noticed over at UD).

IIRC, DaveScot got this point, and commented somewhere that he had told Dembski that this pushes ID back to cosmological ID.

Of course, Dave is a UD outcast too, nowadays.

rni.boh said: IIRC, DaveScot got this point, and commented somewhere that he had told Dembski that this pushes ID back to cosmological ID.

Of all the lunacies of creationists, one of the most baffling is the insistence that the Big G had to have created the Universe, but also has to keep on tinkering with it on an ongoing basis (or put otherwise, the Big G could not have “intelligently designed” evolution).

They will go to every mad length to maintain this logical disconnect, since if that barrier falls they become TEs. “And this would be bad because why?” The only reason I can think of for playing this game is to protect scriptural literalism, but I’m unsure if that’s the only motive.

RodW said: So what has Dembski said in reply to these criticisms?

Over the years, basically nothing, even when he wrote The Design Revolution which had as its subtitle “Answering the Toughest Questions About Intelligent Design”. But he didn’t answer that criticism.

But he very quietly, implicitly, acknowledges it by bringing forth his (and Robert Marks’s) Search For A Search argument: if fitness surfaces are smooth (so that natural selection will work) this is supposedly because they are “front-loaded” to allow this. My own reaction (see here) is that the weakness of interactions at a distance in physics is sufficient to make fitness surfaces smooth. For example a gene affecting my eye pigment typically does not interact strongly with one that affects the length of my big toe, and that is not evidence that some Designer has been tinkering around.

The only reason I can think of for playing this game is to protect scriptural literalism, but I’m unsure if that’s the only motive.

I suggest This as the ultimate objective.

It’s like gibberish for weasels.

The word “like” is superfluous here.

Of course, Dave is a UD outcast too, nowadays.

He was, uproariously, banned from UD in 2009. Apparently his head didn’t explode after Dover, despite his prediction that the school board would win the case because Judge Jones is an ostensible “good ol’ boy appointed by George W. hisself”. (A prediction which unwittingly concedes that the creationists could only win via unethical judicial bias, but incorrectly predicts such bias.)

Not that I pay much attention, but that particular username seems to have disappeared after the UD banning. He may still be out there. He may still be out there, babbling on UD in a more compliant manner, under a new username.

The only reason I can think of for playing this game is to protect scriptural literalism, but I’m unsure if that’s the only motive

Yes, of course that is the only direct motive.

Creationism in public schools has been a religious right obsession since the dawn of America’s decline into a second rate right wing dystopia.

The point of creationism is authoritarianism. Economic decline for most people and a fair degree of blocking the progress of ethnic minorities has been achieved, but pesky women haven’t given up their new rights, and gays have made progress. To justify denying rights to relatively popular and affluent groups, some claim that “you can’t support that even if you think you agree with it” has to be made. You can’t “interpret” the Bible because some other guy can “interpret” it the opposite way. So you have to claim it’s “literally true”. But since Genesis is at odds with modern science, to claim that, you have to claim that modern science is wrong. Hence “creation science”.

But “creation science 1.0” was thrown out in courts, due to occasional respect for America’s vestigial constitution.

ID always was nothing but a “court proof” version of “creation science”. It’s sole content can be summarized as this - “Forget the evidence, here’s a crazy reason why ‘evolution can’t be true’; so what else happened, well, ‘the designer could be an alien, judge, wink, wink’, but of course, kids, it could also be right wing Jesus - you decide kids, which makes more sense, ‘an alien’, or Jesus?”

It was always just “creation science 2.0, now with more weasel words!”.

harold said: Yes, of course that is the only direct motive.

I have this suspicion that there are other motives lurking in there. There are creationists who are not fundies; they are rare, of course, and since it is hard to understand why anyone who wasn’t a fundy would want to be a creationist, they have to be far crazier than anyone could imagine.

Joe Felsenstein Wrote:

So the No Free Lunch Theorem does not model real biology, not even in a simple model of evolution searching for genotypes of higher fitness on a fitness surface.

I would suggest that it doesn’t even model the formation of Jell-O or rock salt.

Mike Elzinga said:

Joe Felsenstein Wrote:

So the No Free Lunch Theorem does not model real biology, not even in a simple model of evolution searching for genotypes of higher fitness on a fitness surface.

I would suggest that it doesn’t even model the formation of Jell-O or rock salt.

The conditions of the NFLT may not model much of anything. But the simple model of genotypes with constant fitnesses is worth investigating (I know you didn’t say it wasn’t). Basically the NFL argument of Dembski was saying that natural selection could not work even in that simple model.

The Orr-Wolpert Defense is basically irrelevant to that point – we have to look at that simple model and see whether natural selection can work there. And in fact, once we look, we see that NS does work as the NFLT imposes an unreal condition on the way genotypes affect phenotypes. In cases with smoother fitness surfaces NS works fine.

Looking over at UD, I thought Elizabeth Liddle was going to point this out – but she diverted to the Orr-Wolpert Defense, alas. So it’s still true that no one over there has answered – or raised –the criticisms mentioned here.

In addition to Dr. Felsenstein’s smoothness argument, it seems to me.….… The peaks in biology are not independent of the organisms’ presence on them, but exhibit non-linear self-feedback. One way to get off being “trapped on a fitness peak” is to have lots of nearly identical offspring. Should they compete heavily with each other in the same niche, propagate disease, etc., average fitness will go down. Their (or some eventual even higher populated generation’s) offspring can then wander around the fitness landscape again after the peak has been “leveled”.

mrg said:

harold said: Yes, of course that is the only direct motive.

I have this suspicion that there are other motives lurking in there. There are creationists who are not fundies; they are rare, of course, and since it is hard to understand why anyone who wasn’t a fundy would want to be a creationist, they have to be far crazier than anyone could imagine.

Technically the point is not that they themselves are fundamentalist, but that they seek to make others fundamentalist, so that unpopular authoritarian policies can be pushed through on “Biblical” grounds.

A Straussian desire to make the population appropriately God-fearing and science denying so that the correct right wing policies can be enforced tends to explain the rare oddballs like Berlinski. I suspect it explains Behe has well.

I challenge you to name even one evolution denier who will openly deny the possibility that the earth could be 6000 years old. Anyone else who thinks they can find an OEC openly stating that the earth is definitively more than 6000 years old is welcome to take up this challenge.

They fall in three clear groups - Openly claim that the earth is 6000 years old (common), privately claim that the earth is 6000 years old when in the right company but try to dissemble in public (common), or consistently claim that they “don’t know” (rare but extant). There may or may not be super-rare members of a fourth category (crackpots who deny young earth openly but also deny evolution). I used to think there were such types but they always seem to turn out to be fundies in the end.

harold said:

Technically the point is not that they themselves are fundamentalist, but that they seek to make others fundamentalist, so that unpopular authoritarian policies can be pushed through on “Biblical” grounds.

You are assigning to them greater logical capability than they possess.

I think I figured out what my puzzlement on the issue of this logical disconnect is: they’ve been telling themselves this isn’t about religion for so long that they honestly believe it, and when confronted with a gap in their thinking that apparently can only be plugged by “scriptural literalism”, they simply pretend the gap isn’t there. It isn’t to fool anyone else, since it doesn’t; it is just to fool themselves.

I see these people as far crazier than you do. If you disagree, all I can say is I hope you’re right, but I don’t believe you are.

https://me.yahoo.com/a/Qjp5enRppdSW[…]SLhOc-#9be9d said:

In addition to Dr. Felsenstein’s smoothness argument, it seems to me.….… The peaks in biology are not independent of the organisms’ presence on them, but exhibit non-linear self-feedback. One way to get off being “trapped on a fitness peak” is to have lots of nearly identical offspring. Should they compete heavily with each other in the same niche, propagate disease, etc., average fitness will go down. Their (or some eventual even higher populated generation’s) offspring can then wander around the fitness landscape again after the peak has been “leveled”.

I think absolute fitness will go down, the question is whether relative fitness of different genotypes changes. This isn’t obvious – depends on how you have the ecology set up.

The NFL argument applies to constant relative fitnesses. To refute it within that model you need to stay within that model and stick to that case.

mrg -

I do assign them more “logic”, in the sense of being consciously or unconsciously able to pursue their authoritarian goals, than some others do.

It’s also true that the human mind is a complex thing.

If a used car salesman is trying to sell someone a car, a very concrete interpretation would be that the salesman is an idiot who sincerely believes everything he says about a lousy car. A slightly more sophisticated interpretation is that he’s a conscious dastardly schemer who is secretly chuckling, and that’s sometimes true. But perhaps more often, even while being driven by his desire to sell the car, he begins to believe his own lines. Nevertheless, he wouldn’t be issuing them if it weren’t for his desire to sell the car.

What authoritarians with sexual, gender, and ethnic obsessions are trying to sell is more complex and abstract, but it’s kind of the same.

Anybody who thinks lunch is free hasn’t been to a fast food place lately. :p

Joe Felsenstein said:

The Orr-Wolpert Defense is basically irrelevant to that point – we have to look at that simple model and see whether natural selection can work there. And in fact, once we look, we see that NS does work as the NFLT imposes an unreal condition on the way genotypes affect phenotypes. In cases with smoother fitness surfaces NS works fine.

I think your point about the smoothness of fitness landscapes is a crucial one; and I think it is more than metaphorical that it has its anchor in fundamental physics.

Deep down in the quantum world of atoms combining with atoms, we may see things snapping into configurations determined by the rules of quantum mechanics. These tend to be at energies on the order of an electron volt or more. Changes at these levels are going to be influenced by energetic interactions with photons of at least that energy.

However, once into the levels of complexity of solids and liquids and above – especially with what physicists refer to as “soft” systems (i.e. systems whose binding energies are comparable to the energies in the thermal baths in which they are immersed) – we almost never see things that snap from one configuration to another; something what would be suggestive of sharp-edged potential wells. If they do snap, there are wells involved that are considerably deeper than the kinetic energies in the ambient environment; and these wells have smooth edges.

As we enter the realms of mesoscopic physics and macroscopic physics in these soft systems; and as these systems interact with the myriad of potential wells in their environment, smooth differentiable potential wells are the rule. Binding energies for the kinds of living systems we know about are on the order of a few hundredths to as much as a tenth of an eV. Action potentials are on the order of a tenth of an eV or less. Soft systems automatically imply smooth wells. You just can’t have a soft system with sharp wells; it would be either all locked together or all apart.

Nothing at these levels is “jagged.” And even though fitness landscapes are not potential energy wells turned upside down, I suspect they are a phenomenological manifestation of the smoothness of the underlying physics (I don’t say chemistry at this point because traditionally we associate chemistry with the combinations of atoms and molecules at energies on the order of an eV and where quantum jumping results in “snapping together or snapping apart,”).

So whatever quantum jumps that may take place in the chemical bonds that make up DNA and other underlying templates, by the time their impact on the higher levels of complexity are felt, it is pretty much all smoothness and only occasional snapping and “bumpiness.”

Dembski, et. al. continue to make the fundamental mistake that atoms and molecules are just lying around to be picked up randomly and plopped into specified arrangements. All of Dembski’s sampling examples expose this misconception about chemistry and physics.

It’s not sequences of ones and zeroes in complex systems. It’s generally a “bumpy slip-slide” in conjunction with an environment that preserves some configurations long enough to replicate themselves into subsequent generations. Soft systems in particular don’t generally behave like ones and zeroes.

mrg said:

AITSE – clearly NCSE turned inside out.

I see Holloway is a computer science masters student now in the Air Force, went to college at Biola University (surprise!). He worked on evolutionary algorithms for his MSc.

In his posting at UD he keeps calling for experts in computer science to defend evolution. I’m not sure that makes sense. Wolpert and Macready’s original NFL theorem is true for general searches, no one is arguing it isn’t. Once you get past that, the issue is whether it applies to biological evolution, and computer scientists won’t know as much about that as mathematical and computational biologists, particularly population geneticists.

Kevin B said:

The bit that seems to get missed with the Weasel program is that there is, in fact, two programs; the “single-stage selection” one that Dawkins only discussed (because it would run for a time that is large relative to the age of the Universe), and the “cumulative selection” one that will complete in the course of a lunchtime, even in interpretive BASIC on an early 1980s Apple microcomputer. It is the comparison of the two programs that is the point that is (perhaps deliberately) glossed over by the Creationist critics.

Dawkins abstracted out one element of the theoretical basis of evolution and devised an experiment to investigate it. This is how science is done, and anyone who does not understand how the Weasel program is merely a small part, rather than the whole, has no right to call themselves a scientist.

There are a number of easily visualized examples from physics that demonstrate how one can map potential wells with random sampling. Probably the most well-known analog example is the sprinkling of iron filings onto a sheet of paper or a clear plastic sheet placed over a magnet.

This illustration carries an important idea that such wells can be quite accurately mapped by noting the probability that one finds a “sampling particle” within a given region of space; and, indeed, that is one of the techniques used by Monte Carlo techniques or other kinds of analog techniques for mapping fields.

There are other techniques that fall under the category of “relaxation methods” that vary parts of a system locally in order to minimize both globally and locally some field such that there is self-consistency that changes less and less as the computing process proceeds.

Hartree-Fock and mean-field computations, for example, converge quite well; and these all fold in fundamental principles that find usually maxima or minima of something like energy or internal stresses or some other physical variable.

In things like the ultrasonic focusing algorithms I wrote many years ago, nothing has to be known about what amplitudes are in any given reflected waveform that has been digitized and stored according to the position from which it was retrieved. One simply scans over the set taking waveforms from within a synthesized aperture and shifting them as a lens would do according to where the waveform appears within the aperture. Any reflected amplitudes in those waveforms that add up coherently then produce a focused image. You don’t have to know what is there; the process automatically pulls it out if it is there in the waveforms.

All of these processes that are run on a computer are simulations of what we understand nature to do.

The thing that impressed me about Dawkins’s Weasel program was its pedagogical simplicity in demonstrating a specific mechanism. At the time I read The Blind Watchmaker, I didn’t have access to a computer where I could play with it. By the time I actually got a computer that I didn’t have to share with everybody else, I had forgotten about Weasel and was doing other things.

Looking back, I regret having forgotten about it, because with a slight change in perspective, it also illustrates some important ideas in physics and chemistry; and I could have made good use of it. I have since run this program on an HP48/49/50 series graphing calculator, and it works just fine; that’s how simple it is.

Let’s take a step back here and look at what is really going on.

1) There is abundant evidence for biological evolution, from multiple fields of inquiry, all of which converges on the same conclusion. This evidence includes numerous incomplete but valid computer science and mathematical models of evolution or aspects of evolution.

2) Dembski starts with the pre-conceived goal of “proving evolution to be false”.

3) Yet he does not familiarize himself with the relevant evidence, let alone address it. Immediately, his work is suspect on this grounds alone. We may compare him to someone seeking theoretical proof, for example, that Jared Lee Loughner did not did not engage is a shooting spree at an appearance by Representative Gabrielle Giffords. Even if he comes up with something plausible, how does he explain away the strong evidence in favor of that which he declares “theoretically impossible”?

4) We are spared the dilemma, fortunately, of a conflict between convincing evidence for the occurrence of evolution, and a compelling theoretical argument that the evidence cannot exist. Dembski does not provide a theoretical argument. In the case under discussion here, he misrepresents the NFL theorem as being a model which “theoretically disproves” biological evolution. However, the reaction of computer scientists, mathematicians, and biologists, including the many scholars cross-trained in two or all three of those fields, is pretty much unanimous. All neutral observers simply agree that Dembski has misapplied and misrepresented NFL theorem.

5) Eric Holloway engages in odd behavior. I think any reasonable observer would agree that the following are true of Eric Holloway - a) He starts with a rigid predisposition to “disprove” biological evolution. b) He ignores all relevant evidence. c) In fact he actively hides from relevant evidence. d) He issues the patently false claim that Dembski’s interpretation of NFL has not been adequately refuted - yet he does so from a venue where he can control feedback, and when feedback to the contrary comes in, he refuses to acknowledge it.

harold said:

Let’s take a step back here and look at what is really going on.

5) Eric Holloway engages in odd behavior. I think any reasonable observer would agree that the following are true of Eric Holloway - a) He starts with a rigid predisposition to “disprove” biological evolution.

He would argue that he’s not rigid and is open to counter-arguments. For example on UD he declared:

Eric Holloway: So, I spent some time reading the critics, and this bore [sic] my frustration. I could not find one author who treated Dembski’s work fairly! If someone could fairly refute Dembski’s work I’d be all over it, but I haven’t found anyone! Instead it’s all passive aggressive ad homineum [sic] and brow beating, with ample burning of strawmen, very tiring to read.

I suspect that his self-image is of someone who is trying to fairly assess the evidence. We might disagree given that

harold: b) He ignores all relevant evidence. c) In fact he actively hides from relevant evidence. d) He issues the patently false claim that Dembski’s interpretation of NFL has not been adequately refuted - yet he does so from a venue where he can control feedback,

Agreed so far.

harold: and when feedback to the contrary comes in, he refuses to acknowledge it.

Well, we don’t know for sure that any feedback has come in to that thread at UD that conveys the main criticism that has been made of Dembski’s NFL argument. Elizabeth Liddle sounded like she was about to do this when she said:

Elizabeth Liddle: The NFL theorems would only apply to evolution if we also considered solutions to the problems of survival that are unconnected with each other. Evolutionary algorithms are far better than blind search algorithms at finding connected solutions, and connected solutions is exactly what they find.

… but then she went on to revert to the Orr-Wolpert Defense of “evolution is more complicated than this simple model”.

Of course the ID types are strangely unable to see the chief criticisms of Dembski’s NFL argument when they’re right before their nose. But a little of the blame attaches to commenters who attack Dembski’s NFL argument by raising all sorts of side issues rather than taking the trouble to understand the main issue – which is that the random association of fitnesses with genotypes produces a situation where one single base change is as bad for the organism as changing every base in the genome. And real biology deals with much smoother fitness surfaces than that.

No one actually raised the main criticism over in UD (except for this brief allusion by Liddle) and if Holloway is unable to find Panda’s Thumb he could claim innocence. Sort of.

Joe Felsenstein said:

No one actually raised the main criticism over in UD (except for this brief allusion by Liddle) and if Holloway is unable to find Panda’s Thumb he could claim innocence. Sort of.

There have been a few times that I have gone over to UD and tried my best to get into the minds of the people making arguments there. I have done this mostly for selfish pedagogical reasons in my attempts to grasp the essence of misconceptions about science.

But I can’t do it for very long over there at UD; it makes me nauseous.

In fact, if any of them could find Panda’s Thumb, or any other textbook that explains scientific concepts, it is highly unlikely they could derive any benefit from it. Their wiring is just too screwed up; it has to be torn out and entirely replaced.

Mike Elzinga said:

But I can’t do it for very long over there at UD; it makes me nauseous.

In fact, if any of them could find Panda’s Thumb, or any other textbook that explains scientific concepts, it is highly unlikely they could derive any benefit from it. Their wiring is just too screwed up; it has to be torn out and entirely replaced.

Threats of violence, eh? ;-)

More seriously, I just reread Elizabeth Liddle’s comments and have to retract something – she never alluded to the main (Richard Wein, Jason Rosenhouse, et al.) criticism of Dembski’s NFL. I misread her remarks which alluded to the coevolution criticism of Orr and Wolpert. So there is a 100% record at UD of opponents of the NFL argument not raising the right points.

harold said:

Let’s take a step back here and look at what is really going on.

Another way of putting this is that the game he is playing is: “You can show you are right in practice but I can prove you are wrong in theory.”

I’ve thrown this at creationists a dozen times and all I get is “dumb looks are still free”.

Joe Felsenstein said:

Threats of violence, eh? ;-)

Maybe a less threatening way of saying that would be: “We can’t patch up the system – we’ll have to install a new OS from scratch.”

Oh here we go. Eric Holloway has posted a long rant essay on the Broader Implications of ID.

It seems it validates everything in a laundry-list of political and social opinions of his. First, it proves that Keynesian economic theory is wrong …

Joe Felsenstein said:

Oh here we go. Eric Holloway has posted a long rant essay on the Broader Implications of ID.

Oh, groan; Aristotle’s (384 – 322 B.C.) efficient and final causes!

Joe Felsenstein said: Oh here we go. Eric Holloway has posted a long rant essay on the Broader Implications of ID.

I just glanced at it and the pomposity of it blasted me right off the page. You could float an airship on that class of rhetoric.

It’s another proof of the principle that, while creationists can say things that sound like they make sense, the longer any one of them talks the more obvious it becomes he’s riding the crazy train.

mrg said:

I just glanced at it and the pomposity of it blasted me right off the page. You could float an airship on that class of rhetoric.

It’s another proof of the principle that, while creationists can say things that sound like they make sense, the longer any one of them talks the more obvious it becomes he’s riding the crazy train.

It is not surprising that ID/creationists retain much of the distorted remnants of early Greek philosophy. The medieval church adopted much of what the Muslims had preserved of the Greeks after the Crusades in Spain during the 11th century. Much of the active (re)translation of these works took place during the 12th century.

Holloway’s rant is sophomoric; as though he just discovered an introductory philosophy textbook and thinks he now has deep insight into how the world works.

You folks are missing the profound implications of ID. It makes lots of other scientific activities unnecessary.

Economists: Do you want to know whether Keynesian theory is correct? Don’t study economics, just read William Dembski!

Physicists: Do you want to know how string theory fits into the rest of physics and gibes with physical reality? Stop messing around with those equations, read Denyse O’Leary instead!

Astronomers: Do you want to know whether intelligent life exists elsewhere and has any prospect of communicating with us? Turn off those telescopes and fire up the browser: Uncommon Descent knows the answer!

and of course

Biologists: Do you want to know about the history of life and the mechanisms that led to it? Stop the sequencing machine, and head to the nearest fundamentalist church where all will be revealed.

Joe Felsenstein said:

You folks are missing the profound implications of ID. It makes lots of other scientific activities unnecessary.

Hence my summarizing Intelligent Design as “It’s all too complicated for us, stupid mortal researchers to ever hope to understand, therefore DESIGNERDIDIT

Eric Holloway has finally responded to the Original Post in a new post at UD.

He cites David Wolpert as saying that the NFL applies to a simple model of evolution. He sees me as saying it does not apply to that model because “the relevant fitness landscape for evolution is not under the domain of the NFL.”

Let me be even clearer. In a simple model of uphill search on a fitness surface NFLT applies. But what does it prove? It proves that this uphill search is on average ineffective, on average when averaged over all possible ways you could associate the fitnesses with the genotypes.

A typical one of these random associations of fitnesses with genotypes is a completely “jaggy” fitness surface. On it, a single base change at a single site carries you to a genotype that has a totally different fitness (randomly-drawn from all possible fitnesses). Now if you instead change all bases in the genome, simultaneously, you also get a randomly drawn fitness. So the two kinds of change should have the same average effect on fitness.

Holloway tries to argue that biology shows that fitness effects of mutations are indeed disastrous. Well, they aren’t perfect but they sure aren’t that disastrous!

So the fitness surfaces we actually have are not at all typical of the ones that contribute the overwhelmingly to the NFLT average. They are much better for natural selection, but that is swamped out in the NFLT average by all the jaggy (“white noise”) fitness surfaces.

This being the case, Dembski’s use of the NFLT does not have the effect of showing that natural selection cannot achieve substantial adaptation.

Holloway has a response for that. See the following comment.

Eric Holloway says that if my point about the invalidity of Dembski’s NFL argument is correct,

In this case Dembski would be indeed wrong about the applicability of the NFL. However, given the high specificity of such a landscape, this would mean that evolution itself is intelligently designed to an extraordinary degree.

That is equivalent to the Dembski-Marks “Search For A Search” argument. As I have argued in posts here and here, if fitness surfaces are far smoother than a random association of fitnesses with genotypes (and they are) that does not prove that any Designer has intervened to do that. Mere physical reality could do that – the weakness of long-range interactions.

Holloway argued in his response to me that Kolmorogov Complexity shows that jaggy fitness surfaces will be overwhelmingly common. I don’t think it shows that because I don’t think there is any proof (or even good argument) that physical reality allows this. If I type on my keyboard (a small physical perturbation of the keys) there are not strong effects on Holloway’s house – the roof does not start leaking, for example. Yet in a physical world that followed this Kolmogorov Complexity rule everything would interact infinitely strongly with everything. Blatantly, physical reality does not work that way – even without a Designer to smooth things out and calm everybody down.

I have given some common-sense arguments to this effect in my 2007 response to Dembski. I am happy that Holloway is at least trying to respond to that – there was no response to it before now.

One footnote: I am off on vacation for five days with my family in the morning. I will try to respond to any further exchanges before then, but may not be able to do so in a very timely fashion.

Eric Holloway Wrote:

Finally, perhaps Dr. Felsenstein is right after all and evolution does happen to possess the extremely rare and valuable fitness landscape whereby algorithmic search is significantly effective to be worthwhile. In this case Dembski would be indeed wrong about the applicability of the NFL. However, given the high specificity of such a landscape, this would mean that evolution itself is intelligently designed to an extraordinary degree.

The highlighted part is not an argument; it is a bald assertion with nothing to back it up. The laws of physics don’t change just because someone doesn’t understand them and then covers up that lack of understanding by tossing around pseudo-scientific jargon.

Joe Felsenstein said:

It was named by its authors, Wolpert and Macready. They showed that when you averaged over all the ways the values of the objective function could be assigned to points in the space, on average no search strategy would outperform simple random search. So there was No Free Lunch, in the sense that to do well, you had to have something extra that makes the surface smooth. In the case of evolution it is just physics, that makes similar genotypes have similar fitnesses quite often.

They were generalizing about general search spaces and general strategies, and were not trying to cast doubt on evolution. Wolpert was quite angry when he found Dembski using his theorem for that.

Ah. Is this effectively like saying that, given an infinite variety of types of data, no data compression method works any better than any other?

Which is true but irrelevant, since given specific types of data some compression methods work very well and some work poorly. Use the wrong compression method on certain types of inappropriate data and you can in principle get data EXPANSION.

One illustration of this is that for a typical “busy” photo, JPG image compression can on an easy bet beat PNG compression. For a simple block drawing, PNG compression will as an easy bet beat JPG compression. “Use the right tool for da job.”

I recall Mike Elzinga pointing out that physical phenomena don’t generally have flat behavior over their range of operation. Indeed, they can be very selective, even when everyone admits there’s no intelligence running them by remote control.

mrg said:

Ah. Is this effectively like saying that, given an infinite variety of types of data, no data compression method works any better than any other?

Which is true but irrelevant, since given specific types of data some compression methods work very well and some work poorly. Use the wrong compression method on certain types of inappropriate data and you can in principle get data EXPANSION.

Yes, there is a good analogy there to the NFLT. The issue is that real genotypes do not live on a fitness surface that is pure white noise, just as images we want to compress are not pure white noise either.

Joe Felsenstein said: Yes, there is a good analogy there to the NFLT. The issue is that real genotypes do not live on a fitness surface that is pure white noise, just as images we want to compress are not pure white noise either.

Well, they CAN be but they are notoriously hard to compress!

Thanks JF. I didn’t want to wade through figuring out the NFLT because it looked like such a handwaving red herring as far as evosci is concerned and I have lots of other things to learn.

But by that analogy I can see that, yes, evolution wouldn’t work if it was rooted in a completely unpredictable chaotic universe. It isn’t, of course, being based on and constrained by a range of phenomena from the rules of chemical bonding to selection pressures.

mrg said:

Well, they CAN be but they are notoriously hard to compress!

Thanks JF. I didn’t want to wade through figuring out the NFLT because it looked like such a handwaving red herring as far as evosci is concerned and I have lots of other things to learn.

But by that analogy I can see that, yes, evolution wouldn’t work if it was rooted in a completely unpredictable chaotic universe. It isn’t, of course, being based on and constrained by a range of phenomena from the rules of chemical bonding to selection pressures.

There are a couple of other examples that can also illustrate the issue. In nature, the surfaces represented by potential wells, or some other kind of surface used to represent a physical variable, are generally continuous and smooth.

A surface that has random “spikes” or delta functions sticking up is not a physically realizable surface.

With real 3-D surfaces one can have issues with saddle points where a surface can have zero slope in every direction yet be neither a maximum nor a minimum. It gets much more complicated in higher dimensional spaces; but if one has something like fitness that is a measure of “how well one is doing,” one can “feel around” in the space to find where this measure maximizes locally.

The white noise example is also useful. As you know, the Fourier transform of white noise is an infinite spectrum of frequencies. White noise is not band limited. Real noise, on the other hand is band limited; there is an upper limit to the frequencies. That already tells us something about the smoothness of the functions we are dealing with; in other words, that the temporal or spatial distribution of the function does not contain spikes anywhere.

I think that one of the ways one can think of “fitness” is that it is analogous to a rate meter or Geiger counter that varies as it approaches a radioactive source. If an organism is doing a good job of producing surviving offspring, that becomes a measure of how close the organism is to “fitting comfortably” within its environment. In physical terms, the myriad of physical stresses on the organism are, in a sense, minimized locally.

But the physical stresses on an organism are ultimately derived from physical processes described in terms of potential wells; and at the level of a complex living organism, all wells are smooth locally.

Zooming way out, however, might produce the appearance that some of them look like “spikes” pointing downward (fitness spikes upward). I suspect that is also part of the problem that ID/creationists are having with their sampling methods. These sampling methods betray their misunderstanding that all of evolution takes place “locally,” with organisms moving on smooth surfaces to nearby peaks. Current organisms are always built on previous templates that get modified or changed in function incrementally.

About this Entry

This page contains a single entry by Joe Felsenstein published on August 23, 2011 9:06 AM.

Prolific Internet Troll Arrested in Montreal and Under Examination was the previous entry in this blog.

Evolution is Both a Fact and A Theory 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