Thursday, June 23, 2011

The Facepalm of Compensation

Readers here and elsewhere will know that Dr Granville Sewell thinks there is a problem with biology and the Second Law of Thermodynamics. He has written several versions of an argument claiming that the Second Law poses a problem for biology, especially the origin and development of life on Earth.

Dr Sewell is not alone in this concern. Generations of creationists have had this concern. However, the answer given is so obvious that even creationist bastions such as the Institute of Creation Research no longer recommend using this issue in debates. That answer is that SLoT only applies in closed systems, and the Earth is not a closed system. The surface of the planet receives energy from the Sun, energy from its core (from radioactive decay and residual heat of friction) and these sources overcome the trend towards higher entropy.

Dr Sewell has attempted to avoid this answer by arguing in several ways. One is to try to apply SLoT to an open system. Another is to attack the idea of compensation that appears in some elaborations of the answer given above.

Dr Sewell's argument is that even an open system MUST rely on passage through the boundary of anything that that is going to increase inside the system.

If an increase in order is extremely improbable when a system is closed, it is still extremely improbable when the system is open, unless something is entering which makes it NOT extremely improbable.

The above quote is from Dr Sewell's Can ANYTHING Happen in an Open System?

One of the biggest problems with this argument, which Dr Sewell has called his controversial tautology, is that it expands SLoT to cover any diffusion problem at all. We can break this down into two sub-problems, expanding SLoT and treating the issue as a diffusion problem.

Can SLoT be expanded to cover anything beyond thermal entropy? Obviously, Dr Sewell says yes here, and in his invocation of "X-order" in the AML paper. At the same time, in a later article he criticizes Dr Dan Styer for allegedly applying SLoT broadly. He has perhaps learned something, since most scientists would agree that you can't, willy nilly, go applying conservation laws wherever and whenever you feel like it.

Secondly, not all of the universe is a diffusion problem. Let's say that my open system of choice is a crowded bar, and I'm interested in the amount of whiskey in the bar as whiskey diffuses across the boundary I've drawn around the bar. Is the amount of whiskey in the bar solely dependent on the amount crossing the boundary? Obviously not, it also depends on the rate at which it is consumed within the bar, the rate at which sober customers (who are not, themselves, made of whiskey) arrive, and the rate at which inebriated customers (partially made of metabolized whiskey) exit.

What is true of whiskey is also true of cosmic rays, high energy photons, radioactive atoms, and many other things. They can enter through a boundary around an open system, but there are significant transformational processes that can occur within the system as well. So Dr Sewell's controversial tautology is neither controversial nor a tautology. It is simply wrong.

This is why Dr Sewell's argument fails at explaining photosynthesis. Similar to the crowded bar, we draw the boundary around the cell wall of a cyanobacteria. Light enters at one frequency, strikes various molecules, is absorbed, its energy is changed into thermal motion and the potential state of various electrons, sugars are produced and eventually a low energy infra-red photon exits the boundary. Sugar did not enter across the boundary. A low energy photon did not enter across the boundary. The quantity of high energy photons inside the boundary has not increased.

This brings us to compensation. We can say that the exit of the low energy photon "compensates for" the sugar. There is an energy difference between the high energy photon that came in through the boundary, and the sugar molecule. If we add in all the thermal motions and escaped photons, we should be able to make the energy equation balance.

But not the order equation. Even though the sugar molecule has lower entropy, the universe as a whole is worse off.

Dr Sewell seems to think that compensation can happen at a distance. It doesn't. Dr Styer says in his article:

Presumably the entropy of the Earth’s biosphere is indeed decreasing by a tiny amount due to evolution, and the entropy of the cosmic microwave background is increasing by an even greater amount to compensate for that decrease.

Does this mean that Dr Styer is engaged in some magical thinking that life here makes the CMB colder via some spooky action at a distance? No. Dr Styer previously wrote:

The Sun heats the Earth through electromagnetic radiation  largely in the visible and near-infrared bands . The Earth radiates electromagnetic radiation largely in the far-infrared band into outer space, where it eventually joins the cosmic microwave background.

So it is clear that the CMB effect Dr Styer is referring to is based entirely on the passage of sunlight through the biosphere of the Earth. Yes, the CMB observed by someone distant from the Earth will have higher entropy than if the same sunlight had struck a dead planet of the same size and location.

Compensation is not action at a distance. You can always trace the interactions back to the point where one high entropy and one low entropy component were created, and see how the high entropy component escaped the open system. In considering the overall accounting for entropy in the closed system (the Universe) within which our open system is embedded, the escaped component is compensation for the low entropy component it left behind. It is only in this overall perspective that we need to worry about compensation, since it is only in this closed system that we need to be concerned with SLoT.

These misunderstandings and logical fallacies have led Dr Sewell to embarrass himself once again, by writing to the journal that published Dr Styer's article, the American Journal of Physics. In a blog entry on Uncommon Descent, Dr Sewell calls AJP a "major physics journal". In fact, it is a journal for articles related to teaching physics to high school and college students. The rejection message he received makes that clear, as well as making clear the overall high crank science level of his writing.

Monday, June 20, 2011

Model Madness: Axe vs. Lynch and Abegg

I recently blogged about a paper by Lynch and Abegg 2010, which I thought was an important paper. It showed that, yes, there was time enough for evolution, mainly because neutral and even maladaptive variations could accumulate in sufficient numbers until they finally coalesced into a new beneficial function.

Douglas Axe, a leading scientist within the ID community, responded to the challenge inherent in this paper. Axe's position and research agenda has long been that there is not time enough for evolution.

At this point, I have to say that even though I disagree with Dr Axe, I give him credit for being the most professional and rigorous pro-ID scientist I have ever read. Yes, professional scientists can work themselves into a corner that eventually becomes crank science as they refuse to abandon a position - witness the Rubin group at the University of Oregon on "birds are not dinosaurs". Yes, I do think Axe is in this position, but he is trying in a principled, scientific way to address the issues. As such, he has demonstrated far more professional integrity than Stephen Meyer or William Dembski, neither of whom is a scientist.

Axe's response is The Limits of Complex Adaptation: An Analysis Based on a Simple Model of Structured Bacterial Populations, which attempts to criticize Lynch and Abegg and to propose an alternative model.

For now, I'd like to focus on the differences in the models. Lynch and Abegg use a model assuming sexually reproducing diploid populations. Axe tries to refute them with an asexually reproducing haploid island population model. Are those differences appropriate?

One way to answer would be to look at the biochemistry that is being discussed, and ask when did it evolve. Is it eukaryotic or prokaryotic in origin? As I pointed out in my last post on this, eukaryotic sexual species with large populations have been around for a billion years. The larger portion of the biochemistry we operate with is eukaryotic - not shared with bacteria.

That calls into question the basic assumption of Axe's model. Another way to look at the issue is to question the realism of the asexual population genetic abstraction. We now know that in real life, as opposed to the test tube or a mathematical simplification, bacteria exchange genes rapidly. The process might be based in conjugation, or it might be through viral infection. In either case, the binary fission model of where genes come from isn't relevant.

Axe's model also incorporates an effective population size that is quite small - 10^9, a billion bacteria. This is justified by appealing to the island model dynamics of Maruyama and Kimura 1980. Even accepting these dynamics, a gene's eye view of the world has to incorporate the reservoirs afforded by other species and viruses. So I would argue that Axe's effective population size is too small.

Friday, June 10, 2011

New York-New Jersey Trail Conference - You Rock!

I recently took a hike in the Ramapo Reservation without a map, and had a less fun time than I hoped for. The natural surroundings were great, but there are many trails, and I could not find what I was looking for.

To the rescue - trail maps by the NY-NJ Trail Conference. Great color and BW maps printed on Tyvek, so they won't rip. NYNJTC does great work for hikers in the area of New York City. Buy their stuff.

(A shoutout of thanks to Danny Chazin, who has worked tirelessly for TC, the Scouts, and his synagogue over the years.)

Thursday, June 9, 2011

Hypothesis: Reordering typical GA operations opens up new opportunities

One of the truisms of the area of evolutionary computation is that time spent on the evaluation of the fitness function dominates the the total resource budget of the run. Therefore, we should want to allocate trials as efficiently as possible, even more so as we are using a method which will inevitably allocate trials to poor choices as part of the exploration of the parameter space.
When GAs are introduced, the fitness function evaluation is typically a single test case, for example evaluating f(x) for some x, which is the phenotype constructed from the individual's genotype. The population might have genotypes of binary strings, these strings then become phenotypes of real numbers, and the phenotypes are evaluated.

For more 'real world' problems, the phenotype has to be evaluated across multiple test cases, perhaps thousands of cases. The outcome of all of these test cases contributes to the overall fitness measure of the individual. As noted in this early paper, the test cases can vary greatly in their discriminant utility. The basic idea here is to break down the test cases into a population of individuals that will co-evolve with the population of possible solutions.

My idea is somewhat simpler. Typically, one individual is tested across all test cases, and a cumulative fitness score generated. In my reordering of operations, all of the new population members are generated, then each member is evaluated on the first test case, then all on the second, etc. We stop the evaluation process at some point to compute a partial fitness score. On the basis of this score, we abandon some members of the population and delete them, replacing them with perturbations of high scoring members.

In terms of the biological analogy, reordering the test case evaluations allows us to select from the population at multiple points in each individuals "lifetime", where the lifetime is the sequence of test cases.

The hypothesis is that this reordering, partial fitness selection, and reward of well performing members will lead to more efficient allocation of trials in problems that are amenable to this reordering.

Wednesday, June 8, 2011

Time Enough for Evolution: part n++

The Rate of Establishment of Complex Adaptations
Michael Lynch, and Adam Abegg 2010

Several points about this important paper:

  • The concerns of evolution critics are addressed in the scientific literature. The paper mentions the critique of Behe and Snoke (2004) as a motivation.
  • The particular set of population genetic models discussed are based on sexual reproducing populations of diploid chromosomes. We are familiar with sexually reproducing organisms as large plants and animals, and therefore the charts in the article which show effective population sizes up to 10^11 individuals may seem alarming. But consider that there are many single celled sexually reproducing plants, animals, and fungi (protists, generally), and that even one trillion (10^12) eukaryotic cells take up less space than the human body.
  • Single cell protist sex has probably been around for a billion years.
  • The evolutionary features we find most striking are at the edges of the vast conserved biochemistry of life. We look at changes in timing and rate of developmental signals that can change a tapir into a giraffe, but these don't require new biochemistry.
Taking these points together, this paper is an important answer to the 'no time for evolution', bignum crowd.

Granville Sewell's SloT article: Zombie or Persistent Vegetative State?

I've been apprised that the 15 seconds of Internet fame alloted to this blog has arrived, in the form of a link from John West's blog entry over at Evolution News and Views. As welcome (in the 'just spell my name right' way) as that may be, it seems that John and/or lawyer Pete Lepiscopo are confused about the issues.

Here's the text of my letter to Dr Rodin, the editor at AML.

Dr Rodin,

I am appalled to see a preprint, apparently from Applied Mathematical Letters, of the often repeated and often refuted nonsense of Granville Sewell on an anti-science web site.

Dr Sewell, whose expertise lies in partial differential equations, has writen several times on the relevance of the Second Law of Thermodynamics to the topic of evolution. Each time he makes poor arguments that do not show any understanding of the physics or chemistry involved, clearly contradicting the philosophy of your journal.

A concise refutation is

The reputation of AML will be harmed by publishing this article by Sewell.

David vun Kannon

You might notice that I don't make any request of Dr Rodin, specifically, I don't ask him to not publish. Nor do I base my argument on my own reputation in the field of evolution or thermodynamic studies (fields where my reputation is at least equal to Dr Sewell's). Rather, I supply Dr Rodin with a link to an article published in 2008.

American Journal of Physics -- November 2008 -- Volume 76, Issue 11, pp. 1031
Entropy and evolution
Daniel F. StyerDepartment of Physics and Astronomy, Oberlin College, Oberlin, Ohio 44074
Quantitative estimates of the entropy involved in biological evolution demonstrate that there is no conflict between evolution and the second law of thermodynamics. The calculations are elementary and could be used to enliven the thermodynamics portion of a high school or introductory college physics course.
© 2008 American Association of Physics Teachers

The above article is far more persuasive than a letter referencing it. While Sewell's lawyer might think an e-mail from a nobody blogger can move an editor to extraordinary actions, I think the credit fairly belongs with the scientific literature that was cited.

West then follows lawyer Lepiscopo into an odd series of quotes and links. For example, West states:
According to the journal's editorial policies, acceptance of an article cannot be rescinded once an author has been notified of its acceptance, and accepted articles are supposed to be withdrawn only "under exceptional circumstances" such as fraud, errors, ethics violations, and the like.
Um, yeah. Follow that "cannot be rescinded" link and you arrive at a page describing the Elsevier editorial process, which is not relevant to the situation after acceptance. Even so, the page states:
Editors with the appropriate EES permission* can rescind (undo) a decision before or after the Author has been notified, or after the final disposition has been set to Reject.
Notice the "can rescind" verb? West and lawyer Lepiscopo appear to have read that as "cannot rescind" for some reason, perhaps a reason favoring their position.

West and lawyer Lepiscopo continue by noting that errors are a valid reason for withdrawing a paper. Well, that is really the point, isn't it? While people familiar with Sewell's Johnny-one-note attack on evolution know he has been singing the same song for years (as Weseley Elsberry showed), more to the point for AML is that this version of the paper does not address the literature, specifically the 2008 paper cited above.

I am not privy to any agreement, certainly less so than John West seems to be. If Elsevier has made a pragmatic choice to buy off a nuisance lawsuit, I understand it from a business perspective. I hope it might make them review the business process that led to this fiasco, a 'rapid review' editorial workflow. I'm glad to hear that Dr Sewell is welcome to submit future articles to AML. This is a privilege that is shared by most of the population, including myself, John West, and lawyer Lepiscopo.

And what of the paper itself? According to West, it can join several of it's kin on Dr Sewell's web site. It might even say 'accepted by AML' on it. But it remains comatose. If John West wants to see how Dr Sewell and I discuss his ideas, he can look at my comments (under the pen name Nakashima) on the various threads on Uncommon Descent where the same thoughts have been floated in the past.