Uncommon Descent Serving The Intelligent Design Community

Turning the 2nd law thermo into a “principle of reasoning”


From Brendon Brewer at Quillette:

I first encountered the second law as a teenager, while reading an issue of the fundamentalist Christian magazine Creation, given to me by my grandmother. Since the article’s author wanted to argue against biological evolution, it claimed that the second law of thermodynamics implies evolution is impossible. Its definition of the second law was that disorder always increases with time. At first glance, this does seem incompatible with evolution by natural selection, which can lead to more complex, “better designed” organisms over time.3 At the time, I thought it was unlikely that mainstream biology would flagrantly contradict mainstream physics, so I remained sceptical of this argument, even though I couldn’t understand the counterarguments I found on the Internet at the time.

As soon as one hears that, be suspicious. The author is saying thathe needs Dawin to be right. Many of these people would trash falsifiability to make Darwin right. Even as top thinkers in biology are quietly departing from him.

The illusion that organisms are well-designed doesn’t have anything to do with heat being transferred.

That idea really bothers you, doesn’t it, Brewer?

If you really want to understand the issues read Michael Denton’s Evolution: Still a Theory in Crisis (2016). But you probably don’t really want to understand the issues; you need to convince yourself that Grandma was wrong and Darwin was right.

So is there a version of the second law that relates to concepts more general than heat and temperature? It turns out the answer is yes, but I had to wait many years before I learned it. Surprisingly, it turns out this more general second law isn’t really a principle of physics, but rather a principle of reasoning. This more general version of the second law not only explains why the Clausius version is true, but gives us a tool for much more general questions, like the evolution question or Muse’s economic musings. It also appears in everyday life, and not just in situations involving heat and temperature. For example, why is darts difficult? Why can’t most men sing operatic high Cs? And why are political polls (somewhat) accurate?

Well, some people are incurably religious, and you have turned the 2nd law thermo into a religion. By the way, there is really no such thing as “somewhat accurate.”

Sure enough, Brewer discovers a “heterodox physicist E. T. Jaynes” (1922–1998) who changes his life and saves Darwin:

It’s worth thinking about whether the second law really does forbid evolution by natural selection. We don’t need to get particularly technical with the concept of entropy in order to have a go at that. All we need to ask is whether it’s plausible that a population of self-replicating organisms will tend to improve their survival and reproductive fitness over time.

The answer is yes, provided the mutation rate is sufficiently low. And if the organisms reproduce sexually, the population’s average fitness will increase even faster.8 This isn’t the Clausius version of the second law, but an example of the Jaynes one: of all the possible deaths, reproduction events, and mutations that could plausibly occur, most would lead to an increase in the average fitness of the population. The probability that a populations’ fitness would decrease is low because, for that to happen, the organisms with worse genomes would have to be reproducing more than the ones with better genomes. More.


Look, if all we are concerned with is “”plausible, all kinds of contradictory theories of the origin of life or of human beings or the human mind are merely plausible. Space aliens are at least plausible. It’s not a high bar, and it certainly isn’t science.

Put charitably, it’s not a high bar, and it certainly isn’t science.

Prediction: Natural selection (Darwin) will survive as a pop cult long after it joins the ether in physics and phlogiston in chemistry.

At this point, it is just too true to be science. Science isn’t about fixing Grandma.

See also: Conclusions: What the fossils told us in their own words

Follow UD News at Twitter!

Robert Sheldon @ 1:
Before discussing the 2nd Law, Granville Sewell should be required reading. He states the problem very clearly, very mathematically, very precisely.
Good lord, no. Sewell does not understand thermodynamics well at all, and as a result makes a hopeless mess of it. For instance, his X-entropies (the only really mathematical part of his work) are basically a mangled version of configurational entropies. They work well enough to get by in the specific case he analyzes, but if you want to apply them to anything other than diffusion through a uniform solid in the absence of gravity they pretty much fall apart. Which means they fall apart a lot. Actually, it's even worse than that: his X-entropies not only have very limited applicability, they're actively misleading in some respects. For instance, their units are all wrong, and as a result Sewell can't understand how one form of entropy could be converted to another. If you use the right formula for configurational entropy, you'll come out with units of energy per temperature (for all forms of entropy), and conversion makes much more sense. If you want to learn about thermodynamics, learn from someone who understands it. NOT Sewell. Gordon Davisson
Robert - I don't know of any Bayesians who insist it needs 2 hypotheses, nad I know quite a few Bayesians. What they do say, though, it that if you have 2 hypotheses then you need to specify likelihoods for both, which is what Dembski is arguing against. A lot of the Bayesian work I do assumes a single model (i.e. a single hypothesis), and this is not uncommon - see Andrew Gelman's philosophy of modelling, for example. You also write "Here’s John Tenant claiming that cladistics separate into “parsimony” versus “Bayesian”." but I can't see where he says that. Please show me exaclty where he claims that. I read him as saying the authors used parsimony and Bayesian, but that doesn't exclude other approaches (e.g. maximum likelihood). Bob O'H
Interesting. My ignorance must be catching. Here's John Tenant claiming that cladistics separate into "parsimony" versus "Bayesian". http://blogs.plos.org/paleocomm/2016/02/17/the-evolution-of-tyrannosaurs/ Looks like Bayesian keeps winning these debates. Wonder when it will address the 2nd law.. Robert Sheldon
BEAST2: Bayesian evolutionary analysis by sampling trees Mung
Bob O'H Well, I've done my best to get educated on the difference between likelihood and Bayesian, discovering that even Dembski's paper http://billdembski.com/documents/2005.09.Fisher_vs_Bayes.pdfsays the difference isn't worth explaining. So I suppose I will just have to endure my ignorance for a bit longer. But it seems that Dembski isn't rejecting Bayesian statistics, he is rejecting Bayesian statisticians, who insist that Bayes Theorem only applies when you have a 2nd example. The reason for Dembski's concern, of course, is that he has endured much criticism for ID not providing a "more likely" alternative to Darwinism. So in this paper he is justifying his use of Fisherian philosophy (but Bayesian analysis!) to eliminate the chance hypothesis (Darwin) in favor of ID. BTW, Dembski makes some asides about Bayesian always is a special case of Fisherian, and I would reply that Fisherian is always a special case of Bayesian. The funny thing about theories, is that they never contain themselves--the theory of everything does not include a theory of theories. The chance hypothesis assumes everything happens by chance except for the experimental protocol we are using, which better be designed carefully. But how do we know that the protocols are not themselves influencing the "exclusion region"? QM says even looking in a particular spot makes that spot special and changes the statistics. This is where I find Bayesian analysis to be so helpful--it isn't phased by QM. Robert Sheldon
Priors are actually 18thC (Bayes & Laplace), although their practical use was developed in the last century. Just as frequentist approaches were, of course. Jaynes was interested in developing objective priors, But we've given up on that because the maths doesn't work in more complicated situations, so they're less use than was hoped. Incidentally, Bill Dembski explicitly rejected Bayesian methods in ID. I hope Robert or Mung will explain the discrepancy. Bob O'H
... a “frequentist” theory of chance is identical to the Boltzmann theory of entropy, which is identical to the “heat formulation” of the 2nd law–all 19th century developments, which is identical to Shannon information. To allow “priors” or “bias” into the equation is what Bayesian, Jaynes, and ID do – all 20th century developments. Thank you E.T. Jaynes Mung
Pretty sure Denton doesn't have anything to say about the 2nd Law. Mung
Robert -
The post on the “cladistics wars” over “most parsimonious tree” was a debate over “frequentist” versus “Bayesian” (or “likelihood”) statistics.
Oh please. This is just utterly, utterly wrong. Cladists reject both Bayesian and frequentist approaches. Evolution isn't built on a “frequentist” theory of chance - one of the main early developers of the mathematical theory (R.A. Fisher) was certainly not a frequentist (as anyone who knows anything about the history of statistics knows). I think Brewer is getting at the connection to the law of large numbers, which is very different to the 2LoT arguments that are usually wheeled out against evolution. Bob O'H
Thanks, Robert Sheldon at 1. I do not say that 2nd law thermo could not be a principle of reasoning. It's more that, running self-evidently through the piece like red ink scattered on the tablecloth, is a need to make Darwinian evolution somehow work despite the 2nd law. The 2nd law - how be this - is not a form of magic. And now you say, it turns out Jaynes wasn't even TRYING to save Brewer's faith or diss Brewer's granny's faith. It will be most interesting to see what happens in pop culture as Darwinism loses influence in biology. It'll take a while to penetrate because almost every pop culture savant who uses the word "evolution" MEANS Darwinism and doesn't know any different. News
News, Before discussing the 2nd Law, Granville Sewell should be required reading. He states the problem very clearly, very mathematically, very precisely. But here I'm just defending the late ET Jaynes, who never denied the 2nd law. Rather, he said that the Darwinian assumptions of randomness that made evolution a product of "frequentist" statistics were never justified. Because Jaynes almost single-handedly revived "Bayesian" statistics, which were first used heavily in the pharma industry where they work better, cheaper and faster than "frequentist statistics". Well then, wouldn't Darwinists love Bayes and Jaynes? Absolutely not. The post on the "cladistics wars" over "most parsimonious tree" was a debate over "frequentist" versus "Bayesian" (or "likelihood") statistics. Evolution is built on a "frequentist" theory of chance, which is identical to the Boltzmann theory of entropy, which is identical to the "heat formulation" of the 2nd law--all 19th century developments. To allow "priors" or "bias" into the equation is what Bayesian, Jaynes and ID do--all 20th century developments. Poor Brewer has stumbled out of the narrow circle of 19th century physics and temporarily lost his bearing in the bright light of 21st century physics. I hope he finds his way. Robert Sheldon

Leave a Reply