Wednesday, February 10, 2010

Cornelius Hunter on Parsimony

The creationist Cornelius Hunter has a post where he attempts to cast doubt on the evolution of placental mammals because he finds the current explanations too complicated. Science values parsimony, he says. This explanation is too complicated and hence we should doubt it.

I really wonder if Hunter has ever studied geology. Explanations for the current configuration of the earth are often extraordinarily complicated. Just study the geology of Idaho. Even the basement rock alone has a description as complicated as:

"The Albion and Pioneer Ranges contain metamorphic core complexes, which expose middle-crustal metamorphic rocks in the center, beneath younger sedimentary rocks along a low-angle normal fault. The Archean and Proterozoic metamorphic rocks contain gneiss and schist that originated as part of the Archean continental crust as well as Paleoproterozoic juvenile volcanic arc rocks. They have been subsequently metamorphosed and intruded, notably in Cretaceous, Eocene and Oligocene time. They are exposed today from Neogene regional extension of the Basin and Range Province."

Why is geology so complicated? Well, largely because it is contingent: it depends on many factors, such as the composition of the early earth. And that, in turn, depended on cosmic bombardment during the formation of the solar system. And that depended on a supernova event prior to the solar system. Add meteorite strikes from outer space, catastrophic floods, and the effects of continental drift and various orogenies, and you've got a recipe for complication.

Well, biology is like that, too. But it's even worse, since in biology we have imperfect self-replication together with contingency. Pretending that this is not likely to result in complex outcomes and a convoluted evolutionary history is an absurd fantasy.


Barry said...

Nothing is more parsimonious (or less scientific) than the creationist explanation: "It's magic!"

Gareth McCaughan said...

Of course the "it's magic" explanation is rubbish precisely because in reality it *isn't* at all parsimonious. "Magic" can do anything at all; therefore, having said "it's magic" you have given no information at all about what happened; to complete the explanation you need to say just what the magic *did*, and doing that is just as un-parsimonious as simply saying "The following things happened: ..." without bothering with the "explanation".

In this respect, "God did it" is exactly the same as "it's magic": neither actually tells you anything. (Except in so far as you have a theory about what sorts of things God does, or what the limits of magic are. If your notions of God or magic have enough predictive detail in them, then you may have a decent explanation after all -- for instance, you might define "God" as "something that ensures that such-and-such laws of nature are obeyed", and fill in the details with whatever the physicists are currently saying. At that point, your theory is exactly the same as conventional physics, the only difference being that instead of saying "These laws are obeyed" you say "God ensures that these laws are obeyed", and then parsimony bites you: the word "God" is performing no real function in your explanation.)

Vladimir Levin said...

"Everything should be made as simple as possible, but no simpler"

cody said...

Reminds me of the alleged Einstein quote, "Everything should be made as simple as possible, but no simpler." (Though looking it up it appears that is just a reinterpretation of a more precise statement he made.)

My roommate has been arguing about this lately, apparently he has always been uncomfortable with the notion of Occam's razor, and it seems this is because he has never thought of it as requiring "all else being equal," but rather as just "the simplest explanation is most likely the correct explanation." Maybe Mr. Hunter is succumbing to the same problem?

Do the creationists have any concept of predictive power?

Miranda said...

The question of "why is the universe this way?" leads to a very parsimonious answer from the atheist, but a very long one from the Creationist.

Filipe Calvario (from Brazil) said...

Even if the explanation of something based on the will of a deity were the most parsimonious, it could not be applicable, for in such a case, it should be applicable in every case! How do you explain apples falling from trees? Laws of physics? No, because God has wished so - it's more parsimonious.

God's Will has being rejected as a scientific explanation of why things have happened for a long time. Science hadn't moved to explain the existence of species, or the how the Universe has evolved to its current state till some century ago, where the big conflict arose. Of course, one can abandon Science in this point, and say: "this does not belong to its realms". However, one cannot expect Science to adopt such a position.

SmartLX said...

He says the explanation is too complicated...compared to what? Is there even a creationist explanation of the origin of this feature (i.e. how God did it) to which we can compare it? Because if not, it's the most parsimonious available explanation by default.

carlsonjok said...

Hunter has sordid history when it comes to conclusions about placental mammals.

RBH said...

Parsimony is a criterion of last resort, deployed only when two hypotheses cannot be distinguished on any other legitimate grounds (accounting for the evidence, consilience with other disciplines and theories, fruitfulness, etc.), not the first test to be applied.

Ty said...

RBH, I love that comment. That goes into my box of useful things to say in response to bad arguments.

Miranda said...

Quotes (or sources) of interest to anyone wanting to follow up:

Kitching, I.J., et al., "Cladistics: The Theory and Practice of Parsimony Analysis," Oxford University Press: Oxford UK, Second edition, 1998. [570.12]


"This phenomenon the separate and independent acquisition of similar features-is known as 'convergence'.
The claim of convergence is the most important objection to the bird-dinosaur link. Convergence is an
extremely difficult problem. When confronted by two superficially similar organisms, it is not al ways easy to
spot whether their similarity is a result of shared common ancestry, or an adaptive response to similar
pressures in otherwise unrelated animals. Ornithologists are very sensitive to this problem, because many
otherwise distantly related birds look deceptively similar because of a common response to the pressures of
flight. In a sense, claims of convergence are unanswerable, as one can always make a case, based on
adaptation, that any feature thought indicative of common ancestry really represents a common response to
similar adaptive pressures in unrelated species. Cladistics has a way of addressing the problem: the
invocation of the principle of parsimony. ... Of course, there is no law that says that evolution is always
parsimonious." (Gee H., "In Search of Deep Time: Beyond the Fossil Record to a New History of Life," The
Free Press: New York, 1999, pp.184-185)

""But make no mistake about the power of scientific materialism. It presents the human mind with an
alternative mythology that until now has always, point for point in zones of conflict, defeated traditional
religion. Its narrative form is the epic: the evolution of the universe from the big bang of fifteen billion years
ago through the origin of the elements and celestial bodies to the beginnings of life on earth. The
evolutionary epic is mythology in the sense that the laws it adduces here and now are believed but can
never be definitely proved to form a cause-and-effect continuum from physics to the social sciences, from
this world to all other worlds in the visible universe, and backward through time to the beginning of the
universe. Every part of existence is considered to be obedient to physical laws requiring no external control.
The scientist's devotion to parsimony in explanation excludes the divine spirit and other extraneous agents.
Most importantly, we have come to the crucial stage in the history of biology when religion itself is subject
to the explanations of the natural sciences. As I have tried to show, sociobiology can account for the very
origin of mythology by the principle of natural selection acting on the genetically evolving material structure
of the human brain." (Wilson, E.O., "On Human Nature," [1978], Penguin: London, 2001, reprint, p.184)

Gareth McCaughan said...

RBH and Ty: it's not quite that simple. Consider your favourite well-evidenced physical theory; quantum electrodynamics, let's say. There's a wealth of experimental evidence in its favour. But of course that doesn't mean that every time anyone does an experiment whose result is predictable in principle via QED the results of the experiment match *exactly*; there's experimental error, and QED (like everything else) is only an approximation anyway. (A stunningly good approximation, to be sure.)

So now consider two theories. Theory 1 is QED as it stands. Theory 2 is QED augmented with a bunch of special-case rules saying "if you do exactly such-and-such an experiment, you get exactly such-and-such deviation from the result predicted by QED". Theory 2, of course, is concocted with perfect hindsight using the results of some experiments that have actually done.

Result: theory 2 "predicts" the available observations better than theory 1. It predicts future observations just as well as theory 1 does, and past observations much better.

None the less, theory 1 is much better than theory 2. I think the reason basically comes down to its big advantage in parsimony. (Which becomes bigger when both theories are embedded in the rest of physics, because the theory-1 stuff "compresses better" on account of resembling the rest of physics in various ways.)

I suggest that the right way to think about this is Bayesian. The experimental evidence provides you with a load of likelihood factors; parsimony is encoded into your prior. A small enough difference in evidence can be outweighed by a large enough difference in parsimony.

RBH said...

Miranda wrote

Cladistics has a way of addressing the problem: the
invocation of the principle of parsimony. ... Of course, there is no law that says that evolution is always

"Parsimony" in cladistics has a somewhat different meaning than the use of the term with reference to competing hypotheses. It can be easily measured by counting; in fact, a computer program can do it.

Gareth commented

RBH and Ty: it's not quite that simple.

Of course it isn't. These are blog comments, not definitive essays. The message of my comment is perhaps weakened a little by your example, but by no means vitiated. Where it counts, predicting novel observations, QED and QED+ are equivalent, and hence parsimony is an appropriate criterion to invoke, particularly since one can also identify QED+ as an instance of special pleading.

RBH said...

Actually, I can put that more simply: Curve fitting doesn't count in comparing theories.

Miranda said...

One more parsimony quote, if you like:

28/04/01 "But Nabokov did not pass over the subject superficially. He gave selection and mimicry a great
deal of careful thought, took issue with the species concepts of the leading evolutionary biologists of the
day (his Harvard contemporaries Ernst Mayr and Theodosius Dobzhansky), and wrote provocatively on
mimicry. It may even be that some of his objections foreshadowed current debate among different schools
of selectionists. But Nabokov withheld his assent for a universe fully ruled by chance, and this may have
permitted his temperamental preference for mystery and convolution to overwhelm the parsimony of natural
selection in wholly accounting for the miracles made manifest in butterflies." (Boyd B. & Pyle R.M., ed.,
"Nabokov's Butterflies," Nabokov D., Transl., Allen Lane / Penguin Press: London, 2000, p.65)

abb3w said...

Gareth McCaughan: I suggest that the right way to think about this is Bayesian.

EG: (doi:10.1109/18.825807)

The trick is that under the mathematically rigorous (proven) sense, parsimony is applied on a two-part combination; loosely, the complexity of the model on the one hand, and the complexity of the inputs that must be assumed for the model to fully reconstruct the experiential data.

So, for example, "It's magic" is very parsimonious in model, but exceedingly non-parsimonious in inputs.

RBH: "Parsimony" in cladistics has a somewhat different meaning than the use of the term with reference to competing hypotheses. It can be easily measured by counting; in fact, a computer program can do it.

Actually, the two are very closely related; in the mathematically formal expression of parsimony, the "model" is a program for a Universal Church-Turing Automaton, and "inputs" that data tape for the program.

What is not easy (in fact, intractible) is creation of an optimal program/input combination for a non-trivial data output.

Tyler DiPietro said...

Taking our cue from Solomonoff, we could point out that "it's magic" provides no information about future experimental data. Encoding all experimental procedures and data as increasingly long finite initial segments of binary strings, you're likelihood of predicting the next digit(s) is never improved. Thus the model is not parsimonious.

diogenes said...

I have recently heard an absurd "origin theory" about the president of the United States-- that the US President is neither black nor white, but BOTH AT THE SAME TIME!


The US population is 1.7% multiracial, so the odds of this dogmatic atheist "Half-Black President" origin theory are 58.8 to 1!

But it gets worse! The atheists' "Obama origin theory" furthermore claims that he was born in Hawaii in 1959!

Now Hawaii is 43rd among US states in population, and the odds against the President being born in Hawaii are 239 to 1!

And, given that the president must be at least 35 and less than say 90 or so, the odds against him being born in 1959 are 65 to 1!

So the combined odds against this dogmatic atheist "origin story" are now 916,375 to 1!

Clearly, the only reason why anyone would believe this "Obama is President" story is a prior metaphysical precommitment to atheism.

But it gets worse! Their "origin theory" furthermore says that he grew up in INDONESIA!

Epicycles upon Epicycles!

Most Americans don't even have a passport! The odds against this unparsimonious origin theory are now MILLIONS TO ONE.

Another falsified prediction of the dogmatic "Obama is President" though police!

Religion drives the "Obama is President" theory, and it matters.

I'm sure glad Mr. Hunter has a less parsimonious theory about the origin of our president!

You do, right? The current theory hurts my brain. So please keep yours simple.

Such as for example, that all the relevant authorities are lying to us.

...And have faked their incomparably vast number of successful predictions.

Don't pay any attention to the VAST NUMBER OF SUCCESSFUL PREDICTIONS of the atheist "Obama is President" theory, like that announcement put in Honolulu newspapers in 1959 when he was born.

Why should a HUGE VAST NUMBER OF SUCCESSFUL PREDICTIONS matter!? The current theory hurts my brain.

Luckily, Mr. Hunter has a more parsimonious origin theory for us, right? One that makes testable predictions that are both MORE SPECIFIC and different from the accepted theory...

...And that is surely not based on countless examples of ad hoc special pleading going back to 1993... oh no, your side would never do ad hoc special pleading over and over and over and over every time their predictions are falsified again and again!

Oh, wait, I forgot. Your theory has never once, not once, made a single successful prediction different from the current theory...

The surfeit of falsified predictions of Intelligent Design, and the total complete absence of even a single distinct successful prediction after 14 years of pretending to do science, drive Intelligent Design and it matters.

If you want some more laughs at Intelligent Design Mysticism, see my new blog post at lampofdiogenes at WordPress.

Cathy S said...

abb3w: I don't get when people say about thinking the Bayesian way...what does it mean? [Even so, I wouldn't use these tools too often!]